Studies on multiple target tracking schemes using multi sensor networks

Abstract

Multiple Target Tracking is a very important task in the surveillance system and is achieved with single or multiple sensors Basic MTT system performs various tasks such as sensor data association tracking and fusion This thesis proposes Zero Back scan MHT scheme to address the data association problem Interacting Multiple Model scheme with velocity and acceleration model filters to perform tracking in Wireless Sensor Network and Combined Gaussian Hidden Markov Model based Kalman Filter scheme for tracking with multiple imaging sensors Existing data association techniques such as Probabilistic Data Association Joint Probabilistic Data Association and Multiple Hypothesis Tracking do not perform well in multiple target environments due to non availability of automatic track initiation Moreover the probability of the presence of the target is computed with reference to previously available targets under heavy pruning To overcome such problems this research proposes a new data association algorithm named Zero Back scan MHT newline