一种改进的级联AdaBoost分类器.pdfVIP

第 27 卷第 12期 计算机应用 V o. 27 No. 12 2007 年 12 月 C ompu ter A pp ications Dec. 2007 : 100 1- 90 81( 2007) 12- 3029- 03 A daBoost 柳 锋, 朱庆生, 杨世泉, 张 敏 (重庆大学计算机学院, 重庆 400044) ( kofsky@ gm ai. com ) : 在基于级联结构的A daBoost分类器中引入 辅助判决函数当样本被某级A daBoost分 类器判为假后, 使用该级的辅助判决函数对其进行二次判决辅助判决函数不仅考虑 当前级分类 器的判决结果, 还考虑前面分类器的历史判决信息实验结果表明, 该方法相对级联 A daBoost分类 器以及二维级联A daBoost具有更好的识别效果 : 级联分类器; A daBoost; 辅助判决函数 : TP 18 : A I m proved cascade A daBoost classifier L IU F eng, ZHU Q ing- sheng, YANG S h -i quan, ZHANG M in (College of Computer Science, Chongqing University, Chongqing 400044, China) A bstract: A ssistant discr mi inant function w as u sed to he p dec ision by the c assif ier in the cascade arch itecture. T he samp es w hich w ere rejected by the A daBoost c assif ier w ou d transfer to the assistant d iscr mi inant function and the assistant dec ision w ou d be m ade. R esu t o f current AdaBoost c ass ifier and h isto ry discr mi inant inform ation has been taken into consideration in the assistant discr mi inant function. T he expermi enta resu ts show the mi provem ent o f the recogn ition per fo rm ance com par ing to the cascade A daBoost c assif ier and 2D cascaded A daBoost c assif ie r. K ey w ords: cascade c assif ie r; A daBoost; assistant d iscrmi inant function , 0 引言 , , [ 1] , A daBoost , , , , , A daBoo st