Point clouds registration based on constant radius features for large and detailed cultural heritage objects

The registration permits to positioning in a single reference system point clouds acquired from different points of view. Since this is typically obtained with an iterative numerical method, it represents an important source of error in the entire reverse engineering process. As all iterative methods, such errors depend on the choice of the initial solution; therefore, this process requires an expert user who, by using dedicated software, choices the sequence of clouds to be registered, imposes for each pairwise the first attempt registration, launches the iterative method, and verifies the final result. With the aim to minimize the error and the user's interaction, some devices are proposed in the market (turntable or the anthropomorphic arm, etc.). The above-mentioned hardware and software tools cannot be used in the cultural heritage applications involving large and detailed objects. In this paper, an automatic alignment method of point clouds is proposed. The method uses as inputs the constant radius features, which are frequently detectable on cultural heritage objects. The automatic alignment of the point clouds is based on the recognition, the segmentation, and the registration of the sweep lines identifiable from these features.