A Review of Image and Point Cloud Fusion-Based 3D Object Detection for Autonomous Driving

With the application of multi-sensor fusion technologies in object detection, Self-driving automobiles have received more and more attention. The complementary advantages of different types of sensors (camera, LIDAR) can provide accurate and reliable environment perception for autonomous driving systems. We categorize the 3D object identification methods based on image and point cloud fusion into early, intermediate, and late, focusing on the camera and LIDAR combination in autonomous driving, furthermore, the characteristics of various methods and the performance of different models are analyzed and compared. Finally, the possible future opportunities and challenges for 3D object detection are evaluated.