| With the rapid development of China’s economy and the sharp increase of demand fortransport, the domestic highway already has a certain scale,but most of the highway havealready used artificial charge. The drawback is that vehicle must stop when through the tollstation, so firstly it will cause traffic congestion of toll station, secondly, parking vehiclefuel consumption, queuing will be a waste of time and automobile exhaust causesenvironmental pollution. This phenomenon has caused more and more people dissatisfied.In order to solve these problems, research and develop ETC system will become the urgentneeds of the traffic in our country. At present ETC system has been used in the bridges andtunnels of electronic automatic charging system in the world.ETC system has been used in the bridges and tunnels of electronic automatic chargingsystem. General owners have installed ID card on window glass and Stored charge, onlywhen the vehicle through a toll station on the specified the ETC lanes, toll fee can bededucted automatically from the ID card, the whole process does not need manual operationand to stop, it takes generally less than two seconds, the system traffic capacity is5~10times of artificial charge lane.In this paper, based on the license plate recognition of ETC system,the author of this paper has made the detailed research on the video of key frame extraction, license plate locali-zation, character segment and character recognition,including：Key frame extraction: Designs a key frame extraction algorithm based on mutualinformation, the method using the mutual information between two consecutive frames inthe color to measure the similarity of two frames of images. The first to use continuous inter-frame difference of mutual information decomposition of the video sequence of thelens, then the average mutual information through a lens in a sequence of quadraticclustering on candidate key frames, finally, the mutual information for candidate key frameaveraging, take the most close to the average frame for the final key frames. Experimentsshow that this algorithm is simple, fast, easy to implement, and it will be better reflect theoriginal video content.License plate positioning: The method based on License plate character and colorfeatures. Firstly, the method uses the color edge operator to extract the image edge, then iteliminates a large number of useless information by density scanning according to thetexture features of License plate. Secondly, using mathematical morphology operation tofind the candidate plate region and mark the area, Thirdly, according to the license plate ofthe inherent features and background features selected in accordance with conditions of thelicense plate area. This method can get relatively satisfactory results in the complex andinterference conditions.Character segmentation: Character segmentation in locating the license plate area,Firstly, the License plate area is converted into grayscale and binarization. Secondly,removed the borders by horizontal and vertical scanning, Thirdly, License plate area isnormalized, Fourthly, according to the prior knowledge of license plate characters withvertical projection method to correct character segmentation.Character recognition: Character recognition is the most key step of license platerecognition system. this paper uses the combination of template matching and feature pointsmatching method. This method design of three template library, including Chinesecharacters, letters, letters and alphanumeric. In order to correct identification ofcharacters,15feature points were extracted from each character and the character similaritycomparison in the template library. This method is simple and effective, easy to implement,improve the character feature extraction time, speed and recognition rate.Based on the above research, this paper designed a license plate recognition system,The experiments show that the license plate recognition system has some real time andaccuracy. |
PDF Full Text Request