Slicing components guided indoor objects vectorized modeling from unilateral point cloud data

Lightweight representation of 3D scene objects is helpful for the effective operation of low-power mobile hardware platforms such as robots and unmanned vehicles. In this paper, we propose a slicing guidance approach efficiently convert the unilateral cloud data of indoor scene into a collection of vector models to reduce the storage space of the scene map. Specifically, we first extract the shape components of the indoor scene by a progressive algorithm based on the cross-section slicing. Then, our approach classify the different primitive shape components according to the curvature of their boundary points, and the primitive shape components are fitted respectively to compensate for the lack parts of original data. Finally, we present a scoring mechanism for component recognition and matching to generate the geometrically faithful model to the input indoor scene. Experimental results demonstrate that our method has better performance in detail description, recognition and vectorized model matching of the unilateral point cloud than existing related methods.