摘要
人脸检测在身份认证、人口统计、实时检测等领域都起着非常重要的作用。作为一种预处理手段,人脸检测和定位可以有效地提高表情识别和人脸识别的识别率。为此提出了一种先定位眼睛再提取人脸的改进方法,首先基于adaboost算法的级联人眼检测器检测到人眼+眉毛的侯选位置,然后根据人眼的地形特征和模板匹配准确定位人眼,最后再利用人脸的拓扑结构提取出人脸,达到有效检测人脸目的的方法。该方法很好地降低了误检率,并大大提高检测速度,同时也解决了由于眼睛和眉毛非常相似而引起的个体混淆。
As a preprocess method, face detection plays a very important role in authentication, demographic statistics and Real - time detection. It can improve the recognition rate in the facial expression recognition and face recognition. The paper presents an improved method which locates eyes, then extracts face. Firstly, the paper shows a novel approach for wide outline of eye and eyebrow based on Adaboost algorithm. Secondly, the topological features and symmetry are applied to eye location. Finally, the simple algorithm based tace topology is used to extract face . By using this method, the rate of false can effectively reduced, the detecting speed can be increased, and the problem of individual diversity due to the similitude of the eyes and eyebrow is also solved.
出处
《计算机仿真》
CSCD
北大核心
2010年第3期211-214,共4页
Computer Simulation
基金
电子信息产业发展基金资助项目(信部运[2007]292号)
关键词
自举算法
级联分类器
人眼定位
模板匹配
Boost algorithm
Cascade classificator
Eye location
Template matching
