Virtual Environment Interface by Sensory Integration for Inspection & Manipulation Control in Multifunctional Underwater Vehicles

Informazioni relative al progetto

ID dell’accordo di sovvenzione: BRPR960150

Progetto chiuso

Data di avvio 1 Aprile 1996

Data di completamento 31 Marzo 1999

Finanziato da

• Specific research and technological development programme in the field of industrial and materials technologies, 1994-1998

Costo totale € 0,00

Contributo UE Nessun dato

Coordinato da

TECHNICAL SOFTWARE CONSULTANTS LTD
United Kingdom

Exploitable results

Virtual Environment Interface by Sensory Integration for Inspection and Manipulation Control in Multifunctional Underwater Vehicles; VENICE

The objective of the VENICE project is to undertake the basic research activities required to produce a self-referencing positioning system for Remotely Operated Vehicles (ROVs) in an underwater environment. This will significantly help the ROV pilot when performing any form of navigation or manipulation task. The VENICE system is based on two sensor systems which are used both individually and in combination, with the sensor positioning results being displayed to the operator via a 3D graphics virtual display of the ROV and its environment. The two sensor systems under consideration in the project are optical cameras and a new 3D acoustic camera. First-stage image processing of the optical images is used to perform an automatic regulation of the camera parameters, i.e. focus, aperture, gains, etc., to ensure that the best possible image is achieved under all environmental conditions. Higher level image processing techniques are then used to filter the images, apply segmentation, extract lines, etc., to recognize features of typical underwater structures. These features are compiled into a symbolic representation of the image which can be used for position calculation. A new 3D camera has been developed as part of the project which generates real-time 3D (x, y, z and intensity) image data from a single acoustic pulse. 3D acoustic data processing algorithms have been developed to perform the filtering, segmentation and recognition stages required to generate a symbolic representation of the features present in the image. The optical and acoustic symbolic representations are then compared against similar information held in a pre-compiled environment database. Efficient matching algorithms have been developed to find the most likely match between the sensor and database symbolic representation. Additional algorithms are then used to estimate the sensor position relative to the database. The sensor data is also integrated at several different levels during the data processing to reinforce the feature recognition steps and to achieve a higher level of confidence in the results than using each sensor separately. For example, edges detected from the acoustic data are projected onto the optical data allowing problems from the obscuration of cylinders due to perspective effects to be identified. Symbolic feature information is held in an environment database which is pre-compiled from CAD drawings of the structure and maps of the sea-bed. The symbolic information can be extracted from this database at various levels of detail to allow very rapid feature matching. The database is also responsible for providing the graphical information, again at several levels of detail, which allows the 3D virtual display of the environment to be created. This display is the primary user interface to the system, and is updated in real-time with the ROV position estimated by the sensor systems. The environment can be viewed from any remote position, allowing the best possible visualization of the task being performed. Alternatively, views from cameras attached to the ROV can be selected allowing direct comparison with video images. ROV simulation software has been developed which correctly simulates the hydrodynamics of the ROV. This can be used for pure simulation or, more importantly, can be used to predict the position of the ROV allowing the scope of the database search for matching information to be restricted to a relatively small volume of space. Manipulation software has also been developed allowing both simulation and real-time control of manipulator arms. A flexible software architecture has also been developed which enables all of the software modules to be used together in a common system. This architecture also allows configuration of the ROV and its attached system to correctly model the use of real ROVs. The system also allows new sensor or control systems to be added in the future without changing any of the existing components.