An Improved GM-PHD Algorithm Based Multi-Target Tracking Method for FMCW Radar

This paper proposes a multi-target tracking method based on an improved Gaussian Mixture Probability Hypothesis Density (GMPHD) algorithm using a 4D millimeter-wave radar. The approach first extracts the target position and velocity data, initializes Gaussian components, and constructs the state and covariance model. The Cubature Kalman filter (CKF) method predicts target states, while the Probability Hypothesis Density (PHD) model dynamically adjusts the target count. A measurement-assisted weight correction strategy refines Gaussian components using maximum likelihood estimation and Kalman gain to enhance the tracking accuracy. Finally, pruning and merging reduce redundant components and computational complexity. Experimental results confirm the improved tracking accuracy with lower complexity in cluttered environments.