大規模歩行データベースのための自動歩行計測システム

全文

(1)Vol.2016-CVIM-200 No.18 2016/1/22. ৘ใॲཧֶձ‫ڀݚ‬ใࠂ IPSJ SIG Technical Report. େ‫ن‬໛าߦσʔλϕʔεͷͨΊͷࣗಈาߦ‫ܭ‬ଌγεςϜ ໦ଜ ୎߂1. ᴳ‫ ݪ‬༃1. େ૔ ࢙ੜ1. ຬ্ ҭ‫ٱ‬1 ୮Ӌ ਅོ1 ീ໦ ߁࢙1. ੨໦ ઍਘ1. ླ໦ Թ೭1. ଜদ େ‫ޗ‬1. ֓ཁɿҰൠతʹาߦσʔλऩू͸ɼืूͨ͠ඃ‫ऀݧ‬ͷาߦө૾ΛऩूऀࡱӨ͢Δ͜ͱͰߏங͞ΕΔɽ͜ͷ৔ ߹ɼඃ‫ूืऀݧ‬ͷࠔ೉͞ɼ͓Αͼऩूऀͷ࡞‫ྔۀ‬ͷ఺͔Βେ‫ن‬໛Խ͕࣮‫͞ݱ‬Εͯ͜ͳ͔ͬͨɽͦ͜Ͱզʑ ͸ɼମ‫ܕݧ‬ͷࣗಈาߦ‫ܭ‬ଌɾσʔλࡱӨγεςϜΛ։ൃ͢Δ͜ͱͰେ‫ن‬໛าߦσʔλϕʔεߏஙΛՄೳʹ ͨ͠ɽ։ൃͨ͠γεςϜ͸ɼ15 ౓ࠁΈͷ 14 ํ޲͔ΒͷาߦσʔλΛࡱӨͰ͖ɼ໿ 11 ͔݄Ͱ 60,000 ਓҎ ্ͷσʔλ͕ऩूՄೳͰ͋Δͱ༧ଌ͞ΕΔɽຊγεςϜͰ͸ମ‫ऀݧ‬ͷา͖ํͷ‫ݸ‬ੑ‫ܭ‬ଌΛߦ͍ɼ࿹ৼΓ΍ า෯ɼาߦ೥ྸͳͲͷ‫ܭ‬ଌΛߦ͏ɽຊ࣮‫Ͱݧ‬͸ɼาߦ೥ྸͱ࣮೥ྸͷධՁΛߦͬͨɽ. 1. ͸͡Ίʹ. σʔλϕʔεͰ͋Δ The OU-ISIR Gait Database, Large. Population Dataset [5] ͸ମ‫ܕݧ‬ͷσϞʹΑΔσʔλऩू. ۙ೥ɼ‫ݸ‬ਓೝূͷํ๏ͱͯ͠‫ݸ‬ਓͷੜମ৘ใΛ༻͍Δੜ. Λߦ͓ͬͯΓɼՊֶ‫ؗ‬΍ΦʔϓϯΩϟϯύεͰ਺೔ؒσϞ. ମೝূ (όΠΦϝτϦΫε) ͕஫໨ΛूΊ͍ͯΔɽόΠΦ. Λߦ͏͜ͱͰσʔλऩूΛߦͬͨɽ͔͠͠ɼ͜ͷํ๏Ͱ͸. ϝτϦΫεʹ͸ DNA [1]ɼࢦ໲ [2]ɼ੩຺ [3]ɼ೒࠼ [4] ͱ. ࡱӨγεςϜͷૢ࡞΍ɼମ‫ऀݧ‬ͷՙ෺؅ཧ΍ɼาߦө૾ղ. ͍ͬͨଟ͘ͷํ๏͕ଘࡏ͢Δ͕ɼͦͷதͰ΋ਓͷา͖ํͰ. ੳͷ݁Ռͷઆ໌ͳͲͷɼऩूऀͷෛ୲͸ղܾͰ͖ͳ͍ɽͦ. ೝূΛߦ͏ɼา༰ೝূ͕ؔ৺ΛूΊ͍ͯΔɽา༰ೝূ͸ɼ. ͷͨΊɼ௕‫ͱ͜͏ߦؒظ‬͸೉͍͠ɽ. ਓ෺͕ηϯαʔ͔Βԕ͘཭Ε͍ͯͯ΋ೝূΛߦ͏͜ͱ͕Ͱ. ଞͷσʔλऩू๏ͱͯ͠ɼ֗த౳ʹΧϝϥΛઃஔ͠ɼࡱ. ͖ɼ๷൜Χϝϥ౳Λ༻͍ͨ൜ࡑ૞ࠪͰͷԠ༻͕‫ظ‬଴͞Εͯ. ӨΛߦ͍ଓ͚Δͱ͍͏ํ๏͕͋Δɽ͔͜͠͠ͷํ๏Ͱ͸ɼ. ͍Δɽา༰ೝূͷ‫ڀݚ‬ͷൃలʹ͸าߦσʔλϕʔε͕ඞਢ. ඃ‫ͱ͝ऀݧ‬ͷηάϝϯςʔγϣϯ΍ ID ͷׂΓ౰ͯ౳Λߦ. Ͱ͋Γɼେ‫ن‬໛͔ͭ‫؍‬ଌํ޲ͳͲͷ༷ʑͳཁҼΛߟྀͨ͠. ͏ඞཁ͕͋Δɽ͞Βʹɼඃ‫ऀݧ‬͸ҰൠͷਓͰ͋ΓɼΠϯ. σʔλϕʔεͷߏங͕๬·ΕΔɽ. ϑΥʔϜυίϯηϯτΛಘ͍ͯͳ͍ͨΊɼྙཧతɼ๏తɼ. ͔͠͠ɼาߦͷσʔλϕʔεߏங͸༰қͳ࡞‫Ͱۀ‬͸ͳ͍ɽ าߦσʔλ͸าߦө૾ΛࡱӨ͢Δඞཁ͕͋ΔͨΊɼը૾Λ ༻͍ΔଞͷόΠΦϝτϦΫεͷσʔλऩू (ࢦ໲΍೒࠼ɼ. ࣾձతॾ໰୊ (ELSI, ethical, legal, and social issues) ͱ͍ ͏‫͔఺؍‬Β໰୊ͱͳͬͯ͠·͏ɽ ͜ΕΒͷ໰୊఺Λߟ্ྀͨ͠Ͱେ‫ن‬໛ͳσʔλऩूΛߦ. ‫إ‬౳) ͱൺֱͯ͠ɼऩूऀͱඃ‫͕ऀݧ‬௕‫ݧ࣮ؒظ‬Λߦ͍ଓ. ͏ํ๏ͱͯ͠ɼզʑ͸าߦө૾ղੳΛ༻͍ͨମ‫ܕݧ‬ͷ௕‫ظ‬. ͚Δඞཁ͕͋Δɽ. ؒͷσϞͱͱ΋ʹࣗಈาߦσʔλऩूγεςϜΛఏҊ͢. ࣮ࡍʹ OU-ISIR Treadmill dataset [6] ΛྫʹͱΔͱɼ· ͣΞϧόΠτ΍ϘϥϯςΟΞͷඃ‫ऀݧ‬ΛืΓɼσʔλऩू. Δɽຊ࿦จͷߩ‫ݙ‬͸ҎԼͷ 3 ͭͰ͋Δɽ. 1. ࠷ઌ୺ͷาߦө૾ղੳʹΑΔΦϯϥΠϯσϞ. ͷ໨తΛઆ໌্ͨ͠ͰΠϯϑΥʔϜυίϯηϯτΛಘΔඞ. ଟ͘ͷਓʹମ‫ͯ͠ݧ‬΋Β͏ͨΊɼັྗతͳσϞͰ͋Δͱ. ཁ͕͋Δɽͦͯ͠ɼτϨουϛϧ্Ͱา͘࿅शΛߦ্ͬͨ. ͍͏͜ͱ͕ॏཁͰ͋ΔɽͦͷͨΊɼओʹ (1) าߦ଎౓ɼา. Ͱɼਤ 1 ͷΑ͏ʹτϨουϛϧɼσʔλࡱӨɼ‫ٸۓ‬ఀࢭϘ. ෯ɼ࿹ͷৼΓɼಈ͖ͷରশੑͳͲͷ‫ݸ‬ੑ‫ܭ‬ଌɼ(2) าߦ೥. λϯΛͦΕͧΕૢ࡞͢Δ‫͕ऀڀݚ‬ඞཁͱͳΔɽͦͷͨΊɼ. ྸਪఆͷ 2 ͭͷཁૉ͔ΒͳΔ࠷ઌ୺ͷาߦө૾ղੳΛ༻͍. େ‫ن‬໛ͳσʔλऩूʹ͸ద͍ͯ͠ͳ͍ɽ. ͨΦϯϥΠϯσϞγεςϜΛ։ൃͨ͠ɽମ‫ऀݧ‬͸ࣗ਎ͷ݁. ͦ͜Ͱɼଟ͘ͷਓ͕ू·Δ৔ॴͰɼาߦө૾ղੳΛ༻͍ Δମ‫ܕݧ‬ͷσϞΛߦ͏͜ͱͰาߦσʔλऩू͕༰қͱͳΔɽ ࣮ࡍʹ 4,000 ਓҎ্ͷาߦσʔλΛ‫ؚ‬Ήੈք࠷େͷาߦ. ՌΛҹ࡮෺ͱͯ࣋ͪ͠‫ؼ‬Δ͜ͱ͕Ͱ͖ɼՈ଒΍༑ਓͱ݁Ռ Λൺֱ͢Δ͜ͱ΋Ͱ͖Δɽ. 2. ࣗಈาߦσʔλऩूγεςϜ ମ‫͕ࣗऀݧ‬෼ࣗ਎ͰσϞΛָ͠Έɼࣗಈతʹσʔλऩू. 1. େࡕେֶ Osaka University, Ibaraki-shi, Osaka, 567-0047, Japan. c 2016 Information Processing Society of Japan. ͕Ͱ͖ΔΑ͏ʹ͢ΔͨΊɼඃ‫ ऀݧ‬ID Λ۠ผ͢ΔͨΊͷ QR. 1.

(2) Vol.2016-CVIM-200 No.18 2016/1/22. ৘ใॲཧֶձ‫ڀݚ‬ใࠂ IPSJ SIG Technical Report. ද 1 ͔Β෼͔ΔΑ͏ʹɼඃ‫࠷͕਺ऀݧ‬େͰ͋Δσʔλ ϕʔε͸ OU-ISIR LP [5] ͷ 4,007 ਓɼ‫؍‬ଌํ޲͕࠷େͰ ͋Δσʔλϕʔε͸ OU-ISIR Treadmill C [15] ͷ 25 ํ޲ Ͱ͋Γɼେ‫ن‬໛͔ͭଟํ޲͔Β‫؍‬ଌ͞Ε͍ͯΔσʔλϕʔ ε͸ଘࡏ͠ͳ͍ɽ. 3. ࣗಈาߦ‫ܭ‬ଌγεςϜ ਤ 1. OU-ISIR Treadmill dataset[ [6]] ͷσʔλऩूͷ༷ࢠɽσʔ. 3.1 ֓ཁ γεςϜͷ֓ཁΛਤ 2 ʹࣔ͢ɽମ‫ऀݧ‬͸·ͣೖ‫ޱ‬ͷύ. λऩूʹ͸ 3 ਓͷૢ࡞ऀ͕ඞཁͱͳΔɽ. ωϧͰσʔλऩूͷ໨త౳ʹؔ͢Δઆ໌Λ֬ೝ͢Δɽ࣍ʹ ίʔυɼମ‫ऀݧ‬ͷಈ͖Λ൑அ͢ΔͨΊͷޫిηϯαʔɼମ. QR ίʔυൃ݊‫( ػ‬ਤ 3 (a)) ͰΑΓৄࡉͳઆ໌Λ֬ೝ͠ɼମ. ‫ऀݧ‬ͷߦಈΛ༠ಋ͢ΔͨΊͷࣗಈԻ੠Λ༻͍ͨࣗಈาߦ. ‫ऀݧ‬ͷ೥ྸͱੑผΛೖྗ͢Δ͜ͱͰ QR ίʔυ͕ൃ݊͞Ε. σʔλऩूγεςϜΛ։ൃͨ͠ɽ. Δɽͦͯ͠า༰‫ݸ‬ੑ‫ܭ‬ଌͷΤϦΞ΁ͱਐΈɼൃ݊ͨ͠ QR. 3. ΠϯϑΥʔϜυίϯηϯτΛಘͨσʔλऩू. ίʔυΛ QR ίʔυϦʔμʔʹ͔͟͢͜ͱͰσϞ͕։࢝͢. ऩूͨ͠σʔλΛ‫ڀݚ‬໨తͰ࢖༻͢ΔͨΊɼΠϯϑΥʔ. ΔɽຊγεςϜͰ͸ɼঢ়‫ʹگ‬Ԡͯ͡ྲྀΕΔࣗಈԻ੠ʹΑͬ. ϜυίϯηϯτΛಘΔ͜ͱ͕Ͱ͖ΔγεςϜͱͳ͓ͬͯ. ͯମ‫ऀݧ‬ͷ༠ಋΛߦ͏ɽ۩ମతʹ͸ɼ௨աͨ͠ޫిηϯ. ΓɼELSI ͷ‫ݖ‬ҖͰ͋Δห‫࢜ޢ‬ͷ‫؂‬मͷ΋ͱγεςϜ։ൃΛ. αʔʹΑͬͯҟͳΔԻ੠͕ྲྀΕɼମ‫ऀݧ‬͸ͦͷԻ੠ʹैͬ. ߦͬͨɽͦͷͨΊɼऩूͨ͠าߦσʔλϕʔε͸ެ։Մೳ. ͯߦಈ͢Δɽาߦ࿏΁ਐೖ‫ޙ‬ɼՙ෺Λ࣋ͬͨ··‫ٯ‬ଆ·Ͱ. Ͱ͋Γɼࠓ‫ޙ‬ͷา༰ೝূͷൃలʹ໾ཱͭ͜ͱ͕‫ظ‬଴͞ΕΔɽ. 2. ؔ࿈‫ڀݚ‬. ਐΈɼՙ෺ஔ͖ʹՙ෺Λஔ͘ɽͦͯ͠͞Βʹ 2 ԟ෮͠ɼՙ ෺Λ࣋ͬͯาߦ࿏͔Βୀग़͢Δɽͦͷ‫ޙ‬ɼࡱӨ͞Εͨ݁Ռ ͱಉҙจΛ֬ೝ͠ɼσʔλอଘʹಉҙ͢Δ͔Ͳ͏͔ΛϘλ. ‫إ‬΍ࢦ໲ͳͲͷଞͷόΠΦϝτϦΫεͷධՁʹ࢖ΘΕͯ. ϯͰ൑அ͢Δ (ਤ 3 (c))ɽ·ͨɼผͷσΟεϓϨΠʹ࠷৽. ͍Δσʔλϕʔεͱൺֱͯ͠ɼ‫ެࡏݱ‬։͞Ε͍ͯΔา༰. ͷ 