Textural Characteristics And Projection Method Based On License Plate Recognition

License plate recognition system (LPR) which is one of the most important components in the intelligent transportation system (LTS), is also the important subjects in the digital picture handle and principles of pattern recognition technology and so on. Technology route of License Plate Auto—Recognition includes:image preprocessing, license plate orientation, character division and character recognition. In license plate locating aspect, is the key stage of the whole system. Accurate positioning license plate area will directly influence on the accuracy of license plate recognition, and it is the key to improve the overall performance license plate recognition system.This paper mainly studied the image pretreatment and license location method, put forward a method based on texture feature and projection method for license plate location. First of all, through CCD digital camera images, and video image, the image of image preprocessing, the color image gray, the image enhancement processing, highlight the interested parts, such as gray stretch, piecewise linear transformation, enhance the contrast, smoothing of the noise, adopt a Otsu method for image binarization processing, finally respectively based on texture feature and combination of projection algorithm for license plate location, we use projection algorithm which based on texture feature and combining method for license plate localization, meanwhile we respectively use the horizontal and vertical projection of image gray level, through the analysis of the characteristics of vehicle license plate area, to horizontal projection of images, can observe the gray level jump characteristic, and the license plate area is in the area of gray more intense, and then to horizontal projection of image, through the study found that the plate region character vertical projection will appear a special of "peak-valley-peak" uniform change continuously, the characteristics of such judgment through the image gray level of the horizontal and vertical projection, can preliminarily determine the license plate candidate area, for license plate localization, select qualified license plate candidate area. Grayscale preliminary determine the horizontal and vertical projection of license plate candidate area, combining the texture feature, morphology processing, carried out in accordance with the license plate characteristics of connecting area selection, combination of two methods can be more accurate for license plate localization, so as to better detect and locate license plate to the correct license plate area.Through the analysis of related theory research and the algorithm, and using VC++software tools to achieve the above thoughts, through a large number of repeated research and testing experiment. By using projection method combining texture features of license plate region to initial positioning, accurate positioning of these two steps gradually filtering method, to effectively remove interference connected regions and similar areas, accurate locating the vehicle license plate area, at the same time we use the method of statistics and performance evaluation to check the result data, summarizes the research of the method has high accuracy to license plate localization, also raised the speed of the location at the same time, enhance the anti-noise performance, strong robustness and stability, achieved the effect of precise positioning plate, for the license plate character recognition provides an important basis.