基于IB-Adaboost算法的人脸检测.pdfVIP

奄a叶拉2017年第30卷第5期 协议·算法及仿真 Electronic Sci．Tech．／May．15．2017 基于IB．Adaboost算法的人脸检测 刘树娟，齐金鹏，卞凤杰，李林鸽 (东华大学信息工程与科学技术学院，上海201600) 摘要针对传统Adaboost算法在光照不均匀、角度确不定的情况下，进行人脸检测出现检测率不佳的问题，文中 匀和人脸角度不定方面能够实现较为满意的检测效果。 关键词人脸检测；IB—Adaboost；Adaboost算法；光照补偿；颜色空间转换；二值化 中图分类号TP391．41文献标识码A 文章编号1007—7820(2017)05—032—04 for FaceDetection BasedonIB--Adaboost Images Algorithm UU Shujuan，QIJinpeng，BIANFengjie，LILinge ofInformation andScience (School Engineering Technology，DonghuaUniversity，Shanghai201600，China) AbstractInviewofthe facedetection ofthetraditionalAdaboost underthecondi． poor performance algorithm tionofunevenilluminationanduncertainof articleforwardan angledetection，thisputs improved illumination and detectionmethods．Themethodof offersilluminationcom- binarization)face ny compensation image to the faceclassifieristrainedbasedontheAda- pensationpretreatmentimproveimagequalitylight；secondly，the boost the transformand binarization are tonarrowthe algorithm；thenYC6C，colorspace image processingperformed facesearch theskinareacolor areobtained color toachievefacedetec— area；andfinally images byR，G，B overlay resultsshowthatthe IB—Adaboostmethodinuncertainunevenilluminationandface tion．Experimental proposed canachievemore testresults． Angle satisfactory face Keywordsdetection；IB—Adaboost；Adaboost；illuminationconversion；bi—