Research Of License Plate Recognition Technology Based On Night Image Enhancement

License plate recognition technology has become the core technology of intelligent transportation and has been widely concerned in recent years.The research direction of license plate recognition is to improve the recognition rate and speed of license plates in various complex environments.Night license plate recognition is a difficult task in license plate recognition.By disassembling the problem of license plate recognition at night,we aim to solve the problems of low brightness and accuracy in license plate recognition.We use night image enhancement algorithms to enhance night images,and then perform license plate recognition on the enhanced images.The specific research content is as follows:In the image preprocessing stage,traditional algorithms are improved.By analyzing the traditional image enhancement algorithms,including Histogram equalization algorithm,gray transformation algorithm,Retinex algorithm,and combining the advantages of these improved image algorithms,a night image enhancement algorithm based on visual perception model is proposed.Aiming at the pain points of low brightness in nighttime images,a relevant nighttime image enhancement algorithm is designed based on the principle of perception cells in the visual perception model.The algorithm utilizes a visual perception model and functionalizes it,designing a bright area enhancement function based on bright area adjustment and a dark area enhancement function.By calculating the median brightness of the image that needs to be enhanced,the two functions are proportionally fused to achieve image brightness enhancement;Then,through local contrast stretching and color restoration,all layers are fused to obtain the enhanced image.After experimental analysis and enhancement,nighttime low illumination images have improved brightness and information entropy while maintaining their original color structure and contrast,making them more suitable for image recognition.Describe the commonly used neural networks and related evaluation standards.The problem of night license plate recognition is decomposed into three units: license plate location,license plate correction and license plate recognition for processing.First,the feature pyramid network of multi-scale feature extraction and the breadth search algorithm are used to locate the license plate in the image.Then,the image is corrected by Affine transformation.Finally,the license plate recognition is realized by adding Convolutional neural network and circular network of attention mechanism.After experimental comparison and analysis,the designed neural network has higher recognition accuracy and recognition processing speed.The algorithm should have corresponding practical use value.In order to make the algorithm have a broader application scenario,a nighttime license plate recognition system has been developed.For images in normal environments,license plate recognition has been implemented,and for images in nighttime environments,the function of image enhancement before license plate recognition has been implemented.The system has good use value and a sense of user experience.