【摘要】 复杂背景中的人脸检测可广泛应用于人脸识别、人机交互等方面。但目前大部分人脸检测方法中存在分类器训练困难和检测计算量大等问题。提出了一种基于层次型支持向量机的正面直立人脸检测方法,在这两方面作了改进。这种结构的分类器由一个线性支持向量机组合和一个非线性支持向量机组成,由前者在保证检测率的情况下快速排除掉图像中绝大部分非人脸区域,后者对人脸候选区域进行进一步确认。在卡内基梅隆CMU等数据库上的实验证明了这种方法不仅具有较高的检测率和较低的误检率,而且具有较小的计算量。更多还原

【Abstract】 Face detection in images with complex backgrounds has many applications such as face recognition and humancomputer interaction. However, most face detection methods have troubles with the training of the classifier and the computational complexity. A face detection method based on a hierarchical support vector machines (SVM) presents improved methods for both of these problems. The hierarchical SVM is a combination of several linear SVMs and a nonlinear SVM. In the detection stage, the linear SVMs are used to quickly exclude most nonfaces in the images while the nonlinear SVM is used to further verify possible face candidates. Experimentals using CarnegieMellon University and Nokia databases demonstrat that the hierarchical architecture of the classifier not only achieves high detection rate, but also reduces the computational complexity.更多还原