Research On 3D Reconstruction Technology Of Hazardous Chemicals Warehouse Based On Two Groups Of Binocular Vision

In the related fields of hazardous chemicals industry,the safety of hazardous chemicals storage has always been a key issue of concern,which is directly related to urban safety and production safety.The traditional supervision methods of hazardous chemicals are mainly video monitoring and manual supervision.However,in the actual supervision process,there are some problems,such as the limited field of vision of the monitoring camera and the occlusion of the stacked goods in the warehouse,which greatly reduces the effect of the safety supervision of hazardous chemicals warehouse.In order to solve the existing safety supervision problems of dangerous chemicals warehouse and realize the intelligent safety detection of the storage distance of dangerous chemicals warehouse,this paper combines the relevant technology of machine vision and image processing to carry out the research of 3D reconstruction and technical supervision of the large scene environment in the warehouse(1)Using Zhang Zhengyou’s calibration algorithm to complete the calibration of the binocular camera and the joint calibration of the two positions,optimize the joint calibration results,and obtain the high-precision binocular camera internal and external parameters and the transformation matrix representing the position relationship between the two binocular cameras.(2)An improved matching algorithm is proposed,and the traditional algorithm is weighted by introducing a two-dimensional Gaussian distribution weighting coefficient with a mean value of zero,so that the matching attention is focused on the center part of the matching block;the introduced statistics are used for the texture characteristics of the image.Description,optimize the initial disparity map to obtain a high-quality disparity map.(3)Based on the calibration results and the disparity map,the warehouse scene is reconstructed from the point cloud to obtain the original point cloud.After the original point cloud is preprocessed,the two sets of point clouds are spliced using the optimized transformation matrix.The splicing error is only It is 22.35 mm,which realizes the rapid unification of the point cloud coordinate system.(4)The color model is used to extract and reconstruct the ground cordon,and the image difference is used to extract and reconstruct the stack.By judging whether the cordon point cloud and the stacking point cloud overlap,it can be judged whether the stacking is out of bounds to achieve safety.Early warning.