Three-dimensional field of view estimation in occlusion-rich environments

Abstract

Human-symbiotic robots will work mainly in daily living spaces for humans. Since generally there are many objects which can cause visual occlusion in these environments, it will often occur that a robot cannot see an object by occlusion while a user can, and vice versa. In such situations, it is desirable for the human-symbiotic robot to be able to interact with a user while recognizing a difference between their fields of view. In this paper, we propose a system for estimating a user's three-dimensional field of view using a camera and depth sensor. First the user's head position is detected using the particle filter, and then its pose is estimated using histograms of horizontal edge components in the head image. Finally, based on the specified head pose, the user's three-dimensional field of view are detected in pixels in the image. Experimental results show that our method works effectively. It is expected that the proposed system enables the robot to recognize regions where the user can see and cannot see from his or her own position, and to behave appropriately while recognizing the difference between their perceptions.