Research And Implementation Of License Plate Recognition Technology

Vehicle is one of the most important tools in modern traffic.This paper studies deeply on license plate recognition technology,from the vehicle license plate location, license plate character segmentation,license plate character recognition three aspects, based on the research to the domestic and foreign literature.To improve the efficiency of license plate recognition,license plate location,character segmentation of respectively have been optimized and improved,which makes the system can be more accurate identification of license plate information.In license plate recognition technology,first of all, to pretreat the collected images.This method can filter out the interference information, improve the image processing speed, the collected images to gray-scale by weighted average method. In the image preprocessing process.Gray-scale image by histogram equalization to improve the contrast of the image, the target image is highlighted in the entire license plate image in favor of the subsequent processing.In the license plate positioning, analyzes the basic characteristics of the license plate, license plate location several research methods.Comprehensive consideration this paper edge detection based positioning method is mainly determined by the detection of each pixel and its neighbors on whether the state of the pixel located at the boundary of an object.The operator to detect horizontal and vertical directions is better,the positioning accuracy is relatively high, while the effect of having a low-noise image is preferably relatively steep gradient.In this paper, the use of horizontal and vertical tilt correction for distortion correction method for processing license plate, license plate character segmentation corrected using a priori knowledge and the method of combining vertical projection.According to the characteristics of license plates, license plate image binarization performed after scanning from right to left, the number of white pixels in each column appears statistical scanning process, if the number of white pixels is equal to or greater than the set threshold, then we determine here is the right boundary of the first seven characters, continue scanning when the number of white pixels is less than the set threshold, we detennined that the previous column here is the left edge of the first seven characters.And so on, to find the seven characters of the license plate.The advantage of this is that first, in the case where the license plate frame is disturbed, and can effectively delimiting character, and the second is to avoid some of the radicals of Chinese characters are mistaken for Chinese characters, increasing the complexity of the process. In the character segmentation process.in order to avoid the second case,the paper set the tlireshold compared with the use of the right border of the distance during the scanning process to determine the character position.Finally,BP neural network segmentation good license plate character recognition processing. BP neural network in the sample is not a lot of cases,the training speed,recognition rate.