摘要
为了提高base haar特征和原始AdaBoost算法的检测率并降低其误检率,提出一种改进型人脸检测算法。该算法采用新增haar-like特征,改进弱分类器选取方式和权重更新方式的AdaBoost算法,构建分类性能强大的级联分类器实现对人脸的有效检测。实验对比证明,与原始AdaBoost算法相比,新增haar-like特征方法检测率提高了1.1%,误检率降低了2.45%;改进AdaBoost算法检测率提高了2%,误检率降低了5.11%;同时新增haar-like特征并改进AdaBoost算法方法检测率提高了4.22%,误检率降低了7.56%。
In order to improve the detection rate of the base haar feature and the original AdaBoost algorithm and reduce its false detection rate,an improved face detection algorithm is proposed.The algorithm uses the AdaBoost algorithm with new haar-like features,improved weak classifier selection methods and weight update methods,builds a cascade classifier with powerful classification performance to achieve effective detection of faces.Experimental comparison proves that compared with the original AdaBoost algorithm,the detection rate of the new haar-like feature method is increased by 1.1%,and the false detection rate is reduced by 2.45%;the improved AdaBoost algorithm method is increased by 2%,and the false detection rate is reduced by 5.11%;At the same time,adding haar-like features and improving the AdaBoost algorithm,the detection rate increased by 4.22%,and the false detection rate decreased by 7.56%.
作者
张彩丽
刘广文
詹旭
才华
刘智
ZHANG Cai-li;LIU Guang-wen;ZHAN Xu;CAI Hua;LIU Zhi(School of Electronics and Information Engineering,Changchun University of Science and Technology,Changchun 130022;Changchun China Optical Science and Technology Museum,Changchun 130117)
出处
《长春理工大学学报(自然科学版)》
2020年第2期89-93,共5页
Journal of Changchun University of Science and Technology(Natural Science Edition)
基金
吉林省科技发展计划项目(20170203005GX)。