Research On Interactive Multimedia System Based On Detection And Tracking Of Human

Multimedia technology occupies an increasingly important position in the daily life of the human in recent years. But lack of human-computer interaction is always one of the problems which need to solve by the Multimedia technology. Detection and tracking of human not only could monitor and track the human body dynamically, but make real-time response to the body action of human and improve the interactivity. Interactive multimedia system based on human vision-based human detection and tracking technology not improves multimedia interactive, but gets rid of the limitations of traditional multimedia systems.So it is important to probe the technology of dynamic detection and tracking of human which is suitable for multimedia interactive system environment.Firstly, this thesis lists four difficulties in the detection and follow-up study of the human body, and analyzed and compares commonly used three human detection method. Take human detection method based on Gaussian mixture model as the basic detection method of the interactive multimedia system. By Introducing Gaussian mixture model-based human detection principle and analysing reasons for the formation of cavitation in the detection process, propose improved target detection algorithm based on Gaussian mixture model which can can reduce the the target internal voids by introduction of pixels before attractions and after attractions switching on Bhattacharyya distance as the threshold. In the experiment of the human body detection algorithm, joins adaptive median filter for image noise reduction pretreatment. At the same time, uses an image morphological dilation and erosion operation to optimize of the phenomenon of empty, and gives the experimental results in improved. In addition, due to light reasons, object detection is susceptible to shadow effects. In order to eliminate the interference of the shadow of the target detection accuracy, introduces the shadow detection method based on color and Gaussian mixture model moving objects and as a basis for it, improves this shadow detection algorithm based on YUV color space to generate new improved shadow detection algorithm based on HSL color space.Then, After comparison of four commonly used moving target tracking algorithm steps and application of environmental, Summs up the human body tracking algorithm flow unobstructed and occlusion. In the unobstructed algorithm section, around two core issues of the algorithm, on the one hand selects the body contour characteristics as matching feature information and introduces several edge detection algorithm to extract the body contour feature, then compares their specific experimental effect, finally chooses extraction algorithm based the LOG operator of edge detection algorithm as the characteristics of the human body. On the other hand, gives the principle of particle filter and Unscented particle filter target prediction algorithm steps. With cover part of the algorithm, firstly, gives a moving target occlusion and end determination condition, then selects the color characteristics as matching characteristics of the body after the end of the block, and uses color histogram quadratic equation to calculate the characteristic distance.Finally, design interactive multimedia simulation system based on human body tracking which is applicable to the actual use of the environment to verify the previously proposed human detection tracking method.