Vehicle Detection And License Plate Location Based On Deep Learning

With the continuous development and improvement of traffic road technology,license plate recognition technology is now widely used in road monitoring and road command system,such as high-speed toll stations,road traffic pedestrain monitoring,residential and parking fees management system.At the same time,it is also an important part of intelligent transportation system,in order to solve the road congestion situation provides a new choice,could help decision makers quickly and efficiently develop the implementation plan,saving labor costs.The vehicle license is the only valid proof of identity,so it is necessary to research license plate recognition technology.This paper introduces the history and background of license plate recognition technology both at home and abroad,presents the machine learning algorithm which has been popular in recent years and its application in image detection and recognition.To overcome some of the limitations of the existing algorithm,this paper used with the machine learning algorithm and proposed a machine learning and image processing technology to deal with license plate recognition system,we used deep learning and Image Processing tools to achieve results.This paper introduces two aspects of vehicle license plate recognition,one is vehicle license positioning,the other is vehicle detection.In this paper,the common algorithms of each part are summarized,the related algorithms are improved,and the knowledge of deep learning and image processing is used to optimize.The main work of this paper is as follows:? In this paper,the basic principles and implementation methods of the two algorithms are introduced for the convolution neural network and the corner density clustering.? In order to solve the existing problems and improve the accuracy of vehicle detection,it is proposed to apply the regional convolution neural network algorithm to vehicle detection.The scheme generates the color hierarchy of the image to obtain the potential vehicle candidate area.The corresponding convolution neural network model is established to extract the local characteristics of the candidate region.The network structure of the convolution neural network is improved,the size of the input image is modified,and the parameters of the network are also adjusted.SVM classifier training is selected for positive and negative samples,which is used to classify the vehicle candidate area,and finally the vehicle information is determined.The experimental datas show that the improved algorithm has achieved better results in vehicle detection and testing.? In order to solve the problem that the low accuracy for traditional license plate location algorithm,a corner density statistical method is proposed to locate the license plate.Firstly,according to the color characteristics of the license plate itself,the image is converted from the RGB color space to the HSL color space,and the HSL license plate image is threshed,and then a series of morphological algorithms are used to filter the image and eliminate the unnecessary information.The corner detection algorithm is used to purify the processed image,and the number of corner points and coordinate information are extracted.Finally,the DBSCAN algorithm is used to determine the accuracy position of the license plate.The experimental results show that the positioning accuracy is high,positioning time is faster,and fulfilling the real-time requirements.