Adaboost人脸检测算法研究及OpenCV实现.pdfVIP

14 5 V o l14 N o5 2009 10 JOURNAL OF HARB IN UN IV ERSITY OF SC IEN CE AND TECHNO LOGY O ct. 2009 Adaboost OpenCV 郭 磊, 王秋光 , 150040) : 针对复杂背景条件下人脸检测的检测率低速度慢的问题, 选用了基于统计的人脸检测 算法. 详 介绍了A daboost人脸检测算法的原理及训练流程. 利用 OpenCV 图像处理软件与A da boost人脸检测算法的结合模块, 选取扩展的H aar特征, 实现了人脸检测并给出了检测结果. 设计 的检测器对 592 个人脸漏检 62 个, 可达倒 20帧 / s 左右. 因此该方法具有较高的准确度和实时性. : A daboost; 人脸检测; OpenCV : TP391. 4 : A : 1007- 2683 2009) 05- 0 123- 04 Resea rch of Face Detect ion Based on Adaboost A lgo rithm and OpenCV mI p lementat ion GUO Le,i WAN G Q iuguang School of E lectrical and E lectron ic Engin eering, H arb in U n iversity of S cience and T echno logy, H arb in 150040, C hina) A bstract: To solve the prob lem that the face detector is slow and its rate is low in com p licated backg round, in this paper, w e choose a face detector a lgorithm based on statistics. The pr incip le and tra in ing flow of A daboost face detector is detailedly introduced. T he m odu le com b ine mi ag ing processing softw are OpenCV and A daboost face de tector algorithm is used to select ex tended H aar characteristics, then the face detector is realized and the resu lt is g iven. 62 faces is undetected w hen detector designed fo r 592 individuals face, and it can reatch about 20 fram e / s. T herefore the m ethod o f m ore accuracy and real t mi e. K ey words: A daboost; face detect ion; OpenCV; com puter vision [ 3] , 0 引 言 , .