为满足人机交互系统中手势识别与理解的需要，提出一种快速、鲁棒的手势分割与指尖检测方法。手势分割通过肤色阈值分割、链码轮廓搜寻、自适应的边缘拟合三个步骤完成，可实现手势由粗到细的逐步精确分割。在手势分割基础上再通过手轮廓边缘的几何特征制定指尖检测算法，可实现多个手指的快速精确检测。实验结果表明，系统方法可行，在各种复杂背景、不同手指指向和运动速度下均能实现快速准确的手势分割和指尖检测，可为下一步人机交互系统中的手势识别和理解打下良好的基础。

To satisfy the demands of hand recognition in computer interaction system, a fast robust hand segmentation and fingertip detection method is proposed. The procedure of segmentation consists of three steps: 1) Skin color segmentation, 2) Contour finding by freeman code, 3) Adaptive spline fitting of the contour. The hand could be segmented exactly step by step. Then, the accurate position of the fingertip was determined through the feature of the hand contour based on the segmentation. Experiments show that the proposed method works very well with high hand segmentation and fingertip detection accuracy in different condition, such as complex background, multiple hand direction and finger motion. It is appropriate for the hand recognition in human-computer interaction system.