Pollution Research And Implementation Of License Plate Location And Segmentation Method

License plate recognition technology plays an important role in modern intelligent transportation systems.This paper studies the location and character segmentation of contaminated license plates.With common location and segmentation algorithms,a large amount of images of vehicles with different degrees of dirtiness,staleness or damages on their license plates have been analyzed.According to the execution result of the program,license plates that can be accurately located and segmented are classified as clean license plates,and the rest as contaminated license plates,with corresponding material libraries established for each.The library of contaminated license plates makes the source of experimental data of this paper.Through analysis of related gray histograms and variance laws,the research scope of the contaminated license plates is further defined and the difficulties encountered in handling these plates are explained in detail.The initial vehicle images have different features from the located license plate images.In order to improve accuracy in license plate location and character segmentation,this paper divides the traditional one-time image preprocessing into initial image preprocessing and license plate image preprocessing,conducted respectively before location and character segmentation of license plates.The location step adopts coarse positioning algorithm combining edge detection and mathematical morphology,forming multiple connected regions in the images before applying precision positioning algorithm to eliminate too-big or too-small local regions or extremely disproportioned connected regions based on prior knowledge to extract the exact license plates.After the second preprocessing,the located license plates can be used for character segmentation directly.This paper proposes a segmentation algorithm based on vertical projection and adaptive template matching with the design of new templates for detecting digital “1” and broken characters,which suppresses the interference of contamination and reduces error rate caused by broken characters to improve accuracy in character segmentation.Finally,a common character recognition algorithm based on template-matching is applied to identify segmented characters and test the feasibility of the location and character segmentation algorithms of license plates in this paper.Seriously contaminated license plates may not be accurately located or segmented.By comparing the gray histograms and variance of the images,this paper objectively analyzes the reasons for such location and segmentation failures.