Robust ground plane detection and tracking in stereo sequences using camera orientation

The research on assisting visually impaired people to navigate in unknown environments commonly employs image processing or computer vision techniques. One key element that needs to be identified in the surrounding environment of a blind user is the ground plane. Its accurate detection is highly important because the user is able to move freely in that area. An assistive device should primarily assist the user in avoiding the obstacles that lie on the ground surface. The images that will be further processed are usually acquired using depth acquisition devices worn by the user. Therefore, in contrast with automotive or robotics applications, which also heavily rely on ground plane extraction for obstacle detection, the camera orientation has more degrees of freedom. This leads to the requirement of designing more complex solutions for the ground plane detection in the case of assistive systems for the visually impaired. In this paper we introduce an algorithm to detect the ground plane taking into consideration the orientation of the camera, namely its pitch and roll. The proposed algorithm is based on an efficient processing of the v-disparity map associated to each frame. A two-step decision making approach to determine the most suited area that corresponds to the ground plane is described. Experimental results with synthetic datasets are provided to prove the efficiency and robustness of the proposed approach.