Adaptive tracking and classification algorithm for multiple extended targets based on irregular shape driven

Abstract

The extended target Gaussian mixture probability hypothesis density (ET-GM-PHD) filter is considered as a promising algorithm for tracking multiple extended targets (METs). However, when the densities of the newborn targets are unknown and some closely spaced targets exist in the tracking scenarios, the tracking accuracy will decrease greatly, even some targets are missed tracking. For these problems, we propose an adaptive tracking and classification algorithm for METs based on B-spline shape driven. Firstly, the B-spline shape features of the estimated MET are extracted and employed for improving the measurement partition, especially for the closely spaced METs. Then shape-feature-based association matrix between the measurements and the predicted targets is calculated for identifying the newborn targets and reducing the time complexity of the Gaussian component calculation. Moreover, the extracted shape features are also used for target state update and classification, which can further enhance the tracking accuracy. Experimental results show that the proposed algorithm has a high execution efficiency and tracking accuracy for the complicated scenarios with irregular-shaped closely spaced METs and unknown newborn target densities.