The Research Of Object Tracking Algorithm Based On Compressed Sensing

With the popularity of computer image processing technology continues to evolve, moving target tracking and analysis is being more and more attention of the people. As an important research direction in the field of computers and advanced technology in many areas, it combines image processing, pattern recognition, automatic control, artificial intelligence and computer together. It has applications in transportation, military, entertainment, medical imaging many areas. Although a lot of target tracking algorithms are proposed, but cannot meet the growing demand applicationsIn this paper we carried out in-depth research on moving target tracking technology. Conscientiously sum up the results of previous studies, in this paper, we have in-depth analysis of the main problem of the moving target tracking research field and Put forward ideas for improvement.Target detection is a very important stage in the process of target tracking. Target detection algorithms, including the inter-frame difference, background subtraction, optical flow method and Gaussian mixture background model and so on. Gaussian mixture background model method is able to detect a moving target, and the ability to adapt detection to meet the real-time requirements of our next target tracking changes in the outside world scene. In this paper, moving target detection using Gaussian mixture background model method for target detection and detected motion, Morphological filtering fill the active area small and medium empty and remove the background noise.The target tracking problem is a key part of this study. Been a variety of target tracking algorithms are proposed, there are still many challenges need to be addressed. Impact of target deformation, motion blur, occlusion, and other factors will give tracking process. In recent years, the theory of compressed sensing sparse expression applied to the target tracking process, Xue Mei et al use of the theory of compressed sensing into the target tracking task tracking problem as a sparse Bayesian framework for solving problem, however, is quite high, in real-time the aspects of the application has been limited. On this basis, this paper based on compressed sensing tracking problem is:the use of compressed sensing theory sparse measurement matrix compression of high-dimensional sparse Surf characteristics from the target sample characteristics dimensionality reduction training Naive Bayeux Adams classifier, and then use the Simple Naive Bayes classifier to classify the target analyses image of the next frame.Through the plurality of video tracking experiments and subsequently the relevant literature mentioned tracking algorithm are compared. The experimental results show that my algorithm has good robustness and real-time. Tracking algorithm based on compressed sensing contrast the similar of11-minimization tracking algorithm, the algorithm of my paper retain the tracking accuracy of the algorithm to improve the tracking efficiency of the algorithm and the robustness of the algorithm, it has a good performance in real-time tracking.