Hyperspectral imaging IED and explosives reconnaissance system

Final Report Summary - HYPERION (Hyperspectral imaging IED and explosives reconnaissance system)

Executive Summary:
The HYPERION system concept is a novel way to perform a forensic analysis after an explosion or to check suspected unexploded Improvised Explosive Devices (IEDs) in a civilian area. A key aspect of the research concept is that it will allow an efficient forensic investigation of the crime scene at safer stand-off distances minimising any unwanted changes of the post-blast area and strengthen the security of the acting police and security services. The concept in the form of a fast forensic analysis could, as compared to the best practices of the forensic procedures today, be a significant improvement since the forensic evidence will be provided on-site to the police. A rapid response from the forensic investigation to the police is a necessity in order to increase the chance of finding the responsible antagonists for the attack thus preventing possible future attacks.

The objective with the HYPERION project was to develop and test a system concept for the on-site forensic analysis after an explosion. The forensic tools and procedures would in majority be used at safer stand-off detection distances. This include tools which can help with the identification of unexploded IEDs. The on-site data provided by the system concept will be the type and amount of explosive used in the attack, the point of origin of the detonation and an assessment of the type of IED. The post-blast scene will be mapped using 3D-registration where the positions in the map that have been analysed in detail will be marked. The quality assured data will be on-site electronically documented and sent to the police timely at the crime scene.

In HYPERION, imaging Infrared detection and imaging Raman detection for stand-off detection of post-blast explosive traces have been used as techniques. For the 3D data acquisition three main components have been used. These are the Kinect camera, a stereo camera and an IMU (Inertial Mesaurement Unit) module. The 3D imaging instruments utilized are capable of generating a depth map of the imaged scene. Furthermore, in the project a software has been developed where the forensic data gathered at a post-blast scene can be used to estimate the charge size of the bomb. This is based on data such as the number and sizes of broken windows due to the shock wave of the detonation, the crater size generated or the distance of particular debris thrown away due to the blast. These data can be entered into the software and the models used in the calculations can make an inverse explosion analysis and thus predict the charge size of the explosive, finally provided as a weight interval. Another part of the project has studied how to make novel sampling with the following purpose to analyse the samples in a conventional forensic laboratory for assessing and identifying the explosive used, i.e. data that can be used in the court. This project part does also support the imaging stand-off detection techniques. Different sampling methods have been studied where the recovery and efficiency of the methods have been thoroughly examined. The influence of packaging method for samples has also been studied where degradation aspects such as temperature and light have been taken into consideration. The analysis of samples have been performed by different laboratory techniques based on chromatography and mass spectrometry. In order to use and demonstrate some of the tools in the field, ruggedisation of some components was also addressed.

Although the outcome of the HYPERION system points to a viable concept for quick forensic results using stand-off tools after a bomb attack, still more work are of importance in order to further enhance capabilities such as lower detection limits, including a larger range of explosives that can be detected, achieving better resolution for 3D registration and creating furthermore accurate models for post-blast calculations for the assessment on the used charge weight.

Project Context and Objectives:
One key aspect in the fight against terror actions with homemade explosives is to prevent the use of ordinary chemicals from being used as precursors to explosives. Home-made explosives (HMEs) are easy to make from readily available materials used for legitimate purposes in everyday life. This availability attracts terrorists and criminals to manufacture and use HMEs since military and commercial explosives are more difficult to come by. The uncontrolled information disseminated via the web and the simultaneous presence of trained persons enables large populations to prepare and build IEDs containing HMEs. This can be realized at home using products that can be bought without any specific authorization (e.g. ammonium nitrate based chemicals and other substances) and using simple equipment present in ordinary kitchens.

HYPERION is a project in which a system concept has been developed and tested for the on-site forensic analysis after an explosion (stand-off post-blast forensics). The forensic tools and procedures used are in majority at safe stand-off detection distances. This include tools which can help with the identification of unexploded IEDs. The on-site data provided by the HYPERION system will be the type and amount of explosive used in the attack, the point of origin of the detonation and an assessment of the type of IED. The crime (post-blast) scene will be mapped using 3D-registration and the positions in the map that have been analyzed in detail will be marked. The data will be on-site electronically documented and can be transferred to the end user timely at the crime scene. In summary, the tools of the HYPERION system will be used on-site and give a timely and quick response to police forces in order for realizing an appropriate initiation of the continued investigation. All of the actions need to be documented in a secure and tamper-proof way.

At a crime scene, due to the detonation of an IED or the possible presence of an unexploded IED, first responders and police, ambulance, fire brigade and forensic teams are units that are present. One of the first measures taken at a crime scene due to a detonation of an IED are the initiation of rescue actions. The responsible person is the rescue leader that has to start rescue persons and at the same time fence in and establish a crime scene safety zone.

The HYPERION tools will be used for investigating the type and amount of explosives used in the attack. This is performed using imaging Infrared (IR) and Raman detection techniques. The 3D registration of the crime scene provides a documentation of the area and the developed inverse explosion analysis (IEA) tool can be used for assessing the charge weight. The data acquired helps in the assessment of the point of origin for the detonation and the assessment on the type of IED used. The forensic data will be tamper-proof documented and digitally stored on a computer.

After the crime scene area has been secured, the laboratory forensic sampling and analysis can be started. In HYPERION, sampling protocols are used that takes into consideration new sampling materials, stability of samples, optimized packaging procedures and the exclusion of any contamination of the samples used as evidence. The purposes are to further strengthen the forensic evidence obtained from the stand-off analysis and also to support the results obtained from the new forensic technology in order to show that the data are reliable.

The crime scene is finally left to the responsibility of the rescue leader for clearing up the area.

Figure 1. The HYPERION concept. (see separately attached file for all figures)

A rapid response from the forensic investigation to the police is an absolute necessity in order to increase the chance of finding the perpetrators of the attack or for the possibility for the police to be proactive in the case of a series attack such as the London Underground (2005) or Madrid train bombings (2004). For the police, the first 24 hours is of importance for a successful outcome of the crime investigation. This means that the forensic investigation and analysis of the post-blast scene of the attack has to be carried out quickly. In addition, it is of importance that the analysis data of the crime scene is of a sufficiently high quality so it can be used as evidence in a trial.

The information the police authorities need to know initially for facilitating the investigation is the type and amount of explosive that has been used in the attack. The type of explosive will reveal what kind of threat the authorities are facing and will give a hint about where the explosives could have been obtained. Explosives that are of the home-made type require the utilization of a clandestine laboratory, a “bomb factory”, for the production. This would allow the police the opportunity to use intelligence and other observations for the localization of the bomb factory that may finally lead them in the direction of the perpetrators of the attack. The planning and financing, obtaining equipment and the preparation and production of the home-made explosives are normally events that will take from weeks to months for terrorists leaving plenty of various traces for their activities. The amount of explosive will significantly influence the timeline of the terrorist events since a larger amount can require longer time to produce. In the possible utilization of military explosives or civilian explosives the bomb factory may be less advanced, however still needed for building the IED. The information will nonetheless be important for the police since certain areas or pathways can be more important to investigate than others.

The point of origin for the detonation is needed primarily for assessing the charge size of the bomb and type of IED. It is important for the crime investigation to assess if the IED is of e.g. VBIED (Vehicle Borne IED), PBIED (Person Borne IED) or LBIED (Left Behind IED) types. The assessment on the type and the position of the IED and also the damage effects the IED was intended for would from the perspective of building up the case and the evidence chain be beneficial for the trial.

The crime scene area also needs to be well documented using ordinary highly resolved 2D photographs but most important using a 3D registration. In this 3D registration the hot-spots that have been analyzed using the forensic stand-off detection tools as well as the areas that have been sampled for the laboratory forensic analysis can be marked. The 3D registration contributes to the forensic investigation since it is possible to observe the crime scene from many angles facilitating the investigation and evidence presentation in the trial. The 3D crime scene registration can also be used to register the typical damage patterns in the direct vicinity of the post-blast scene. On-site electronic documentation of forensic data is important in order to preserve the chain of custody and also to be clearly presented to the police.

The HYPERION project started 2012 and ended technically by the demonstration in September 2015. The system is a research prototype where the concept of stand-off post-blast forensics have been evaluated. The HYPERION tool points at many advantages if realised such as a system that can be used with no contamination of the post-blast scene and a guidance for where the later manual samplings can be done since the stand-off technique can locate explosive traces. Furthermore, the compiled set of forensic data presented in one software gives the police an easier overview of the scene and what has been found. In addition, the HYPERION system can be deployed and provide many important results within a few hours after arrival at the post-blast scene.
The main objectives are listed as below:
- Collection and evaluation of end user requirements via workshops and meetings.
- Identification on threat substances and substances of importance for the project.
- Characterization of home-made explosives and studies on the dispersion of post blast explosives.
- Analyses of real post-blast samples using conventional techniques and stand-off detection tools.
- Planning for the final validation and development of realistic scenarios.
- Identification of the 3D-registration tools and the methods and apparatus to use.
- Compilation and development of tools for inverse explosion analysis where a tool for quick estimations is included. Specification on information needed to collect for the on-site software.
- Analyses of explosives traces and post-blast samples using stand-off imaging Raman and IR spectroscopy as well as commercially available handheld devices.
- Compilation of a report on present sampling techniques and initiation on the development of sampling procedures.
- Compilation of a specification and architectural design on the software system for information management and data fusion.
- Initiation and specification on the hardware ruggedisation design. Drawings and thermal simulations for the imaging Raman apparatus.
- Planning and preparation of the final project demonstration and the type of explosive to be used.
- Demonstration of the imaging Raman equipment for some European end users.
- Testing and validation of subsystem tools such as stand-off detection devices, sampling protocols, inverse explosion analysis tool and information management tool.
- Testing and validation of the HYPERION system including interface between subsystems.
- Performance demonstration of the HYPERION system concept for end users and stakeholders.
- Final symposium for end users and stakeholders.

