Research On Automatic Vehicle License Plate Recognition Adaptive For Real-time Implement By DSP

Vehicle License Plate Recognition (LPR) is a combination of the computer vision technology and the pattern recognition technology, and it plays an important role in Intelligent Traffic System (ITS). Automatic LPR systems can be used broadly in such practical traffic systems as toll gates of highways and bridges, parking lots supervision, urban traffic monitor systems, ports and airdromes. It can improve the automatization degree of vehicles monitor and supervision, so it promises great practical and economic value.A LPR system generally includes four parts: image preprocessing, license plate location, characters segment and characters recognition. Methods and algorithms already existed of each part are discussed and compared, and then new methods adaptive for Real-time implement by DSP are proposed and emulated on PC and DSK.To meet the real-time request, the license plate is first located and segmented from the grey level image after preprocessing and then binaryzation of the license plate area only is implemented, sparing the long time spent on binaryzation of the whole image. A new location method based on improved line scan and mathematical morphology is provided: Line scan is first applied for cursory location taking full use of the jumping feature of gray levels in plate area, and then mathematical morphology is adopted for accurate location. Experiments show that this method can search and locate the license plate rapidly and with high accuracy.Usually the vehicle license plate got from the original image is somewhat slantwise, so first the plate is emended. Then individual characters are segmented form the characters series for further recognition. According to the positional sequence of the plate characters, four different classifiers are designed specially for Chinese characters,numbers,alphabet,admixture of numbers and alphabet. For Chinese characters,numbers,alphabet, Wide-girdding characters are extracted to match with the standard characters, the minimum Euclid distance is taken as the recognition rule; For the recognition of admixture of numbers and alphabet, because of the similar characters between alphabet and numbers, classifiers series-wound...