Depth information extraction of on-board monocular vision based on a single vertical target image

Calculating obstacle distance for on-board monocular vision system on intelligent vehicle was investigated when the obstacle has been detected. An extraction method of depth information only using a single target image without any internal camera parameters was developed. The mapping relation between image row pixel values and the actual imaging angles was established with the image of vertical target, which was placed in the front of a camera. The obstacle depth information was extracted in real time by combining the projection geometry model. Given the characteristic of vertical target image, an algorithm of sub-pixel corner detection and location was designed, which includes region of interest setting, template matching, candidate points clustering and screening and precise location, etc. Experimental results show that the method has high precision and real-time performance. Compared with the method of putting reference on the road, it does not need large calibration site and could avoid the data fitting error. And the method also has a simple calibration procedure with a single image, which is suitable for practical application.