Model-Free Segmentation and Grasp Selection of Unknown Stacked Objects

摘要: We present a novel grasping approach for unknown stacked objects using RGB-D images of highly complex real-world scenes. Specifically, we propose 3D segmentation algorithm to generate an efficient representation the scene into segmented surfaces (known as surfels) and objects. Based on this representation, next grasp selection which generates potential hypotheses automatically selects most appropriate without requiring any prior information or scene. tested our algorithms in scenarios live video streams from Kinect publicly available object datasets. Our experimental results show that both proposed consistently perform superior compared state-of-the-art methods.