Reasearch On Historical Figures Interactive Platform Based On Body Recognition

Nowadays with the development of economy, people’s living level becoming higher and higher, meanwhile people’s spiritual pursuit growing fast, so nowadays more and more people pursue their own spiritual enjoyment, such as more and more people like to read poems, more and more people like to visit science and technology museum, history museum and other places, in the paper, the interaction platform of historical figures is mainly used in science and technology museum, exhibition hall and the historical museum, this platform can let people who interact with the platform have a true feeling.In this paper, this platform is mainly using the knowledge of computer vision to achieve the function of human-computer interaction, this method is mainly use the video acquisition device to gather the picture of people’s real-time operation action when he interact witu the platform, then object segmentation, object detection, motion tracking, human orientation, behavior recognition. The platform can obtain the audience’s operation intention, then give the feedback to the platform, then the platform makes the corresponding interaction results.In the paper, the platform in the computer vision is mainly covered this following aspects:human body detection, human body tracking, gesture recognition of human-computer interaction and human body positioning.This paper’s main research work as follows:First, this article introduces the basic knowledge of digital image, and some basic process of digital image processing, then introduced the way of DirectShow to obtain the image sequence. Then this paper introduces the three algorithms which widely used in the human body detection, at the same time, introduce the three algorithms’s advantages and disadvantages, then this paper select the target detection algorithm based on gaussian mixture model as this platform’s human detection algorithm, but when this platform in the place which have a larger changes in the light, this algorithm will appear dectection error problem. In order to avoid this kind of situation, this paper proposes a modified target detection algorithm based on gaussian mixture model, this algorithm using real-time background subtraction and gaussion mixture model, then combine this methods, when the light changes very big, use the background subtraction for foreground detection and real-time update the gaussian mixture model, then achieve the effect of accurate detection, and when the light change small, when update the paramters of the gaussian model every two frame image to make better real-time performance of the algorithm.Then the Mean Shift tracking algorithm is discussed in the article, the basic principle of Mean Shift algorithm was introduced in detail, and introduces the principle of target tracking based on Mean Shift algorithm and use this algorithm have a experiment in the platform to prove this algorithm used in the platform is better. In order to better implement and historical figures interactive platform of interactive function, this paper proposes a simple binocular vision orientation method, the method combining space geometric projection principle of linear positioning and the camera can avoid complex calibration process which in the traditional positioning algorithm, so the practicability of this method is higher.An the end of this article, combined the human body detection, tracking and position algorithm which proposed in this paper to design the historical figures interactive platform, and this platform applied in museum.