DIMUS will investigate the advanced technology of multisensoring in order to obtain correct and complete scene information in monitoring and control applications with safety-critical constraints. The objective is to design and develop a system supporting a human operator in the interpretation of situations occurring in a metro (subway) station environment. A prototype system will be developed containing a set of tools for the integration and fusion of visual and non-visual data provided by different kind of sensors.
Special tools for the acquisition and merging of multisensor data, for low-level image and signal processing and for adaptive data fusion will be developed and used in the construction of a monitoring and control system for subway stations. The concept o f adaptive and robust control will be investigated, and a tool allowing flexible adaptation of the multisensors to a changing scene will be implemented.
The interpretation processes will be optimised in order to recognise objects by the parallel execution of search, detection and matching tasks.
DIMUS will incorporate several existing systems produced by the partners for stereo, motion and range-finding, and will also use visual-tactile sensor systems. A moving camera subsystem for focusing attention on particular regions of the subway environme n t will also be developed. The exploitation of knowledge-based methods for the recognition of moving objects will be investigated.
The demonstration of intermediate results concerning the adaptation of the multisensors to the environment is scheduled to take place after 17 months. This will show the fusion of visual and non-visual information in order to support a focus of attention on an area of interest. The final demonstration will be performed in a real metro station environment shortly before the end of the project.
Combining several types of sensorial data obtained from many sources of information (data fusion) has the advantage of providing robust problem solving in the presence of missing or faulty sensorial data, due to information redundancy and diversity. Several applications of this technique have been investigated.
Information provided by magnetic resonance angiography and digital subtraction angiography has been integrated to give a 3-dimensional view of the reconstructed cerebral blood vessels surrounding an aneurysm, to assist in the division processes of medical diagnosis and therapy.
Fuzzy logic has been implemented to aid the decision making process for a robot gripping device, and to run petrochemical processes in a large production plant.
Odometric data and a laser range scanner have been combined to provide a global view of the environment for semi autonomous robots, capable, for example, of remote surveillance tasks.
Image processing and interpretation tasks of various types have been made easier by the use of multisensor visual information. Applications include surveillance of traffic and pedestrians.
The combination of an active sensor (a lidar time resolved laser fluorosensor) and a passive multispectral scanner has been implemented for the purpose of environmental monitoring of river or ground water.