Design And Implementation Of License Plate Recognition Algorithm Based On High-definition Image

License plate recognition technology is an important research subject at presentin the automatic control domain and intelligent transport domain,which realizes thetraffic management by the use of computer vision and pattern recognitiontechnology.Therefore,it has a high value in academic research and practicableapplication.With the rapid development of automatic control,computer andcommunication techniques,new methods and technique have been published in resentyears.They are applied widely in real projects,and have Significant effect.In general,the license plate identification algorithm is the core of license platerecognition technology,which mainly includes three parts:license platelocalization,character segment and character recognition.This paper studied licenseplate location technology,license plate character segment technology and license platecharacter recognition technology,and improved the three parts.High-definition andintelligence is the developping direction of the license plate recognition technology,therefore,this paper is researched and implementation based on the of high-definitionimages.The experimental results show that the license plate recognition method in thispaper can be effective to identify plate quickly.This paper designs a fast algorithm of the license plate location using priorknowledge of the license plate and morphological method.The algorithm firstly doesvertical edge detection on vehicle images before which is processed by using themathematical morphological method, and screening plates according to the licenseplate prior knowledge.But,since the noise around the license plate area can easily leadto license plate adhesion,it may be that we can’t find the correct license platelocation.Therefore,this paper proposes a method of using linear density to suppressthe noise around the license plate area.The linear density method can enhanceanti-jamming of the license plate localization.Experiments show that it can obviouslyimprove the license plate location accuracy by this method.This paper studies and applies a method based on the extraction of connecteddomain.The method is based on the concepts of credible character and secondcredible character,and it can gradually find the corresponding characters by the twoconcepts.Also,it obtains the precise location of character with the horizontal andvertical projection method and slove the problem about the lean of the license plate,and the precise location of character can eliminate the edge burr of thecharacters.This paper adopts the BP neural network method. This paper designs Separatelythree classifiers about the first Chinese character,the second English letter and the lastfive mixed character.Specially for the similar characters of mixed characters,wedesign the corresponding secondary classifier.According to the characteristics ofdifferent characters,there summarizes four feature extraction methods,includingline-pixel-to-byte method which is applied to Chinese character classifier andDifferences block matrix method which is applied to secondary classifier.Finally,thepaper compares the BP neural network with support vector machine (SVM) methodby experiment.This paper realizes the license plate recognition algorithm by usingVC2005.Experimental results verify the accuracy and adaptability of thealgorithm,and license plate character recognition real-time can be also improved.