Research On Moving Target Tracking Method In Cobra Reconnaissance Attack System

A new moving target tracking method in Cobra Reconnaissance Attack System is researched. The system can automatically track,lock and attack the moving target to greatly improve attacking and monitoring capability of Cobra System after equipping the CAPF.The thesis analyses all kinds of moving object tracking algorithms,according to the demands and characteristics of Cobra Reconnaissance Attack System. Firstly the kernel-based MeanShift algorithm is discussed well. MeanShift algorithm uses the color histogram of nuclear function to represent the target model,and its robustness is good in the normal circumstance. But it exists tracking failure if the target is all covered or its speed is fast. That is to say,the algorithm fails when the target is similar as the background and the target appears again.Finally a new Improved MeanShift object tracking algorithm is proposed, which is combined with Kalman filter. The method uses not only the position database and the feature database, but also the Bhattacharyya coefficient between the target and the target candidate model to find out optimal estimation of the object movement. From this point Mean Shift algorithm starts to search the target in the neighborhood and tracks again as it is fully covered or its speed is too fast. According to the experiments,the new algorithm has good results and high accuracy. This thesis presents the experimental results of Improved MeanShift object tracking algorithm, which provides theory foundation and experimental data for hardware implementation of Cobra Reconnaissance Attack System.