摘要 : In the development of autonomous driving technologies, 3-D car detection based on light detection and ranging (LiDAR) has been recognized as a key research topic. However, the performance of LiDAR-based methods is limited by the i... 展开 In the development of autonomous driving technologies, 3-D car detection based on light detection and ranging (LiDAR) has been recognized as a key research topic. However, the performance of LiDAR-based methods is limited by the incomplete point cloud problem caused by signal missing and object occlusion. To this end, we introduce a novel early fusion method called key supplement to tackle this challenge. Specifically, we supplement the key area, where there is a high probability of incomplete point cloud problem, by using the pseudo point cloud generated by images. In addition, we propose a method that takes full advantage of multimodality to reconstruct incomplete point cloud. The experiment results show that the performances of the examined state-of-the-art LiDAR-based methods are improved by incorporating our proposed key supplement method. 收起