Intersection image recognition for driving-view images with edge information

Abstract

While a vehicle driving in the street below tall buildings or viaduct, GPS receiver may receive the satellite reflected signals or may not receive satellite signals. That will cause serious positioning errors. This is called as multipath transmission error of global satellite positioning system in the metropolitan area of a city. With using the driving-view image fetched from the video of vehicle event data recorder, a road subimage is derived for intersection image recognition in this paper. Based on edge ratio of road sub-image, the decision of intersection is determined in the proposed algorithm. Upon completion of intersection image recognition, and then with comparison of geographic information system, the multipath transmission error of GPS will be corrected. Experiments are carried out for some VEDR's videos to demonstration the computational advantage of the proposed method.