Stereo Vision-Based Start-Inhibit for Heavy Goods Vehicles

Abstract

This paper presents a robust method for obstacle detection with stereo cameras. Arbitrarily aligned cameras are calibrated using a dense grid; a direct mapping between image pixels and world points is made to remove lens distortion and perspective in the same pass. Cubic splines are used to recover unknown points not present in the grid. After the transformation phase, left and right images are compared and the differences are analyzed using a polar histogram to detect vertical structures and to reject noise and small objects. World coordinates of detected objects are recovered and fed to the sub-system for further processing and to take appropriate actions. According to experimental results, the proposed algorithm can be useful in different automotive applications, requiring realtime segmentation without any assumption on background. In particular the system has been tested to investigate presence of obstacles in blind spot areas around heavy goods vehicles (HGV). The system presented in this paper is currently installed in a Volvo truck