摘要
人脸检测是自动人脸识别系统的基础。为提高人脸的检测速度,采用一种基于两级分类器和支持向量机的人脸检测方法。该方法中,第一级分类器采用特征基方法,对待检测区域进行粗筛选,过滤掉大量非人脸背景信息,然后在剩余区域,用第二级分类器支持向量机进行验证,若是人脸,给出标识。支持向量机有效克服了神经网络中可能遇到的局部极小值和过学习问题,是统计学习理论的基础上新发展起来的机器学习算法。实验结果表明该算法提高了检测速度,系统更有效率。
Face detection is base of building system of automatic face recognition. To improve the speed of detection, the paper presents a novel method of two classifiers and support vector machine based face detection. The first classifier uses feature-based method to detect the input image, and finds poison where there may have faces. Next, the support vector machine classifier checks whether there is or not. The support vector machine can overcome the short of neural network, such as local minimization and over learning, and is a new general method of machine learning based on statistic learning theory. The experiment results show the method can improve the speed of detection and effectively detect face.
出处
《计算机与数字工程》
2008年第11期143-145,157,共4页
Computer & Digital Engineering
关键词
人脸检测
分类器
机器学习
支持向量机
face detection, classifier, machine learning, Support Vector Machine
作者简介
尚凯，男，硕士，研究方向：数字图像处理，模式识别。
徐东平，男，博士，研究方向：多媒体技术，图像处理，模式识别等。
于红芸，女，研究方向：飞行器控制系统设计与仿真。
