基于曲率的指尖检测方法

指尖检测是基于视觉的徒手人机交互系统的关键环节,由于背景的复杂性和系统的实时性要求,导致指尖的精确定位在处理速度和准确性方面存在很大问题。本文针对这一问题,提出了一种简单高效的基于曲率的指尖检测方法。该方法首先将输入视频流基于肤色空间进行二值化,并将二值化后的视频序列作为输入数据;然后采用边缘检测算法提取出肤色区域的边缘(肤色区域的轮廓),在肤色区域的轮廓上根据曲率来对类指尖的点进行检测,并且根据类指尖点的位置关系来判定一个肤色区域是不是手;最后通过过滤算法过滤掉误判手臂点。实验结果表明,该方法在不同的应用背景下都具有很好的检测效果,对光照的鲁棒性也较高,并且能够达到实时检测的效果。

Fingertip Detection Method Based on Curvature

The fingertip is an important feature whose detection is the key connection for many vision-based barehanded human computer interaction(HCI) systems. A novel method for fingertip detection is presented.The input data is the binary image segmented from the input video stream by the skin color space.Then,the edge detection algorithm is used to extract the edge of skin region.On the contour of skin region,fingertip-like points are detected using curvature information.The positions of these fingertip-like points are used to determine whether a skin region is a hand or not.Finally,it filters out some incorrect fingertip points.Experimental results show that the proposed algorithm performs well in different backgrounds and is robust to the influence of illumination in a real-time system.