Augmented Reality System Of Human-computer Interaction Technology Research And Application

Augmented Reality(AR) is one technology that registers virtual objects generated by computer on real scene acquiring by video capture device in order to present combined image. AR technology is widely used in military, medical, education, culture and entertainment fields.As the development of Human-Computer Interaction (HCI), people are no longer confined to traditional interaction ways, but trying to explore more intelligent and natural interaction ways. Movements of hands and gestures are the most intuitive way of interpersonal communication, which contain a wealth of information. Movements of hands and gestures could be directly used in interacting with the machine. Occurring as an new means of HCI, AR technology has widely extended interaction ways of HCI systems and greatly improved the scope of application of AR technology.In this paper, based on the binocular vision principle, three parts, including segmentation gestures, finger positioning and interactive methods, are studied.Firstly, in segmentation stage, for solving the misrecognition of skin-like region, a segmentation algorithm combining skin color and background mode to preclude some skin-like regions is proposed.. Skin color and the background model are established and then image is segmented by these two methods respectively. The background model is updated according to the segmentation result of skin color model. Logical calculation on two segmentation results is conducted. At last, the biggest connected region is selected as the hand region. Experimental results indicated that this method is valid and effective. Disturbance of skin-like regions(hands are not overlap skin-like regions) could be filtered in real time.Secondly, in the finger positioning stage, based on morphological characteristics of the fingers, a recognition algorithm is proposed. By tracing the contour of segmented hand image, ROI(region of interest) of gestures is getting. During scanning the ROI, by calculating the zero-order moment and the first moments, center of image gravity can obtained. Comparing the gesture boundary points with the center of gravity, location of the fingertip could be determined. Experimental results show that this method could achieve better processing speed.Thirdly, interaction aspect, an AR system based on double camera and fingertip interaction is designed and realized., By using fingertip to drag the virtual image, the virtual image could be zoomed in and zoomed out. To judging whether the finger touches the virtual objects or not, geometry method and binocular vision principle are adopted. When the fingertip touched events of zooming in or zooming out are triggered. This system makes the HCI more natural and convenient.