Study on automatic target identification method in substation 3D scene model

Abstract Implementing safety control at substation sites is an important way to ensure the safety of operators and the normal operation of equipment, and recognizing objects in the three-dimensional scene of a substation is a new direction to improve the level of safety control at substations. The article utilizes LiDAR technology to obtain the 3D point cloud data of a substation, and pre-processes the point cloud data through statistical filtering and voxel downsampling. Then, we use the improved ICP algorithm to align the 3D point cloud data of the substation, and realize the 3D scene modeling of the substation by fusing the point cloud data and inputting it into Unity3D software. In order to realize the automatic detection of targets in the 3D scene of the substation, this paper takes the YOLOv5 algorithm as the basis, introduces the multi-scale feature fusion BiFPN module to obtain more accurate positional and high-dimensional semantic information, and then combines with the CAM mechanism to further improve the model’s recognition accuracy of targets in the 3D scene of the substation. After verified by the self-constructed substation dataset, it is found that the improved YOLOv5 model has a smaller model volume on the basis of satisfying the automatic recognition of substation 3D scene model targets, which helps to realize the safety monitoring of substation operations and improve the safety control level of substations.