基于改进的AdaBoost算法的人脸检测与定位  

Face detection and location based on improved AdaBoost algorithm

作　　者：徐前[1] 赵德安[1] 赵建波[1] 

机构地区：[1]江苏大学电气信息工程学院,江苏镇江212013

出　　处：《传感器与微系统》2010年第1期94-97,共4页Transducer and Microsystem Technologies

摘　　要：针对传统AdaBoost算法在训练过程中出现的退化现象和检测率低的问题,提出了一种有效的解决方法。该方法在传统AdaBoost算法的基础上,对样本的权值参数和弱分类器的加权参数加以改进,有效地抑制了困难样本权值的过分增大,加强了分类器对样本的识别能力,并提高了系统的检测率。实验证明:使用该方法训练的级联人脸检测器具有良好的性能。Aimed at the phenomenon of degradation and the issue of low detection rate in training process of traditional AdaBoost algorithm, an effective method is presented. On the basis of the traditional AdaBoost algorithm,this method effectively restrains weights of hard samples not to expand largely and strengthens the capacity of classifier for recognition of samples by improving the parameters of sample weights and the weak classifier weighting value. The experimental results show that the face detector established by the new method has good performance.

关 键 词：人脸检测 权值参数 级联分类器 检测率 

分 类 号：TP274.2[自动化与计算机技术—检测技术与自动化装置]