The Research And Development Of Vehicle License Plate Recognition System

Vehicle License Plate Recognition is an important part of the Intelligent Traffic System,which has important application value in management and monitoring of traffic road. VehicleLicense Plate Recognition system is a machine vision system，which takes license plate imageas an object recognition. It involves digital optics, digital image processing, computer vision,artificial intelligence and some other fields. This paper makes a comparison and analysis ofthe relevant algorithms of Vehicle License Plate Recognition, and clarifies the algorithm ofeach link combined with the process of license plate recognition. The function of the fourmajor components in the MATLAB software is implemented eventually. License plate imagepreprocessing, license plate location, license plate character segmentation and license platecharacter recognition compose the four major components. And the software design ofVehicle License Plate Recognition System is accomplished in this paper.1. In the link of image pre-processing, firstly, the weighted average method is used toconvert the RGB color image into gray scale image to improve the processing speed of thewhole system. Secondly, gray stretch, histogram equalization and median filtering processingare correspondingly conducted in view of the images in different qualities. Finally, Robertsoperator is used to detect the edge of license plate.2. In the link of license plate location, firstly, the License plate location is determined byscanning the horizontal and vertical direction projection after the processed image based onmathematical morphology. Secondly, Radon variation method is used to correct the inclinedplate. Finally, row license plate of binarization image is scanned. License plate frame andrivets is removed by referring a priori knowledge.3. In the link of license plate character segmentation, the characters segmentation isrealized according to characters projection of the vertical direction and the prior knowledge ofsingle character. This paper also performs characters normalization processing by usingbilinear interpolation algorithm.4. In the link of license plate character recognition, firstly, it extracts character feature asinput of BP Neural Networks by means of using the grid pixel statistic feature extraction. Secondly, the sample character to form characters recognition network is trained via buildingstructure of net topology. Finally, it tests whether the BP Neural Networks will succeed byusing the test sample.On the basis of the above study, the Vehicle License Plate Recognition system isdeveloped based on MATLAB and the system function is tested through the experiment. Theresults show that the Vehicle License Plate Recognition designed in this paper can orientate,segment and recognize characters accurately. Moreover, the algorithm of the whole system issimple with the characteristics of real-time and robustness at the same time.