HOG 特徴量と差分ステレオによる複数識別器を用いた人物検出.pdfVIP

G22 HOG 鐗瑰敬閲忋仺宸垎銈广儐銉偑銇倛銈嬭鏁拌瓨鍒ュ櫒銈掔敤銇勩仧浜虹墿妞滃嚭 涓ぎ澶у锛廕ST CREST 鈼嬫湁姹熻獱锛屻儮銉偄銉儍銈点兂銉夈儹锛屾槦宸濅綉纾紝鐢熷舰寰癸紝瀵烘灄璩㈠徃锛屾鐢板拰鏄 Human Detection Using Multiple Classifiers Based on Subtraction Stereo with HOG Features Chuo Univ. / JST CREST Makoto ARIE, Alessandro MORO, Yuma HOSHIKAWA, Toru UBUKATA, Kenji TERABAYASHI, Kazunori UMEDA It is important to detect humans from an image automatically for the measurement of human flow with a surveillance camera. We have proposed a stereo vision method called Subtraction Stereo which focuses distant measurement on foreground regions. We construct a human detection method with little incorrect detection using multiple detectors. A fast and stable human detection is achieved using the subtraction stereo with Histograms of Oriented Gradients (HOG) features. The effectiveness of our method is verified by experiments. 锛戯紟搴忚珫 锛擄紟锛 銈︺偅銉炽儔銈︺伄璧版熁鎵嬫硶 Dalal 绛夈伅锛孒OG 鐗瑰敬閲忋倰绠楀嚭銇欍倠銈︺偅銉炽儔銈︺倰妲樸€呫仾銈点偆 鐩ｈ銈儭銉┿亱銈夈伄浜烘祦瑷堟脯銇仧銈併伀锛屻偒銉°儵鐢诲儚銇嬨倝鑷嫊 鐨勩伀浜虹墿銈掓鍑恒仚銈嬨亾銇ㄣ亴閲嶈銇с亗銈嬶紟銇撱倢銇俱仹銇垜銆呫伅锛 銈恒仹璧版熁銇楋紝鐢诲儚鍏ㄤ綋銇嬨倝浜虹墿妞滃嚭銈掕銇ｃ仸銇勩倠[2]锛庛仐銇嬨仐锛 鍓嶆櫙闋樺煙銇伩銇窛闆㈣▓娓倰闄愬畾銇欍倠宸垎銈广儐銉偑銈掓彁妗堛仐銇 銇撱伄鎵嬫硶銇с伅锛屽ぇ閲忋伄瑷堢畻鏅傞枔銇屽繀瑕併仺銇倞銉偄銉偪銈ゃ儬銇 銇汉鐗╂鍑恒伅闆ｃ仐銇忥紝瑾ゆ鍑恒倐澶氥亸銇倠锛 銇勩倠[1]锛庛伨銇燂紝浜虹墿銇ぇ銇俱亱銇舰銈掕〃鐝俱仚銈嬨亾銇ㄣ亴鍙兘銇 鏈爺绌躲仹銇紝宸垎銈广儐銉偑銈堛倞寰椼倝銈屻倠鍓嶆櫙闋樺煙銇伩銇 Histograms of Oriented Gradients (HOG) 鐗瑰敬閲廩2] 銈掑樊鍒嗐偣銉嗐儸 銈倛銈婂緱銈夈倢銈嬪墠鏅牁鍩熴伄銇裤亱銈夌畻鍑恒仐锛屽墠鏅牁鍩熴伄璺濋洟鎯 銇楋紝銈︺偅銉炽儔銈︺倰璧版熁銇欍倠锛