| Automatic vehicle plate identification plays an important role in the intelligent control and management of the modern traffic. As the key technology of automatic vehicle plate identification, the license plate recognition (LPR) has attracted more and more attention.In this paper, some characteristics are gotten based on automobiles vehicle plate analysis. The key technologies of LPR are researched and discussed based on these characteristics by using the technique of digital image processing and Neural Networks.Preprocess of image, license-plate locating, image binarization , character segmentation and character recognition are mainly expatiated in this paper. After that, the LPR prototype system is developed.After the experiments of various vehicle images, it has been found out that preprocess of image is more appropriate. These preprocess includes grey, stretching the gray of image, median filtering, edge detection. Through the processing, the noise on the image is eliminated effectively and vertical texture is clearer. Location of the license plate is the key step in the LPR technology. A method of the window search and shift scan is proposed based on vehicle analysis. According to the general style of automobile license plate, a method based on Hough transformation is proposed to acquire the parameters of four edge of the license plate. Incline license plate is corrected by distorted image. The method of whole dynamic iterative is to get the threshold value. Using the threshold value, plate image is processed to binary of image. Modulus is discussed how to effect the binary of image.A method is proposed to acquire the location of plate character based on vehicle plate model characteristics. The method also uses window search and shift method. Because the locations of characters have relation to template characters, the template character is classified in the first. The character recognition is based on Neural Networks classifiers integration. |
PDF Full Text Request