Research on Robust Detection System of Vehicle Attitude Fusion Features Based on Multi-Sensor Fusion

Abstract

With the rapid development of artificial intelligence technology, the field technology of intelligent vehicles is becoming more and more mature. In order to solve the problem of difficult to extract the positional information due to the vestigial point cloud of moving vehicles acquired by LiDAR in complex traffic scenes, a real-time positional estimation method for dynamic targets based on point normal vector features of multi-sensing fusion is proposed. The least-squares fitting algorithm is used to solve the point normal vectors on the target surface, the principal component analysis is used to solve the number of cluster centers of the point normal vectors by adaptive clustering algorithm on the high-dimensional data, the clustering point fitting plane is obtained, and the mutually perpendicular planes representing the point normal vectors are selected as the X and Y axes of the coordinate system to which the target to be measured belongs, and the multi-target sensing fusion feature technology is combined with the sensing features of radar laser and vision, and the point cloud position calculation is completed. The experimental results show that the method can estimate the posture parameters of the moving vehicle in real time and exhibit good robustness.