6 ਓͷ݁Ռ͕දࣔ͞Ε͓ͯΓɼ‫ݸ‬ੑ‫ܭ‬ଌͷ݁ՌͷൺֱΛ. σʔλϕʔε͸ඃ‫͕਺ऀݧ‬গͳ͍ɽ‫إ‬σʔλϕʔεͷඃ‫ݧ‬. ߦ͏͜ͱ͕Ͱ͖Δ (ਤ 3 (d))ɽ. ऀ਺͸ 30,000 ਓҎ্Ͱ͋Γɼࢦ໲͸໿ 1,600 ສਓͰ͋Δɽ ද 1 ʹ୅දతา༰σʔλϕʔεͷৄࡉΛࣔ͢ɽ. 3.2 σϞϯετϨʔγϣϯ. า༰σʔλϕʔεߏஙʹ͸ɼඃ‫ऀݧ‬ͷଟ༷ੑɼඃ‫ͱ͝ऀݧ‬. ຊσϞ͸ɼา༰‫ݸ‬ੑ‫ܭ‬ଌͱา༰Λ༻͍ͨ೥ྸਪఆͷ 2 ͭ. ͷาߦঢ়‫( گ‬෰૷΍าߦ଎౓ɼՙ෺Λ͍࣋ͬͯΔ͔Ͳ͏͔ͳ. Λߦ͏ɽͦΕΒΛߦ͏ࡍɼࠨ͔Βӈ΁ͷ໿ 2.5 ඵͷา༰ը. Ͳ)ɼࡱӨঢ়‫؍( گ‬ଌํ޲΍৚݅มԽͳͲ) ͳͲΛߟྀ͢Δඞ. ૾ྻ *1 Λ༻͍ΔɽࡱӨ͸ USB Χϝϥ (PointGrey, FMVU-. ཁ͕͋Δɽྫ͑͹ɼඃ‫ऀݧ‬ͷา͘৔ॴ΍র໌৚݅ɼःณ͕ͳ. 13S2C-CS) Λ༻͍ͯɼ640×480 ըૉ͔ͭ 30fps Ͱߦ͏ɽา. ͍Α͏ʹͳͲɼίϯτϩʔϧ͞Εͨঢ়‫گ‬Λ૝ఆͨ͠৔߹ࡱ. ༰γϧΤοτΛ༻͍ͨ෼ੳΛߦ͏ͨΊɼγϧΤοτநग़͸. Өঢ়‫گ‬͸௨ৗ‫ݻ‬ఆͰ͋Γɼࣨ಺ͳͲͷೝূ͠΍͍͢‫؀ڥ؀‬. ॏཁͳཁૉͰ͋ΔɽͦͷͨΊɼ(1) ؔ৺ྖҬ (region of in-. ‫͋Ͱڥ‬Δ͜ͱ͕๬·͍͠ɽඃ‫͕਺ऀݧ‬ଟ͍͜ͱ΋ॏཁͰ͋. terest, ROI) ͷઃఆɼ(2) ըૉ୯ҐͷΨ΢γΞϯഎ‫ܠ‬ϞσϦ. ΔͨΊɼίϯτϩʔϧ͞Εͨঢ়‫گ‬Λ૝ఆͨ͠৔߹ɼSOTON. ϯάɼ(3) ৭৘ใΛ༻͍ͨըૉ୯ҐͷӨআ‫( ڈ‬ਤ 4 (a)(b)*2 ). Multimodal [7]ɼOU-ISIR LP [5]ɼTUM-GAID [8] ͳͲ͕. Λߦ͏ɽROI ΍ᮢ஋ઃఆ͸ GUI ্ͰεΫϩʔϧόʔͳͲ. ద͍ͯ͠Δͱ͍͑Δɽ. Λ༻͍ͯ༰қʹઃఆՄೳͰ͋Γɼॳ‫ظ‬ઃఆ࣌ͷΈσϞͷ৔. ·ͨɼ൜ࡑ૞ࠪΛ૝ఆͨ͠ΞϧΰϦζϜ։ൃͷ৔߹ɼର ৅ө૾ͷࡱӨঢ়‫گ‬͸௨ৗ‫ݻ‬ఆ͞Ε͓ͯΒͣɼඃ‫ऀݧ‬͸‫ྗڠ‬. ॴʹԠͯ͡ઃఆΛߦ͏ɽ ‫ݸ‬ੑ‫ܭ‬ଌ. తͰ͸ͳ͍͜ͱ͕૝ఆ͞ΕΔɽͦͷͨΊɼඃ‫ऀݧ‬ͷଟ༷ੑ. า༰‫ݸ‬ੑ‫ܭ‬ଌʹ͍ͭͯઆ໌͢Δ (ৄࡉ͸ [26] Λࢀর͞Ε. ͱಉ༷ʹาߦঢ়‫گ‬΍ࡱӨঢ়‫گ‬΋ॏཁͳ఺Ͱ͋ΓɼSOTON. ͍ͨ)ɽา༰ͷಛ௃ͱͯ͠ɼาߦ଎౓ɼาߦप‫ظ‬ɼา෯ɼ଍. Large Database [9] ΍ USF HumanID [10] ͳͲ͕ద͍ͯ͠. ͷӡͼͷࠨӈରশੑɼ࿹ͷલৼΓɼ࿹ͷ‫ޙ‬ΖৼΓɼ࿹ͷৼ. Δͱ͍͑Δɽ͔͠͠ɼSOTON Large Database [9] ΍ USF. Γͷࠨӈରশੑɼഎ‫ے‬ͷ৳ͼͷ 8 ͭΛ࢖༻͢Δɽ·ͣϞϧ. HumanID [10] ͸ͦΕͧΕͷཁૉͷछྨ͕গͳ͘ɼͦΕͧ. ϑΥϩδʔॲཧΛߦ͍ɼ࠷େྖҬͷϑΟϧλϦϯάΛߦ͏. Εͷཁૉʹॏࢹͨ͠σʔλϕʔε (‫؍‬ଌํ޲ʹಛԽͨ͠. ͜ͱͰγϧΤοτը૾ྻɼͦͯ͠ମ‫ऀݧ‬ͷۣ‫ྖܗ‬ҬΛऔಘ. CASIA B [11]ɼาߦ଎౓ʹಛԽͨ͠ OU-ISIR Treadmill. ͢ΔɽΧϝϥΩϟϦϒϨʔγϣϯΛࣄલʹߦ͍ɼ଍Լ఺Λ. A [12]ɼ෰૷มԽʹಛԽͨ͠ OU-ISIR Treadmill B [13] ΍. ‫ج‬४ͱ͢Δ͜ͱͰɼࡱӨ։࢝࣌ͱऴྃ࣌ͷ଍Լ఺ͷ‫͔཭ڑ‬. ࡱӨঢ়‫گ‬ͷछྨʹಛԽͨ͠ WOSG [14]) ͕ΑΓద͍ͯ͠Δ. *1. ͱ͍͑Δɽ. *2. c 2016 Information Processing Society of Japan. ࡱӨ։࢝ͷλΠϛϯά͸ޫిηϯαʔΛ༻͍ͯ൑அ͢Δ (‫ޙ‬ड़)ɽ จ‫[ ݙ‬25] ͷํ๏ͷ؆қ൛Λ࣮૷.. 2.

