No abstract

No abstract

We address the problem of tracking targets from distributed sensors in a cluttered environment (see Fig. 1). In a paper which will be presented at IEEE ICASSP-92 we introduced a new method relying on Bayesian theory and HMMs, which is theoretically dedicated to the single target case. We here present a second approach, which is valid in the multiple target case, thanks to the use of Dempster/Shafer's theory of evidence. Our applicationis, to ourknowledge, an original one for this theory.The sensors measure ranges or dopplers, so that a sensor taken alone provides only partial observability of a target: 3 sensors at least are necessary to achieve total observability. The targets are characterized by their state parameter vectors at any time t, composed of their position and velocity: (X(t),Y(t),VX(%VY (tal We suppose that the target parameter space can be partitioned in adjacent cells, each of these cells containing at most one target, and being centered on a quadruplet: (Xi, Yj, Vxk, Vyl) 0-44, O

No abstract

The problem of target tracking from distributed sensors in a cluttered environment is addressed. In recent publications [ 11, [2] w e have introduced a new and original approach in target tracking and target motion analysis. Two variants of this method have been developped, one using the bayesian probabilities theory, and one using the theory of evidence. These variants and their specificities with regard to classic tracking methods are here recalled, with some extentions to the previous work. Special attention is here given to the comparison between these two variants, and to difficult cases as multiple maneuvering targets o r crossing targets.