Support Vector Machines for Multiple Targets Tracking with Sensor Networks

Abstract

In this paper, we address the problem of tracking multiple targets trajectories that potentially cross each other. We propose a solution to this problem by using the support vector machines (SVM) to predict the position of the targets from the past history of the measurements. We will predict the dynamic behaviour of the targets using the SVM method, after turning off the sensing mode. By this way, we will avoid wrong measures that might take sensors and which are due to the superposition of physical signals. The simulations results demonstrate the potential advantage of this approach.