摘要： 提出了一种基于灰度信息的人脸检测系统的结构和方法。该系统是由两级分类器组成,这样既加快了系统检测速度,又保证了其检测精度。两级分类器均是由RBF神经网络构成,并且第二级分类器通过有效的机器学习方法: Adaboost,提高了其识别能力。该方法根据一定的规则把多个弱分类器组合成一个强的分类器,从而达到提高识别率的目的。 实验表明,通过Adaboost增强后,分类器检测能力要优于SVM分类器。另外,由于采用了旋转采样窗口,该系统可以检测到轻微平面旋转(<±20°)的人脸。

关键词: 人脸检测, 机器学习, 旋转采样窗口

Abstract: This paper presents view-based face detection system structure and its algorithm. Due to its two level structure, the system detecting speed is accelerated and its recognition rate is also improved. The two classifiers used in the system are made of RBF network and the second one is boosted by the Adaboost algorithm for high performance. The main idea of this algorithm is to form a strong learner by a simple combination of weak learners in order to improve recognition rate. The experiments show that the boosted classifier achieves better results than the SVM classifier. In addition, the system can detect small rotation (<±20°)in plane faces via rotation sampling window.

中图分类号: 

• TP391.41