YOLO and K-Means Based 3D Object Detection Method on Image and Point Cloud

抄録

LIDAR based 3D object detection and classication tasks are essential for automated driving(AD). A LIDAR sensor can provide the 3D point cloud data reconstruction of the the environment surrounding a vehicle. However, the detection in a 3D point cloud still needs a strong algorithmic challenge to bring the AD to real life. This paper consists of three parts focuses on the realize the 3D object detection function. (1)LIDAR-camera calibration. (2)YOLO-based detection and Point Cloud extraction, and (3) k-means-based point cloud segmentation. In our research, we used a camera that can capture an image to achieve real-time 2D object detection by using YOLO to transfer a bounding box to a node whose function was to make 3D object detection on LIDAR-based point cloud data. A high-speed 3D object recognition function in GPU can be achieved by comparing whether the 2D coordinates transferred from the 3D points are in the object bounding box or not and by performing a k-means clustering