Research Haze Plate Image Processing Technology Based Gigabit Ethernet Cameras

With the development of economy and the improvement of people’s living standard, the number of vehicles is increasing rapidly, so the vehicle management is put forward higher request. The vehicle license plate recognition system is an important part of vehicle management system and it is bound to become one of the hotspots of current research. Images used in this system are mostly taken in clear weather condition. While images under the interference of the foggy and haze are not clear and will seriously impact on the post processing of license plate. So in this paper, software platform based on the VS2008, mainly completes acquiring the images, enhanced defogging, the license plate location and character segmentation.In the image acquisition, first, through the gigabit network cameras and other cameras are compared, it selects the gigabit network camera which has faster speed in the network transmission and isn’t restricted in the length of transmission. Second, it researches the GigE Vision interface standard and TCP/IP protocol, analyses the format of each function command. Finally, it achieves many functions of camera, such as the various parameters of camera setting, image acquisition and displaying and so on.The paper mainly research on color image defogging and denoising algorithm of binary image in the image pretreatment. In the aspects of image defogging, it researches the dark channel prior theory algorithm, the global Retinex algorithm, the local single scale and multiscale Retinex algorithm, and summaries the advantage and disadvantage of each algorithm; then the improved Retinex algorithm is elaborated and completes the clear processing of image. The paper researches the neighborhood method and square window denoising algorithm in Binary image. While the experiment finds that they couldn’t keep the lines of character intact, so it designs a method of multi direction line window.In the stage of license plate location, the paper mainly researches and achieves the edge detection algorithm based on gray-scale image. Then the double edge algorithm based on color image is introduced and uses the large quantity of information of color image. Finally horizontal and vertical projection is used to realize the license plate location.In the projection of character segmentation, first, the paper introduces the type of image binarization algorithm, and focuses on the bimodal of histogram and the Otsu algorithm. Second though the method of Hough detects tilting angle of the license plate, it uses the angle to correct the license plate region. Then the connected domain algorithm removes frame, rivets and other interference information, and can remove the broken borders which aren’t removed by the scanning method. Finally projection and prior knowledge of the license plate accomplish the character segmentation.