Design And Implementation Of Target Detection System For Steel Coil Logistics Warehouse Based On Point Cloud Data

The rapid and accurate detection of the steel coil target in the unmanned logistics transportation of the steel plant is the basis for improving the logistics transportation efficiency and operation schedule.In view of the current situation of inaccuracies in positioning and slow operation caused by manual operation of steel coils in the steel plant logistics warehouse and the in and out transportation,as well as the large reliance on the operator’s experience,there are problems such as delayed message delivery and poor accuracy.Therefore,the traditional steel coil target detection has been unable to meet the high-efficiency operation of the steel coil logistics library.The intelligent steel coil logistics target detection construction is very important.Therefore,this article conducts research based on the target detection project of the steel coil logistics warehouse in the unmanned logistics transportation of a steel plant,mainly including the following research content:Aiming at the current situation of poor manual detection accuracy and low efficiency of the steel coil logistics library,a target detection scheme for the steel coil logistics library based on point cloud data is proposed.The system adopts three-dimensional laser scanning technology,non-contact and all-weather real-time scanning,combined with three-dimensional point cloud processing method to classify the detection and identification of the target,and solves the inaccurate detection of the steel coil saddle in the scene of cars,trains,and warehouse areas in the steel coil logistics library.The problem of realizing low-error and efficient positioning of the target in the steel coil logistics library.Aiming at the characteristics of the actual steel coil logistics library with many types of targets,diverse characteristics,and different shapes,a height difference jump classification method,angle judgment classification method and point cloud position relationship classification method based on target point cloud data are designed.The loading and unloading targets in the steel coil logistics warehouse are analyzed.And the length,width and height characteristics of the object are judged by the height difference jump of the visual blind spot of the point cloud target,the target contour of the target adjacent point angle,the distribution of the point cloud position and the normal vector direction.The target posture successfully classified the steel coil logistics library into car scene,train scene and storage area scene for classification detection and processing,which solved the problems of the large variety of steel coil logistics library targets and the unsatisfactory detection results.Aiming at the defects of poor target positioning accuracy and high detection error in the steel coil logistics library,a set of point cloud data target detection algorithm is designed.The characteristics of different target types are comprehensively analyzed.The algorithm first uses the iterative closest point algorithm(ICP)for stitching.And then combines the voxel grid down sampling algorithm(Voxel)to simplify.And then uses straight-through filtering and spherical radius filtering to remove noise to eliminate burrs.And then uses European-style The clustering algorithm finds the region of interest(ROI),judges the edge with the maximum angle measurement criterion,and finally uses the random sampling consistency algorithm(RANSAC)to fit the target model of the steel coil saddle,and obtains the detection and positioning results from the model parameters.The high-efficiency output of the characteristics of the steel coil saddle in the three scenarios of classification is realized,and the problem of poor detection accuracy relying on manual experience is solved.Finally,the system made detailed scheme design,functional analysis and functional module design,and finally designed and implemented the target detection system of the steel coil logistics library.Obtain the target point cloud data on the hardware,and realize the detection feature output on the target detection software.The function test,communication test,performance test and trial operation verify the accuracy and efficiency of the system.