Multiple target tracking using Support Vector Machine and data fusion

In this paper, same target is being sensed by multiple sensors and the main objective is to classify the information into set of data produced for the same target. Once tracks are initialized and confirmed, the number of targets can be estimated; the future predicted position and target velocity can be computed for each track. Fusion is necessary to integrate the data from different sensors and to extract the relevant information of the targets. Support Vector Machines (SVMs) are generally binary classifiers and the multi class problems are solved by combining more than one SVM. This paper proposes a novel scheme for multiple target tracking using SVM classifier. The proposed scheme achieves classification by finding the optimal classification hyperplane with maximal margin. Also Kalman Filter (KF) and 1 Backscan Multiple Hypothesis Tracking (1 BMHT) are used for filtering and association respectively.