Project Results:
WP2 Dissemination:
The objective of WP 2 is to disseminate the results and findings of the HYPERION project to the partners, to interested Stakeholders, including end-users, the European Commission, Governments, legislative and inspection bodies, law enforcement agencies and the public. It is also for the intentions to encourage a market for the HYPERION system components that can be useful and applicable for the end-users.
A public website was set-up and delivered Month 3 in the project. The website has been continuously updated during the project. (Website: www.hyperion-fp7.eu.)
In February 2013, a workshop in Madrid was held and mainly organized via the HYPERION project. There were four EU FP7 funded projects represented (HYPERION, EMPHASIS, ENCOUNTER and FORLAB). The workshop gathered many end-users, project partners and researchers. A final HYPERION demonstration and symposium for end-users and stakeholders were performed in September 2015 where the developed tools and capabilities were shown.
Research results from the activities within the HYPERION project were published within a number of scientific publications and conference talks as well as in posters and information brochures distributed at Fraunhofer exhibition stands. A press release covering activities in the final HYPERION measurement campaign was published and HYPERION was mentioned within a documentary film produced by Deutsche Welle hat covered Fraunhofer IAFs activity in stand-off detection of explosives. The HYPERION system concept was also documented in a story by Discovery channel and aired in November 2015.
WP3 Crime scene requirements analysis and forensic strategy:
With the information gathered within the FINEX (Forensic International Network for Explosives Investigation) group and the detection community, supplemented by the results of the Madrid workshop organised within the HYPERION project a comprehensive overview on the needs and requirements in the forensic field could be established. This information has been reported within the project.
The generated information on the defined threat substances and precursors were further supplemented by results from various smaller blast tests and measurements on detonation velocity to confirm the results of the literature research. The identified threat substances and precursors were additionally characterised with IR- and Raman Spectroscopy as well as X-Ray diffraction where possible. With this generated information a report was made. Some detonations have been performed with IEDs of the home-made type. One detonation was performed with the bomb being placed in a garbage bin in the field without any further objects at the scene. Another detonation was performed using the same type of bomb but the inclusion of a car close to the detonation point as well as the inclusion of a rack of plates close to the detonation point. The plates were of different type of materials such as ceramics, fabrics, plastic and metal. The purposes of the presence of the plates was to get samples with post-blast explosives traces on in order to check the concentration of explosives and the possibility for stand-off detection on the plates. After the detonations, samples were collected. The samples were both fragments and ground samples. Also sampling on the car was performed; inside and outside surfaces. The samples were analysed using conventional analytical techniques such as gas chromatography or liquid chromatography coupled to UV detector or mass spectrometer.

Figure 2.

Figure 3.

A conclusion that can be made from the post-blast studies is that measurable concentrations of post-blast explosive traces are available at the scene both on fragments and on ground. Furthermore, the study indicates where to start make stand-off detection at a post-blast area in order to have the largest possibility to detect explosives. Also, the inclusion of objects at the scene will largely influence the distribution of post-blast explosives in the area.
With the contacts to EU-Groups (NDE and EU-Matrix Group) and the forensic community (FINEX) established the involved End user defined three applicable realistic scenarios for the HYPERION device. These scenarios were also matched with the demands from the community of bomb disposal specialists and explosive investigators. The first scenario is a post-blast vehicle borne improvised explosives device, as it will allow testing the HYPERION device on many different difficulties, typically encountered in real cases. For example this setup will generate the possibility to evaluate the stand-off detection methods on a crime scene which involves a big area. Further the crater can vary quite much depending on the positioning of the charge. At last the matrix in which the explosive substances have to be found tends to be diverse and quite near to real cases. The second scenario is a search in a clandestine laboratory. In this case a multitude of different substances have to be differentiated and classified. The whole setup is a demanding task for the scene reporting routines, as evidence has to be located and correctly labelled. This work has typically to be done in a very cramped environment with little space and the danger of sensitive materials present. The third scenario defined is the discovery of a second device at a scene of a successful attack. The successful attack has been described by one of the end users in the project after their real case experiences. The detection of the second device would be in addition to routine work of the bomb disposal / forensic team. The described difficulties of a post blast scene in the middle of a big city, with the typical hazards of large scale destruction in an urbanized area also have to be taken into account.
On the 24/25th of June 2015 the BKA hosted an End-user workshop to get direct input on how the Stand-off detection system of the HYPERION Project could be used in a realistic environment. The participants were invited from the law enforcement, forensics and military community. At the 25th of June around 30 specialists in the different fields of expertise joined the meeting. The Workshop was structured in three parts: An introduction to HYPERION at first, followed by a practical demonstration and in the end a feedback session and discussion with the attending end-users.
The first part was an introduction to the aim of the HYPERION Project. The main forensic goals were explained and which partners are responsible for the different work packages. It was stressed that one of the commitments of the team was that there will be practical results, which could be used by the security community within a short timeframe after the project. This was appreciated. On the other hand it was made clear that this would only be possible if there will be enough feedback from the end-users to steer this research in the right direction. The developments achieved so far in the project were summarized shortly. After this introduction, which was mainly through presentations the practical session started. These hands-on experiments were the main aim of the event.
The practical session were a demonstration of the stand-off detection abilities of the Raman Device, developed by the FOI. Three different setups were prepared to demonstrate the versatility of the Stand-off detection device. The experiments were conducted at the 100m shooting range of the Forensic Science Institute of the BKA therefore, all 30 participants could not only follow the three prepared detection tests but could also inspect them personally.

Figure 4.

The following three Setups were prepared. The explosives samples were located on a clean metal plate in drill holes with different diameter radius. This was to demonstrate the minimal amount needed to get a positive detection result. The smallest hole in the plate was just visible with unaided eyesight. The next sample was set on a painted Car door, which was previously used for shooting tests. This was to show that different (also realistic and dirty) back-grounds are possible.

Figure 5.

The last samples were positioned at the maximum distance of the system, approx. 25m, on a textile bag and two transparent vials.

Figure 6.

This experiment was to show the maximum effectiveness and the potential of analysing through a closed container and on day to day items.
All visitors could freely inspect the test sites and comment on the preparations. They were also free to operate the system themselves to get a feeling how it would be operated and if it would be easy enough for the typical operator in the field.
The detection experiments went smoothly. All samples in the setups could be detected within a short timeframe. After the successful trials the participants were invited to talk to the partners and inspect the equipment in depth. There was a lively discussion and a fruitful exchange of information.

Figure 7.

After the successful practical part all participants joined again in the lecture room for the feedback session. The event and the practical experiments were discussed between the organizers and the end-users. A first feedback from all end-users was, that they were happy to see an EU-research project working and in particular a stand-off detection device for explosives.
The following further comments were made to steer the research the direction of the end-users:
- A detection device used in real life has to be smaller,
o it has to be transportable by one man (police)
o could be that big but then has to be very rugged, especially air transportability would be important (military)
- There has to be the possibility to add substances to the databank of the device
- Three to five meters distance as detection range would be enough. More is always welcome
o The military end-user like to have as much distance as possible
- It would be good if there was the possibility to put the device on a robot
- Measuring a point in contrast to an area would also be enough
- There should be a good reliability of the identification of the explosives
- The device has to be decontaminated, there should be a practical routine for this
- The measurements should be short in time
- If there would be a possibility of a fast screening this would be appreciated
- Safety of the measurement is paramount
Meaning: no initiation possibility by the laser measurements
The event was well received from the attending end users as well as from the project partners. All regarded this kind of demonstration as beneficial for all participants. It is envisaged to use the feedback in the HYPERION Project and coming similar projects in the area. Also the experts expressed their willingness to join an event like this again in the future.
WP4 3D-registration of the crime scene:
3D Reconstruction System Review

The HYPERION 3D reconstruction system is depicted in Figure 8. Three main components form the acquisition part. These are Kinect camera, stereo camera and Inertial Measurement Unit (IMU) module. The first two are the imaging instruments and the last one provides information about the orientation of the system during data acquisition. Both imaging instruments are capable of generating a depth map of the imaged scene. While Kinect camera provides the depth information directly, the same information has to be computed from stereo camera output. Availability of depth information makes it possible to use several approaches for 3D reconstruction. We also decided to add an inertial sensor to the imaging instruments. IMU module provides the rotation angles of the data collection system with respect to world coordinate axes at each frame acquisition. Using these values as constraints for visual odometry (the problem of finding the location and orientation of the camera at each frame) enables us to compute the pose of the camera more accurately, thus to perform better reconstructions. The data collected by the acquisition subsystem is fed to reconstruction which generates the desired output.

Figure 8.

Hardware Components of the Prototype System
We have built the prototype data collection system so that it can be carried by one person only. The prototype system can be seen in Figure 9.

Figure 9.

The components of the data collection system are explained below:
• The main components of the data collection system can be seen in the right hand of the user. The stereo camera is on the top (yellow), Kinect camera is at the bottom. The IMU unit is located on top of the stereo camera. They are assembled through a custom made handle on which the components are screwed in tight so that their relative location and orientation do not change during acquisition.
• The bag strapped on the chest of the user contains a portable computer. Both cameras are commercial-of-the-shelf (COTS) products. Both cameras are bundled with acquisition software written for Windows. We used the portable computer not only as the acquisition interface, but also as the storage medium. Since the regular hard drives are slow for saving high resolution images at a relatively fast frame rate, we used a solid-state hard drive as the storage medium.
• The carrier case of the portable computer is also custom made. It enables the user to open the screen and use the computer on the run. It can be folded and strapped on the chest during acquisition (as seen in the picture), and it has all the necessary gateways for the cables of both cameras and IMU unit.
• An external battery was used to power the cameras. In order to make the system last longer, a regulator circuit was designed and produced on a PCB. It was integrated with the battery and its case.
• The user has to see the acquired images during the acquisition so that he/she can ensure that the relevant parts of the scene are in the field-of-view at all times. While it is possible to accomplish this by observing the screen of the portable computer, this is rather uncomfortable for the user. Sunlight during the day and its associated glare considerably reduce the visibility of the scene. Moreover, walking around the scene while the head is lowered toward the screen is not safe, since the crime scene is usually littered with debris. Thus, we decided to use a display unit attached to a goggle. The mechanical interface needed to attach the unit to the goggle was designed and produced. The whole screen of the computer is projected on the display unit, which enables the system prototype easier to handle and operate.
Software Components of the Prototype System
In order to use the prototype system on the field, two software modules have been developed: Acquisition and reconstruction. Graphical user interfaces of those modules can be seen in
Figures 10, 11 and 12.

