Feature-aided GM-PHD Algorithm for Sea-surface Target Tracking

Target tracking in sea clutter is always a challenging problem due to the weak target returns, especially in terrible sea conditions. Since the radar returns are frequently submerged by strong clutter, conventional tracking algorithms, which only use the dynamic characteristics of the target like position, speed, etc., failed to achieve satisfying tracking performance. Hence, additional features could be introduced in order to distinguish target and clutter effectively. Specifically, this paper introduces three features in the time domain, frequency domain, and time-frequency domain respectively, and combines them into the conventional Gaussian Mixture Probability Hypothesis Density (GM-PHD) multi-target tracking algorithm. Moreover, we utilize Multiple Hypothesis Tracking (MHT) to manage the track for further improving the tracking results. The experimental results on the CSIR public set verity the effectiveness of the proposed method.