A Millimeter Wave Radar Target Detection Method in K-distributed Clutter Background

With the rapid development of radio frequency (RF) and integrate circuit (IC), the millimeter wave (mmWave) radar has received a lot of attentions, due to its 4D sensing ability, high resolution, and high integration. By transmitting frequency modulated continuous waves (FMCWs) and receiving their echo signals, mmWave radar obtains the raw data of target sensing. A routine data processing course is first using detecting and clustering to form point cloud and then performing machine learning methods on the generated point cloud. The key to improve the target recognition performance is reducing the missing detection probability (PMS) and the false-alarm probability (PFA) simultaneously in detection method for point cloud generation. However, the current target detection algorithms, e.g. constant false-alarm rate (CFAR), fail to take both the PFA and the PMS into account. In this paper, to address this problem, we design a novel mmWave radar target detection algorithm based on multi-channel sequential detection theory, where a more realistic scenario that mmWave radar would meet is considered: the target detection tasks are on the range-velocity-angle 3-D grid, and the clutter is K-distributed. The experimental results show that the proposed algorithm can achieve much better target detection performance compared with the conventional one, with preset PMS and PFA embedded in the detector.