Improved GM-PHD-Based Underwater Multi-Target Tracking Algorithm

To address the issues of incorrect estimation, false alarms, and missed detections that occur in traditional Gaussian Mixture Probability Hypothesis Density (GM-PHD) filters when tracking closely spaced targets, an improved algorithm is proposed in this paper. First, a weight redistribution scheme is introduced, which adjusts the Gaussian component weights of the targets to enhance the tracking accuracy of the GM-PHD filter when targets are in close proximity. Next, target label information is incorporated into the target state estimation to prevent tracking errors, such as target swapping, during multi-target tracking when target crossings occur. Simulation results show that, compared to the traditional GM-PHD algorithm, the proposed improved algorithm exhibits better tracking performance.