P‐2.22: Research on 3D Target Detection Algorithm in Automatic Driving Scenario

Abstract

As a research hotspot in computer vision, 3D object detection has wide application prospects. In the field of autonomous driving, the perception system with three-dimensional object detection function enables the vehicle to perceive the surrounding environment, make the vehicle more intelligent, and play a role in assisting driving. Aiming at two commonly used sensors: monocular camera and lidar, combined with deep learning, this paper separately studies the 3D target detection algorithm of PointPillars based on attention mechanism and the 3D target detection method based on monocular vision. The results show that the detection performance based on monocular vision still has advantages and can be used as an effective supplement to Lidar based three-dimensional target detection, improving the robustness of the sensing system.