摘要
该文提出了一种基于改进的Adaboost算法的人脸检测方法。Adaboost是一种构建准确分类器的学习算法,它将一族弱学习算法通过一定规则结合成为一个强学习算法,从而通过样本训练得到一个识别准确率理想的分类器。但是,Adaboost在有高噪音样本的情况下,有可能发生过配现象,该文在Adaboost算法的基础上,对其权值更新规则做了改进,并结合PCA进行人脸检测。仿真试验表明,该方法具有良好的性能,同时可以在一定程度上有效防止过配现象的发生。
This paper presents a new algorithm based on improved Adaboost for face detection. Adaboost is a learning algorithm for constructing accurate classifiers. It can obtain a strong learning algorithm by combining a series of weak learning algorithms through some rules, but it tends to overfit in the presence of highly noise samples. In this paper ,we improve the weigh -update rules,and use PCA and our improved Adaboost to detect face images. Experiment results show that this approach has good performance , and the modified weigh - update rules can effectively a- void overfitting.
出处
《计算机仿真》
CSCD
2005年第9期167-169,253,共4页
Computer Simulation