Rear obstacle detection system based on depth from kinect

This paper presents a rear obstacle detection system based on depth information obtained from Kinect sensor. The proposed system can be used for parking assistance applications and backup aid. It improves the false detection rate in single view based algorithms by using depth information. In addition, real distance to obstacles can be calculated by depth information. It is possible to alarm to driver effectively according to distance. Although the proposed system is based on conventional edge detection algorithm, we confirm the feasibility of combining depth map with color image by improving the accuracy of detection. Experimental results show that depth fusion based obstacle detection algorithm is more accurate than single view based algorithms in terms of false detection rate.