An optimization method for point cloud registration based on the rate of point cloud density variation

An optimization method for point cloud registration combined with the spatial density change rate of point clouds is put forward. This method is utilized to address the issue where the registration optimization process is deviated due to a large number of mismatched point pairs, resulting in the algorithm converging to the wrong position. Adaptive voxel down-sampling is carried out based on the density difference of voxel grids. Mismatched points are filtered through the density change rate. The ratio of the spatial volume change to the point cloud quantity change under different radius coefficient weights is restricted within an appropriate range. This enhances the probability of inner points being selected on the basis of reducing the sampling set, thereby enhancing the accuracy and robustness of point cloud registration.