A novel occlusion-free active recognition algorithm for objects in clutter

Duofan Jiang, Hesheng Wang, Weidong Chen, Ruimin Wu

Abstract: In this work we propose a new active recognition algorithm for objects in a highly cluttered environment. In contrast to conservative active recognition methods, where robots observe the scene from different viewpoints, our robots are capable of rearranging objects for better perspectives. The whole system bases on an analysis for the occlusion relationships between objects in the scene. After that, a moving strategy is presented to free the objects from occlusion. Experimental evaluations are conducted using a robot arm with an RGB-D sensor mounted on it.