Research On Data Registration Method Of Terrestrial 3D Laser Scanning

With the socio-economic and scientific and technological advances constantly,as well as the ground three-dimensional laser scanning technology is increasingly mature,making the scanning technology can be gradually and widely used in reverse engineering,graphics and other fields.This technology can make three-dimensional scanning of the measured entity directly without touching the measured entity,get the 3D point cloud data and the texture information of the solid surface,so that the measured entity is digitized and the three-dimensional reconstruction is possible.In the three-dimensional reconstruction,the final three-dimensional model not only contains the spatial geometry of the measured entity,but also contains the spectral radiation information of the solid surface,which can restore the measured entity information to the greatest extent.In the scanning process,because of the characteristics of the technology itself and the limitations of the measured entity,often using a multi-site scan to obtain a complete entity of three-dimensional information.And different stations point cloud data in which the coordinate system is different,which requires the point cloud registration.And the registration of the multi-point cloud can be completed by the two points of cloud registration,so the study of the three-dimensional reconstruction process is the key to the registration of two points of the cloud registration,which can directly affect the accuracy of three-dimensional reconstruction.This paper focuses on the registration of 3D point cloud data,which mainly includes the following aspects:(1)This paper introduces the classification of the point cloud registration method and the source of the error and the point cloud registration propagation model.(2)From the existing registration algorithm proposed in this paper used in the registration algorithm.In this paper,the initial registration and then fine registration,the initial registration using the main direction of the paste method,this algorithm can be faster and more accurate to reduce the two point cloud data between the rotation and translation of the dislocation,For the subsequent fine registration to provide more accurate registration data;fine registration using the improved ICP algorithm,in the ICP algorithm,the search for the nearest point is used in MATLAB built-in function delaunay,the optimal analysis of the use of is the SVD algorithm,through which the improved ICP algorithm can more quickly iterate out the optimal registration parameters.(3)In this paper,we use the Leica scanstation2 type three-dimensional laser scanner to scan the staircase handrails on the third floor of the survey building and a washing machine,and obtain the three-dimensional point cloud data of two stations and three stations respectively,and then use the collected data to the proposed algorithm is validated.That is,the use of these two sets of data to verify the improved ICP algorithm and the initial registration after the registration method is feasible,and with the standard ICP algorithm for comparison.The experimental results show that the proposed algorithm has a significant improvement in the accuracy and speed of point cloud registration.