Motion detection system for HRI based on image parameters and HRI context

Motion detection is foundation of service robot vision in Human-Robot Interaction (HRI). Foreground detection is an important step of the motion detection. If the foreground detected is not correct, the motion detected will be wrong. To improve the efficiency and robustness of motion detection for HRI, a motion detection system for HRI based on image parameters and HRI context (MDIPC system) is proposed, which composes image pretreatment, image evaluation, foreground detection, foreground evaluation, and motion detection. At first, the framework of the MDIPC system is introduced. Secondly, evaluation of the Image Mean Filter in HLS color space is proposed. And thirdly, parameters representing the image feature and foreground in the HRI interaction are introduced, and these parameters include changing features of image´s edge and the foreground objects´ size. Experiments of typical interaction scenes validate that the proposed method provides much improved results. Quantitative evaluation shows that the proposed method is suitable for motion detection in HRI.