Real-Time 3D Object Detection Algorithm Based on the Fusion of Image and Point Cloud

3D object detection based on the fusion of cameras and LiDAR has now made significant breakthroughs. There are many schemes that utilize different features of cameras to assist LiDAR for 3D object detection with promising results, but most of these network architectures consume a lot of computational resources. In this paper, we propose an IoU-Aware fusion network based on point cloud and image, IAOR for short, which is a simple network framework that significantly improves the performance of point cloud 3D object detection and further solves the problem of inconsistent classification and prediction of point clouds. We evaluate our proposed fusion algorithm on the KITTI object detection dataset. The results show that our algorithm can significantly improve the object detection accuracy of most existing 3D detection algorithms, especially for small objects such as pedestrians and cyclists, and the proposed fusion algorithm is much faster than most fusion algorithms. The proposed fusion algorithm is also environmentally adaptive as verified by a real vehicle platform.