Research On Key Technologies For Point Cloud Registration Using Depth Camera In The Safety Distance Detection Of Hazardous Chemical Warehousing

Stacking safety distance detection is an important step in ensuring the safety of hazardous chemical storage.Traditional methods such as video surveillance and manual inspection and supervision are difficult to quickly detect safety hazards.By using two-depth cameras to collect point clouds from different locations,it is possible to obtain three-dimensional spatial coordinate information of hazardous chemical warehouses while solving the problem of view occlusion between stacks and completing intelligent detection of safe distances.Among them,point cloud registration,as a key link,faces technical challenges such as a large amount of storage point cloud data,similar cargo characteristics,and a low overlap rate of registration scenarios.The paper focuses on the point cloud registration algorithm in the safe distance detection process of hazardous chemical storage,and has done the following work:In order to solve the problems of large data scale and similar features of multiple cargo stacks when extracting 3D point cloud features in warehousing,the H-SURF feature extraction algorithm is proposed.Combining the RGB image,depth map,and obvious corner points of the rectangular stacking of the target scene,Harris corner points are extracted using grayscale difference optimization and fused with SURF feature points to construct an H-SURF feature descriptor to enhance the feature description of the RGB image and complete feature matching in two-dimensional space.The algorithm improves the accuracy and timeliness of feature extraction from warehouse point clouds.In order to solve the problems of multiple mismatched points and poor algorithm efficiency in point cloud registration under the conditions of large storage area and low overlap rate of hazardous chemicals,a point cloud registration algorithm based on H-SURF feature regions is proposed.By combining depth maps with camera parameters to construct a mapping relationship,the matched H-SURF feature descriptors are mapped to the 3D point cloud space to form a 3D H-SURF feature descriptor for point cloud coarse registration.Then,the H-SURF feature regions are intercepted to improve the scene overlap rate for point cloud fine registration,solving the problem of the ICP registration algorithm easily falling into local optimal solutions in low overlap scenarios and reducing the registration time of point clouds.In the visualization stage of storage safety distance,this paper proposes a calibration-based way for large-scale coordinate system transformation to solve the problem of excessive accumulated errors in the coordinate system transformation process of point cloud.The camera calibration in the warehouse is used to obtain the corresponding feature points,and the constraint function is constructed using the unit orthogonality of the matrix.Then,the gradient descent method is used to optimize the transformation matrix to reduce the cumulative error,realizing the display of the three-dimensional information of the warehouse in the world coordinate system.In the last stage,the Background-Diff algorithm is used to extract the stacking and complete the safety warning of the warehouse stacking.