Research And Implementation Of Algorithm Of Vehicle License Plate Recognition System

With the rapid development of road traffic, vehicle ownership is increasingly, and how to manage the vehicle becomes the urgent problem for solving in the field of intelligent transportation. Vehicle license plate(VLP) recognition as the key technology in the intelligent transportation field, has widely used in crossroads, high-speed road traffic management, parking fees management and other scenes. Taking advantage of modern computer processing technology and image processing technology, vehicle license plate recognition system achieves the access to vehicle license information. The paper mainly discusses and researches the vehicle license plate recognition technology. It includes three sub-modules researches on the algorithm of license plate area detection, character segmentation and character recognition.1) License plate region detection algorithm. According to the captured image sequences of different field scenarios, this paper presents the license plate region detection algorithm based on multi-information fields voting. The algorithm uses the Gaussian mixture model background modeling for the image sequence to detect the moving vehicle target and real-time determine the target vehicle, while conducts the license plate region detection on the vehicle target image. Using the multi-information feature of plate, selecting more stable edge features as the foundation, making the license plate region background color and character traits as the basis, the license plate region detection makes use of k-means clustering segmenting characters and background to improve the robustness and accuracy of license plate detection and reduce false.2) License plate character segmentation algorithm. According to the license plate region is detected in the first module, the paper attempts to make the character segmentation step to reduce dependence, and puts forward innovatively an adaptive license plate character segmentation method. The methods sets multiple criteria judged to quickly and easily search for character spacing 2, 3 positions based on the projection characteristic and license plate aspect ratio. At the same time use the brightness differences of the license plate characters and background, and combine the projection and spacing characteristics to search the spacing of adjacent characters, and then segment the character.3) License Plate Recognition Algorithm. This paper does not segment the character directly in actual use segmentation algorithm, but tries to improve the accuracy of single-frame and multi-frame license plate segmentation based on the character segmentation, then proposes an innovative new license plate character recognition algorithm framework. The algorithm obtains the same license plate of several test results through judging the target vehicle. For the single-frame license plate character image, conduct the sliding window search character on the center of character segmentation, and for consecutive multi-frame image, conduct the time window search. Making the final single character recognition is the result of search optimization and optimal target character in voting.Algorithm proposed in this paper is tested to verify on vs2010 platform, many experiments made with different scenes captured image field, and experiments show that the proposed algorithm is a good ability of robustness and accuracy.