Based PCL No Target 3D Point Clouds Registration Of Building

3D point cloud modeling is widely used in civil engineering,cultural relics protection,digital urban construction,settlement deformation,disaster analysis and many other fields.Point cloud registration is the foundation and key technology of the above model.3D point cloud registration refers to the process of correcting the multi view coordinates of complex objects acquired from multiple stations to a single view coordinate system.The accuracy and speed of registration has always been the focus and difficulty of point cloud registration research.This paper uses the no target registration ideas of“first coarse registration,after fine registration”to study the registration process deeply.The classification of registration algorithm is summarized,and the registration of the feature space,similarity measure and search strategy in the planar image registration is extended to 3D point cloud registration.Coordinate correction can be achieved by using the orientation histogram feature(SHOT)as a descriptor and using random sample consensus(RANSAC)to estimate the corresponding points and transform matrix in the process of coarse registration.In the process of fine registration,the parameter rational setting of the Iterative Closest Point(ICP)algorithm is explored,some improvement of the algorithm is realized.For the uneven distribution of point cloud,the minimum space point is presented to assess the accuracy of registration reasonably.Depth camera Kinect and terrestrial 3D laser scanner are used to collect building data,and the computer experiment environment is setted up by combining point cloud database PCL with Microsoft Visual Studio 2013.Finally,the experimental results show that the registration is accurate and reliable,has strong robustness,but the speed needs to be improved.