基于层次支持向量机的人脸检测  

Face Detection Based on Hierarchical Support Vector Machines

作　　者：简国强[1] 黄竞伟[1] 秦前清[2] 覃志祥[1] 

机构地区：[1]武汉大学计算机科学与技术学院,武汉430079 [2]武汉大学测绘遥感信息工程国家重点实验室,武汉430079

出　　处：《计算机工程》2005年第22期181-182,188,共3页Computer Engineering

摘　　要：针对彩色图像人脸检测,提出了肤色模型和层次支持向量机相结合的人脸检测方法。检测时首先利用调节的肤色模型提取出人脸候选区域,然后对这些候选区域用线性支持向量机和主成分与非线性支持向量机相结合的层次支持向量机进行验证,获得真正的人脸区域。实验表明,该方法对图像偏色有一定的鲁棒性并可以用于灰度图像的人脸检测,而且检测正确率和速度比基于肤色和模板匹配的方法有了一定的改进。This paper proposes an approach for fast and accurate face detection, which is based on skin color model and hierarchical support vector machines. The approach starts from a coarse segmentation to obtain face candidate areas using adjusted skin color model. Then those candidate areas are verified by hierarchical SVMs which are combined by linear SVM and PCA＆ nonlinear SVM in order to get real faces. Experiments with a detection system show that the detection algorithm is robust to color distortion and is also able to be applied to face detection in gray images, and the algorithm can speed-up face detection and can promote detection correct rate to some degree compared with the detection algorithms based on skin color model and template matching.

关 键 词：人脸检测 肤色模型 支持向量机 主成分分析 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]