The Research And Development Of The Virtual Scene System Mutual Technology Based On The Somatosensory

With the in-depth study of computer graphics technology, virtual reality technology is becoming a popular technology. It is widely applied to many areas, for example, science museums, entertainment and aerospace. With the development of computer science, there are a lot of new technologies for virtual interactive, however, many of these cannot control a virtual reality system naturally and without restriction. Besides, the control process is susceptible to the environment, body, clothing and other factors, which seriously affected the interactivity of the virtual reality system. Now, moving hands or feet, the somatosensory technology can capture the body’s movement thread through the identification device without using other devices. This technology is still in development, so it has a great significance of using the unlimited somatosensory technology to control the virtual scenes.Virtual interactive technology based on somatosensory and virtual reality system was researched from two parts, which are researching on posture classification method and the building of virtual interaction platform. This paper solved the problem that the virtual interactive technology is susceptible to environment, body, clothing and other factors. First, bone image provided by the Kinect was used to get the three-dimensional space coordinates of body, therefore collected the information of body. Second, in order to improve the recognition efficiency, the support vector machine method which need small sample space was chosen for classification, besides, in building multi-class support vector machine, this paper combined the Half-against-Half and one-against-one classification methods, which reduced the number of nuclear function. Third, PCA method was used to analyze the main component of posture eigenvectors, and the highly efficient posture recognition was realized.For the problem that the experiencer is easily interfered with other moving targets, this paper combined the inter-frame difference method with the LK optical flow method hence strengthened the anti-interference capability of the system. First, obtain activity area using inter-frame difference method for its rapidity. Second, considering the large calculation and bad real-time performance of optical flow equation, this paper used motion estimation method based on block to predict the direction of movement. Third, compared the impact of different block sizes on real-time performance, this paper chose the block size which has less impact, besides, compared the forecast results with moving target area identified by the inter-frame difference method to improve the anti-interference capability.On the aspect of building platform, this paper used3DMax2011for virtual environment modeling and optimized the result, then, imported this to the OpenGL. This method achieved the virtual scene control based on the results of posture recognition. The experiments proved the feasibility of the posture identification methods and anti-interference methods in this paper.The results showed that the posture recognition method for virtual reality system and anti-interference method can improve system interaction in complex environments.