Research of Face Detection Classifier Based on Support Vector Machine

Abstract

Face Detection is a process that the face of all including the location, size, and configuration is identified when the image is input,在一个图象环境中，人脸在肤色、姿势、表情等方面千差万别，有时还会存在部分遮挡，这使得人脸检测算法非常复杂，存在训练、识别算法设计困难和计算量大等问题。 In an image environment, the face’s color, position, facial expressions, and other aspects are so diversity and sometimes there will be some shielding that the detection algorithm becomes so complex, and also make many problems in the computation.针对为一问题，许多研究者提出了各种解决方法，其中支持向量机(SVM)较好地解决了小样本、非线性、高维数等实际问题，引起了研究者的重视。 To address the problem, many researchers propose a variety of solutions, including Support Vector Machine (SVM) which can solve some actual problems such as the small sample, nonlinear, and high-dimensional, and now many researchers pay attention to it.

The traditional classification on face detection includes the face patterns matching method and the PCA method, and so on. There are a lot of deficiency in the face of the pass rate、inhuman face filtration rate and the computational complexity.

支持向量机（ SVM ）方法是基于结构风险最小化原理的统计学习方法。 Support Vector Machine (SVM) is based on structural risk minimization principle of statistical learning methods. 利用线性SVM和非线性SVM分类器可以获得理想的人脸检测分类效果，但非线性SVM计算复杂度较高，速度较慢。 linear and nonlinear SVM classifier can be get ideal face detection classification results, but the nonlinear SVM method have a lack of higher computational complexity.

为了提高人脸检测性能，同时适当考虑复杂度，减短检测时间，提出一个基于SVM的人脸检测系统。 In order to improve face detection’s performance, and reducing the complexity and the detection time, a SVM-based Face Detection System is proposed. 该系统以支持向量机为工具，使用适用于复杂环境下灰度图像的人脸检测算法，可以有效地提高检测速度，降低虚警数。This system support vector machine can raise the checking speed and lower the number of false alarm effectively by using the appropriate face detection algorithm.

该算法首先用“双眼-平均脸模板对”进行人脸定位粗筛选，然后分别通过线性SVM粗分类器检测和非线性SVM分类器完成人脸检测。 This algorithm first locate face and do filtration roughly by using the "eyes - the average face template ", and then finish the face detection through a rough classification of SVM linear and nonlinear detection SVM classifier.

Key words: Face Detection , Support Vector Machine ,Classifier , Linear