Abstract: A method is proposed which is based on curvature scale space to detect people’s head. The presented algorithm analyzes the head shape, and gets the features which are different from other parts. The curvature of each point on the contour is calculated under the curvature scale space frame, combines with the special features of head shape to detect the head. Experimental results show that the presented method can detect people’s head effectively in complex background and provide a new approach for people’s head detection.

Key words: head detection, curvature scale space, pattern recognition

摘要： 利用人头部轮廓的形状特征，提出一种基于曲率尺度空间的人头部检测算法。算法通过比较分析得出了人头部轮廓区别于其他部分轮廓的形状特征，在多曲率尺度下，计算物体轮廓曲线上每一个点的曲率，结合形状特征信息进行人头部检测。实验结果表明，算法有效解决了复杂背景下的人头检测问题，为人头部检测提供了新的途径。

关键词: 人头检测, 曲率尺度空间, 模式识别