The task of the acquisition module is to collect the data and record on the solid-state hard drive. We utilize COTS products as data collection equipment with their APIs that run on a PC. Thus, the personal computer in the prototype system acts as the acquisition and storage unit. The acquisition module GUI are written in MATLAB and APIs of the cameras and IMU are properly called from there. Acquisition module user interface (Figure 10) allows the user to choose several options such as camera type (stereo or Kinect), choose navigational (IMU) data, output folder, and the file format of the acquired images. This is the interface that is projected on the eye display unit during acquisition, so that the user is able to see what is in the field of view during data acquisition through the two image fields on the top of the GUI window.
The reconstruction module essentially implements all the algorithms used/designed in the course of the project. The main ones are depth from stereo (in the case of stereo images are acquired), visual odometry, and 3D reconstruction. All methods are implemented in MATLAB and called from GUI designed and implemented in MATLAB. Some of the subroutines with high computational complexity are written in C and called from MATLAB through mex-function utility. There are two modes of the reconstruction module user interface which allows the user to perform two separate operations necessary for the 3D reconstruction. The first mode is the visual odometry (Figure 11). The second mode is the 3D reconstruction itself (Figure 12). The user interface allows the user to choose several options for visual odometry and 3D reconstruction such as input data, visual odometry parameters, stereo depth map construction parameters, and 3D reconstruction parameters.

Figure 10.

Figure 11.

Figure 12.

Algorithm Design – Depth from Stereo
The prototype system uses two imaging units with depth map generating capability. Kinect camera contains a depth sensor which utilizes the concept of structured light. It consists of an IR projector and a monochrome IR camera. Using a triangulation technique, it directly outputs the depth map. In contrast, stereo camera provides two images of the same scene acquired from slightly different viewpoints. Then, a stereo reconstruction algorithm is employed to compute the disparity map of the scene along with certain smoothness requirements. The literature is quite rich in terms of stereo reconstruction algorithms. We have chosen to implement a method proposed by Geiger et al (A. Geiger, M. Roser and R.Urtasun “Efficient Large-Scale Stereo Matching”.), which approaches the disparity computation problem from a Bayesian perspective, thus providing robust outcomes. The method starts with finding support points (the correspondences that we are sure about) in both acquired images. Triangulation, matching, left/right consistency and post processing steps follow and produce the final depth map. Both imaging systems has their stong and weak points in terms of depth map generating capability. While Kinect fails for transparent, specular, flat dark surfaces, the stereo depth map is incomplete for low texture regions. The depth map of the scene is meant to be an intermediate product towards 3D reconstruction. Sample depth map results can be seen in Figure 13.

Figure 13.

Algorithm Design – Visual Odometry
Once the depth map of a frame (we use frame in this context) as the image pair acquired at the same instance by the stereo camera) is computed, the 3D coordinate of the point imaged at each pixel with a valid depth value can be evaluated using the camera calibration values (focal length etc.) and associated disparity value. These 3D coordinates will be represented in the frame of reference of the camera defined by the location and orientation of it at the acquisition of that particular frame. 3D reconstruction requires the repetition of this process for every frame and somehow combine all that information together. To achieve this, one has to know the orientation and the location of the camera at each frame acquisition, so that all the information can be represented in the same frame of reference. The process of extracting this information from acquired images is called Visual Odometry (VO). We utilized two VO approaches from the literature which provides stable and satisfactory results (A. Geiger, J. Ziegleri C. Stiller, “StereoScan: Dense 3d Reconstruction in Real-time), (P. F. Alcantarilla, C. Beall and F. Dellaert, “Large-Scale Dense 3D Reconstruction from Stereo Imagery”). VO basically finds corresponding points between consecutive frames so that it can compute the incremental translation and rotation of the camera. To achieve this, stable features has be detected and stable descriptors has to be used to find correspondences. Blob and corner detectors are used for detecting the features and Sobel filter based sparse descriptors are used to find correspondences. After all consecutive frames are processed, the camera path is calculated by accumulating the incremental paths. We also supported the VO solution by providing the rotational parameters from IMU module. A sample result of VO can be seen in Figure 14.

Figure 14.

Algorithm Design – 3D Reconstruction
As mentioned in the previous section, both imaging instruments we utilize are capable of generating a depth map of the imaged scene. Availability of depth information makes it possible to use several approaches for 3D reconstruction.
The straightforward way is to project each pixel to 3D world using the associated depth value and the calibration parameters of the stereo camera. This can be done for every pixel with a valid depth value in every acquired frame. Provided that the pose of the camera (location and orientation) is known correctly at each frame, this should provide the desired 3D reconstruction. The result is a so-called point cloud, a collection 3D points with their associated color information. If desired, a triangular mesh can be constructed from such collections. The problem with this approach is the storage requirement. The utilized stereo camera acquires images of size 1280x960. Depth from stereo algorithm, in general, manages to compute valid depth values for roughly 0.66 of the image pixels in a frame. One has to record 6 values (3 coordinates, 3 RGB values for texture overlay) per pixel with a valid depth value after projecting it to 3D. The stereo camera operates up to 15 frames/sec. For the reconstruction of a modest scene, such as 10 minutes of image acquisition, the amount of generated data quickly runs out of hand. Clearly, since consecutive frames have significant overlap in terms of area they image, the amount of redundancy in the reconstruction described above is significant. We utilized this redundancy when designing our 3D reconstruction method.
KinectFusion is an alternative approach for 3D reconstruction (R. A. Newcombe, S. Izadi, O. Hilliges, D. Molyneaux, D. Kim, A. J. Davison, P. Kohli, J. Shotton, S. Hodges, A. Fitzgibbon, “KinectFusion: Real-Time Dense Surface Mapping and Tracking”). Instead of directly forming a point cloud, it attempts to recover the surface (the face of the objects that we see) in the scene. In order to accomplish this, it essentially introduces an additional dimension to the mathematical definition of the problem. It defines a real valued function of 3 variables (x, y, and z coordinates) whose zero level set defines the surface in question. This is called signed distance function (SDF). The function takes positive values for locations in front of the surface and negative values that are occluded by the surface. For each depth acquisition, it updates the already computed SDF in a weighted average manner. Note that, the locations that need to be updated for each frame acquisition are not just the ones that constitute the surface. All the 3D locations that happens to be in the field-of-view of the camera have to be updated. This is a very memory intense operation. In order to achieve this task, KinectFusion algorithm defines a cubical area in front of the camera at the beginning of the acquisition, and divides that cubical area into a regular grid. SDF is defined and updated on that grid whose extent is finite. Obviously, the extent and the resolution of the area that can be reconstructed depend on the size of the cube and the density of the grid, respectively. In addition to that, KinectFusion uses a so-called “Truncated Signed Distance Function” (TSDF) which is defined in a neighborhood of the zero level set of SDF. This significantly reduces the number of grid points that need to be updated in each iteration. We utilize this “reconstruction on a predetermined grid” approach in our method.
Our 3D reconstruction method is equally applicable to stereo or Kinect cameras, while depth map has to be computed in the former, it is readily available in the latter. Similar to KinectFusion approach, we define a cubical grid in 3D world, in the frame of reference of the first acquired frame. Every pixel is projected onto 3D world and the cell of the cube that contains the projected point is marked. This procedure is repeated for every frame and every pixel with a valid depth estimate. If the camera location and/or orientation significantly deviate from its initial value during acquisition, we save the results of the cubical grid and reset it using the frame of reference of the current frame. When all the pixels are processed, each cube cell is put through consistency checks to decide whether it should be included in the final reconstruction (one of the checks is that it has be filled with enough pixels since total number of pixels are much greater than the number of cells). If a cube cell is deemed valid to be included in the final reconstruction, its coordinate and average color values of contributing pixels are recorded.

Figure 15.

WP5 Explosive effects and reverse event analysis:
Crime scene registration and determination of the explosion source
The explosive charge amount involved in an explosive incident is a useful piece of information that can speed up the criminal investigation to find the perpetrators. To determine the charge amount TNO has generated two specific deliverables:
1. A ‘Guideline for crime scene registration’ (D5.1) that helps the crime scene investigator to register the information needed for the assessment.
2. A calculation tool for quick first order estimate of the size of the explosive charge involved, the so-called TNO Inverse Explosion Analysis tool (TNO IEA tool).

The Guideline for crime scene registration has been based on existing post blast investigation protocols, historic post blast investigations, and end-user requirements. The Guideline contains an overview of the most reliable sources of information, and the data that needs to be registered.

The TNO IEA tool can perform inverse calculations for explosions. This means that, given a type of damage, the size of the explosive source is calculated. Such calculations are not standard in the field of explosive blast and damage. Existing calculation methods start from the explosion source, and then calculate stepwise first the explosion effects and then the consequences. It is then an iterative process to find the amount of explosives involved. For the IEA-tool, TNO has developed specific inverse models to calculate the charge directly. This has been done for some specific effects and damage, namely: fire ball, window breakage, building damage, crater dimensions and debris throw.

The IEA tool assists a forensic analyst to estimate the explosive charge mass and point of origin based on damage at and around the post blast scene. The inverse calculations lead to a set of charge mass estimates with varying reliability. Furthermore, some estimates give just a lower or upper bound. A statistical method has been developed to combine the various types of data, and to determine an overall charge mass distribution.

The quality and the validation of the IEA tool have been controlled and tested through the following actions:
• An active bug- and wish list and version management was maintained. And the software was tested actively.
• Two hands-on-sessions with the tool have been held with two different end users.
• The performance of the tool was tested against three real case studies, i.e the Khobar towers attack (1996), the Enschede firework disaster (2000), and the Oslo bombing (2011). Reasonable results were obtained.
• The performance of the tool was tested and validated during the HYPERION end demo (23rd of September 2015) under real conditions. Three different scenarios were assessed by an independent investigator, with satisfactory results, both concerning the quantitative charge calculation as well as concerning the use of the tool.

Figure 16.

Figure 17.

It has been seen that the IEA tool predicts reasonable and realistic charge masses, although with a large spread. Such a spread however is realistic for forensic studies due to large uncertainties. If more accurate predictions are needed, e.g. in complex situations with blast shielding and/or focussing, Computational Fluid Dynamics (CFD) simulations can help.

Figure 18.

WP6 Imaging stand-off trace detection:
Within the HYPERION project Fraunhofer IAF developed an imaging IR backscattering spectroscopy system for detection of trace amounts of various explosive substances and their precursors (Figure 19). A tunable dual core Quantum Cascade Laser source emitting in the range of 7.5 µm to 10 µm was used as a spectrally selective illumination source to illuminate the target. This spectral range is known as the molecular fingerprint region of the electromagnetic spectrum, as most organic compounds exhibit highly characteristic spectral features in this range (Figure 20). Another important benefit of this spectral range is that a fully eye-safe operation is ensured, as light in this spectral region is not transmitted through the human eye onto the retina.

