Research And Application Of Coal Hopper Material State Detection Technology Based On 3D Reconstruction

Coal hoppers in power plants have problems with material sticking,material blocking,and reasonable distribution of feeding and discharging time.It is necessary to detect the status of coal hopper materials in time.Due to the limitation of the scene,the traditional manual method is difficult to observe and the detection accuracy is low.Therefore,designing an automated coal hopper measurement system has great economic and application value.The method based on binocular vision has the characteristics of low cost and wide range,which is suitable for coal scuttle measurement system.In the actual scene,due to the single characteristics of the coal pile,insufficient light,and large coal scuttle volume,there are holes and weak edge information in the disparity map after binocular vision stereo matching;The generated point cloud has many outliers,low registration accuracy,holes,missing edges and other problems.In order to solve the above problems,this paper designs a coal scuttle material state detection system based on three-dimensional reconstruction,and conducts state measurement verification through a small coal scuttle.The main research and results are as follows(1)Aiming at the problem of weak edge information and holes in the disparity map of coal piles,this paper improves the disparity optimization algorithm based on Semi Global Matching(SGM)algorithm.This method uses the canny operator to calculate the edge cost and retains the edge information of the coal scuttle.According to the disparity map,the occlusion area and the mismatch area are judged,and the second smallest disparity and median are filled.Experiments show that compared with the original SGM algorithm,this method can effectively detect the edge information of the coal scuttle and has a good filling effect on the holes in the disparity map of the coal pile(2)Aiming at the problems of low accuracy of coal piles point cloud registration and a large number of noise points,this paper improves the point cloud registration method based on Iterative Closest Point(ICP).This method performs statistical filtering,voxel filtering and down-sampling on the point cloud,and uses the consistency estimation to perform coarse registration and ICP fine registration.Experiments show that compared with common registration algorithms,this algorithm has the best speed while satisfying accuracy.(3)Aiming at the holes and missing edges of the coal pile point cloud,this paper designs a hole repair algorithm based on spatial grid division.This method divides the point cloud into voxel grids,interpolates and iterates for each grid,uses the edge grid information to continuously expand outward,repairs holes and missing edges,and obtains a complete coal pile point cloud.Experiments show that this method can effectively repair the holes and missing edges of the coal pile point cloud.This paper builds a coal hopper material state detection system based on 3D reconstruction.The system reconstructs the three-dimensional point cloud of the coal scuttle through binocular vision and detects the state of the coal scuttle material.Experimental measurement shows that the system is easy to operate and has good repeatability.It can detect the feeding,discharging,and half-storage states of the coal hopper material to meet actual production requirements.