Research Of Multi-license Plate Location And Recognition Under Complex Scenes

License plate Recognition(LPR) is the kernel resolution of Intelligent Transport System(ITS).It is widely used in the toll station,automatic payment highway fee, parking lot management,etc.LPR contains three sections-license plate localization, char extraction and char recognition.It integrated technologies such as pattern recognition,artificial intelligence,computer vision and image processing,etc.The main researches about Multi-license Plate Location and Recognition Under Complex Scenes of vehicle license plate and its corresponding technique in this paper are as follows:(1) The pre-processing of vehicle image is before character recognition.After gray-scale for vehicle image and comparisons to a variety of clear methods,this paper proposes a a combination method of interpolation self-adapted histogram equalizing and the Gaussian filtering.The method makes the vehicle image clearly in complex environment.Finally paper executes the edge detection by vertical operator, which is benefical to the multi-license plate localization.(2) An algorithm of multi-license plate automatic locationing based on the ROI is proposed.First of all,the plate was crude located according to the distance of its edges spacing characteristic,and the result is used as ROI.Secondly,according to gray-scale changes and texture characteristic of license plate,paper carries multi-layered detection on ROI,and confirms license region further.Finally paper executes operator by skew correction with the located license plat.The algorithm can locate a number of license plate in image,with high locating accuracy and speed.(3) A method based on an improved elastic model matching is proposed and used in identifying the digits and letters of the license plate.The algorithm execute each image to be recognized in appropriate movement and distortion in a small range,then calculates the minimum Euclidean distance between both.When all the models are marched by the same algorism,the marching which has the minimum Euclidean distance is recognized as the best marching and as the final marching result.The algorism can resolve the problem bring by letter deform and defects in practical application,and simultaneously it can effective recognize the character of license plate by combining with the Gabor Chinese character feature extraction recognition methods.(4) At last,we design a prototype system of on vehicle license plate recognition system based on object-orient methods and prove the effectivity of above algorithms of some experiments.