License Plate Positioning And Segmentation Algorithm

Vehicle license plate recognition is one of the important applications and the core technology of computer vision and pattern recognition in the intelligent traffic field. Besides it's the important part of enforcing intelligent traffic management. The complete automatic license plate recognition system is composed of five parts, namely vehicle image acquisition, vehicle license plate location, plate tilt correction, character segmentation and character recognition. Meanwhile, image segmentation arithmetic plays an important role in the complete automatic license plate recognition system and often used. In a manner of speaking, the stand or fall of it directly influences function of automatic license plate recognition system. However, the target image gained in the VLP recognition has abundant natural background and automobile body background information. It is also affected by factors such as lighting condition (daylight,night), weather condition (cloudy, rain, fog, snow), undercutting and vague. Therefore, it impacts seriously quality of obtained image and brings about large difficult dealing with following vehicle license plate location and character and character recognition. So the results of segmenting complex license plate using regular image segmentation arithmetic are not good. As a result, we ought to using the appropriate way in segmenting license plate image to reduce the quantity of error segmentations. The main contents related in the paper are as follows:1. Explained the purpose and significance of the subject topics, Describes the situation at home and abroad in recent years and study work in the paper2. Aiming at complex image, improvement theory of the transition region of image segmentation algorithm is proposed, which can solve well the problem of the regional extraction error. Experimental results show that the algorithm has good noise immunity, precise extraction of transition region and the ability to maintain the license plate image details, while it also has good robustness for the poor illumination and fuzzy segmentation of license plate image. 3. Introduce and analyze regular vehicle license plate location arithmetic. On the basis of merits and demerits of current vehicle license plate location arithmetic,a kind of vehicle license plate location arithmetic is put forward utilizing geometrical characteristic and textural features. In this arithmetic, firstly, image is filtrated by adopting the pyramid way in order to denoise and keep details and textures of image effectively simultaneously. Secondly, geometrical characteristic and textural features are excavated to the full by adopting Canny operator and mathematical morphology. Thirdly, the candidate region of license plate image is obtained using corner detection method. Experimental results show that this arithmetic can denoise and keep details and textures of image effectively, eliminate non-plate region and then precisely locate the license plate image.4. Combine morphological erosion, dilation method and vertical, horizontal projection to segment the license plate image. Then get the final results of license plate character segmentation and provide the good basis for following process of license plate character segmentation and recognition.