Practical and Robust MLS-based Integration of Scanned Data

Abstract

The paper proposes a set of techniques for improving the quality of MLS surfaces reconstructed from point clouds that are composed by the union of many scanned range maps. The main idea of those techniques is that the range-map structure should be exploited during the reconstruction process and not lost in the uniform point soup that is usually fed into reconstruction algorithms; on this purpose a set of per-range-map weighting schemes are proposed. The presented weighting schemes allow to cope with some of the various issues that usually arise during the integration of point clouds composed by set of range maps, like tangible alignment errors, anisotropic error on sensor data and sensible difference in sampling quality.