Research Of Hand Positioning And Gesture Recognition Based On Binocular Vision

when interacting with virtual objects in virtual reality environment, the operator using the input devices about traditional two-dimension and keyboard could not achieve immersion and funny. In the usual process of interpersonal communication, people can not only use language and word to communicate, but also can communicate with gestures. Human hands have not only contained a large number of interactive information with human cognitive habits but also the three-dimensional space position information from the continuous movement of hands, so we can combine the location and gestures information as a whole. The optical positioning equipment, with its easy installation, relatively inexpensive, non-contact measurement, is increasingly subject to attention.a system, both locating the three-dimensional position space of hand and simultaneously identifying the appropriate gestures, is designed, which can serve as the interface to driver virtual hand completing manipulation of grasping, moving and releasing virtual objects. The system can be used as a natural, efficient interface driven with immersive virtual hand in training, virtual assembly, virtual surgery, robotic arm control and so on.In this paper, we first analyze the system's requirements and design the system's functional. The whole system is divided into six parts:the operator, the image acquisition module, hand positioning module,3D registration module, gesture recognition module and the virtual operating module, the system's framework is given combined with six modules. Then the implementation of several modules are introduced. In the positioning module, the author uses CamShift algorithm to track the targets, and improves it to track targets automatically. The paper also proposes a practical method of extraction feature points of hand, which makes the center of mass of hand as the match point and calculates the location information of the target based on Mathematical model of binocular visual positioning; In the gesture recognition module, in order to avoid the influential of information of brightness, we first convert RGB color space to HSV color space;then we use two constraints of position and area of contours to determine hand contour form HSV image. Finally, according to the convex hull theory, we can extract the convex hull of hand contour, calculate the ratio between hand contour area and convex hull points area, and recognize gestures by differentiating ratios of different gestures.Finally, the paper introduces the system environment of the hardware and software, shows the core flow for this system, and gives an application in the virtual chemistry experiments as the specific examples.