Research And Improvement Of Key Algorithms In License Plate Recognition

License plate recognition system has been widely used in vehicle access management,highway toll management,electronic police and so on.License plate recognition system plays an important role in the regulation of vehicles,and can realize the urban traffic supervision.In order to prevent traffic jam,it is of great significance in real life.At present,there are many license plate recognition systems,but in complex environment(such as illumination conditions,deformed license plate,dirty license plate,etc.),the recognition rate of license plate will be greatly reduced.Therefore,how to improve the license plate recognition rate in complex environment is of great significance.This paper mainly introduces four aspects of license plate preprocessing,license plate character location,license plate character segmentation and license plate character recognition.Some of the algorithms are improved.The main innovations of this paper are as follows:1.In order to enhance the binarization effect of OTSU algorithm,the attribute weighted naive Bayes algorithm is used to improve the OTSU algorithm because the pixels in the image are discrete data.In other words,attribute weighted naive Bayes algorithm is used to find an optimal threshold to distinguish foreground and background as far as possible.2.In order to improve the accuracy of license plate character location,the RGB color image is converted into HSV color image,and then the HSV color image is constrained by the color of license plate.The total number of white pixels in the image is found by gray and binarization of constrained images.In the experiment,it is found that when the total number of white pixels is less than or equal to 4200,The license plate candidate region extraction algorithm based on projection and color synthesis features is more accurate.When the total number of white pixels is more than 4200,The algorithm of license plate candidate region based on gray level is used to extract the license plate candidate region accurately.Finally,the support vector machine is trained by using these license plate regions and non-license plate regions.The trained support vector machine model is used to distinguish the license plate area,and the accuracy of license plate location is improved.3.The related principle of neural network is introduced.The principle of BP neural network is introduced emphatically.Finally,three kinds of BP neural network in reference [26] are used to improve the optical character recognition(OCR)engine,and then the improved optical character recognition engine is used to carry out the license plate character recognition to verify its recognition effect.Through the experimental comparison between the above improved scheme and other algorithms,it is found that the localization accuracy of the improved localization algorithm in this paper has been improved in the environment of poor illumination conditions,and the experimental data show that the accuracy of the improved localization algorithm has been improved.The success rate of location has reached95%,and the improved optical character recognition engine has also improved the recognition effect of license plate,and the recognition rate of license plate character has reached 97%.