Design Of License Plate Detection System In Haze Scene

Recently,the number of automobiles in my country has been growing rapidly,and the haze problem caused by exhaust emissions and sandstorms has become more and more serious,causing huge troubles to the license plate detection and recognition module of the intelligent transportation system.Cameras are used as monitoring equipment in traffic scenes.The captured images are affected by haze and light.Traffic control departments need to manually read the illegal license plate numbers in the captured images,which is time-consuming and laborious.The object detection technology in deep learning has the advantages of fast speed and high accuracy,and has good performance when detecting small targets such as license plates.Combining the existing single image defogging technology and object detection framework,three innovative points are proposed to realize the detection and recognition of vehicle license plates in complex traffic scenes.The main research work of this paper is as follows:1.This paper designs a license plate information location method based on light-yolov3.This method uses deep separable convolution,pool Ted pyramid network,channel feature rearrangement,and scale enlargement modules to restructure the original network,which can locate the license plate information in the image at a faster speed and higher accuracy.2.This paper proposes a dual-channel fusion single image defogging algorithm based on Retinex to defog and optimize the details of the image collected by the camera.This method uses the multiscale Retinex algorithm in the RGB channel to dehaze the image,and uses adaptive saturation stretching and guided filtering to optimize different components in the HSV channel to further optimize the details in the image,and finally dehaze the two channels The resulting image is weighted and corrected.3.This paper designs the license plate location recognition system under the haze scene.By judging whether the collected image contains fog,and judging whether the image needs to be defogged according to the returned results,finally,the license plate is located and cut using LightYolov3,and the cut information is recognized using OCR,effectively improving the work efficiency of traffic control staff.