Point cloud registration using a viewpoint dictionary

摘要: The use of 3D point clouds is currently much interest. One the cornerstones cloud research and applications registration. Given two clouds, goal registration aligning them in a common coordinate system. In particular, we seek this work to align sparse noisy local cloud, created from single stereo pair images, dense large-scale global representing an urban outdoors environment. approach keypoint-based registration, tends fail due sparsity low quality cloud. We propose here new approach. It consists creation dictionary smaller using grid synthetic viewpoints over then perform via efficient search. Our shows promising results on data acquired