基于级联式Boosting方法的人脸检测

Face detection based on cascaded boosting algorithm

提出一种基于级联式Boosting方法的人脸检测算法。先用PCA方法对人脸图像进行特征参数的提取,在此基础上,利用算法中的每一个Boosting分类器学习的历史信息,基于线性回归特征消除(RFE)策略,消除AdaBoost中的冗余,据此判别一幅图像是否为人脸图像。在ORL人脸图像库的仿真实验结果显示,这种方法明显提高了检测性能,证明了该算法是有效的。

An algorithm based on Cascaded Boosting for human face detection was proposed, Firstly, the PCA was used to extract the feature from human face images, and integrate historical information into successive boosting learning. Based on linear recursive feature elimination（RFE） strategy, the redundancy of AdaBoost was removed. The ORL face database was used to test the proposed method. Experiment results shows that the method is effective and good detective performance is achieved.