Tracking Algorithm Research On Video Moving Object Under Complex Background

With the internationalization process forward, national defense and security issues get more and more attention. Intelligent monitoring system plays an irreplaceable role in safeguarding national security and social stability, and the video moving target tracking plays a critical role in normal operation technology of intelligent monitoring system.At present, the existing target tracking algorithms have some limitations. In order to improve the stability and accuracy performance of tracking algorithms, and to meet the needs of the targets tracking in different scenarios, this paper proposes the improved CAMShift algorithm. Through deep analysis of the traditional CAMShift algorithm, this paper mainly improved CAMShift algorithm from the following three aspects:first, when the images was interference by a small amount of background pixel or effected by light, the size of the tracking window will not be stable, so in this case backproject can be transformed to binary image, and the binary image will be geometry and morphological processed, this method can furthest reduce interference. Second, when the image sequence is dynamic background, because of CAMShift algorithm based on static color model, namely the denominator of the Bayes law remains unchanged. The fixed background template can’t match with the dynamic background pixels. Eventually lead to the inefficiency of traditional CAMShift algorithm of tracking. this article through update color model of the search windows in each frame, rather than just update the background color histogram, ensure that the obtained results consistent with tracking target. So, we can use the color model which is continuously updated to replace the denominator of the Bayes law, get the real-time matching of the target model and dynamic model of color.Third, when occlusion happens, target was apt to lost. Aiming at this problem, this paper took the Kalman filter into the automatic updated CAMShift algorithm through calculating Bhattacharya coefficient to judge whether occlusion. This method can achieved tracking under the condition of occlusion.For the static background interference, static occlusion, dynamic background interference, illumination change and so on, using OpenCV visual function library in C++6.0environment programming, and the tracking results are analyzed and compared. Tracking results show that compared with the traditional CAMShift algorithm, improved tracking algorithm is practical stronger, higher accuracy and more stability. Provide the underlying data and theoretical basis for the target of the recognition and interpretation of higher level, it has important theoretical and practical application value.