Research On Key Technology Of License Plate Recognition Of Highway Based On Image Preprocessing

Intelligent Transportation System（ITS）is the advance subject in the research ofglobal communication. The development trend of road traffic in21stcentury is ITS, andthe license plate recognition system is an important aspects of the research direction.License plate location, license plate character segmentation and license plate characterrecognition are three key components of license plate recognition system and a focus ofresearch as well. The result of the first two technologies, namely license plate locationand license character segmentation will affect accuracy of character recognition directly.At present, the license plate recognition technology has been widely application inhighway operations and management. This paper mainly studies plate location andcharacter segmentation for expressway license plate recognition system.License plate location is to locate plate region. A license plate location method isachieved which combines8-Direction sobel edge detection with morphological in thispaper，on the basis of traditional sobel edge detection. It can get better connected region,remove pseudo-plate region effectively, and locate the license plate region better, afterimproving the quality of plate image is with pre-process. The method mentioned in thispaper can surmount difficulties of candidate regions selected effectively which is causedby sunlight, pollution and other reasons in the locating process.License plate character segmentation is to cut up a standard individual.Animproved character segmentation method is proposed which combines character widthwith vertical projection. Firstly,2-D Otsu image binarization method is proposed whichis based on GAPSO to obtain better binarization image by striking the optimal threshold.Then, the plate rivets, upper and lower border and useless region are removed in thelight of brightness transiting, after rectifying the plate. Finally, individual characters aresegmented with the improved character segmentation method which combines characterwidth with vertical projection, according to a priori knowledge such as standardcharacter distribution and arrangement rules of plate. The method solves the characterfracture, adhesion, and other complex phenomena better.The experiments show that the license plate location method is proposed in thispaper locate the plate region more accurately and the character segmentation isproposed can segment individual characters effectively, when the quality of an originalwas poor. It was a good basis for the subsequent character recognition.