基于凸包算法的指尖识别方法

针对现有指尖检测方法容易受背景的干扰，并且指尖误判点较多的情况，提出了一种基于凸包分析的指尖检测算法。首先，为减少类肤色背景和光照对手部轮廓提取的影响，采用YCbCr肤色模型和背景差分法相结合的方法提取手部轮廓；然后采用快速凸包算法获得手部轮廓的凸包，并利用凸包顶点和缺陷点计算出掌心的坐标，根据凸包缺陷深度和手指几何特征及曲率特征检测手指指尖；最后采用有灯光和类肤色背景干扰的环境，而不是背景单一的环境进行实验，验证了所提算法的鲁棒性。实验结果表明，该算法可以有效识别出指尖数目和手掌的位置，实现简单的数字手势（0~5）的识别，具有较强的鲁棒性。

Fingertip recognition based on a convex hull algorithm

Since the conventional fingertip detection method is easily influenced by background and there are too many finger misjudgment points, this paper proposes a detection algorithm based on the analysis of a convex hull. Firstly, in order to reduce the influence of approximate skin background color and light intensity on the extraction of the hand area, we use the YCbCr skin model combined with the background difference method to extract the hand area. We then derive the hand contour convex hull and, finally, use the defects and hull points to obtain the coordinates of the hand palm. Based on the depth of the convex hull defects, the geometric features of fingers and curvature allow detection of the fingertips. In order to verify the robustness of the algorithm developed in this paper, we employed a light and colored background, rather than a single background environment; the results showed that the method can effectively identify the number of fingertips and the location of the palm, and can also recognize numerical (0~5) hand gestures, demonstrating its efficacy.