Figure 19.

Figure 20.

While the laser is tuned across the available tuning range, the backscattered light is collected by a high performance MCT camera. The sensor produces a hyperspectral image where every pixel vector represents the Infrared spectrum at a specific location in the scene. This hyperspectral image is analyzed for traces of hazardous substances like homemade explosives or their precursors using custom data analysis software. Along with the spectroscopic measurement data, a visual camera is used to capture a spatially calibrated visible image of the scene. The detection result can therefore be presented to the user as color-coded detection map overlaid with the visual image to ease user interpretability. In addition, the output is relayed to the central sensor network server for data fusion and further analysis.
For the final HYPERION demonstration campaign, the sensor was built into a customized trailer that can be operated as a fully autonomous stand-off spectroscopy laboratory (Figure 21). In this configuration, the sensor can be towed into the post-blast scene and is measurement-ready within 20 to 30 minutes after arrival. The operational measurement range was 7 m to 25 m.

Figure 21.

Two detection scenarios were demonstrated: to prove system applicability for forensic sampling in a real world post-blast scenario, an Ammonium Nitrate based IED was detonated in proximity of a parked car. In addition a bag pack contaminated on the outside with Potassium Chlorate (KClO3) was placed in the scene, acting as a simulant for an unexploded IED.

Figure 22.

Figure 22, shows a typical measurement result of the IR standoff spectroscopy system, successfully detecting and identifying a KClO3 contamination on one of the bag pack's straps serving as a simulant for an unexploded IED. The detection algorithm successfully detected the contaminant without generating false alarms for the competing substances in the library. The bag was placed in 10.2 m distance of the sensor.
To demonstrate the application in forensic sampling several measurements were performed on a car exposed to the fallout of an Ammonium Nitrate based IED detonation described above (Figure 22).

Figure 23.

Several positive detection results of Ammonium Nitrate contamination were obtained by the IR standoff spectroscopy system (Figure 23), while none of the measurements produced false alarms for the competing substances.
Our findings suggest that a thin film of Ammonium Nitrate was fairly uniformly spread on the surface exposed to the fallout of the detonation. As IR radiation is fully absorbed by solid substances within the first few microns of a particle, this spatially spread contamination is beneficial for high accuracy detection in contrast to a contamination of the same volume 'stacked' on a single spot for the IR backscattering spectroscopy measurement method. However, for the case of detection of a hygroscopic chemical, such as Ammonium Nitrate, detection sensitivity showed to decay over time after detonation. Shortly after the detonation the system detected the target substance Ammonium Nitrate with a considerably higher confidence compared to measurements performed around two hours after the incident. We tentatively attribute this observation to the adverse weather conditions (slight rain, high humidity) and the hygroscopic nature of Ammonium Nitrate. We therefore hold system mobility for fast response time as a key feature to strive for, within development of a commercially mature product.
Below, we show details of a measurement on one of the front trim strips in Figure 24. As can be seen, the system indicated positive detections of Ammonium nitrate in the resulting detection map. We also give the spectra in the areas classified by the detection system as 'contaminated' and 'background' respectively, as well as the spectrum gained by linear un-mixing the background spectrum from the contaminated spectrum. From the library spectrum we follow, that the most distinct feature of Ammonium Nitrate in the considered wavelength range is located around 1040 cm-1. This spectral feature is well observable both in the mean spectrum of the contaminated area as well as – to a lesser extent – in the mean spectrum of the area classified as 'background' suggesting, that this area was also contaminated with a – yet thinner – film of Ammonium nitrate. Nevertheless, the latter nicely serves, to reconstruct the library spectrum closely, using the mean spectrum of the area classified as contaminated and linear un-mixing.

Figure 24.

Optical system configuration
To meet system requirements a variety of different types of detectors have been suggested. Vigo System S.A. considered both flat and immersed detectors in quadrant configuration, as well as single element detectors. Particular models of these detectors and configurations were created and simulated. Vigo System S.A. have used optical modelling in ZEMAX software to validate the particular designs.
Application of hyperhemispherical lens concentrate the incoming radiation and increases detector detectivity by an order of magnitude in comparison to flat detector configuration. There were two solutions with immersion lens considered. The first approach was to assemble four single element detectors with immersion lenses on a sapphire carrier in a quadrant arrangement as in the Figure 25 below. The marked red squares show the single element optical area. It can be noticed that there are huge spaces between neighbouring detectors. Due to large insensitive gaps this solution was rejected from further analysis.

Figure 25.

The second solution utilized four detectors with one immersion lens as in the figure below. This solution has been considered as the most promising one, as featured both, high sensitivity due to application of immersion lens and high fill factor as the 4 elements were very close to each other.

Figure 26.

Development of detector for spot detection - Detector system concept, adaptation, manufacturing process
Implementation of dedicated detector for stand-off detection system is based on the concept of IR photodetector (Figure 27) developed at Vigo System S.A. that relies on integration of optical and detection function in a monolithic heterostructure chip. The functions are concentration of radiation, enhanced absorption and efficient and fast collection.

Figure 27.

The key features of the detector are listed below:
▪ cut-off wavelength: 10.6 μm
▪ normalized detectivity D*: >2.5∙109 cmHz1/2/W
▪ the detector technology is based on MCT photoconductive detector (PCQI) in a quadrant arrangement
▪ the detector is stabilised in temperature by a 3 or 4 stage Peltier element
▪ immersed optics increase voltage responses

The basic detector material is HgCdTe graded gap alloy, characterized by outstanding photoelectric and structural properties. Since lattice constant is weakly dependent on composition, the material can be used for complex 3D chips whose regions serve as radiation absorber, minority and majority carrier contacts, optical concentrator, passivation and other purposes.
Hyperion detectors are produced from HgCdTe (MCT) grown by MOCVD AIX-200 epitaxial system for II-VI compounds on GaAs substrate. The complete process flow of manufacturing IR detector consisted of: MOCVD epitaxial growth, material characterization, processing (etching, passivation, metallization, dicing), lens micropolishing, wire bonding, desiccant assembly, window sealing, hermetization and device testing. Detectors were measured using FTIR spectrophotometer, blackbody, impedance analyser, I-V characterization tools and noise measurements. Once all test were completed they were integrated with pre-amplifiers and thermoelectric cooler (TE) controller into detection module.
Two generations of detector modules have been fabricated at VIGO company. The first generation was a four- quadrant (4Q) detector operating with an immersion lens. Main work was focused on optimization absorber thickness in order to suppress possible interferences inside detector. New design should ensure almost full absorption of radiation in a single pass of the absorbing layer. Structure for photoconductor was optimized with analytic simulations and later on grown in VIGO's epitaxial system. Figure 28 shows bottom view of quadrant photodetector which consists of four square photoconductors (92 x 92 μm each).

Figure 28.

Active element monolithically integrated with hyperhemispherical microlens was mounted directly at the top of TE cooler (Figure 28). In order to achieve minimum temperature of active element sapphire chip carrier was removed. Cooling to 210K was necessary to get required detector performance. This implicates the use of 3- stage Peltier cooler. One of the biggest challenges (from the mounting point of view) was centring the microlens and making of wire connection between metallization pads and TO8 pins.

Figure 29.

Although the 4Q detector yields a high detectivity it was found during the project that the match of the spot detection result to the pixels in camera image was very difficult. The optical area of 1000 µm x 1000 µm of the 4Q detector corresponds to 25 x 25 Pixels in the camera. Typical distributions of traces of explosives, e.g. by thumbprints only cover some pixels. Thus, the gain in sensitivity is lost versus the better spatial resolution of the camera. Therefore it was decided to fabricate a 4Q detector with the detective area compatible to the infrared camera.
The second-generation detector matrix was designed to have detector elements without immersion lens. The size of the individual detector element is 40μm x 40μm with a pitch size of 80μm (Figure 29). These dimensions are compatible with the pitch of the two available IR cameras with 40 µm x 40 µm and 28 µm x 28 µm, respectively. Figure 29b presents a magnification of the quadrant detector after processing. The image shows 4 detectors, Au contacts to particular detectors, common Au ground line and etched area (GaAs visible). In order to increase the detector performance a miniaturised four stage thermoelectric cooler was used.

Figure 30.

Two generations of the quadrant detector were integrated in a compact detection module (Figure 30) that consisted of:
▪ detector element
▪ ZnSe window with AR coating
▪ preamplifier
▪ Peltier cooler
▪ thermoelectric cooler controller

Figure 31.

Quadrant detector was equipped with custom designed four channel preamplifier, where each channel consists of two amplifying stages. First of them is a transimpedance amplifier, with ability to control the detector biasing voltage and compensate the bias current (to avoid amplifier saturation). The second stage is a noninverting voltage amplifier, with variable gain and output voltage offset compensation. The mentioned above parameters are set independently for each channel. Designing the preamplifier, it was mainly focused on: reduction of cross talks among the channels, achieving low noise level, low input impedance, stability, in band gain flatness, and reaching sufficient cutoff frequency > 100 MHz.
It was decided to use the 2 stage preamplifier with the variable detector bias voltage and the bias current compensation. Including two stages, instead of one, has kept the flexibility of the design in terms of the gain together with the bandwidth. Top frequency of 100 MHz was reachable starting from 750 V/A of transimpedance, up to 30 kV/A. Low gain is necessary, when the high signal dynamic range is required. Otherwise, especially detecting the low signals, higher gain is convenient.
First preamplifier gain stage consists of ultrafast opamp.

Figure 32.

In the current version of the circuit, used in HYPERION project, the detector is AC coupled with the preamp. However, when DC coupling is needed (PV detectors, IR DC components is existing), the DC regime is available by applying short circuit instead of the C34 capacitor.
Each channel contains n-p-n transistor (T4) buffer for supply voltage filtering. In the given OPA847 application (the detector resistance is around 60 Ohm) the bandwidth limit according to the opamp GPW value is 120 MHz. The crosstalk reduction is provided by usage of the 4-layer FR4 PCB laminate.

The second gain stage

Figure 33.

The further gain is provided by the second stage - voltage preamplifier. The preamplifier is based on the current feedback operational amplifier, fitting in the SOT-23 package.
The preampifier second stage allows to trim the gain within 1.3 V/V up to 5 V/V. Also the output voltage trimming is provided. The gain adjustment provided by the R35 was effective with no impact on the preamp stability, as well as the offset, which remained unchanged even though the gain was manipulated.

