Point Cloud Based In-vehicle Occupancy Detection Method by Using 77GHz mmWave Radar
Xinda Cao, Jiangang Liu, Yulin Wu, Fengzhi Shao, Yübo Wang, Guolong Cui · 2024
In recent years, with the introduction and development of vehicle-to-everything (V2X) and child presence detection (CPD), there’s an increasing demand for in-vehicle perception systems. Millimeter-wave (mmWave) radar has become one of the mainstream sensors in this field for high accuracy, small size, and low power consumption. In this paper, we develop a real-time in-vehicle occupancy detection method based on 77GHz mmWave radar. Firstly, a pre-processing framework based on dual-path detection is used to obtain more point clouds from weak targets. Then density-based spatial clustering of applications with noise (DBSCAN) is applied to classify the occupants’ point clouds. Finally, we propose a novel probabilistic population-assisted occupancy detection algorithm based on the long and short-term feedback results, which can suppress missed detection and false alarms well. Our algorithm has been successfully deployed onto the radar board and achieves an average accuracy rate of 99.10% under various complex scene experiments.