Obstacle detection in stereo sequences using multiple representations of the disparity map

Object detection represents an important task in various application fields, such as automotive or assistive technologies. In this context stereo cameras are among the most common devices used for acquiring images and information from the environment. Efficient processing of the disparity maps computed from stereo images represents a crucial step in obstacle detection algorithms. In this paper we provide an application-independent framework for obstacle detection based on disparity processing. In order to identify regions in the 3D environment that can be classified as obstacles we employ multiple representations of the disparity map: V-disparity, U-disparity, θ-disparity. We provide a comprehensive overview of those representations and their use in obstacle detection algorithms. We evaluate the proposed framework in the context of automotive and visually impaired assistive applications, using data from both real and virtual environments.