Design And Implementation Of License Plate Image Enhancement System Based On Dehazing Algorithm

As the rapid development of machine vision and artificial intelligence high-real-time outdoor clear image acquisition systems are widely required in many fields,such as license plate recognition,unmanned driving,real-time traffic monitoring,maritime military target recognition,and space-based system target recognition and positioning.In the case of foggy weather,the image quality of outdoor shots is gray and white,and the information in the image scene is damaged,These lead to the dissaisfactory results of key information extraction based on clear images not very satisfactory.Therefore,the research of high real-time outdoor image acquisition and processing system is of great significance.This thesis has carried out the following research around the existing defogging technology:(1)The imaging mechanism of images under foggy weather conditions is studied.Then the degradation mechanism of the image and the atmospheric scattering model of the image are revealed.And,the refraction law of natural light in foggy weather is analyzed,and the specific reasons for the deterioration of image quality are summarized.On this basis,study the histogram equalization algorithm in the classic defogging method,Retinex Scale Single algorithm,Retinex Scale Multi algorithm,convolutional neural network defogging method,color attenuation bright defogging method and dark channel prior defogging method in classic defogging methods are listed.(2)Further study of in-depth of dark channel priori defogging methods and color attenuation inferior defogging algorithms is carried out,analysis and comparison of their respective advantages and disadvantages and scope of application are conducted.On the basis of analyzing the defogging effects of various defogging methods,the defogging images are subjectively and objectively analyzed according to the indice as following,mean square error(MSE),structural similarity(PSNR),peak signal-to-noise ratio(SSIM)and Three evaluation indicators based on visible edge contrast(new visible edge ratio e-;visible edge gradient ratio-r;black/white pixel ratio-?).To improve the halo effect of the sky at the edge of the scene caused by the dark channel prior defogging method,optimize the guided filter transmission map,and calculate the calculation time of each part,and specifically design and implement seven images including six of the seven methods of improved algorithms.The image quality evaluation after defogging of two indicators is obtained from the evaluation index data to improve the defogging effect of the image on the original basis.(3)The Raspberry Pi platform is used to realize the VNC remote connection between the Raspberry Pi and the PC.An improved dark channel priori defogging algorithm is implemented on the hardware defogging platform;the hardware-based raspberry Pi4 B defogging platform is conducted,including the platform's operating system,file configuration and development environment,and use Open CV to implement the improved defogging algorithm.The improved defogging algorithm is transplanted to the built defogging platform,the design of the defogging system for license plate images based on the defogging algorithm is realized,also as the license plate recognition and verification of the defogging license plate image,and Qt Designer is used to build the PCside UI interface.