(3) Vol.2016-CVIM-200 No.18 2016/1/22. ৘ใॲཧֶձ‫ڀݚ‬ใࠂ IPSJ SIG Technical Report. ໊લ. ද 1 ༗໊ͳา༰σʔλϕʔεͷৄࡉ ඃ‫ ਺ऀݧ‬γʔέϯε਺ ৚݅มԽ. CMU MoBo( [16]). 25. Georgia Tech( [17]) HID-UMD( [18]). ‫؍‬ଌํ޲. ࣨ಺ (I)/ࣨ֎ (O). 600. ͋Γ. 6. I (treadmill). 15. 268. ͋Γ. -. O. 18. 20. ͋Γ. -. -. 25. 100. ͳ͠. 1. O. 55. 222. ͋Γ. 2. O. SOTON Small Database( [19]). 12. -. ͋Γ. 3. I. SOTON Large Database( [9]). 115. 2128. ͋Γ. 2. I/O. SOTON Multimodal( [7]). >300. >5000. ͋Γ. 12. I. SOTON Temporal( [20]). 25. 2280. ͋Γ. 12. I. USF HumanID( [10]). 122. 1870. ͋Γ. 2. O. CASIA A( [21]). 20. 240. ͋Γ. 3. I. CASIA B( [11]). 124. 1240. ͋Γ. 11. I. CASIA C( [22]). 153. 1530. ͋Γ. 1. O. OU-ISIR, Treadmill A( [12]). 34. 612. ͋Γ. 1. I (treadmill). OU-ISIR, Treadmill B( [13]). 68. 2764. ͋Γ. 1. I (treadmill). OU-ISIR, Treadmill C( [15]). 200. 200. ͋Γ. 25. I (treadmill). OU-ISIR, Treadmill D( [23]). 185. 370. ͳ͠. 1. I (treadmill). OU-ISIR, LP( [5]). 4,007. 7842. ͳ͠. 2. I. TUM-IITKGP( [24]). 35. 850. ͋Γ. 1. O. TUM-GAID( [8]). 305. 3370. ͋Γ. 1. O. WOSG( [14]). 155. 684. ͋Γ. 8. O. ਤ 2 γεςϜͷ֓ཁ. Βาߦ‫཭ڑ‬Λ‫͢ࢉܭ‬Δ͜ͱ͕Ͱ͖ΔɽࡱӨ࣌ؒΛ 2.5 ඵͱ. ͧΕ࠷΋઀ۙ͢Δ୯‫࣋ࢧ٭‬૬ (ਤ 4 (d)) ͱ࠷΋཭ΕΔ྆‫٭‬. ઃఆ͍ͯ͠Δ͜ͱ͔Β؆୯ʹาߦ଎౓Λ‫͢ࢉܭ‬Δ͜ͱ͕Ͱ. ࢧ࣋૬ (ਤ 4 (e)) Ͱ͋Γɼ྆‫࣋ࢧ٭‬૬ؒͷҠಈ‫ܦͱ཭ڑ‬ա. ͖Δɽ. ͔࣌ؒΒҰาͷา෯Λ‫͢ࢉܭ‬Δɽશͯͷฏ‫ۉ‬ΛऔΔ͜ͱͰ. ࣍ʹɼαΠζͷਖ਼‫ن‬ԽΛߦͬͨγϧΤοτը૾ྻ (ਤ 4. (c)) Λ࡞੒͠ɼ࣌ؒ࣠Ͱਖ਼‫ن‬Խࣗ‫ݾ‬૬ؔΛ࠷େʹ͢Δ͜ͱ ( [27]) Ͱาߦप‫ظ‬Λ‫͢ࢉܭ‬Δɽ·ͨɼาߦ͸࿹ͱ‫͕ͦ٭‬Ε. c 2016 Information Processing Society of Japan. ͦͷମ‫ऀݧ‬ͷา෯ͱ͠ɼ෼ࢄΛ‫͢ࢉܭ‬Δ͜ͱͰ଍ͷӡͼͷ ࠨӈରশੑͱ͢Δɽ ͞Βʹɼ୯‫࣋ࢧ٭‬૬ʹ͓͍ͯମͷલ໘ɼഎ໘ͷ‫ڥ‬քΛ‫ܭ‬. 3.

(4) Vol.2016-CVIM-200 No.18 2016/1/22. ৘ใॲཧֶձ‫ڀݚ‬ใࠂ IPSJ SIG Technical Report. ͯ͠༻͍ΔɽGEI Λಛ௃ϕΫτϧ x ∈ RM ͱͯ͠ల։͢ Δɽ͜͜Ͱ M ͸ಛ௃ϕΫτϧͷ࣍‫ݩ‬ɼͭ·Γ GEI ͷը ૾αΠζͰ͋Δɽͦͯ͠Ψ΢εաఔճ‫( ؼ‬GPR, gaussian. process regression) ʹΑΓ೥ྸਪఆΛߦ͏ɽ ֶशσʔλͱͯ͠ D = [X, y ] Λߟ͑Δɽ͜͜Ͱ X = (a) QR ίʔυൃ݊‫ػ‬. (b) QR ίʔυϦʔμʔ. [x 1 , · · · , x N ] ͸ N ਓͷา༰ಛ௃Ͱ͋Γɼy = [y1 , · · · , yN ] ͸ͦΕͧΕʹରԠ͢Δ࣮೥ྸͰ͋Δɽͦͯ͠ɼ2 ਓͷา༰ ಛ௃ x i ͱ x j ͷྨࣅੑ͸ RBF(radical basis function) Χʔ ωϧͰҎԼͷΑ͏ʹఆٛ͢Δɽ. ||x i − x j || , k(x i , x