Research And Implementation Of Hand Gesture Recognition Based On Binocular Vision

Nowadays, hand gesture recognition especially the recognition based on the computer vision is widely used, which has become the trends of gesture recognition technology, because it can supply a kind of Human Computer Interaction (HCI), which is natural and harmonious. Because of its diversity, ambiguity, complexity, spatial and temporal differences and many other features, gesture recognition can always be a research topic of the domain in computer vision with a lot of activity and challenge. At the same time, gesture recognition as a hot issue among the new technologies of HCI, is concerned by researchers and scholars domestic and aboard.After the analysis and summary of the relevant research, this paper makes some efforts on the systemic research of hand gesture recognition based on the binocular vision. The main content and results reflect in the following respects:Firstly, I do some researches on the coordinate system and camera models, which are often used in camera calibration; and discuss the calibration method of the internal and external parameters of the single camera. A new fast and simple camera calibration method is aroused with Zhengyou Zhang plane template algorithm, and a calibration method for the binocular three-dimentional vision is established on the basis of the method mentioned above.In this section of gesture segmentation, the common methods are compared and analyzed together. And the research of gesture identification method is given, which is based on the complexion modeling and a complexion modeling method is established in YCb’Cr’space for special imaging equipment after nonlinear transformation to eliminate the impact of partial illumination combined with the theory of Hse R L complexion modeling.Finally, template matching and the profile analysis are used to detect the fingertip. According to the comparative analysis of the results, a new fingertip detecting method is established with the advantages of simpleness, high precision and fast calculating speed. On account of the parallax principle, the fingertips coordinates can be calculated, and smooth continuous gesture trace and good experimental result are obtained with Kalman filter to process the fingertip trace.