The supply
The preamplifier is protected against the supply voltage overload and the voltage inversion. On both: positive and negative line, there are LDO voltage regulators with minimum voltage drop around 0.5 V.

WE 6.3 Stand-off spot detection with enhanced sensitivity
During the first 18 month of the project IAF conducted measurements in order to compare resonant thermal heating and polarisation sensitive detection for enhanced spot detection. It was found that the resonant thermal heating effect is too weak and too slow to make use of it in a real world detection system, while polarisation sensitive detection yield very promising results. In summary, for a scene where enough backscattering intensity is available, the polarisation contrast typically gives larger modulation of the signal, thus better detection sensitivity. If the signal strength becomes very weak, the depolarisation component may be obscured by other noise sources. For this case the unpolarized backscattering spectrum yields the better result.
Therefore it was decided to concentrate on the polarisation sensitive detection scheme. IAF decided to upgrade the IR camera system with a rotating filter set which enables polarisation sensitivity IR imaging. In parallel, VIGO System worked on modification of the detector process in a way that grating structures could be deposited onto the active area of the sensor. The idea was to manufacture small detector matrix equipped with polarisation sensitive detector elements and detector elements without polarisers (used as reference). In the first generation of four- quadrant (4Q) detector operating with an immersion lens there was not possible to deposit grating structure. Unfortunately, later work with grating structures onto flat detectors were also not successful. The spacing between the evaporated lines had to be very small, typically a fraction of the wavelength and this was a main technological problem.

Additional activities within HYPERION project

Optimization of PV device
The following main requirements for photodiodes were determined: ≥10 µm cutoff wavelength, ~100x100 µm electrical active area and zero bias operation. However, fabrication of such devices is challenging because they suffer from weak absorption of IR radiation and a low junction resistance comparable to a parasitic, built-in or external series resistance. The resulting reduction of their responsivity and hence sensitivity can be minimized by a proper design and optimization of the detecting structure. The efforts were concentrated on N+pP+n+ Hg1-xCdxTe devices, with a p type absorber and wide bandgap contacts type N+ and P+.
Additional measurements of test structures revealed highly nonlinear resistance of contacts made by various examined metals to wide gap (x>0.4 of Hg1-xCdxTe) P+ layer. Their zero bias resistance varied in the range from 10-3 Ω×cm2 for evaporated Au to 2×10-2 Ω×cm2 for evaporated indium (Table 1). Therefore, P+ layer was interfaced with metallization by the n+ layer, with composition x~0.2 which was slightly greater than that of the absorber to make use of double pass of infrared radiation, incident from the N+ common contact side. The n+ layers allowed for low (4×10-6 Ω×cm2) resistances of n+-Ti contact.
A HgTe layer, with a nominal (i.e. assuming no atomic diffusion between adjacent layers) thickness of about 0.1 µm, was deposited prior to growth of the n+ layer in order to reduce composition in the vicinity of the P+-n+ interface. In fact, such HgTe layer is changed to HgCdTe by interdiffusion during growth. However, its nominal thickness had to be carefully adjusted in order to avoid formation of a parasitic p+-n+ junction, which acts as a nonlinear series impedance with significant resistance and diffusion capacitance, especially for zero bias. Table 2 presents nominal layer diagrams of the photodiode structures from two wafers: Hyperion-1 and Hyperion-2. They differ from each other in the nominal thickness of this HgTe layer (coloured table cells). Their room temperature characteristics of product of differential resistance Rd and photodiode area A along with dark current density J as a function of bias V are demonstrated in Figure 34. The Rd-V curve of the structure Hyperion-1 shows an increase in Rd at forward bias close to 0 V, resulting from the too thin HgTe layer between the P+ and n+ layers. This is not the case with the structure Hyperion-2, where the minimized series resistance allowed for obtaining the spectral sensitivity shown in Figure 35.

Table 1..

Table 2.

Figure 34.

Figure 35.

Improvement of adhesion Au contact in quadrant photodetector
The quality of common Au contact is very important in terms of crass talks and detector noise reduction. The improved process consisted of ionic cleaning of semiconductor surface, evaporating of Titanium layer (thickness 11 nm) and finally evaporating of Au layer (thickness 400 nm) – all done during one vacuum process.
As a result, resistance of Au contact was 2 times smaller, noise of detector was reduced by 20% and cross talks were significant reduced in comparison to previous technology. All these effect improve S/N at the detector output.

Optimization of wirebonding
Wirebonding is a method used to attach a fine wire, in this case 25 μm in diameter, from one connection pad (metallized bond sites on interconnection substrates) to TO8 pin, completing the electrical connection between detector and TO8 pins.
In most cases, all steps are handmade which are time consuming and operator depended. Wirebonding optimisation was an attempt to establish bonding parameters in semi-automatic wirebonder and improve quality of wire connection between detector and TO8 header pin.
This work was divided into two subtask. At the beginning, aluminium wire was used. One of advantages of using aluminium wire is its bondability to thermoelectric cooler (TEC) ceramics substrate which means that TEC do not require additional modifications.

Figure 36.

This work showed that there is a possibility to make a semi-automatic wire bonding connections for VIGO detectors using Aluminium wire. It also occurred that quality of such wire connection strongly depends on ceramics surface. In some cases aluminium wire did not stick to ceramics surface or detached from it (after Pull test).
Because of that, there was a decision to prepare wire connections using gold wire. Unfortunately, gold wire do not stick to ceramics substrate and it required special Au patterns on particular TEC stages. Such design was prepared and thin layer of Au was deposited on TEC stages (Figure 37).

Figure 37.

In the next step bonding parameters was adjusted and wire connection was performed. In comparison to connections with aluminium wires, ball bonding with gold wires resulted in more reliable contacts.

Vacuum encapsulation technology
Vacuum encapsulation technology significantly improves the IR detectors performance, (especially LWIR detectors). The main advantages of vacuum encapsulation is possibility to obtain much lower detector temperature comparing to standard encapsulation (with an inert gas). The second advantage is the lack of convection inside detector housing which is important for some application.
VIGO used resistance welding technique to join TO8 header and custom designed stainless steel detector cup. Helium leak rate measurements confirmed vacuum tight connection. Measured leak rate was lower than 2*10-10 mbar L/s which is very good result.
The second important issue is to ensure vacuum tight connection between detector cup and detector window. This part was not finished and require more R&D work in the future.

Summary
VIGO System S.A. developed and manufactured two IR detection modules for Hyperion stand-off detection system. Detection modules were based on two types of photoconductive quadrant detectors: monolithically integrated with hyperhemispherical lens and flat one. Newly developed photodetectors and detection modules extended VIGO System products line. Additionally, several other technological changes were introduced which influenced on better quality of detectors and IR detection modules.
Imaging Raman stand-off detection
FOI has performed a realistic scenario test in order to evaluate imaging Raman spectroscopy as a forensic analysis tool for detecting traces of explosives at a post-blast scene. The scenario chosen was a left-behind IED, based on ammonium nitrate and a TNT (2,4,6-trinitrotoluene) booster, which was detonated in a plastic garbage bin. Sample plates were mounted vertically on a holder approximately 6 m from the point of detonation. Minutes after the detonation, samples were analyzed with stand-off imaging Raman spectroscopy from a distance of 10 m. Trace amounts of both the secondary explosive (ammonium nitrate) and the booster (TNT) were detected (Figure 39). The findings indicate that it is possible to determine the type of explosive used in an IED from a distance of 10 m, within minutes after the attack, without tampering with physical evidence at the post-blast crime scene. That, in turn, allows for accurate laboratory forensic sampling of the area after it is secured, to further strengthen the evidence collected by stand-off imaging Raman spectroscopy. The most important conclusion from the study and stand-off detections is that there are enough concentrations of post-blast explosive traces that can be detected using imaging Raman stand-off detection. This shows that this part of the HYPERION concept is a viable approach.
Experimentals
The imaging Raman setup (Figure 37) which the measurements were conducted with uses a 532 nm Nd:YVO4 laser (Spectra Physics Explorer XP 532-5) with a 50 kHz repetition frequency and 5 ns (fwhm) pulse length to illuminate a sample of interest. The back scattered light from the interesting area is collected by a 200 mm Smith-Cassegrain telescope (Celestron C8-A-XLT) and passed through an optical system consisting of: a 532 nm notch filter (F1) to suppress the reflected laser light and the scattered Rayleigh light, a f=-100 mm lens is used to triple the magnification of the telescope, and finally a polarizing beam splitter used to split the light into two paths, one towards the ICCD chip and the other to a CMOS camera. The s-polarized light, directed towards the CMOS camera, is focused by a f=60 mm lens and the CMOS acquires a white light image of the target. The p-polarized light is directed through a liquid crystal tunable filter (LCTF) (Varispec VISR) with a bandwidth of 0.25 nm and an optical density of four outside the transmitted wavelength.

Figure 38.

The light is finally captured by an ICCD-camera (Andor iStar 334T). The field of view at 10 m distance measures 28 mm × 28 mm with an image resolution of approximately 70 µm. In the figure above, additional components in the instrument can be seen. These are however not commented here since they are not essential for the scope of our present experiments.

Results
The sample plates that were produced from the work in WP 3 were analyzed from a distance of 10.3 m from our instrument (Figure 38), measured with the integrated distance meter. The results are shown in Figure 40.

Figure 39.

Figure 40.

Development of eye-safe Raman device
In the first part of the project the imaging Raman device was based on a 532 nm YAG laser. This laser is not eye-safe. During the very last part of the first period in the HYPERION project significant development progress has been made in order to realise an eye-safe apparatus. The key component is a filter that allows the use of the UV-laser that need to be used for realising eye-safe operation. This filter has now been tested and works. Further improvements was conducted on the filter and from 2015 a functional eye-safe detection device was assembled.

Figure 41.

The detection limit was tested by filling a pre-drilled Aluminium plate with ammonium nitrate. Amounts of 30 μg could be detected.

Figure 42.

Aiming and aligning equipment
The pan and tilt head has been assembled and found to be working satisfactorily in order to aim the Raman equipment in measurements. Further tasks that have been worked on and implemented include remote operator console for aiming and aligning the stand-off instrument.

Figure 43.

