Research On License Plate Recognition Algorithm Based On Machine Learning

With the rapid promotion and development of wisdom city strategy in the country's major cities,the core of each operating systems in a city will be analyzed,integrated a nd handled by a new technology.As an indispensable part of city life,public transport i nformation,to a large extent needs to be managed and operated quickly and efficiently.Based on the research of the requirements of the extended functions of a public tra ffic information management system,this thesis processes images containing license plate information in the system with machine learning technology,and analyzes and discusses intelligent recognition in license plate recognition algorithm for the domestic and foreign literature.Considering the features of the system,this thesis realizes the automatic identi fication of license plate character.The concept of license plate recognition algorithm mainly comes from the Open So urce system EasyPR.And combining with the requirements of this system and the EasyP R,this thesis makes analysis and improvement.In this paper,license plate recognition al gorithm is based on OpenCV module(Open Source Computer Vision Library,namely Op enCV)cross-platform Computer Vision Library,and has the feature of light weight,high efficiency.License plate recognition process consists of two parts:1)License plate location:Through analyzing and processing a picture containing a li cense plate,a primary license plate tiles is obtained.After the judgment of the primary li cense plate tiles,the blocks that only containing license plate information are obtained at the end.This positioning methods in this phase uses HSV color space model and Sobel edge detection algorithm to make an integrated positioning.License plate judging process is mainly based on Support Vector Machines(SVM)algorithm for training and learning,and judges whether the block diagram is true.2)License plate character recognition:Through character segmentation and character recognition of the license plate tiles obtained by positioning the vehicles,and then order the output character registration information.In this stage,character segmentation is main ly according to the methods of edge contour extraction and the external rectangular.And character recognition process is mainly through the two layers of Neural Network model in Artificial Neural Network(ANN)for learning and classification.The major algorithm of license plate character recognition module in this paper is based on OpenCV open-source vision library.It is written by c++ and uses VS2013 to operate and test.The test result shows that the automatic license plate recognition algori thm of this thesis has a certain positioning accuracy and recognition rate,and basically meet the requirements of the system a number of conditions.