Vehicle Target Detection and Tracking Method Based on Depth Data

Abstract

Existing geometric feature-based target detection and tracking methods have high false detection rate, and missed detection in target tracking process easily leads to false target correlation. To solve these problems, this paper proposes vehicle target detection and tracking method based on LiDAR depth data. According to lidar depth data characteristics, a grid-based parameter automatic clustering (PAG) algorithm is used to process the original data. In addition, target segments are extracted from each cluster to obtain target features. On this basis, vehicle targets are identified and position information of the targets is calculated. Kalman filter algorithm is adopted to formulate filter management strategy to complete target association and state estimation. Finally, an experimental vehicle equipped with forward laser radar is used to verify the proposed method. Experimental results show that the proposed method can accurately identify and track multiple vehicle targets and avoid erroneous target association.