Research On Gesture Detection Based On Variational Level Set Model

Gesture recognition is an important research field of intelligent human-computer interaction. Gesture detection is a crucial step in gesture recognition. How to obtain the exact contours of hands is the bottleneck to improve the recognition rate and extend gesture set. Most gesture recognition systems use skin color model to segment hand. The segmentation results are affected by to the light, background and other conditions. Aimingat gesture detection, level set is applied to get the exact contours of hands in this paper. Specifically, the main contents of this paper are as follows:1.Firstly, the skin color model is used to segment the hand area. Secondly, the variational level set method proposed by Li and the adaptive variational level set method are applied to gesture segmentation. The segmentation results of these three methods are compared. The adaptive level set method can indeed achieve better detection results.2.Some gestures have sharp corners (such as one standing finger).Both Li variational level set method and the adaptive level set method cannot resolve such problem because of the restrictions of their edge indicaor functions.Aiming at this shortcoming, convex variational level set method is applied to gesture segmentation. Aadaptive variational level set method and convex variational level set method are integrated to resolve these two problems at the same time:the dependence on the initial contour, and the unsatisfactory sharp corner segmentation. Thus the application of variational level set method in gesture detection is promoted.3.Finally, the system based on improved adaptive variational level set is designed and implemented.