RIGID POINT CLOUD REGISTRATION BASED ON α-REGION TRANSFORMATION
With the enrichment of 3D data sets in recent years,3D rigid point cloud registration plays an important role in environment recognition,indoor reconstruction,and surgical operations.Although there are many researches on rigid point cloud registration,these researches cannot deal with the local deformation caused by shooting.In view of this situation,the problem of rigid point cloud registration for α-region transformation is proposed,which can deal with the local deformation of the point cloud by allowing different regions to have different rigid transformations.In order to solve the newly proposed problem,a two-stage rigid point cloud registration algorithm RPCA was designed.For the first stage,the maximum density sub-graph was used to improve the point correspondence to obtain the global transformation.For the second stage,an adaptive algorithm was designed to quickly solve the α-region transformation based on the global transformation.A large number of experiments show that the two-stage algorithm reduces the registration error by at least 48％.Compared with the current method with the best registration performance,its execution time is reduced by 86％.
3D point cloudPoint cloud registrationRigid transformation
国家科技期刊平台
NSTL
万方数据
