| Moving object detection and. tracking, as an important part of the field of computer vision, has aroused widespread concern of scholars in recent years, and has made considerable progress in the field, there has been a lot of classic algorithm in target detection and target tracking, however, in practical applications, there are still some problems to be solved. In this paper, we focused on two aspects of target detection and target tracking to study and discuss, the main work is as follows:For the moving target detection:we have introduced optical flow method, the frame difference method and background subtraction method of three commonly used target detection method, after analyzing the merits and demerits of the three detection methods, a background reconstruction algorithm based on pixel intensity classification is researched,and a background update method is presented based on the background gradual change and mutation,and verified the detection effect of the method by experiments.For the moving target tracking:this paper focuses on Mean Shift algorithm and its application on moving target tracking. The algorithm is used in normalized color histogram as the target template, in the subsequent frames, the best target candidate areas are searched iteratively by Mean Shift algorithm based on the Bhattacharyya coefficient. The method is real-time and robustness to target deformation, however, due to a single color histogram modeling the way, does not contain the target space, making the tracking is not accurate when the target exist obscured or similar to the background.Aiming at the shortcomings of the traditional Mean Shift tracking algorithm, the target of texture features and color features are combined to establish the target model, and introduce the Kalman filter to the Mean Shift tracking framework, experiments show that the improved method can effectively solve the target is blocked or inaccurate tracking problem under complex background. |
PDF Full Text Request