Human Hand Information Acquiring And Modeling Based On Computer Vision

Virtual Reality technology is one of the hottest computer simulation technologies, and provides an immersion sense in the 3D environment which is generated and interactive by using the computer graphics systems and a variety of display and control interface devices.Highly precise hand modeling technology based on computer vision is the key to human-computer interaction, and it mainly includes: acquiring hand’s data through the unconnected collecting device(binocular camera, Kinect camera, Leap Motion etc.), driving virtual hand in 3D scene by the motion mapping between the hand and the virtual hand, and realizing interaction between virtual hand and 3D model in the virtual scene.This paper deeply studies theories and key technologies about virtual hand interaction technology, and designs interactive demo on the Unity3 D platform, and verifies the algorithm.Detail contents are shown as follows:(1)The paper puts forward a template matching algorithm of updating successive frames by recognition for human skeleton point with Kinect camera. The algorithm uses successive frame to update the hand template and realizes the prediction for the bent and hiding finger.By using GVF Snake algorithm, the hand-shaped profile acquired from Kinect camera can be extracted. After positioning palm point from the initial contour model, accurate and complete target boundary can be obtained through the iterative search.(2)This paper builds the virtual hand model according to analyzing the physiological structure of hand and the characteristics of bone motion. It also establishes the mesh of virtual hand with aesthetic and transformational requirements by using 3ds Max software. Five fingers of the virtual hand are performed respectively motion modeling on the basis of virtual hand’s hierarchy.(3)This paper achieves the motion correspondence between virtual hand model and hand by scripting on Unity3 D platform, and studies the mapping of virtual hand model in the scene.According to vision device is unable to get accurate joint angle, this paper completes the bent pose of virtual hand by combining with gesture recognition points. On this basis, the paper further writes interactive scripts demo, reproduces the hand gestures and completes interactive in the scene.