Obstacle detection with 3D camera using U-V-Disparity

Obstacle detection has been one of the most critical features for reliable driving scene analysis. This paper presents an approach for an automatic obstacle detection system. The proposed system makes use of depth information generated by a 3D camera mounted on the front of a moving vehicle. Obstacles projected as line features in the V-U-Disparity map can be extracted to detect the road surface and obstacles. A Steerable Filter is employed at an early stage to dramatically lower the noise. Furthermore, a modified Hough Transform is placed to extract the straight line feature from the depth map with improved accuracy. The system is robust in dealing with fault detection caused by roadside features which is a commonly shared problem in many other obstacle detection approaches.