The Research Of Moving Target Recognition And Tracking Technology Based On The Visual

Moving target tracking is an important research field of computer vision, which has broad applications prospects in public safety, weapons guidance, education, medical treatment and so on. The identification and tracking of moving objects is a process of detecting,identifying, and tracking the moving targets which are in image sequence, and it’s a combination of target recognition and tracking method. With the development of computer vision technology in recent years, moving target tracking has become a hot research issue.Although there are many excellent target tracking algorithms, in practical applications they still faces many factors which affect the tracking, such as the lighting changes, moving target deformation, scale changes, background interference, target occlusion, and so on. Therefore,the study of robust object tracking algorithm is still an issue of great significance.There are many excellent tracking algorithm in academic fields, and CSK tracking algorithm is one of them. The advantage of CSK lies in the use of cyclic shift dense sampling methods and classifiers training with fast Fourier transform, which make its tracking speed be outstanding. However, CSK only uses the gray feature of targets, and lacks the capacity of describing the target appearance. What’s more, the update of classifier parameters is so linearizing that it cannot adapt to large changes in tracking the objects.So in this paper I will improve the CSK tracking algorithm to remedy the defects we have mentioned. I will describe the moving target with its color characterization, while reducing the dimensions of feature with the principal component analysis method and removing the redundant information of characteristic to insure the accuracy and robustness of target characterization. Introducing weight parameter β in the update pattern of original parameter and making the updating scheme be nonlinearized, so that the classifier’s training and updating can become more stable and accurate.After that, we will expanding the algorithms set and test data set of the benchmark test platform, then comparing and analyzing the improved algorithm with the current outstanding algorithm by using the new test platform. The comparison including these aspects: tracking accuracy, success rate and time robustness (TRE) and space robustness (SRE). And we will do the same thing to the new algorithms.The experimental result proves that the new algorithm can track better for many complex sequences, such as illumination. background clutter,occlusion, deformation, motion blur, and so on.In our experiment UAV is a control target, we will transfer the images we have collected to the ground through wireless network. And we will send the control instruction which is obtained by the algorithm to UAV through data transmission module. By doing this, we can control UAV to track the objects on the ground and validate the capacity of the algorithm to handle interference factors(that is,the various potential situations). Eventually, the result shows that the improved algorithm has faster tracking speed and better capacity of handling interference factors.