Incomplete vehicle information completion using generative adversarial network to enhance the safety of autonomous driving

Autonomous driving is the key technology of intelligent logistics in Industrial Internet of Things (IIoT). 3D vision using Light Detection And Ranging (LiDAR) under vehicle industrial standard is the rigid demand in autonomous driving due to its lower cost, more robust, richer information, and meeting the mass-production standards. However, the appearance of incomplete point clouds losing geometric and semantic information is inevitable owing to limitations of occlusion, sensor resolution, and viewing angle when the LiDAR is applied. The emergence of incomplete point clouds, especially incomplete vehicle point clouds, would lead to the reduction of the accuracy of autonomous driving vehicles in object detection, traffic alert, and collision avoidance. Therefore, the point fractal network (PF-Net), a precise and high-fidelity 3D point cloud repair network based on Generative Adversarial Network (GAN), is first applied to repair incomplete vehicle point clouds in autonomous driving. To evaluate the performance of the GAN-based point cloud repair network, an autonomous driving scene is created, where three incomplete vehicle point clouds are set for different autonomous driving situations. Experimental results demonstrate the effectiveness of the PF-Net for challenging vehicle point cloud completion tasks in autonomous driving.