Image Processing-based License Plate Recognition System

Automatic vehicle license plate recognition system,LPR(License Plate Recognition),is a special system that applies computer vision treatment and technology to automatic recognition of vehicle license plates.It's also a comprehensive application of computer vision treatment and partern recognition technology in intelligent transportation field.The system involves image processing, artificial intelligence,information theory,communication technology,engineering,and therefore it is both of important theoretical research significance and of practical application value.Based on digital image processing-related theory and combining computer vision and pattern recognition technology,a thorough research and analysis is made of vehicle license plate recognition system.The major work completed here includes the following:It is a virtual loop-based video detection algorithm that puts virtual loops in video images for vehicle detection.Its basic principles are similar to physics coil detection method that sets an artificial top virtual loop and a bottom one in the appropriate location of the image waiting for detection.When vehicles pass by pre-set virtual line areas,image characteristics will change,but when no vehicles pass,image characteristics remain unchanged.Thus,pressing line detection of vehicles can be made on these two detection lines.Original images of vehicles should be preprocessed and using differential image of the processed image,mathematical space dimension is reduced.Gauss transform and wavelet transform of the data related to mathematical space dimension reduction are carried out to locate the horizontal position of a license plate and then mathematical space dimension reduction and gauss transform of the plate image are carried out once more.Lastly,morphology expansion and corrosion treatment are made so as to achieve precise positioning of the plate.Two character segmentation methods are discussed in detail.One is a character segmentation based on method integration and the other is character segmentation based on fast algorithms of regional labeling and large space locating method of plates.The former uses Laplace transform and regional growth method for regional character detection and using priori knowledge about license plates,character region is located.The latter obtains candidate regions of characters through rapid regional labeling algorithm and based on large spacing characteristic,the final character region can be achieved.After segmentation,the characters have different sizes,brightness and varying clearness,so the size of the characters should be normalized.First,based on through-line number recognition method,the characters of license plates are classified roughly and then template matching method is used for detailed classification and recognition.Construction of License Plate Recognition System Software Platform:if an abstract and fragmented algorithm can be shown through an intuitive form of expression,it will be more convincing.At first,an algorithm was developed under Matlab,the language used was scripting language,but the implementation efficiency of Matlab is not as good as c/c++,Moreover,the algorithms developed under Matlab can only be achieved through self-compiling programs. Therefore,based on c/c++language,an intuitive platform for the demonstration of the effectiveness of license plate recognition should be constructed.