Research On Hand Recognition Oriented To Virtual Sculpture

With the development of virtual reality technology, mouse, keyboard and other traditionalhuman-computer interaction technology has not been able to meet current needs, gesture recognitiontechnology Based on visual is an important part of multimodal man-machine interface technology, hasbeen used more and more widely. How to control the virtual world and improve the virtual systemnaturally and effectively has become a research hotspot in this area.In this article gesture recognition technology has been applied to virtual sculpture system.it cannot only expand the function of the system, but also reflect its authenticity.The content of this paperare as follows:The preprocessing of gesture image. Pretreatment is the first step after gestures capture bycamera and the processing quality directly affects the efficiency and accuracy of gesture recognition.The capture of the RGB image should be converted to HSV color model, and use H channel forsegmentation of gesture, including: median filter smoothing, Otsu, morphological processing, grayprojection method for cutting and contour mapping.The extracting of gesture features. According to the requirements of virtual sculpture of gesturerecognition system,gestures should meet three invariances of rotation, translation and scaling. Basedon this, this paper choose Ratio of the circumference and area as one of features, it has rotationaland translational invariance. Then select7Hu moments of the gesture and the numbei of extendedfinger as a gesture features, they do not change with gestures’s Rotation, translation and scaling.Gesture recognition. This paper introduce two methods of template matching and support vectormachine. Firstly, it introduces the support vector machine theory and its application; Secondly, itanalysize and compare four methods of one-to-one, one-to-multiple, directed acyclic graph, andsupport vector machine tree, and introduces the composition and selection of kernel function. Finally,selected one-to-multiple support vector machine based on RBF kernel in function, and got theoptimum parameters through calculating.