Demonstration of Raman equipment for end users
Further tests of the detection performance of the device has been an work task for the second period of the project. It can be concluded that the UV laser based device operating at a wavelength of 355 nm provides a sensitivity that makes it hard to detect post-blast traces of explosives. The device used earlier in the project based on a green 532 nm laser provides a sensitivity that can be used for certain scenarios of post-blast explosive detection. However, the part crucial for getting a detection equipment working in the UV region of 355 nm is a tunable filter. The present tuneable filter used has a transmission of the order 4-5 %which is too low to provide the needed sensitivity for many detection targets. For some post-blast explosives a positive detection has been accomplished while for the majority of post-blast samples a negative result has been achieved. The research focus at FOI has now been shifted to alternative techniques, still based on Raman spectroscopy that seems very promising in terms of fast and sensitive modes of detection.
As a further part of this work package, a demonstration of the Raman equipment has been performed for end users. This has been done in Wiesbaden at BKA facilities in Germany (see WP 3) and at Home Office, Centre for Applied Research and Technology (CAST) at Sandridge, UK (Figure 44).

Figure 44.

WP7 Forensic sampling and analysis:
A post blast scene of IED is usually chaos. A rapid, well-planned and systematic response is essential in order to minimize the loss of information. After the crime scene area has been secured, the laboratory forensic sampling and analysis can be started. An IED explosion causes a series of devices and residues which are essential to the correct analysis of the incident. The efficient recovery of the explosive residues will directly affect the progress of the investigation. After collecting the samples, they must be secured in appropriate containers and packaged in order to be sent to the laboratory for forensic examination, all that, with the proper establishment of the chain of custody.
In HYPERION project, new and validated protocols have been developed to fulfil with the above requirements. The research work was focused towards two main areas: one concerning the sampling pattern for evidence collection; the other covering the evidence physical collection for both field and laboratory sampling.
Two are the main outputs or foregrounds from the work performed at WP7 in Hyperion:
1) A guide for explosive sampling in post-blast scenes. The three main purposes of the protocol were: to have a sampling pattern for the collection of evidences at difficult scenarios, to describe improved and validated methodologies for sampling in the field and for the preservation of samples in their way to the laboratory, and to have an appropriate methodology for the validation of the stand-off equipment developed during Hyperion project.
Sometimes explosives scatter in such a way that there are no clear evidences or suspicious objects containing their traces. In this case, if extensive sampling is not applied, possible existing traces may not be found, although there are obvious signs of an explosion: the initial collection of only few samples may finally result in the non-identification of the explosive causing the blast.
This protocol provides a methodology for an exhaustive sampling that can be applied after collection of suspicious evidences or when there are no objects suspicious of containing explosive traces. It is important to collect and properly store the samples in the shortest time after the blast happens. Samples from exhaustive sampling will only be analysed if other suspicious evidences do not probe positive identification of explosives.
Due to the relatively high number of samples involved –depending on the range of the blast-, the entire use of this protocol may take a lot of time for explosive qualitative identification. The suggestion is to minimize the investigation task by a logical prioritization of samples analysis. Once all the samples proposed in the protocol are collected, the sequence of analysis will be defined based on the highest probability to detect explosives in the samples. Thus, samples taken in the epicentre, near to the epicentre or in the wind direction will have higher probability of containing explosives than others located in more remote areas. A randomly analysis of some samples taken on these sites will be the first step of the process.
In a post-blast scene, the residues or explosive traces can be found in powered particles or in the initiated piece. The contaminated debris can be different in size and shape, and sometimes it is not possible to bring them to laboratory for their analysis. The objective of the protocol is to suggest different collection methods for both type of debris, those that can be transported to the laboratory and those that cannot. Also, recommendations for container selection and preservation during transportation have been defined in order to avoid the explosive degradation or volatilization before analysis at the reference laboratory. In order to fulfil this objective a state of the art review was performed and experimental research was carried out with different explosives, containers and storage conditions. Also advices and suggestions based on the expertise of the work team and of the end users have been included in the protocol.
2) Application of DART for laboratory direct analysis of evidences collected in the field. Experimental work was conducted to look into a technique aimed to be a “direct sampling” technique, avoiding the most time consuming separation techniques. The speed of intervention is especially important in an explosion scene, having results regarding the type of explosive as soon as possible are quite relevant. In order to speed up the investigation, an analytic screening of the evidence in laboratory takes a special significance. A direct mass-spectrometry of evidence or samples without a chromatographic separation was chosen to be investigated. For this approach the Direct Analysis in Real Time (DART) was used.
The results for the two main outputs of WP7 in the project were collected in the associated deliverables which consists of a document divided in two main parts.
- The first one “Report on validated protocols for forensic sampling of explosives traces at post-blast scene” is dedicated to the description of the new field sampling protocol in an explosion scene and the experimental work necessary for the validation of this protocol. The report contains two chapters. In the first one, a protocol with a pattern for the sampling according to the range of the blast is proposed. The second chapter addresses the protocol for samples collection, including the selection of containers and conditions for proper preservation of the samples. In addition, the methodology applied and the results obtained in the validation of both mentioned protocols are explained in detail in an annex.
SAMPLING PATTERN
A grid sampling is proposed to be used, in order to ensure that the target population is fully and uniformly represented in the set of the n samples collected; then the samples are taken at regularly spaced intervals over the sampling area. The proposed sampling design comprises the sampling grid sizes and the number of single samples to be taken in each sampling grid. To ensure a representative sampling in the post-blast sites which ranges can be highly different, three different protocols are proposed for three different range intervals. In each protocol the number, shape and size of all sampling grids, and the number and location of single samples to be taken in each grid to obtain the corresponding composite sample, are set.
In all sampling pattern designs, additional samples are proposed to be taken. Thus, sampling in the epicentre area requires special attention. Other important issues such as wind direction or building/object presence in a post-blast scene can modify the explosives dispersion affecting to the sampling strategy. Also, taking blank samples must be a routine in any sampling procedure, allowing minimization of false positives.
The segmentation suggested in this chapter can also be used to facilitate the location of the suspicious objects into the explosion area. Finally, all samples collected into the explosion area must be correctly labelled and preserved for their analysis aiming to find traces of explosives used in the bomb attack.
For sampling pattern protocol validation, a controlled blast was performed in May 2015. The explosion was conducted in The Technological Institute “La Marañosa” (ITM), an organisation under the Directorate-General of Armament and Equipment of the Ministry of Defence of Spain. The validation showed that: high concentration of the explosive was located in the epicenter and in the area around it.; the grids located in the wind direction had very high nitrate values and these concentrations were higher than those obtained in the opposite direction; taking several subsamples (integrated in only one sample for analysis) in each grid allowed getting higher and more representative values than those obtained by a traditional pattern.

Figure 45.

EXPLOSIVES TRACES COLLECTION
Sampling of explosives residues is classified in two methodologies depending on the characteristics of the object/debris or the location where the sample shall be taken. For the purpose of the present protocol, these methodologies are called “crime scene swabbing sampling” and “crime scene direct sampling”.
The swabbing sampling methodology is applied for large, or permanent, objects that cannot be moved from the explosion area. This methodology is a technique that allows taking the samples in a support called swab. Finally the swab, with the collected sample, is stored in a sealed container, identified and safe-kept until analysis is initiated at the laboratory. In the below Figure 46, some results are shown where the tests were carried out with ammonium nitrate. The obtained recovery percentage for the seven tested surfaces with the three solvents and with the different swabbing materials are shown.

Figure 46.

Concerning swabbing solvents, no significant differences in the extraction capacity of the three solvents tested were appreciated. In terms of swabbing material, recoveries with cotton and alcohol wipes are similar while polyester shows a slightly lower yield. The higher changes are observed depending of the surface where the explosives are deposited: recoveries from non-porous surfaces are always good (above 80 %), however worst recoveries (20 to 40 %) are obtained from porous surfaces, even though ammonium nitrate was always clearly detected. Tests with PETN, TNT, RDX and NG are under progress.
The crime scene direct sampling methodology is carried out when it is possible, or adequate, to take samples from explosion area and to transport them afterwards to the laboratory. Depending of the debris nature, some recommendations for the type of container to be used are set. Also, parameters related with the sample storage and preservation conditions are fixed.
- The second part consists of a “Report on new or improved sampling techniques in laboratory for forensic analysis” which is devoted to the description of the experimental research carried out in terms of sampling techniques in laboratory, with special emphasis in direct analysis of evidences without previous treatment or separation steps.
Fundamentally all prime sources of evidence in an explosive attack can be analysed by this screening technique without deleting the possibility of further investigation with a proper laboratory method. The screening process takes in most circumstances around a minute per sample. This possibility of fast information gathering is especially important if the number of evidence is very high as it normally is if the device had a larger charge. The sensitivity is, while not as low as in a trace analysis method, still adequate to give solid information to the investigation officers. This DART-MS screening procedure will speed up analysis time in general and takes forensics quite a step forward to resolve the problem of fast preliminary identification of the explosive used in the event of a larger attack.
WP8 Information management and data fusion:
Several tasks in the WP was addressed. One was to determine the data exchange format of the post-blast scene between the information management software and the 3D modeling system. It was important that the format of this 3D model could be correctly interpreted by the information management software handling the post-blast scene registration. Various platforms including Blender suite, ArcGIS (the ESRI’software suite) and CapaWare were evaluated. The 3D registration system in the project provided data samples in order to eventually come to an agreement on a common format that would avoid format conversions, which is a risk of losing information and compatibility.
It appeared clearly in the project that a difficulty would be to locate the different elements in the scene relative to each other and the sensors did not have a native absolute localization system. A GPS system will not be sufficiently reliable for the resolution needed and differential GPS, which is much more expensive will be only functional for outdoor scenes. The choice of solution was to make a project specific location system. The system developed consisted of several tags manually placed directly in the scene and presented in the 3D scene modeled. The sensors pointed these tags before the analyses started. This allowed to locate the sensors and thus the target analyzed in the scene and enabled the system to place the target in the analyzed scene modeled by the cameras. The formalism of the location system was then defined.

Figure 47.

For a secured encrypted communication during the crime scene registration, it was proposed to implement an Ad-hoc Wifi connection AES 256. For the preparation of datasets, it was proposed that the encryption of the workstation as zone central and users sessions management by OS, and a by hand delivery to the concerned agent. For the storage and transportation to the court, it was proposed a cryptographic protocol against truncation attack and a time stamped log of each process during the registration. Then, it was proposed an on-site backup of the crime scene registration on an external locked drive.
Registration of the forensic analysis with the 3D model included the following:
- A real time display
- The display of the probabilities on the type of IED detected and the epicenter in the 3D picture of the post-blast scene
- The position of chemical elements, addressing where (which point) in the 3D picture was the stand-off detection performed.
- The possibility to present details on the numerical dataset.

