Research On Video DE-FOG Algorithm For Traffic Monitoring

Nowadays, the traffic monitoring system has become an important part of China's intelligent transportation construction. Most cities in our country have established a relatively complete traffic monitoring system which is based on the monitoring cameras.With the worsening of the atmospheric pollution, the foggy weather heavily affects the quality of the pictures taken by the cameras and leads to the poorer of the image resolution and color distortion degree. The low quality photos which are taken in the foggy weather has been completely unable to meet the needs of practical application.Hence, the removing haze of the foggy videos has attracted more and more researchers'attention and concern.This study mainly aims at the traffic surveillance videos captured in the foggy weather. Based on the traditional atmospheric scattering model and the dark channel theory which is proposed by He Kaiming, a defog algorithm for traffic surveillance video is put forward. The main research content is in the following.Firstly, for the limited application of the dark channel transcendental theory in the bright area, quadratic function is introduced into the tolerance mechanisms to propose a tolerance mechanism transmitsive rate correction algorithms based on the quadratic function. In addition, from the point of the transmissive nature, this paper presents a transmittance correction algorithms which combines the dark channel and inverse dark channel together. By merging the two algorithms, more real bright regional transmittance is achieved without identifying the bright and dark areas or setting the experimental parameter values for the foggy pictures. This method has a strong applicability in dealing with various kinds of foggy photos which contains the bright area. In order to improve the efficient of the defog handling, this paper adopts the adaptive fast minimum filtering method to achieve dark color channel of the foggy photos. The cartridge-optimized guided filter is also used to refine the rough transmittance. At last, an optimization algorithm based on the atmospheric light is raised aiming at the traffic monitoring videos.This paper widen the defog method of the single image. A video defog method is proposed bases on the objective fact that the depth information of traffic monitoring video background is the same. The first step is getting the background of the traffic monitoring video by using the mean background fitting arithmetic. The following steps include calculating, amendment, refining the transmittance of the background and taking the transmittance of the background as the transmittance of all the frame within a period of time. As the change of the fog concentration and the influence of the complex traffic situation, the transmittance will change a little after a period of time. In order to keep the continuity of the color and structure character after the defog handle, an background update arithmetic is put forward which is combining Kalman filtering with classification method in this paper. The background image will be re-fitted within a certain period of time. Then, combined with transmittance generic ideas, the traffic surveillance video is de-fogged. The experimental result shows that this method can work well in real-time defogging by dealing with the foggy and poor quality traffic monitoring videos.