Automatic 3-D digitization using a laser rangefinder with a small field of view

Abstract

We address the problem of the automation of surface digitization using a precision 3-D laser rangefinder. Because of their small field of view, such sensors navigate closely to the digitized object and are subject to collisions. Unlike previous techniques that addressed only the exhaustiveness of digitization, this work focuses on collision avoidance. To safely identify empty space, shadow and occlusion phenomena are carefully analyzed, and so is the effect of sampling. Then, the planning problem is solved using a hierarchical approach. At low level, we digitize a single view and address collision avoidance using path planning techniques. At high level, the problem becomes the choice of the next best view. Some implementation details and experimental results are presented.