The project decided to make a specific location system. Work on 3D software simulations in order to manage the post-blast scene registration were performed. Many ways were tested to locate sensors into the 3D scene digitized and also for adding manual input and various information types into the 3D digitized scene.

Figure 48.

In order to integrate the 3D scene stored in a specific data file it was also needed to develop a specific plugin since the chosen format being not supported by the chosen registration platform which were Blender suite. This software is a free and open source 3D suite that supports the entirety of the 3D pipeline (modeling, simulation, rendering) and the Blender’s API can be employed for Python scripting to customize the application and write specialized tools. Using Blender allows the integration of input like data coming from sensors, the development in python of specific scripts and it is quite easy to use for users. It was the most modular platform to date.

Figure 49.

Furthermore, it was also developed plugins for registration of conclusions of the post-blast investigation for generation of automatic reports. During the final demonstration, the software was able to manage many 3D models files of the same scene simultaneously and it was also able to locate all the sensors and measurements automatically and to add them in the interface as the files arrive in the secured container available by wireless connection for the sensors.
A simulation of the registration scenario was realized in a joint experimental activity between a few partners late May 2015 in order to validate the entire process and to assess the precision of the coordinate system used. In order to ensure this, three markers were placed in the area to be reconstructed. The user who collected the data for reconstruction had to ensure that markers were in the field of view so that they would be in the final reconstruction result. Multiple data collection and processing efforts was performed in order to reach a workable configuration. In the figure below, a sample image from a data collection is shown where there are more than 700 images in the collection. All three markers are visible (two of them yellow, the closest one white). Experimentations were performed about the ideal placement of these markers and their height relative to the ground. It was important to make sure that there w no degeneracy in the mathematical equations that would register the measured points to the reconstruction frame of reference. The trial session was a success and confirmed the feasibility.

Figure 50.

Figure 51.

The data collected for reconstruction of an area even of modest size is very large. The reconstruction process essentially manages a data reduction. It distils the information contained in all the collected images and represents it in a 3D model. There is no industry standard about how to represent a 3D model in which format and in the project it was experimented on several of them. In order to keep the details of the scene as much as possible, and to make physical measurements on the model as accurate as possible, it was decided against using any kind of triangulation (surface mesh). Consequently, the representation of the model as a point cloud turned out to be most natural choice for this effort. As the format, it was decided to use PLY (Polygon File Format). It can be written as either ASCII or binary format and it is one of the most common formats in the industry, particularly for three dimensional data as output of 3D scanners. Software development efforts in order to provide this kind of output was also worked out under this work package.

Figure 52.

A plugin to add the point of origin was also developed. It allowed to register more information in the final report.

Figure 53.

The use of the information management system containing the entire outcome of the final demonstration in the project has been documented in the associated classified deliverables within work package 10.
WP9 System integration for hardware:
In this WP the activities for integration the hardware for the system validation of the stand-off forensic detection tools has been carried out. In particular, the tasks has been focused on the design of the package for the stand-off Raman and IR detection sensors. For the Raman sensor, the mechanical design has been developed and combined with thermal simulations of benefit for the detection components inside the hood.
FOI has been developing an eye-safe stand-off detection system based on the imaging Raman detection principle. It will be working in open air when mounted on a tripod. The design and realisation of the package for the Raman sensor included the following aspects:
An aluminium box with electrical connectors, a visible and UV transparent Window for the primary laser giving rise to Raman scattered light from targets. An anti-reflection window coated in the wavelength interval of use (355 -700nm) and protection for operation environmental conditions. The solution for the temperature control included fans and aperture in the box for the thermoregulation of the laser, detector and other components.

Figure 54.

The communication of the spectra measured and the command control for the Raman instrument were based on USB serial connection. The proposed system integration platform for the Raman apparatus can be seen in the below Figure 55.

Figure 55.

The features for the integration also included a power supply battery unit for the autonomy of operation in the scenario and a temperature sensor that protects the laser device by limiting the operation. The thermal simulation has allowed assessing the thermoregulation technique based on fan cooling and air ventilation. The maximum operating temperature +42°C of the critical components was defined and based on the operating environmental temperature of +4 to +35°C and on the components heat dissipations. The mechanical drawing of the aluminium box and crown glass window was finalised in March of 2014. USB serial electrical interfaces have been defined for signals as well as +12 V unique power supply voltage.
The main laser module for the stand-off IR system has been redesigned in order to enable the envisaged field trials. A new small and ruggedized design of the broad band tunable quantum cascade laser module was realized. Besides the more compact design of the housing of the external cavity laser, a special process has been developed for the mounting of the first collimation lens at the output facet of the laser chip. The position of the lens is fixed by a photosensitive glue, while the laser beam is controlled with an infrared camera under operation. This procedure results in a fixed, thus ruggedized unit, comprising the sub mount for thermal management, the laser chip, and the collimation lens. By these measures maladjustments due to vibration and thermal effects is minimized and a more stable operation of the laser in real world conditions becomes possible.
The thermoregulation of the Raman apparatus high power laser and detector has improved the usability of the devices for field use in an outdoor environment and in a real post-blast scenario. The main implemented features for the Hyperion hardware tool are a rugged packaging including a suitable cover and dust proof protection and an air cooling system allowing the VIS or Near UV laser source and Imaging cameras to survive the external environmental conditions at an outdoor post-blast scene. The transportability and deploy ability includes the item of that all Peli cases are robust, wheeled and allow the transport of the parts of the Raman sensor and sampling tools by a single operator at the scene. For the power autonomy a power generator and battery pack are easy to be operated and suitable for > 2 h operation of the sensor. Based on COTS power fuel based generator and supply unit (PSU) the RAMAN sensor has been made autonomous. The transportability has been achieved based on a wheeled rugged box for the tripod, the Raman pan and tilt head and also the power supply units. The materials and tools for the sampling tasks were also based on Peli cases.

Figure 56.

Figure 57.

The imaging system designed to collect the data for 3D reconstruction contained several components. The efforts of putting together all these so that one person can perfrom the data collection required for 3D reconstruction were also performed in this work package. The final 3D imaging system and its components can be seen below.

Figure 58.

The initial idea was to use the screen of the PC as the main display unit for the user while collecting data. Because of this, a custom made case was designed and produced. However, sun glare made this very difficult to work. Consequently, it was decided to add a display unit on a goggle so that the user would be able to see the imaged area under all conditions. The case of the laptop is kept closed and carried on the chest of the user. The battery we used to power the cameras had issues in terms of keeping the charge for long durations. Thus, a regulator circuit was designed and produced on a PCB and added to the battery. In order to keep all the measurement sensors together and make the operation of them smooth, a special handle was designed and added to the system.

Figure 59.

WP10 Systems tests and validation:
The work in wok package 3 was one of the cornerstones for input to the development of the scenario to use in the final validation. Also, the research on inverse event analysis from work package 5 was of importance. The final validation has been continuously discussed during the project meetings. Communication with WP leaders provided details for the planning of the validation event in the end of the project.
The final validation of the tools to be developed in the HYPERION project meeting was done at the test site Grindsjön at the Swedish defence research agency on September 24, 2015. The test site is located at a secure area 40 km south of Stockholm. The home-made explosive of choice in the IED was some of the prioritised explosives selected in WP3. Also, an IED that was not detonated, without ignition system, was placed on the scene after the detonation of the first charge. There was structural elements such as walls containing windows, cars, road signs and other material at the scene in the surrounding of the bomb that was detonated. Personnel from Swedish police helped with using the HYPERION tools at the final validation. The stand-off detection equipment, reverse event analysis and 3D-registration devices were used after the detonation. The final step was the post-blast samplings.
The Hyperion demonstration was divided into three scenarios whereof only scenario 1 was fenced in at all after the detonation with all of the HYPERION tools used after detonation. The scenarios were:
Scenario 1: Detonation with charge placed in a waste-bin 0.5m above hard ground.
Scenario 2: Detonation with charge buried 0.5m down in soft ground.
Scenario 3: Detonation with charge placed indoor of a freight container with windows.
In May 2015, there were a rehearsal between the WP 4 (3D-registration) and WP 8 (Information management) and WP 6 (Imaging Raman detection) carried out. The idea was to test the communication and interface between the tools. The tests were carried out over a two-day activity and resulted in a good outcome. The communication protocol worked between the sensors. The imaging Raman device could send data to the information management system as well as the 3D-registration system. One item remained after this activity and this related to that the actual data sent by the Raman device ended up a few decimetres incorrectly in the vertical position although the horizontal position was accurate. This was later mitigated at the final demonstration.
The final rehearsals for the actual demonstration was done the week before September 24-25. The test plan based on a report correlating to the Milestone 14 in the project proved to work well besides some minor changes. For example, it was noticed that the charge size was too large in relation to the windows of 4 mm thickness. This made all the windows to break during the rehearsal detonation. The fact was mitigated by using less charge size in combination with somewhat thicker windows, 6 mm. Since windows are a “pressure sensor” for the shock wave this was an important part in order for the inverse explosion tool to have any data for the calculations.
Other items related to the positioning of sensors and items in the scene. The demonstration was challenging since that the HYPERION tools needed to be validated and tested at the same time as the audience needs to have the place and chance to understand the actions taking place. In total three detonations with different purposes was made as described above. This made it possible to use some of the HYPERION tools in different ways.
In summary, the demonstration could show a viable concept for the HYPERION tools where all stand-off data could be used, sent and compiled in one software. Still much more research is needed for better 3D-reconstructed images, more sensitive detection equipment and further work on the models for prediction of charge weights. A future system working in an optimised manner would help police forces in many respects such as safer forensics with less risks for contamination of the scene. The details of the demonstration are reported in the associated classified deliverables in WP 10.

Figure 60.

Potential Impact:
Hindsight always makes solutions to problems so much clearer than at the time of their occurrence. For example, today we are aware of many risks including the high jacking of planes and their purposeful crashing into civilian buildings, with devastating effects to public well-being. But in the morning of September 11, 2001, no one would foresee that such an event was just about to happen. This was simply a failure of our imagination.

