An Improved Gaussian Mixture-Probability Hypothesis Density Filter for Underwater Multiple Target Tracking in Dense Clutter Scenario

Unknown data association between measurements and targets is the fundamental problem of underwater multiple target tracking (MTT) in passive sonar. Fortunately, owing to its high computational efficiency, the Gaussian mixture-probability hypothesis density (GM-PHD) filter provides sufficiently good performance for solving this data association problem. Nevertheless, when the intensity of newborn targets is unknown or incorrectly initialized, the conventional GM-PHD filter will have a false trajectory or even fail to track the target. Additionally, the existing methods utilize the current measurement set or previous and current measurement sets to estimate newborn target intensity, which makes it difficult to achieve satisfactory tracking results in dense clutter scenarios. Hence, in this study, a novel underwater MTT algorithm based on the enhanced GM-PHD filter is established. Moreover, a sliding window iterative Hough transform algorithm is proposed to eliminate clutter interference and extract newborn target measurements. Furthermore, to estimate newborn target state based on newborn target measurements, the least squares algorithm is employed. In conjunction with the sliding window iterative Hough transform algorithm and least squares algorithm, the improved GM-PHD filter can eliminate clutter interference and improve tracking accuracy. Simulation experiment results reveal the effectiveness of the algorithm.