Finding empty volumes inside 3D models of storage furniture

Abstract

Placing objects inside household storage furniture by a robot requires visual perception capable of detecting empty space inside furniture suitable for object storage. The volume of this empty space must be sufficient to contain objects of interest and it must be located immediately above a flat horizontal surface, which could support these objects. In this paper, an algorithm for finding empty volumes suitable for object storage inside furniture is proposed. The algorithm takes as input 3D models of furniture aligned with axes of a known reference frame and produces 3D bounding boxes of empty space suitable for storage of objects of predefined size. The proposed method is based on the 3D distance transform and connected component analysis. Experimental validation of the algorithm is performed on a dataset with 3D CAD models of storage furniture.