Visual Feature Extraction and Topological Analysis for Inferring the Configuration of Hanging Cloth
This paper presents algorithms for inferring the configuration of a hanging rectangular piece of clothing from a single range image. This is the first step towards unfolding a towel using a dual arm robot. The algorithm employs feature detection followed by a representation of the surface by means of several layers of overlapping 3D polygons. By analysing the topology of these layers, their mapping to the original (unfolded) rectangle is achieved. The final mapping allows reconfiguration on the garment or grasp planning to achieve unfolding.