摘要： 提出一种基于肤色信息的改进水平集分割算法，给出整个算法的推导和实现过程，实现复杂背景下的精确人手轮廓分割，在进行人手跟踪时，使用粒子滤波对手的位置和大小进行跟踪，并用跟踪结果初始化水平集函数，以此加快轮廓曲线的收敛速度，获得手的轮廓后，对指尖位置进行定位。实验结果表明，该算法能够在复杂背景下实时、准确地跟踪人手轮廓和指尖位置。

关键词: 人手跟踪, 粒子滤波, 水平集, 肤色

Abstract: This paper proposes an improved level set segmentation algorithm based on skin color feature to implement accurate hand contour segmentation from clutter background. It introduces the deduction and implementation of the proposed total algorithm. When applying the improved algorithm to hand tracking, the location and size of hand obtained by particle filter is taken to initial the level set function, which can speed up the convergence speed of curve iteration. The extracting contour is used to locate fingertips. Experimental result shows that the algorithm can track accurate hand contour and fingertips from clutter background in real-time.

Key words: hand tracking, particle filter, level set, skin color

中图分类号: 

• TP18