A Joint Detection and Tracking Paradigm Based on Reinforcement Learning for Compact HFSWR

Due to its limited transmit power and smaller receiving antenna array, compact high-frequency surface wave radar often encounters increased challenges in detecting and tracking sea-surface targets continuously. In tracking scenarios with dense clutter or multiple targets, weak target signals are often missed due to improper detection thresholds, leading to track fragmentations during target tracking. In order to improve target detection probability and enhance target tracking continuity, a joint detection and tracking (JDT) paradigm, which establishes a closed loop between the detector and tracker, is proposed. When a target of interest is tracked, the tracker sends its predicted range, Doppler velocity, and azimuth back to the detector, then the detector builds a detection gate centered at the predicted range and Doppler velocity on the range–Doppler map. Within the detection gate, an optimal detection threshold dependent on the detection background and tracking environment is determined using reinforcement learning. In this way, a potential target plot may be detected with a higher detection probability and the detected plot is provided for track update. The proposed paradigm employs tracking information to provide adaptive detection parameters for specific targets through reinforcement learning to enhance both target detection and tracking performance. Experimental results with field data demonstrate that compared with traditional detection before tracking scheme, the proposed JDT paradigm achieves a superior performance with the average tracking time on target being increased by 13.33 min and the average missed detection rate being reduced by 0.8<inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$\%$</tex-math></inline-formula>.