Research On The Key Technologies Of Virtual Interaction Based On Image Recognition

Human-computer interaction is one of the most key component parts of virtual reality. With the rapid development of virtual reality technology, human-computer interaction becomes more and more important. Now, people are no longer satisfied with the traditional interactive tools based on mouse and keyboard because they can not express themselves directly to the application environment. When using our hands to interact with virtual environment generated by computer, we can put our daily life skills directly to the interactive activities of virtual environment, which will greatly improve the maneuverability of the computer.Firstly, the paper introduces the content and development of virtual interaction. Then, this paper studies several key technologie of binocular stereo vision based on the theory of binocular stereo. The binocular stereo vision system is composed of following technologies:capture of images, camera calibration, image pre-processing and feature detection and 3D calculation. In this paper, CCD camera and video card are used to capture images, also studies the calibration of camera. So this paper uses Zhengyou Zhang's camera calibration method which is based on plane calibration that the way is between traditional calibration and self-calibration. It picks up Harris corners as features, calculate the parameters of the cameras. By calibrating the camera stereoly, the parameters between the two cameras are obtained. After image processing, stereo calibration is used to make the two image plane to be aligned by line. Then searching the images in one-dimensional and finish the stereoly matching according to the epipolar constraint conditions. Finally, the 3D coordinates are obtained according to the theory of stereo vision. The depth information of the object is also restored, and then the depth image is obtained.Visual C++ and OpenCV are used in the paper to test the algorithm.