The Research And Implementation Of Virtual Reality System Based On Intelligent Video Technology

With the continuous development of virtual reality technology and the mature of display technology, using virtual reality technology to experience the simulation in virtual environment has not only been applied unprofessional fields such as military, medicine and film production, but also has extended to people’s daily life. At the same time, the rendering of virtual reality has been greatly improved because of the way-beginning of computer processing level especially in graphics processing ability and 3D display devices’ expansion. Virtual reality technology has been applied in various industries as its direct presentation and realistic rendering. Naturally, the applications of virtual reality and related technologies have become a research hotspot.This paper analyses the requirement for virtual reality experience system’s function and performance as the background of the virtual household experience system for application. Put forward to use Kinect as the video collection equipment and propose a motion controlled human-machine interaction scheme in virtual reality by extracting the depth information and gesture recognition. Using Unity 3D game engine to construct virtual reality scene and improve scene rendering by defining the shader, I complete immersive three-dimensional scene presentation with Oculus Rift DK2.In the human-computer interaction module, I use open natural interaction framework Kinect for Microsoft SDK to extract the depth information and obtain the controller’s virtual node location and finally implement the gesture recognition of human body. I compare three different gesture recognition algorithms and choose the gesture recognition algorithm based on joint angle detection. I also propose an improved joint angle detecting algorithm to decrease the measuring errors that caused by unstable spatial position of joint detection and improve the accuracy of gesture recognition.To improve the accuracy of joint detecting, I introduce the initialization process before gesture recognition, this process read the operator’s joint spacing as standard value, combine with the characteristics of human body activities to identify the reliability of key points during gesture recognition and set the joint reliability. At the same time, I use Kalman filter algorithm to reduce the effects of noise and interference in the joint inspection process and set the observation and estimation values’ weight dynamically in the Kalman filter according to the reliability of joint in current frame.As for virtual reality scene construction, I mainly focus on improving the effect of virtual scene rendering and increase the experience of scene reality. In this paper, I custom the Shader control rendering process to simulate the specular reflection and bump diffuse of real world objects, and complete the Shader of secular reflection and bump diffuse to improve the display effect. Binocular cameras invoke the OVR plugins to accomplish the adaptive of Oculus Rift DK2 in Unity 3D settings.