摘要
详细分析了Adaboost 算法的原理,依次说明了Adaboost 算法的Harr-like 特征提取方法,并通过积分图的方法实现了Harrlike 特征的快速提取,大大提高了人脸的特征提取速度,然后说明了Adaboost 算法的人脸与非人脸的阈值提取方法.并编写c＋＋程序对基于Adaboost 算法进行了实现,采用的人脸库为MITEx 人脸库,通过多轮的训练得到了-个级联的人脸检测分类器,最后进行了实验并对实验结果进行了分析.
The principles of Adaboost algorithm are analyzed in detail, And, in turn, illustrating the Harr-like feature extraction method ofAdaboost algorithm and realizing Harr-like characteristics extraction rapidly, greatly improving the human face feature extractionspeed. Then illustrates the face and non-face threshold extraction method. Writing a vc＋＋ program to implement Adaboostalgorithm. Using MITEx face library. Through several rounds of training, getting a cascade-classifier. Finally making a experimentand the experimental results were analyzed.
出处
《信息工程期刊（中英文版）》
2015年第3期90-96,共7页
Scientific Journal of Information Engineering
基金
该成果得到北京市教育委员会科技发展计划面上项目（KM201410016016）和北京建筑大学科学研究基金（No.00331614021）资助
