The System Of License Plate Recognition Under Complex Background

LPR (License Plate Recognition) is the core of intelligent transportation system, which is widely used in violation detection, highway automatic charge and intelligent parking management, etc. LPR includes five aspects—image pre-processing, plate location, slant correction of plate, character segmentation, and character recognition; and covers multi-subject fields, such as:pattern recognition, human intelligence, computer vision and digital image processing. In this essay, the author does research on the technology of plate location and recognition under complex background, and the main findings are presented as follows.(1) image pre-processing. First, gray color images by using weighted average method. Then adjust the image intensity distribution to improve the quality of images with the histogram equalization technique. Finally the disturbing noises in images will be removed with median filtering technology.(2) Plate location. First of all, the author deals with the plates by applying mathematic morphological opening and closing operation to get the possible candidate regions, and then screens those candidate regions to get the initial location of plates. If there are several candidate regions after initial location, we shall employ the color space locating technology based on HSV to re-screen those regions through its color and color difference.(3) Slant correction of plate. The author makes slant correction of plate mainly through rotary projection. After the slant correction, the peak-width of horizontal projection drawing is the narrowest, and the row of vertical projection drawing which the project value is zero, is the most. We can make horizontal and vertical slant correction individually according to these two features.(4) Character segmentation. First of all, we remove its horizontal frame by using the texture, space and width of character, then partition characters through vertical projection, and finally unify the size of segmented characters for convenience of recognization.(5) Character recognization. We recognize characters on the basis of ANN (Artificial Neural Network), because the characters on plates have its own specific range, the author designs three kinds of neural networks for the respective recognization of Chinese character, letters, number and letters.In the last part of this essay, the author employs VC+OpenCV to design and implement LPR, the system can successfully identify vehicle license under complex background.