基于改进ICP的建筑物点云配准算法OA

To solve the problems of large errors and slow speed in point cloud registration,this paper proposes an improved ICP point cloud registration method based on the fusion of normal vector constraints and Euclidean error.Firstly,based on the use of the Internal Shape Feature(ISS)algorithm to extract key-points,the Binary Directional Histogram(BSHOT)is adopted to quickly and efficiently match the binary feature descriptors of key-points,in order to improve the efficiency of point cloud registration.Sec-ondly,it uses dual quaternions to solve the rotation matrix and translation vector to achieve rough registration of point clouds.Final-ly,the ICP algorithm with normal vector constraints is used to achieve overall registration of point clouds.The experimental results show that using open-source point cloud clusters as the dataset,the method improves the registration efficiency by 33.38%and the registration accuracy by 18.32%compared to the SAC-IA+ICP algorithm in building point cloud scenarios.Compared with other point cloud registration methods,the method provides a new approach for high-precision 3D reconstruction of buildings.