Even if we could create the perfect bomb-mitigation system, the terrorists might eventually find a way around that system. This is the foundation for dealing with the problem in multiple fronts, i.e. considering the entire terrorist timeline, from the bomb material supply chain to the detection of bomb factories and from the handling of an identified bomb-threat to the forensic reconstruction of a crime scene. The earliest phase of them all relates to the possibility for discovering terrorists and stopping bomb attacks. This, in turn, relates to the issue: who is a terrorist and what can be done to prevent people from becoming terrorists? Today, there is no clear and widely accepted definition on terrorism, other than it apparently involves the use of violence or the threat of violence for the purpose of obtaining control, power and obtaining political goals: effective ways of releasing fear and causing destructive consequences for society and citizens globally. A significant amount of research effort has been expended on the issue of why people choose to become terrorists. Nonetheless, convincing groups and people to feel and develop destructive thoughts and terrorist actions should have a significant priority and should be a long-term preventive method.

The first known large-scale use of IEDs was during World War II and ever since IEDs have been used both for unconventional warfare in military theatres of operation such as in Iraq and Afghanistan and in terrorist attacks such as Bali (2002), Madrid (2004) and London bombings (2005) and more recently in Oslo (2011) and Boston (2013). The threat of terrorist attacks is consequently a very real concern for citizens in many parts of Europe and elsewhere in the world. IEDs can be manufactured from military and commercially based explosives or home-made explosives and often a combination of these are used in terrorist attacks. Fundamentally, terrorists use whatever is easily available for the preparation of bombs and due to the wide range of options in this respect there are numerous parts that comprise the threat to the society.

The Bali bombs, 2002, were manufactured in a two-story house a few kilometres away from the terrorist attacks resulting in hundreds of dead and injured people. The bomb factory had not been disturbed since the bomb makers departed and forensic experts could make a vast amount of analyses. The bombs were found to be based on a combination of potassium chlorate and aluminium powder. Both of the substances used in the Bali bombs are non-volatile and it is likely that these can be detected as traces in the vicinity of bomb factories or at the post-blast scene.

In March of 2004 in Madrid, ten bombs against the commuter train system exploded due to a coordinated terrorist attack resulting in hundreds of killed and injured people. Shortly after the attacks the police identified an apartment in Leganés in south of Madrid as the likely bomb factory. The suspected people trapped in the apartment set-off their explosives resulting in killing themselves, nevertheless, forensic investigators later found out that the explosives used by the suspects were of the same type as those ten bombs used on 11 March. The bombs were found to be made of Dynamite probably bought from a retired miner who still had access to blasting equipment.

It is evident that there exists an urgent need for new tools in the extensive work of preventing future terrorist attacks and this is a large scientific challenge in the security domain today. Equally challenging is the development of novel forensic methods that can provide high quality forensic data that can be used in a court of law. For the law enforcement services there is a large need for tools that contributes to reliable early forensic data after bomb attacks.

The preparation of an IED attack involves different phases, where planning and financing the operation is the first phase. Thereafter precursors are obtained either illegally or purchased legally and transported to the location where the preparation of the IED is performed. The phase before execution of the operation takes place in a so-called bomb factory, for example a kitchen or a garage, where equipment and chemicals are manipulated in order to create the IED. The discovery of bomb factories is of primary importance in the prevention of terrorist activities, where investigations can be conducted with fewer time constraints and with greater accuracy than in later stages.

Due to the large variety of HMEs that can be prepared, currently there are no specific commercially available sensors to survey the presence of precursors or the transformation of such chemical compounds into IEDs inside a suspected environment. The search, monitoring and identification of suspicious substances are also complicated by their different physical properties that they possess (vapour, solid, liquid). The use of different sensors can normally be used at different phases of the timeline where the objective can be the same and thus identification and detection of precursors and explosives is required.

A timeline for the terrorist preparations include several steps. In the planning and financing phase, terrorist plots may mainly be revealed through conventional intelligence gathering methods, e.g. surveillance of known terrorist groups, informants, monitoring of money transfers or intercepting the group in a robbery. In many cases, necessary household chemicals or chemicals readily available in hobby stores can be obtained at low cost and little effort. Monitoring trade of restricted materials may be one possible way to reveal terrorist plots at this stage if large amounts of material are purchased or stolen.

The preparation and production phase can by necessity be quite long (weeks to months). Unless the bomb-makers are quite professional they will need to test and practice the procedures and in many cases the chemical manufacturing process in itself takes considerable time. Detection during transportation is much more disadvantageous than detection during production. First of all, the detection itself is more difficult and the time frame for intervention activities is also much less as compared to the production phase. Furthermore, even in case of successful detection in the transportation phase, the interruption of the attack is not without risk. If the person carrying the bomb is a suicide bomber, the bomb may be detonated with severe consequences for law enforcement personnel as well as innocent bystanders. When the plot has reached the last phase - the attack itself - it will be too late and severe damages and casualties will be a fact. At this phase it is of high importance to quickly get early and reliable forensic data that can guide the investigations in the right way.

The outcome from a system such as HYPERION would be to provide police forces on-site with accurate information on the type and weight of explosives used in the attack. In addition, the assessment of the type of IED and the point of origin for the detonation are very useful forensic data. The data produced from a system such as HYPERION should be seen as early forensic data that that needs to be complemented later with conventional forensic investigations. The main benefit for achieving an early reliable forensic information would be a larger chance of finding the responsible antagonists and thus possibly preventing further attacks. The capability to stop further bomb-attacks is naturally a clear positive impact for the security of society and citizens.

A system like HYPERION would allow law enforcement agencies to have rapid forensic evidence and be able to start the crime investigation at a very early stage thus facilitating the chance of finding the perpetrators of the attack. In addition, the secure documentation of the chain of custody will provide a larger possibility for conviction in a trial. The Guideline for crime scene registration and the inverse explosion analysis tool contributes to both of these goals of the HYPERION tool. Thanks to the ease in use, the deliverables can be used by any crime scene investigator. It is planned by TNO to study, together with NFI (Netherlands Forensic Institute), how the deliverables can be introduced in the forensic world. Options to be studied vary from introduction in the education of police men and forensic investigators to application in the field. The following quote of Mr Koeberg of NFI, who has used the tool during the HYPERION demo, explains why: “The calculation tool allows for on-the-spot application. This will help in giving faster and better first estimates of charge sizes, with full traceability and reproducibility, which are important issues in the forensic field.” The tool has been disseminated at different security and explosives conferences.

The HYPERION activity has increased the Selex ES capabilities to install and to design an autonomous land deployable sensor for explosive traces and other chemicals detection, moreover it has increased the knowledge about the active hyperspectral detection.

Computer vision has been one of the fastest growing fields of the last decade. The dramatic improvement in computing power, the fact that almost everyone started to carry a camera in their cell phone with them and other factors played role in this growth. It is clear that, 3D reconstruction, being a way to represent the real world closer to reality than a single projective snapshot, also gathered quite a bit of interest along with other applications of computer vision. ASELSAN have utilized two of the more popular imaging equipment’s to achieve this purpose (stereo and Kinect). The main advantages of our choice is cost and data collection speed. All the equipment we gathered for the prototype system were COTS products (though, we built some custom mechanical interfaces) and their total cost is order of magnitude less than the main competitor technology, laser scanner based systems. We believe an affordable system with easy deployment and data collection capability would prove crucial, especially for local law enforcement. When a crime committed which gathers interest on national/international scale, usually no cost is too much for the analysis of the crime scene and subsequent analysis. However, for crime scenes for which local law enforcement has to take the lead, budgetary concerns are always present. In such cases, having a low cost and fast deployable system will be very crucial to achieve the desired results out of forensic analysis.

The infrared laser based backscattering spectroscopy technique for stand-off detection of explosives proved well suited both for forensic sampling in post-blast scenarios and for detection of residues on unexploded IEDs within the HYPERION final demonstration campaign.
As described in section Scientific and Technological Results our finding suggests that a thin film of the explosive used for IED manufacturing was deposited on the objects in vicinity to the detonation. Therefore, successful detection is not dependent on measurement on a specific spot in the scene, but will yield an identification of the target substance within the first few measurements.
In summary, infrared laser based standoff detection of explosives has been demonstrated to be a promising tool for fast forensic investigation for first responders. The technique is eye safe and has the potential for further development.
In future projects, larger detection distances can become feasible by increasing the laser power and by optimization of optics. In addition, real time hyperspectral image acquisition and target detection can be addressed as well as development of a smaller man-portable version of the standoff sensor.

It was shown in the beginning of the HYPERION project by FOI that the imaging Raman system, based on a 532 nm laser, could detect residues of different threat substances on surfaces in post-blast scenarios. One drawback with this system was the fact that eye-safe operation cannot be accomplished due to the laser source required where people needed to wear laser safety googles or keep outside a rather large safety zone. This was one of the basis for the decision to develop an imaging Raman system based on the same technique and with similar performance, but at in an eye-safe way.

Summing up the results and experiences the performance of the eye-safe apparatus falls a bit short, mainly due to the difficulties in developing high-performance tunable filters in the UV-range but also due to the fact that some substances doesn’t emit Raman signal when excited at 355 nm. This has been seen to be the case when it comes to solid substances (for gaseous or liquid forms of the same substances this is not the case). The result is an eye-safe system that is quite close at reaching the needed performance for post-blast residues, but is not all the way there in real life measurements although some positive post-blast measurements have been made during the course of the HYPERION research work.

The stand-off detection research at FOI now aims at increasing the performance in terms of sensitivity and speed. Inherent in the “tunable filter- based” design are the large light losses, mainly due to sequential imaging at different wavelengths, leading to effective transmission often well below 1 %. FOI presently investigates the possible benefits of using compressive sensing and digital micromirror devices (DMD) applied to imaging Raman spectroscopy for stand-off detection of trace amounts of explosives. In doing so an experimental setup have been built and designed to test compressive sensing concepts in realistic scenarios. The total light throughput of this type of setup compared to tunable filter based imaging systems is orders of magnitudes higher, and this is expected to be reflected in increased sensitivity or decreased measurement times.

The great flexibility in this type of setup in combination with compressive sensing techniques can in turn make possible combining imaging with non-spatially resolved fluorescence suppression techniques, such as Kerr gating.

In summary, although the outcome of the HYPERION system points to a viable concept for quick forensic results using stand-off tools after a bomb attack, still more work are of importance in order to e.g. further enhance the detection limits, including a larger library of explosive compounds that can be detected, achieving better resolution for 3D registration and creating furthermore accurate models for post-blast calculations for the assessment on the used charge weight.

Final Report Summary - HYPERION (Hyperspectral imaging IED and explosives reconnaissance system)

Share this page

Download