Stereo Vision Based Parking Assist System

This paper presents semi-automatic parking assist system, which performs stereo vision based recognition of free parking site and parking guidance with automated steering operation. Pixel structure classification and feature based stereo matching extracts the 3D information of parking site. Parking site marking is separated by plane surface constraint and is transformed into bird's eye view, on which template matching is performed to determine the location of parking site. Obstacle depth map, which is generated from the disparity of adjacent vehicle can be used as the guideline of template matching by limiting the search range and the orientation. Proposed method using both the obstacle depth map and the bird's eye view of parking site marking increases the operation speed and the robustness to visual noise by effectively limiting the search range. Using this map information, geometrical relation is tested and efficient parking path is generated. Desired steering angle data correspondent with wheel travel is transmitted to electric power steering via serial communication. This system was implemented on real passenger car and tested in practical parking situation.