一种改进的AdaBoost算法——AD AdaBoost

被引:53
作者
李闯
丁晓青
吴佑寿
机构
[1] 清华大学电子工程系
关键词
AD AdaBoost; 目标检测; 级联结构; 弱分类器; 加权参数;
D O I
暂无
中图分类号
TP391.41 [];
学科分类号
080203 ;
摘要
目标检测问题是计算机视觉领域最普遍和关键的问题之一.基于级联结构的AdaBoost算法目前被认为是较有效的检测算法,但是其在低FRR端的性能仍需改进.文章提出了一种针对目标检测问题的改进AdaBoost算法———AD AdaBoost.AD AdaBoost采用了新的参数求解方法,弱分类器的加权参数不但与错误率有关,还与其对正样本的识别能力有关.该算法能够有效地降低分类器在低FRR端的FAR,使其更适用于目标检测问题.新旧算法在复杂背景中文字检测的实验结果对比证实了新算法在性能上的改进.
引用
收藏
页码:103 / 109
页数:7
相关论文
共 9 条
[1]  
Neural network-basedface detection. Rowley H A,Baluja S,Kanade T. IEEE Transactions on Pattern Analysis and Ma-chine Intelligence . 1998
[2]  
A general framework for object detection//. Papageorgiou C P,Oren M,Poggio T. Proceedings of the6thInternational Con-ference on Computer Vision . 1998
[3]  
Astatistical method for3D ob-ject detection applied to faces and cars. Schneiderman H,Kanade T. Proceedings of the IEEE Conference on Computer Vision and Pattern Recogni-tion . 2000
[4]  
Robust real ti me object detection. Viola P,Jones M. Pro-ceedings of the2nd International Workshop on Statistical and Computational Theories of Vision . 2001
[5]  
Fast and robust classification using asym-metric AdaBoost and a detector cascade. Viola P,Jones M. Advances in Neural Information Processing System14 . 2002
[6]  
I mage parsing:Unifying segmentation,detection,and recognition. Tu Zhuo-Wen,Chen Xiang-Rong,Yuille A.L,Zhu Song-Chun. Proceedings of the International Conference on Computer Vision . 2003
[7]  
Adacost:Mis-classification cost-sensitive boosting. Fan Wei,Stolfo S,Zhang Jun-Xin,Chan P. Proceedings of the16th International Conference on Machine Learning . 1999
[8]  
Adecision-theoretic generalization of on-line learning and an application to boosting. Freund Y,Schapire R E. Journal of Computer and SystemSciences . 1997
[9]  
Real-ti me rotation invariant face detection based on cost-sensitive AdaBoost. Ma Yong,Ding Xiao-Qing. Proceedings of the IEEEInternational Conference on I mage Processing . 2003