摘要
AdaBoost算法要提高检测精度,需要级联更多的强分类器,这样会降低检测速度。针对这个问题,在AdaBoost级联分类器中引入加权判决函数,对其中相互独立的级联分类器判决结果进行信息融合,不增加级联的强分类器个数,提高了检测率。实验结果表明,该方法在保证检测速度的同时,提高了检测率,在CMU+MIT人脸测试库上取得较好的效果。
This paper develops a novel cascade classifier structure. Weighted decision function is used to hal Pdecision by the classifier in the cascade architecture. It remedies defects of the slow detecting speed, which caused by cascaded a lot of strong classifiers. In or- der to obtain the complementary and effective and comprehensive information, it is necessary for the front of classifiers' results to fuse information. Experimental results on face detection show that the improvement of the recognition performance comparing to traditional cascade AdaBoost classifier. Finally, experimental results on CMU + MIT dataset demonstrate that the algorithm is efficient.
出处
《微计算机应用》
2009年第9期39-42,共4页
Microcomputer Applications