Panorama Based Point Cloud Reduction and Registration

To reconstruct environments 3D point clouds acquired by laser scanners are registered. This is an important but also time consuming part of any mapping system for mobile robots. The time needed for mapping is drastically reduced when the size of the input data is reduced. This paper examines different ways of reducing the size of point clouds without losing vital information for the matching process. We present novel point cloud reduction methods on the basis of panorama images. It is shown that the reduced point clouds are ideally suited for feature based registration on panorama images. We evaluate the presented reduction methods based on their effect on the performance of the registration algorithm.