Classification information-assisted adaptive target detection and tracking method

To address the challenges of target omission and data association errors in radar multi-target detection andtracking in cluttered environments, we respectfully propose a framework that integrates joint classification information, target detection, and target tracking. The classification information of targets and clutter based on distance-Doppler spectrum is utilized to assist detection and tracking. In the detection stage, we have designed an adaptive detection method to adjust the detection threshold based on the clutter background where the missed target is located, enabling re-detection. Furthermore, the classification information has the capability to filter the clutter measurements generated by re-detection. In the tracking phase, a neural enhancement message passing algorithm is used to unify the data as -sociation and measurement classification module. The Dempster-Shafer rule is used to fuse the probabilistic data as-sociation confidence and classification confidence, which may help to better allocate the confidence between different information sources, effectively solve the problem of information conflict, and improve the accuracy of data associa -tion. It is thought that this framework could be an effective way of linking classification information, target detection and target tracking. Experimental results combining radar measurement data and simulation data suggest that the pro -posed method could help avoid the problems of incomplete target tracking caused by target omission and the problem of erroneous shutdown between clutter and target. This method could improve the performance of radar target detec-tion and tracking in cluttered environment.