j ; r) = exp − 2r2 (d) ࠷৽ͷ 6 ਓͷ݁Ռ. (c) σΟεϓϨΠ & ϓϦϯλʔ. (1). ͜͜Ͱ || · || ͸ L2 ϊϧϜΛද͠ɼr ͸ RBF Χʔωϧͷϋ ΠύʔύϥϝʔλͰ͋Δɽ. ਤ 3 లࣔʹ͓͚Δॏཁͳཁૉ. ೖྗͷา༰ಛ௃ x ∗ ʹର͠ɼਪఆ೥ྸ y∗ ͷࣄ‫཰֬ޙ‬෼෍ Λ GPR ʹΑͬͯ‫ٻ‬ΊΔɽࣄ‫཰֬ޙ‬෼෍ P (y∗ |x ∗ , D) ͸Ψ ΢ε෼෍ N (y∗ ; µy , σy2 ) ʹΑΓఆٛ͞Εɼฏ‫ ۉ‬µy ͱ෼ࢄ σy2 ͸ҎԼͷΑ͏ʹఆٛ͞ΕΔɽ. µy = k T∗ (K + S)−1 y σy2 (a). (b). = k(x ∗ , x ∗ ) −. k T∗ (K. (2) + S). −1. 2. k∗ + σ ,. (3). ͜͜ͰɼK ͸ (i, j) ཁૉ͕ k(x i , x j ) Ͱ͋Δ N × N ͷਖ਼ํ ߦྻͰ͋Γɼk ∗ ͸ i ߦ͕ k(x i , x ∗ ) Ͱ͋Δ N ࣍‫ݩ‬ͷϕΫτ ϧɼS ͸ (i, i) ཁૉ͕ σ 2 Ͱ͋Δ N × N ͷର֯ߦྻͰ͋Δɽ. (c). σ 2 ͸೥ྸͷ‫؍‬ଌ‫͋Ͱࠩޡ‬Γɼ࣮‫ݧ‬ઃఆͱͯ͠ σ 2 = 0.25 ͱ ͍ͯ͠Δɽ. (d). (e). (f). (g). ࣜ (2)(3) ʹ͓͍ͯɼ࠷΋͕͔͔࣌ؒΔ‫ࢉܭ‬͸ N × N ߦ. ਤ 4 า༰‫ݸ‬ੑ‫ܭ‬ଌͷྲྀΕΛࣔ͢ɽROI(੺ઢ) Λࢦఆͨ͠‫ݪ‬ը૾ (a). ྻͷ‫ྻߦٯ‬ɼͭ·Γ (K + S)−1 ͷ‫͋Ͱࢉܭ‬ΔɽO(N 3 ) ͷ. ͔Β (b) ͷΑ͏ʹγϧΤοτ (ന) ͱӨ (փ৭) Λநग़͢ΔɽӨ. ͕͔͔࣌ؒΓɼΦϯϥΠϯͷ‫ܭ‬ଌͷো֐ͱͳΔɽͦͷͨ. আ‫ޙڈ‬ɼը૾αΠζͷਖ਼‫ن‬ԽΛߦͬͨา༰ը૾ྻΛ (c) ʹࣔ. Ίɼಈత Active Set Λ༻͍ͨ GPR Λ༻͍ɽೖྗϕΫτϧ. ͢ɽ͜ͷา༰ը૾ྻ͸ 3 ຕ͓͖ʹબ୒ͨ͠΋ͷͰ͋Δɽ·ͨɼ. x ∗ ͷۙ๣ͷֶशσʔλ͚ͩΛ༻͍ΔɽೖྗϕΫτϧ x ∗ ͷ. (d)ɼ(e) ͷΑ͏ʹ୯‫࣋ࢧ٭‬૬΍྆‫࣋ࢧ٭‬૬͸શͯͷମ‫ʹऀݧ‬ ͍ͭͯଘࡏ͢Δɽ࿹ͷৼΓʹ͍ͭͯɼ(f) ʹࣔ͢Α͏ʹɼ୯‫٭‬ ࢧ࣋૬ʹ͓͍ͯ‫ܭ‬ଌ͢Δ (྘͕લํ޲ɼ੺͕‫ޙ‬Ζํ޲)ɽ·ͨɼ എ‫ے‬ͷ৳ͼʹ͍ͭͯɼ಄͔Β্൒਎ͷϥΠϯ͸ (g) ͷ੺͍ઢͷ Α͏ʹ‫ܭ‬ଌ͢Δɽ. ࢉ͠ɼͦͷ‫ڥ‬քΑΓલํ޲ͷ࿹ৼΓͱ‫ޙ‬Ζํ޲ͷ࿹ৼΓΛ ͦΕͧΕ࿹ͷલৼΓɼ࿹ͷ‫ޙ‬ΖৼΓ (ਤ 4 (f)) ͱ͢Δɽͦ ͯ͠ɼ୯‫࣋ࢧ٭‬૬ؒͷ࿹ৼΓͷ෼ࢄΛ‫͢ࢉܭ‬Δ͜ͱͰ࿹ͷ ৼΓͷࠨӈରশੑͱ͢Δɽ࠷‫ʹޙ‬ɼ಄͔Β্൒਎ͷϥΠϯ ͔Βഎ‫ے‬ͷ৳ͼ (ਤ 4 (g)) Λ‫͢ࢉܭ‬Δɽ ೥ྸਪఆ ࣍ʹ೥ྸਪఆʹ͍ͭͯઆ໌͢Δɽৄࡉ͸ [28] Λࢀর͞Ε. K(≪ N ) ۙ๣Λ࢖༻͢Δ৔߹ɼ‫ྻߦٯ‬ͷ‫ࢉܭ‬͸ O(K 3 ) ͷ ࣌ؒͱͳΓɼO(N 3 ) ͱൺֱͯ͠େ෯ͳ࣌ؒͷ୹ॖͱͳΔɽ ͦͷҰํͰɼK Λখ͗͘͢͞͠Δͱ೥ྸਪఆͷਫ਼౓͕ ௿Լ͢ΔͨΊɼਫ਼౓ͱ‫ؒ࣌ࢉܭ‬ͷτϨʔυΦϑΛߟྀ͠ ্ͨͰ࠷΋ద੾ͳ K Λઃఆ͢Δඞཁ͕͋ΔɽͦͷͨΊɼ. OU-ISIR Large Population dataset, Camera 1, Version 2 (OULP-C1V2) [5] ͷҰ෦Λ༻͍࣮ͯ‫ݧ‬Λߦͬͨɽ͜ͷσʔ λϕʔε͸ N = 1, 678 ͷֶशσʔλͱ 2,257 ͷςετσʔ λ͔Β੒Δɽͦͷ݁ՌɼK = 10 ͕࠷దͱ൑அ͞Εͨɽ͞ ΒʹɼRBF Χʔωϧͷύϥϝʔλ r ͸બ୒͞Εͨۙ๣͔ ΒҎԼͷΑ͏ʹఆٛ͞ΕΔɽ. ͍ͨɽ ·ͣɼαΠζͷਖ਼‫ن‬ԽΛߦͬͨγϧΤοτը૾ྻΛ̍प ‫Ͱظ‬ฏ‫ۉ‬Խ͢Δ͜ͱʹΑΓɼฏ‫ۉ‬γϧΤοτ [29] Ͱ͋Δ. GEI(gait energy image) [30] Λ࡞੒͠ɼGEI Λา༰ಛ௃ͱ. c 2016 Information Processing Society of Japan. r=. K 1 X ||x kN N ID(k) − x ∗ ||, K. (4). k=1. ͜͜ͰɼkN N ID(k) ͸ೖྗ x ∗ ͷ k ൪໨ͷۙ๣Λද͢ɽ. 4.

