基于骨架运算的动态手势定位及识别算法

为满足在不同背景条件下用户的手势识别需求,提出了一种对于动态手势定位及识别的算法。在摄像头固定、目标移动的前提下,采用ViBe算法对背景建模后结合肤色检测,对手势目标区域进行检测。文中提出了掌心的定位方法,算法利用形态学骨架运算提取图像特征,为解决用户不同衣着和动作造成判断区域的多样性和不确定性,进行有效区域选取后定位掌心坐标,根据掌心坐标通过似圆度和掌心极坐标进行手势定位和识别。实验结果证明,所提算法能够在复杂背景下准确地对手掌高于手肘的手势判断出掌心点,分割出手掌并进行手势识别,该算法同时拥有较好的鲁棒性和实时性。

Dynamic Gesture Localization and Recognition Algorithm Based on Skeletonization

A novel algorithm for dynamic hand gesture localization and recognition based on skeletonization is proposed, which satisfies the requirements of the user under different background conditions. With the presumption that pictures of the moving objects are taken by fixed camera, the ViBe algorithm combined with skin-color detection is used for detecting the hand area. In order to eliminate the variety and uncertainty caused by user＇s various dresses and motions, an innovation in this article is stated in the method of palm center extraction, which is skeletonizing the hand area, finding out the effective area and then the coordinate of palm center, and positioning and recognizing the hand gesture referring to the roundness and palm center polar coordinates. Experimental results demonstrate that, in the situation which the position of hand is higher than that of elbow, the algorithm works precisely under the complex background with good robustness and real-time performance,