Research on Environmental 3D Object Detection Technology Fusing LIDAR Dot-Cloud and Vision Image

With the rapid development of autonomous driving and industrial automation, 3D object detection plays a crucial role in understanding and interpreting the surrounding environment. Single sensors, such as LiDAR or cameras, often face issues of data missing and insufficient information in complex scenarios. To address these limitations, This study designed a 3D object detection method that fusingLIDAR dot-cloud and Vision Image. Firstly, a joint calibration of LiDAR and camera is conducted to unify the coordinate systems, followed by Beidou timing to achieve synchronization of fusion signal time reference. Next, images captured by the camera are input into an improved YOLOv8 algorithm for 2D object detection, generating 2D bounding boxes for the objects. These 2D predictions are then used, along with the joint calibration parameters, to deduce the 3D conical regions of the objects. Finally, the DBSCAN clustering algorithm is employed to obtain accurate 3D information of the objects. The feasibility of this method is demonstrated through hardware platform construction and testing.