摘要
文中实现了彩色序列图像的人脸检测和识别的系统 .所用到的关键性技术是肤色运动分析、主成分分析 (PCA)和支撑向量机 (SVM) .首先根据彩色序列图像中人脸的色度特性、运动特性、几何特性和灰度分布特性完成人脸的检测工作 ;其次通过人脸模式之间的相关性进行主成分分析 ,提取并且选择特征 ,将所选择的特征训练SVM ,最后用已经训练好的SVM完成对人脸的识别任务 .系统算法结构遵循以下原则 :先使用运算量少的简单方法尽可能减少搜索空间 ,然后在已经大大减少的空间中再用复杂方法处理 ,可以在保持高的检测和识别率的同时 。
A human face detection and recognition system on color image series is implemented in this paper. The key technologies involved are skin hue analysis, motion analysis, primary component analysis (PCA) and support vector machine. The human faces can be detected through their hue, motion, geometry and gray distribution features. PCA analysis can be done through the relation between face patterns, and then the features can be extracted and selected. Using the selected features to train multiple SVMs, which can finally classify human faces. The system structure is according to the following principle: Firstly simple methods are used to reduce the search space, and then more complicated methods are used in the reduced space. So the system can have a quick response speed as well as holding high detection and recognition rate.
出处
《电子学报》
EI
CAS
CSCD
北大核心
2003年第4期544-547,共4页
Acta Electronica Sinica
基金
国家自然科学基金 (No 30 1 70 2 74)
