A license plate localization method based on region narrowing

Present a license plate localization method about narrowing region gradually. According to the smooth curve shape of gray-scale histogram, distinguish between daytime and nighttime for license plate images. For daytime image, with longitudinal difference, iterative binarization, and removing-noise erosion, give prominence to the transverse lines of vehicle body; with the longitudinal projection smooth curve, determine the left and right borderlines of vehicle body; eliminate the both sides of the ground, and narrow the image region transversely. For the nighttime image or the daytime image after transverse narrowing, with transverse difference, transverse projection, curve smoothness, curve zoom, and curve segment attribute analysis, determine the top and top borderlines of license plate region, to narrow longitudinal image region. Then, with the longitudinal projection smooth curve of transverse difference image, determine the left and right borderlines of license plate region, to narrow transverse image region. Hereinto, utilize the methods about three-point average curve smoothness and curve proportional zoom for many times. Experiments show that the methods in the text provide a solution to the license plate localization of small-size daytime and nighttime images of the over-speed vehicle on freeway, and the methods have high practicality.