The issues of combining three-dimensional surfaces and three-dimensional point clouds are important for solving a number of applied problems: three-dimensional reconstruction of the underlying surface of aircraft, construction of three-dimensional models of objects and a number of others. The report deals with the registration of point clouds, in particular, the algorithm of rigid registration of three-dimensional clouds. A software implementation of an iterative algorithm for combining three-dimensional point clouds is described, its main advantages and disadvantages are analyzed, and a modification of the ICP algorithm is proposed, including preliminary combination of the corresponding parts of the clouds using an approach based on three-dimensional descriptors of key points. The use of a combined algorithm for combining three-dimensional clouds allows you to build dense three-dimensional models of objects based on their separate disparate stereo images. The accuracy and time estimates of the proposed approaches are presented.
Keywords:
3D models, three-dimensional point clouds, rigid registration algorithm, iterative nearest point algorithm, combined point cloud matching algorithm
