Research On Overhead Crane Positioning System Of Steelmaking Workshop Based On Vision And Lidar Fusion

This paper mainly studies a target detection system for steelmaking workshop base d on camera and lidar to improve the positioning and tracking ability of overhead cranes in complex environments.The system first collects data from the environment in the ste elmaking workshop through the camera and lidar,real-time detection of the overhead cr ane in the steelmaking workshop through the Yolov5 deep learning detection network,a nd object detection on the point cloud,and then fuses the information to finally achieve target tracking and positioning.The main directions of thesis research include:(1)Design the overall scheme of the system,respectively carry out the system arch itecture from hardware and software,and select sensors according to actual needs.(2)Study the internal parameter calibration of cameras and the joint calibration met hods of cameras and lidar to lay a foundation for information fusion.(3)Based on the Yolov5 model,the target detection and tracking of the overhead c rane is carried out on the image,and the dataset is established according to the site situat ion to simulate the respective working environment to increase the robustness of the trai ned model.(4)First,the point cloud is segmented based on the depth map of the point cloud,a nd then the target recognition of the crane in the point cloud is carried out based on the CVFH descriptor,and the system also synchronizes the positioning and tracking of the t hree-dimensional data of the crane hook,and fuses the positioning of the hook and the p ositioning of the crane with Kalman filter,which optimizes the positioning accuracy of t he crane.(5)The camera-based target detection results and lidar-based target detection result s are fused,and the fusion strategy is designed according to the site situation.The experimental results show that the target detection and positioning system has good performance in terms of accuracy,detection speed and positioning accuracy,and c ompared with the single sensor system,the system can make full use of the advantages of multiple sensors to improve the positioning accuracy and robustness of overhead cran es in complex environments.This paper studies the key technologies including Yolo V5 model,depth map point cloud segmentation,CVFH descriptive sub-point cloud recognition,Kalman filter data f usion,camera and lidar fusion technology,and the new overhead crane positioning syste m developed through the research of these key technologies provides an important auxil iary role for the fine management and intelligent operation of the overhead crane in steel making workshop,which has important practical significance for material tracking man agement and production efficiency improvement of steelmaking enterprises.