Research On License Plate Location Technology In Foggy Conditions

License plate location plays an important role in automatic license plate recognition. It has direct impact on the accuracy of license plate recognition. Therefore, the result of license plate location is significant for license plate recognition. This thesis mainly focuses on the algorithm of license plate location in fog to improve the accuracy of license plate location under complicated surroundings. It mainly includes:(1) In order to improve the degradation for image acquired in fog, this thesis proposes automatic haze removal algorithm for single image. Firstly, the haze image is transformed from RGB to YCbCr color space. Then, the luminance component is operated by multiscale Retinex. Finally, the haze removal image is obtained by using image quality assessment. The experimental results show that the proposed algorithm can effectively improve the degradation of haze image and improve its clarity.(2) Homorphic filtering algorithm based on FFT and wavelet transform are analyzed respectively, which have been simulated by MATLAB. The experimental results show that FFT based method can improve the global contrast of image while its effect on local is not ideal. Wavelet transform based method can effectively eliminate the impact brought by non-uniform illumination. The local contract has been enhanced significantly, especially along image edge.(3) To overcome the drawback of traditional median filter, the adaptive median filter is proposed. The impulse noise as well as its magnitude is determined by the noise detection. The filtering window size is selected according the magnitude. The adaptive median filter is utilized to filter the detected noise. The experimental results show that the adaptive median filter can retain the image detail while removing noise. Compared with the traditional median filter, the adaptive median filter has obvious advantages.(4) In consideration of the plenty of corners in the license plate, color license plate location algorithm based on Harris corner detection is proposed. Firstly, the image is transformed from RGB into HSI color space and performed filtering. Then, the corners are detected with the help of hue and saturation information. Finally, the location for license plate is completed by coarse and fine searching with consideration with the character of license plate itself. The experimental results show that the proposed method can locate accurately license plate under complex weather conditions and the cost has been reduced significantly compared with the traditional method.