(5) Vol.2016-CVIM-200 No.18 2016/1/22. ৘ใॲཧֶձ‫ڀݚ‬ใࠂ IPSJ SIG Technical Report. ਤ 6. (a) ҹ࡮͞ΕΔମ‫ऀݧ‬ͷ݁Ռ. ΧϝϥηοςΟϯά. (ਤ 2 ͱ 6 ͷΧϝϥ 1 ͔ΒΧϝϥ 7)ɽ·ͨ‫ؼ‬ΓͰ͸ɼӈͷ ԣ͔Β‫ޙ‬Ζํ޲ͷΧϝϥ 6 ୆͓Αͼલํ޲ͷΧϝϥ 1 ୆͔ ΒࡱӨ͞ΕΔ (ਤ 6 ͷΧϝϥ 1’ ͔ΒΧϝϥ 7’)ɽ ͦͷ݁Ռɼՙ෺Λ͍࣋ͬͯΔঢ়ଶͰ 7 ํ޲͔ΒࡱӨ͞Εɼ ඪ४าߦͰ߹‫ ܭ‬14 ํ޲͔ΒࡱӨ͞ΕΔɽͦͷͨΊɼଟํ ޲ͷา༰ೝূ΍ํ޲͕ҟͳΔ৔߹ͷา༰ೝূɼՙ෺Λ࣋ͬ ͍ͯΔ৔߹ͱ͍࣋ͬͯͳ͍৔߹ͷา༰ೝূͳͲ͕Մೳͱͳ Δɽ·ͨɼੑผਪఆ౳ʹ΋࢖༻ՄೳͰ͋Δɽ ωοτϫʔΫΧϝϥ͸։‫ؗ‬தઈ͑ͣาߦ࿏ΛࡱӨ͍ͯ͠ (b) ࠷৽ͷ 6 ਓͷ݁Ռൺֱ ਤ 5. ମ‫΁ऀݧ‬ͷ݁Ռදࣔɽ(a) Ͱ͸ɼϨʔμʔνϟʔτͰ‫ݸ‬ੑ‫ܭ‬ଌ ͷ݁Ռ͕දࣔ͞Ε͓ͯΓɼ8 ֯‫ܗ‬ͷ಺ଆɼਅΜதɼ֎ଆ͸ͦΕ. ΔͨΊɼମ‫ऀݧ‬ͷ ID ͝ͱʹө૾ͷ੾Γग़͠Λߦ͏ඞཁ͕ ͋ΔɽͦͷͨΊɼೖ‫ޱ‬ͷ QR ίʔυͰ ID Λऔಘ͠ɼޫి. ͧΕ࠷খ஋ɼฏ‫ۉ‬஋ɼ࠷େ஋Λද͢ɽ(b) Ͱ͸ɼମ‫ʹऀݧ‬Αͬ. ηϯαʔΛ௨աͨ࣌ؒ͠΋ϩάͱͯ͠อଘ͍ͯ͠Δɽޫి. ͯҟͳΔ৭Ͱ݁Ռ͕දࣔ͞Ε͍ͯΔɽ. ηϯαʔ͸ɼਤ 2 ͷΑ͏ʹ (1) ೖ‫ޱ‬ɼ(2) ߦ͖ͷελʔτ ϥΠϯɼ(3) ‫ؼ‬ΓͷελʔτϥΠϯɼ(4) ग़‫ޱ‬ͷ 4 Օॴʹઃ. ମ‫΁ऀݧ‬ͷ݁Ռදࣔ. ஔ͓ͯ͠ΓɼࢀՃऀ͕Կԟ෮໨ͷาߦΛߦ͍ͬͯΔͷ͔Λ. า༰‫ݸ‬ੑ‫ܭ‬ଌͱਪఆ೥ྸͷ݁Ռ͸ 2 ௨Γͷํ๏Ͱମ‫ऀݧ‬. ൑ఆ͢Δ͜ͱʹ࢖༻͢Δɽ2 ԟ෮൒͢Δ͜ͱ͔Βɼ߹‫ ܭ‬5. ʹදࣔ͞ΕΔɽ1 ͭ͸ɼମ‫ؼ͕ͪ࣋ऀݧ‬Δ͜ͱ͕Ͱ͖Δɼ. ճ෼ͷาߦΛࡱӨՄೳͰ͋ΔɽωοτϫʔΫΧϝϥ͸εέ. ҹ࡮͞ΕΔ݁ՌͰ͋Δ (ਤ 5 (a))ɽࡱӨ͞Εͨ‫ݪ‬ը૾΍γ. δϡʔϥʹΑͬͯ։‫ࣗʹؒ࣌ؗ‬ಈతʹࡱӨΛ։࢝͠ɼด‫ؗ‬. ϧΤοτը૾ʹՃ͑ͯ‫ݸ‬ੑ‫ܭ‬ଌ΍ਪఆ೥ྸͷ݁Ռ͕දࣔ͞. ࣌ؒʹࣗಈతʹࡱӨΛऴྃ͢ΔΑ͏ʹઃఆ͞Ε͍ͯΔɽ. Ε͍ͯΔɽ΋͏ 1 ͭͷํ๏ͱͯ͠ɼ࠷৽ͷ 6 ਓͷ݁Ռൺֱ ͷը໘Ͱ͋Δ (ਤ 5 (b))ɽମ‫ʹऀݧ‬Ԡͯ͡ҟͳΔ৭Ͱ݁Ռ ͕දࣔ͞Ε͓ͯΓɼ6 ਓͷ‫ݸ‬ੑ‫ܭ‬ଌͷ݁ՌͱࡱӨ͞Εͨө ૾͕දࣔ͞Ε͍ͯΔɽ. 3.4 ΠϯϑΥʔϜυίϯηϯτ զʑͷ‫ڀݚ‬ͷͨΊʹาߦө૾Λऩू͢Δͱ͍͏໨తʹՃ ͑ɼาߦө૾Λ༻͍ͨา༰ೝূͷ‫ڀݚ‬ͷͨΊɼσʔλϕʔ εΛ࡞੒ͯ͠ެ։͢Δͱ͍͏໨త͕͋ΔͨΊɼࡱӨͨ͠. 3.3 าߦσʔλऩू. σʔλΛ‫͢༻࢖ʹڀݚ‬ΔࡍΠϯϑΥʔϜυίϯηϯτ͸ͱ. ମ‫ऀݧ‬ͷาߦσʔλ͸าߦ࿏ͰͷσϞମ‫ݧ‬தɼԁ‫ܗ‬ͷ 4. ͯ΋ॏཁͰ͋ΔɽͦͷͨΊɼ๷൜Χϝϥʹؔͯ͠ ELSI ͷ. ෼ͷ 1 ্ʹ 15 ౓ࠁΈʹઃஔ͞Εͨ 7 ୆ͷωοτϫʔΫΧϝ. ‫ݖ‬Җͷ 1 ਓͰ͋Δห‫ͱ࢜ޢ‬૬ஊ͠ɼσϞͷ໨త΍σʔλͷ. ϥ (AXIS Communications, Q1614) ͰࡱӨ͞ΕΔ (ਤ 2)ɽ. ར༻ʹؔͯ͠ͷઆ໌΍ΠϯϑΥʔϜυίϯηϯτͷऔΓํ. ԁͷத৺͸าߦ࿏Ͱ͋Γɼ൒‫ܘ‬͸໿ 8 ϝʔτϧɼΧϝϥͷ. Λܾఆͨ͠ɽ. ߴ͞͸໿ 5 ϝʔτϧͰ͋Δɽલํ޲ɾ‫ޙ‬Ζํ޲ΛࡱӨ͢Δ. ͞ΒʹɼΠϯϑΥʔϜυίϯηϯτʹؔ͢Δදࣔ͸ 3 Օ. Χϝϥ͸ɼձ৔ߏ଄্ͷཧ༝ʹΑΓը֯಺ʹःณ෺͕өΓ. ॴ͋Γɼ1 ͭ໨͸ೖ‫ޱ‬ͷύωϧͰ͋Δɽาߦө૾Λ༻͍ͨา. ࠐΜͰ͠·͏ͨΊɼΧϝϥ 7 ͷΈ‫ʹ޲ํٯ‬഑ஔͨ͠ɽ. ༰ೝূͱ͍͏σϞͷ໨త΍σʔλऩूʹؔ͢Δઆ໌͕ॻ͔. અ 3.1 Ͱड़΂ͨ௨Γɼମ‫ऀݧ‬͸ՙ෺ஔ͖·Ͱา͍ͯՙ෺. Ε͍ͯΔɽ2 ͭ໨͸ QR ίʔυൃ݊‫͋Ͱػ‬ΓɼλϒϨοτ. Λஔ͖ɼͦͷ‫ ޙ‬2 ԟ෮͢ΔɽΧϝϥͷࡱӨ֯౓͸ɼԟ෮ͷ. PC ʹಉ༷ͷઆ໌͕දࣔ͞Εɼಉҙͨ͠৔߹͚ͩ QR ίʔ. ߦ͖ͱ‫ؼ‬ΓͰ൓ରํ޲ͱͳΓɼߦ͖Ͱ͸ࠨͷԣ͔Βલํ޲. υ͕ൃ݊͞ΕΔɽ͞Βʹɼମ‫ڙࢠ͕ऀݧ‬ͷ৔߹ɼอ‫ऀޢ‬ͷ. ͷΧϝϥ 6 ୆͓Αͼ‫ޙ‬Ζํ޲ͷΧϝϥ 1 ୆͔ΒࡱӨ͞ΕΔ. ํ͕આ໌Λ֬ೝ͢Δ͜ͱͰಉҙ͢Δ͔Ͳ͏͔Λ൑அ͢Δɽ3. c 2016 Information Processing Society of Japan. 5.

