Research Of The Key Techniques On License Plate Recognition

The research and development of license plate recognition system are hot topics in modern traffic development and determine the intelligentization and modernization of traffic system.The LPR system is a special computer vision system in the real-time case.The LPR system research involves the techniques such as digital image processing, computer vision, pattern recognition, artificial intelligence and so on.The main research work of this thesis is presented as follows:1.Image segmentation technique is adopted to license plate location.A new location method on license plate color is proposed in this thesis, which locates license plate area using the considerable projection information and H component of HSV color model. The experimental results show that the algorithm is able to quickly and efficiently extract license plate field and not sensitive to the unconstrained illumination condition and complex scenes.The accuracy is over 98.5% in location.2.Character segmentation is a special application in printed character segmentation. Image is segmented into many subimages to facilitate character recognition with only one character in each subimage in the course of character segmentation. The improved scan-line algorithm proposed in this thesis performs well on detecting the up and low boundaries of characters. Single character is picked out by global partition with adjustment in local areas.3.Chinese characters, letters and numbers in the license plate are perceived in the character recognition process which is the kernel part of LPR system. The key to recognition is the selection and organization of classifiers. As far as Chinese characters in license plate are concerned, a new recognition method combining texture feature with images' feature of segmented Chinese characters is presented. Considering letters and numbers,two methods are adopted in this thesis.One on multi-classifier integration performs much better.The recognition rate is 98.6%. The other one applying improved BP neural networks carries out the recognition of letters, numbers, letters and numbers respectively .The experimental result is good and the whole recognition rate is 97.2%.4.A real-tune license plate recognition system is designed and implemented on these research, which consists of three parts: license plate location, character segmentation and character recognition.The system makes a quite satisfactory performance. It is applied in traffic peccancy visual monitoring system and makes the confirmation of peccant vehicle more automated and intelligent.