基于Adaboost算法的人脸检测技术的研究与实现

摘要

人脸检测是人脸识别技术的基础,首先提出人脸检测系统的构成,分析Adaboost算法对图像进行人脸检测的基本原理.根据Adaboost算法形成了简单的矩形特征作为人脸特征,即Haarlike特征,然后由多个Haarlike特征相当于一个弱分类器,由多个弱分类器级联成为一个强的分类器,并将级联分类器用于动态人脸检测中,从截取的每一帧图像中进行检浏.经过实验验证,采用这种方法和步骤进行人脸检测达到了比较好的精度和速度,为接下来的人脸识别提供了前提条件.%Face detection is the basis of face recognition. The structure of the face detection system is introduced and the basic principles of Adaboost algorithm is analyzed inthis paper. Based on Adaboost algorithm, a simple rectangular feature is formed as a facial feature, whch is Haar-like features. A weak classifier is formed by a number of Haar-Iike features, and multiple weak classifiers are cascaded into a strong classifier. The cascade classifier is used in dynamic face detection to detect faces captured from each frame image. Experimental results show that this method and process of face detection can achieve a relarively good accuracy and high speed, and provide preconditions for the next face recognition.