Detectionandtrackingofmoving Targets Under Complex Background

As an important part of the video monitoring system and intelligent transportation system, moving target detection, especially under the complicated background, has become an cardinal direction of current study. In current research, the hotspot and difficulty problem is how to improve the accuracy and efficiency in the moving target detection and tracking, because the results of its processing quality directly affect the understanding of the high-level video.This article embarks from the various complex scene in moving target detection. It deeply analyzes the various complex scenarios for the influence of the moving target detection. Aiming at the shortcomings of the existing in the classical algorithm, it improved and the enhanced the classical algorithm and proposed a moving target detection and tracking algorithm which based on an improved gaussian mixture model.The main work is as follows:Firstly, after researching the moving target detection algorithm and the common background modeling method, the research advanced the moving target detection based on mixture gaussian model and analyzed the detection and extraction of moving targets under the complex background. Secondly, by introducing the background for determining parameters, the research advanced an improved gaussian mixture model and prospected for the noise of the improved algorithm based on the eight neighborhood area. The above improvement could solve the rustling leaves and the interference of complex background such as illumination change very well. At the same time, it could have a good solution for the moving targets being misidentified as "empty" in a static video image. Thirdly, it analyzed several commonly used moving object tracking algorithm, including tracking based on feature point, tracking based on regional correlation, tracing based on optical flow, tracking based on the average drift. Then, the moving target tracking algorithm could be realized based on kalman filter algorithm and on Meanshift algorithm. Through the forecast of the location of the motion target in the next frame, the search became narrow. The tracking trajectory were detected after matching with the target.The article also solve the problem of how to eliminate shadows, the algorithm is based on the luminance information. It has a nice effect in eliminating shadows.Through the use of the above modules, the research could be able to deal with the moving target detection and recognition with the illumination variation or the jitter interference background, and the accuracy could reach above93%.