The Research Of License Plate Lociation For License Plate Recognition

License plate recognition deals with many research fields such as image processing, pattern recognition, computer vision and meteorology and so on. With the development of computer image processing technology, license plate recognition research has made great progress, but the automatic vehicle identification is still a complicated issue.Plate location is an important pre-processing part of the automatic vehicle identification.Entering the license plate, separate the object and the image background by image processing and computing and remove a variety of Interference region with Characteristics of a license plate.The accuracy of license plate location directly affects license plate recognition accuracy. As the complexity of the background images collected,there are a large number of complex and rich natural background as well as the vehicle body background in the license plate images, at the same time, light changes, climatic conditions changes, fuzzy, worn, deformed and tilt of the license plate images,brings great difficulties to the license plate location algorithm, making the license plate location be a challenging problem in the automatic license plate recognition system.Traditional license plate positioning technology, mostly based on gray-scale image texture features, but because many characteristics and limitations of texture features, there are many deficiencies in dealing with complex background license plate images.With the development of computer image processing technology as well as the escalating expansion of the computer hardware, computer processing power greatly improved, as a result many scholars plunge themselves into license plate recognition technology studies based on the color image, and achieved some results, but research is still immature.Fast and efficient and accurate identification of license plates under a variety of complex environments is still a difficult problem to solve.This paper analyzes and summarizes the features of image segmentation algorithms commonly used and defects of applications in the license plate location, for the license plate image characteristics, we propose a license plate location algorithm based on a combination of color and texture features.Innovative introduce the quaternion algebraic method to the plate positioning technology, using it to describe the color model, to solve the problem that color model been unable to accurately described.In this paper, the license plate location algorithm is divided into two steps.The first step is to do the original image preprocessing to remove noise and uneven illumination effects.After image preprocessing, using the vertical edge detection by Sobel operator to get its edge image.The edge image will be processed by the connection method to get the candidate region image containing license plate, thus the first step is completed.The second step,restore the image of the candidate region to the original image, and extract color image of the candidate region containing license plate region. With the quaternion algebraic method to describe the color image, and then use quaternion principal component analysis to extract the characteristic parameters candidate region image.Then get the precise positioning of the plate area using k-means clustering method to classify the image area, and finally through the license plate prior knowledge to achieve automatic recognition of license plate. One of the most critical technology is the quaternion description of the image and the quaternion principal component analysis for image feature extraction.Through algorithm analysis and experimental verification, the proposed algorithm can overcome the defect of traditional license plate location algorithm that location is not accurate when dealing with complex background license plate images.Algorithm is more accurate in plate positioning, and can effectively deal with there are lights around the license plate and license plate has the same color with the auto body and other issues. It ensures the input data is accuracy in follow-up steps such as character segmentation and recognition.