The Research On Traffic Video Compression Technology Based On Compressed Sensing

The traffic video that is an important part of intelligent transportation system has been widely used. The huge data of traffic video is continuously produced, which brings great challenge to the store of traffic video. How to compress the traffic video has been turned into an emerging research topic. However, the traditional statistical compression coding method, which has been used currently, to compress the traffic video, do not make full use of the characteristics of traffic video. The traffic video has the features of stable background, certain sensitive area, complex image texture, which is also affected by outdoor illumination change and the weather’s change, because the cameras are normally installed in outdoor. Taking full advantage of the characteristics of traffic video and researching compression method which is suited to the special traffic video has become an important research topic.Compressed sensing theory provides a useful way for the compression of traffic video. Because of the large spatial and temporal redundancy of traffic video, traffic video can be compressed effectively by using compressed sensing theory. According to the above ideas, the compression of traffic video based on compressed sensing is researched in this paper. The specific works are as follows:(1) On the basis of understanding the theory of compressed sensing and the relevant theorems, the compression and reconstruction of traffic image based on K-SVD algorithm are focused on. In view of the shortcomings of high time complexity and general image quality, a new K-SVD algorithm based on wavelet tree and variable iterative times is proposed in this paper. The simulation results show that the PSNR of new K-SVD algorithm is increased by 2 dB, and the running time of algorithm is reduced by about 15%, compared with the original K-SVD algorithm.(2) The traffic video preprocessing is the basis of the design of traffic video coding framework. In the preprocessing part, first of all, the background of traffic video is modeled. Gaussian mixture model is used to extract the background, which is more clean and clear compared with the average method. Secondly, a background updating method based on block classification is used in this paper. The background update algorithm obtains the difference image by using three frame difference method, and determines the threshold of classification by using the adaptive iterative threshold method so that the background is updated with the extracted background. Third, to enhance image at night, the traffic video scene is divided into day or night. Finally, in order to increase the rate of video compression and video quality, a variable sampling rate calculation model is put forward in this paper. According to block compressed sensing theory, the experienced fitting function of variable sampling rate algorithm is used, and on this basis, a variable sampling rate observation compression process that is suitable for GOP(Group of Picture) is described.(3) A compression coding’s framework of traffic video based on compressed sensing is designed. The simulation experiment shows that the framework of traffic video has good compression performance.