The goal is to understand the kind of occlusions that arose by the scanning. Our input is a range data set that provides 3D spatial information related to the sensor position. The data contains depth information that we exploit for detecting occlusions.
In order to detect occlusion is more convenient to switch from true 3D point-sets to a 2.5D representation, i.e. a range image. In this way, we observe the scene from the sensor viewpoint, that is, we look at occlusions the same way the sensor does.