Sky recognitions for driving-view images

Most of the cars today have facilities of GPS and vehicle event data recorder (VEDR) while driving on roads. When a car drives through underpass, tunnel, or under the viaduct, etc., GPS will not be able to receive the satellite signals and makes the error navigation. In order to resolve this problem, a new sky recognition technique for driving-view image from the video of VEDR is proposed in this paper. Based on the theory of entropy, the road conditions of T-junction, skewed road, and shaded trees are determined in the sky subimage of driving-view image. Then, with analysis and statistics of gray scale image from the sky subimage of driving-view image, the road conditions of tunnel, underpass, and viaduct are separated from the road conditions of sky. Experimental results are carried out with some varied conditions of driving-view images, and show our proposed algorithm succeed in recognition of these road conditions.