Abstract: The recognition and description of human hand’s biological characteristics in human computer interaction assembly system is the basis of the description of the hand assembly action and the behavior understanding. The template contour coordinates are given by the skin color recognition results based on the skin color modeling and identification, feature model learning on the knuckles. The local rotation rate and point by point vector product of hand contour chain are calculated. Describe the contour features of hand knuckle for point by point comparison method modeling. The position of fingertip and finger root is detected and inferred based on the experience of finger biological structure. The preliminary identification of finger contour feature is realized. The preliminary orientation of the finger is completed by the hand finger shape fitting, and the position and orientation of the finger are analyzed. Hand image information fusion is carried out on the basis of multi feature. Combined with knuckle position detecting by the hand knuckle learning model and adaptive threshold knuckle recognition, the finger shape positioning information is modified to improve the accuracy of detection. Experimental results show that the proposed method is feasible to detect the finger contour.

Key words: hand finger outline, human-computer interaction assembling, finger shape fitting, detection method, picture processing

摘要： 人机交互装配系统中人体手部生物特征的识别与描述是手部装配动作描述与行为理解的基础。在肤色建模与识别、指节特征模型学习的基础上，通过肤色识别结果给出的模板轮廓坐标，计算轮廓链的局部转角比率和逐点向量积，描述基于逐点比较法的指节处的轮廓特征，结合手指生物结构的经验判据，对指尖与指根位置进行检测与推断，实现手指轮廓特征点的初步识别，通过指形拟合完成手指初步定位，为指形平面位姿分析提供基础。在手部多特征检测的基础上进行图像信息融合，结合指节学习模型与自适应阈值识别的指节位置检测，对指形的定位信息进行修正以提高检测的准确性。实验测试结果表明，该手部指形轮廓信息检测方法可行。

关键词: 指形轮廓, 人机交互装配, 指形拟合, 检测方法, 图像处理