Research on Target Detection and Tracking Algorithm Based on 4D Millimeter Wave Radar and Machine Vision

In response to the suboptimal accuracy of vehicle detection in complex environments using current millimeter-wave radar and traditional machine vision fusion schemes, this paper proposes a novel algorithmic improvement strategy based on the study of conventional target detection methods. Firstly, the traditional millimeter-wave radar is replaced with a 4D millimeter-wave radar, and an adaptive Kalman filtering algorithm is employed to effectively filter out radar clutter and achieve precise target tracking. Secondly, the enhanced deep visual network MobileNetV2+SSDLite is trained using vehicle datasets to improve the accuracy of visual vehicle recognition. Thirdly, a decision fusion scheme is adopted to integrate millimeter-wave radar signals with visual signals. Experiments conducted in various scenarios demonstrate that the new improved scheme can effectively estimate and track target vehicles, exhibiting excellent vehicle recognition performance in practical application environments, along with higher reliability and robustness.