Moving Human Detection And Tracking Based On Adaptive Background Updating And Color Information

With the rapid development and maturation of computer networks and multimedia technology, as a new educational mode, E-Learning plays a more and more important role in education. Now the number of people who take education through networks is explosive increasing every year. E-Learning is becoming the global trend in education and training region."Smart Classroom"is an important concept in E-Learning. In smart classroom, teachers can use many high-tech products (such as mouse, keyboard, laser pen, display, touch screen, projection and so on) to teach the students both in the room and out as comfortablely as in traditional classroom. To let students in long-distance feel as natural as they are in a real classroom, it needs the system to track the teacher and broadcast the vivid video about the teacher to students in long distance.After reseach about the moving human detection and tracking, this paper proposes an effective moving human detection and tracking algorithm which based on adaptive background updating and color information. After analyzing the smart classroom's traits, a multimodal fusion algorithm is designed. Based on above algorithms, a real-time moving human detection and tracking system is implemented.The main work in this paper includes:1) Study and compare the methods of moving human detection. Moving human detection based on adaptive background updating method was proposed. This method adopt the Gaussian Mixture Model (GMM) which establishs multiple guassians to represent the pixels of the background. The parameters and weights of the guassians are updated automatically over time so that the background are always the newest. This method solves the problem about the dynamic changing scene.2) Study and compare the methods of moving human tracking. Object recognition method based on color information is proposed. This method computes and normalizes the color information of the object region, then compare with that of the historical moving objects to recognize the multiple objects. Of course, the color information of the object is updated over time.3) Design a multimodal fusion algorithm of the smart classroom. According to the traits of smart classroom, this algorithm uses audio signal, raw devices (including mouse, keyboard, touching screen and so on) signal and video image from the panoramic camera to orientate the positon of the main speaker. This algorithm not only selects the main speaker automatically, but also solves the hiding problem. By the way, this algorithm ensures the quick self-rectification capability when the system tracks wrong human.4) Using the above algorithm, a real-time moving human detection and tracking system is implemented. It is running stably and reliably.