Change detection on points cloud data acquired with a ground laser scanner

Ground based laser scanning is now well settled in the fields of cultural heritage, as-built modelling and monitoring applications. We intend to use laser scanning to detect changes on building sites or inside facilities, mainly for monitoring purposes but also to be able to provide synthetic information in case of an emergency. In both cases, the main constraint is time. A minimum precision on the measure of movements is also required, depending on the type of application. Laser scanner seems to be the perfect tool for such applications as it quickly acquires a large amount of accurate 3D data. But the comparison of so huge datasets implies the use of appropriate structures and ad-hoc algorithms. A specific octree structure is described, and then several simple cloud-to-cloud comparison techniques are presented. The best one, based on the Hausdorff distance computation is improved on various points. Also, as a full automatic process seems still unachievable, a software framework has been developed. It intends to minimize human intervention and therefore prevents from wasting the LiDAR speed in time-consuming post-processing operations.