Probabilistische modellen voor het volgen van meerdere objecten met meerdere sensoren (ANN.5312)

The use of distributed systems with active cameras which can detect and track moving objects is getting more and more common in surveillance and safety and traffic monitoring applications. Advanced image processing techniques can be used to detect moving objects in real time to extract features from these objects. To track the objects, the information from the different camera systems will have to be combined. A number of fundamental problems has to be solved: the identification problem ('is the object which is observed by one camera the same as the object observed by some other camera some time ago?'), the representation problem ('how can probabilistic information be integrated in a distributed system?') and the control problem ('how to control the cameras such that maximal information is obtained?'). In such a distributed observation system there will be both dependencies as independencies between measurements from different sensors. Recently Bayesian networks have shown to be able to represent such dependencies and to exploit them.
This project will study to which extent Bayesian network can be applied to other than the traditional areas such as diagnosis, and to which extent these networks can be used in a multi-agent system. We will combine our expertise on probabilistic models for representing sensoric data with our expertise on other formalism for the representation of uncertainty.
The methods will be implemented and tested on a system consisting of a network of intelligent cameras, communication devices and actuators. Such a system is currently being developed at TNO-FEL, with which we will closely cooperate in this project. In cooperation with the other users we will shape our final configuration and testing program.

Er zijn nog geen resultaten bekend.

Four companies are involved in this project.