(6) Vol.2016-CVIM-200 No.18 2016/1/22. ৘ใॲཧֶձ‫ڀݚ‬ใࠂ IPSJ SIG Technical Report. ͭ໨͸าߦ࿏ͷग़‫͋ʹޱ‬ΔσΟεϓϨΠͰ͋Δɽ͜ͷσΟ. ໭ΔΑ͏ʹઃఆ͍ͯ͠Δɽ. εϓϨΠͰ͸ɼ(1) ҹ࡮͞ΕΔ݁Ռɼ(2) 7 ୆ͷωοτϫʔ. ࠷‫ʹޙ‬ɼզʑ͸ମ‫ܕݧ‬ͷ௕‫ظ‬లࣔΛՊֶ‫ؗ‬ͷ‫ྗڠ‬ͷ΋ͱ. ΫΧϝϥͰࡱӨ͞ΕΔαϯϓϧը૾ɼ(3) σʔλͷར༻΍. ߦ͍ͬͯΔɽମ‫ݧ‬தʹ‫͜ى‬Δ໰୊ΛϦηοτϘλϯͳͲΛ. ‫ݸ‬ਓ৘ใอ‫ͮ͘جʹޢ‬σʔλ؅ཧํ๏͕දࣔ͞ΕΔɽ͜ͷ. ༻͍ͯղܾͨ͠ΓɼϓϦϯλʔͷτφʔަ‫׵‬΍ҹ࡮༻ࢴͷ. σΟεϓϨΠͷલʹ͸Ϙλϯ͕͋Γɼମ‫ऀݧ‬͸ಉҙ (੨Ϙλ. ௥ՃͳͲͷ࡞‫ۀ‬͸Պֶ‫ؗ‬ͷελοϑʹ͓‫͍͍ͯ͠ئ‬Δɽ. ϯ) ͱෆಉҙ (੺Ϙλϯ) ͷ൑அΛ͢Δ (ਤ 3(c))ɽಉҙϘλ ϯΛԡͨ͠৔߹͸݁Ռ͕ҹ࡮͞Ε (ਤ 5 (a)) ɼ݁Ռൺֱͷ σΟεϓϨΠʹදࣔ͞ΕΔ (ਤ 5 (b))ɽෆಉҙϘλϯΛԡ ͨ͠৔߹ɼ݁Ռ͸ҹ࡮͞Εͣɼ݁Ռൺֱʹ΋දࣔ͞Εͳ͍ɽ. 4. ݁Ռͱߟ࡯ 4.1 ‫ؒ࣌ࢉܭ‬ ମ‫ܕݧ‬ͷσϞΛߦ͏্ͰɼΦϯϥΠϯͰ݁Ռ͕දࣔ͞Ε Δ͜ͱ͸ॏཁͰ͋Γɼ‫ݸ‬ੑ‫ܭ‬ଌͱ೥ྸਪఆͷ 2 ఺ʹؔͯ͠. 3.5 γεςϜͷ࣮૷. ‫ؒ࣌ࢉܭ‬ͷධՁΛߦͬͨɽ10 ճͷฏ‫ۉ‬Λ‫͜ͱͨ͠ࢉܭ‬Ζɼ. ࣗಈͰγεςϜΛ੍‫͢ޚ‬ΔͨΊɼา༰σʔλऩू༻ʹ 7. ‫ݸ‬ੑ‫ܭ‬ଌ͕ฏ‫ ۉ‬1.5 ඵɼ೥ྸਪఆ͕ฏ‫ ۉ‬0.2 ඵͰ͋Γɼ߹. ୆ͷ PCɼ‫ݸ‬ੑ‫ܭ‬ଌͷσϞ༻ʹ 4 ୆ͷ PC Λ࢖༻͍ͯ͠Δɽ. ‫ܭ‬໿ 1.7 ඵͷ࣌ؒΛཁ͢Δɽ‫ܭ‬ଌ͕։࢝͢Δͷ͸ମ‫͕ऀݧ‬. σʔλऩू༻ͷ 7 ୆ͷ PC ͸ PoE ϋϒΛ௨ͯ͡‫ݸ‬ผʹ 7 ୆. ՙ෺Λஔ͖ɼ࠷ॳʹંΓฦ͢ͱ͖Ͱ͋Γɼาߦ͕ऴྃ͢Δ. ͷΧϝϥͱ઀ଓ͓ͯ͠ΓɼλΠϜαʔόʔʹΑͬͯ࣌ؒͷ. ·Ͱ࢒Γ 1.5 ԟ෮Ͱ͋ΔɽͦͷͨΊɼาߦऴྃ·Ͱ໌Β͔. ಉ‫ظ‬Λߦ͍ͬͯΔɽσϞ༻ͷ PC ͸݁ՌදࣔͱϓϦϯλʔ. ʹ 1.7 ඵΑΓ͕͔͔࣌ؒΔͨΊɼาߦऴྃ࣌ʹ଴ͭ͜ͱͳ. Λ੍‫͢ޚ‬Δ PC1ɼQR ίʔυϦʔμʔ༻ͷ PC2ɼ‫ݸ‬ੑ‫ܭ‬ଌ. ݁͘Ռ͕දࣔ͞Ε͍ͯΔঢ়ଶͰ͋Δɽͦͷ݁ՌɼΦϯϥΠ. ༻ͷ PC3ɼޫిηϯαʔ༻ͷ PC4 ͷ 4 ୆Ͱ͋Δɽ4 ୆ͷ. ϯͷσϞ͕Մೳͳ‫͍͑ͱؒ࣌ࢉܭ‬Δɽ. PC ͸࿈ಈͯ͠ಈ࡞͢Δඞཁ͕͋ΔͨΊɼ‫ڞ‬༗ϑΝΠϧΛ ༻͍Δ͜ͱͰηϚϑΥʹΑΔ੍‫ޚ‬Λߦ͏ (ਤ 7)ɽ ମ‫ ͕ऀݧ‬QR ίʔυϦʔμʔʹ QR ίʔυΛ͔ͨ͟͠ ͱ͖ɼPC2 ͸ ID ϑΝΠϧΛ࡞੒͠ɼPC3 ͱ PC4 ΁ͷ ID. 4.2 ೥ྸਪఆͷਫ਼౓ධՁ QR ίʔυൃ݊‫Ͱػ‬ମ‫͕ऀݧ‬ೖྗ࣮ͨ͠೥ྸͱͷൺֱΛ ߦ͏͜ͱͰਪఆ೥ྸͷਫ਼౓ධՁΛߦ͏ɽ. ϑΝΠϧΛ࡞੒͢ΔɽPC3 ͱ PC4 ͸ ID ϑΝΠϧͷଘࡏΛ. ‫ࡱࡏݱ‬Өͨ͠ମ‫ऀݧ‬ͷ಺ɼ1,573 ਓΛֶशσʔλɼ755 ਓ. ֬ೝ‫ޙ‬ɼID ΛಡΈࠐΈɼϑΝΠϧΛ࡟আ͠ɼࡱӨ଴ͪͷঢ়. Λςετσʔλͱ͍ͯ͠ΔɽQR ίʔυൃ݊࣌ɼମ‫͕ऀݧ‬. ଶͱͳΔɽମ‫ॳ࠷͕ऀݧ‬ͷંΓฦ͠ͰηϯαʔΛԣ੾ͬͨ. ࣗ෼ͷ೥ྸͱੑผΛೖྗ͢Δ͕ɼඞͣ͠΋ਖ਼͍͠৘ใΛೖ. ͱ͖ (ਤ 2 ʹ͓͚Δηϯαʔ 2 Λ 2 ౓ԣ੾ͬͨͱ͖)ɼPC4. ྗ͢Δͱ͸‫ݶ‬Βͳ͍ɽ·ͨɼࣗಈͰࡱӨΛߦ͍ͬͯΔͨΊɼ. ͸ࡱӨ։࢝ͷϑΝΠϧΛ࡞੒͠ɼPC3 ͕ࡱӨ։࢝ϑΝΠ. ΤϥʔʹΑΓλΠϛϯά͕ͣΕͯ͠·ͬͨΓɼഎ‫ࠩܠ‬෼ʹ. ϧͷଘࡏΛ֬ೝ͢Δ͜ͱͰࡱӨΛ։࢝͢Δɽ࠷‫ʹޙ‬ମ‫ऀݧ‬. ΑΔγϧΤοτ͕ਖ਼͘͠ੜ੒͞Ε͍ͯͳ͍৔߹΋ߟ͑ΒΕ. ͕ୀग़ͨ͠ࡍ (ਤ 2 ʹ͓͚Δޫిηϯαʔ 4 Λԣ੾ͬͨͱ. ΔɽͦͷͨΊɼֶशσʔλ΍ςετσʔλͱͯ͠ར༻͢Δ. ͖)ɼPC4 ͸ PC2 ༻ʹୀग़ϑΝΠϧΛ࡞੒͢ΔɽPC3 ͸. ࡍɼར༻Մೳͳσʔλ͔Ͳ͏͔Λਓͷ໨Ͱ֬ೝ͢Δඞཁ͕. ‫ݸ‬ੑ‫ܭ‬ଌ͕ऴྃ‫݁ޙ‬ՌΛอଘ͠ɼ݁ՌอଘϑΝΠϧΛ࡞੒. ͋Δɽ‫ ࡏݱ‬30,000 ਓҎ্ͷσʔλ͕ऩूࡁΈͰ͋ΓɼͦΕ. ͢Δ͜ͱͰ PC1 ʹ఻͑ΔɽPC1 ͸݁ՌอଘϑΝΠϧΛ֬. ΒશͯͷσʔλΛ֬ೝ͢Δ৔߹ɼͱͯ΋͕͔͔࣌ؒͬͯ͠. ೝ‫ޙ‬ɼΠϯϑΥʔϜυίϯηϯτͱ‫݁ʹڞ‬ՌΛදࣔ͠ɼମ. ·͏ɽ‫֬ࡏݱ‬ೝ͕ऴ͍ྃͯ͠Δσʔλ͕·ͩগͳ͍ͨΊɼ. ‫ऀݧ‬͸σʔλར༻ʹؔͯ͠ಉҙɼඇಉҙΛϘλϯͰ൑அ͢. ࢖༻͍ͯ͠Δσʔλ͕গͳ͍ͱ͍͏ঢ়‫͋Ͱگ‬Δɽ. Δɽͦͷ‫ޙ‬ɼPC1 ͸ PC2 ʹϘλϯԡԼΛ఻͑ɼPC2 ͕ୀ ग़ͱϘλϯԡԼΛ֬ೝ‫ظॳޙ‬ঢ়ଶ΁ͱ໭Δɽ. ೥ྸਪఆਫ਼౓͸ਪఆ೥ྸͱ࣮೥ྸͷฏ‫ۉ‬ઈର‫( ࠩޡ‬MAEɼ. mean absolute error) ͰධՁΛߦ͏ɽͦͷ݁ՌɼMAE ͸. ·ͨɼຊγεςϜͰ͸ࣗಈԻ੠ʹΑͬͯମ‫ऀݧ‬ͷ༠ಋΛ. 8.7 ࡀͱͳΓɼจ‫[ ݙ‬28] Ͱใࠂ͞Ε͍ͯΔϕʔεϥΠϯͷ. ߦ͍ͬͯΔɽମ‫͕ऀݧ‬Կԟ෮໨ʹͲͷޫిηϯαʔΛ௨ա. ݁ՌͰ͋Δ 8.2 ࡀͱൺֱͯ͠গ͠‫͕ࠩޡ‬େ͖͍ͱ͍͏݁Ռ. ͔ͨ͠ʹΑͬͯҟͳΔԻ੠Λ࠶ੜ͠ɼ૝ఆ֎ͷߦಈʹର͠. ͱͳͬͨɽ. ͯ͸‫ࠂܯ‬จΛ࠶ੜ͢Δɽ‫ࠂܯ‬Λແࢹͨ͠৔߹ɼਖ਼͍͠ߦಈ. ߟ͑ΒΕΔ‫ݪ‬Ҽͱͯ͠ɼจ‫[ ݙ‬28] Ͱ࢖༻͍ͯ͠Δγϧ. Λͱͬͯ΋Β͏Α͏ελοϑ͕༠ಋ͢Δ͜ͱͰɼମ‫͕ऀݧ‬. Τοτ͸खमਖ਼͕Ճ͑ΒΕ͍ͯΔͱ͍͏఺Ͱ͋Δɽຊγε. ࣮֬ʹ 2 ԟ෮൒าߦ͢ΔΑ͏ʹ͍ͯ͠Δɽ. ςϜ͸ΦϯϥΠϯͷσϞΛߦ͏ͨΊɼखमਖ਼ΛՃ͑Δ͜ͱ. ͜ͷγεςϜ͸૝ఆͨ͠ঢ়‫Ͱگ‬ਖ਼͘͠ಈ࡞͢ΔΑ͏ʹߏ ங͍ͯ͠ΔͨΊɼମ‫͕ऀݧ‬૝ఆ֎ͷߦಈΛͱͬͨ৔߹ʹΤ. ͕Ͱ͖ͣɼγϧΤοτͷ඼࣭͕ྼΔ͜ͱͰ‫͕ࠩޡ‬େ͖͘ ͳ͍ͬͯΔͱߟ͑ΒΕΔɽ. ϥʔঢ়ଶͱͳΔ৔߹͕͋Δ (ྫ͑͹ɼલͷମ‫͕ऀݧ‬ମ‫ݧ‬த ʹࢠ‫͕ڙ‬าߦ࿏ʹਐೖ͢ΔͳͲ)ɽͦͷͨΊɼγεςϜͷ࠶. 4.3 าߦσʔλऩूͷਐ௙. ‫ى‬ಈΛߦ͏ϦηοτϘλϯΛ༻ҙ͓ͯ͠Γɼలࣔελοϑ. զʑ͸లࣔ։͔࢝Β 144 ೔ؒͰ 31,090 ਓͷาߦσʔλ. ͕ඞཁʹԠͯ͡ϦηοτϘλϯΛԡ͢͜ͱͰॳ‫ظ‬ঢ়ଶ΁ͱ. Λऩू͍ͯ͠Δɽ‫ެࡏݱ‬։͞Ε͍ͯΔา༰σʔλϕʔεͷ. c 2016 Information Processing Society of Japan. 6.

(7) Vol.2016-CVIM-200 No.18 2016/1/22. ৘ใॲཧֶձ‫ڀݚ‬ใࠂ IPSJ SIG Technical Report. ਤ 7 ηϚϑΥʹ‫ͮ͘ج‬ϑϩʔνϟʔτ. ͠ɼ༧ࢉ΍ਓ݅අΛߟ͑Δͱɼॳ‫ظ‬ίετͱҡ࣋ίετ͸ τϨʔυΦϑͷؔ܎ʹ͋Δɽେ‫ن‬໛ͳาߦσʔλΛऩू͢ Δ͏͑Ͱɼॳ‫ظ‬ίετ͕ߴ͘ɼҡ࣋ίετ͕௿͍ͱ͍͏γ εςϜ͕࠷దͰ͋Δͱߟ͑Δɽ. 5. ݁࿦ ຊ࿦จͰ͸ମ‫ܕݧ‬ͷσϞͱ‫ʹڞ‬าߦσʔλऩूγεςϜ ʹ͍ͭͯड़΂ͨɽମ‫ऀݧ‬͸࠷ઌ୺ͷาߦө૾ղੳͰ͋Δ‫ݸ‬ ੑ‫ܭ‬ଌɼาߦ೥ྸਪఆͷΦϯϥΠϯσϞΛָ͠Ή͜ͱ͕Ͱ ͖ɼಉ࣌ʹऩूऀଆ͸ΠϯϑΥʔϜυίϯηϯτͷͱΕͨ ਤ 8. ମ‫ऀݧ‬ͷ౷‫ܭ‬৘ใ. าߦσʔλΛऩू͢Δ͜ͱ͕Ͱ͖Δɽ‫ࡏݱ‬ɼ೔ຊՊֶະདྷ ‫ؗ‬ͷ‫ྗڠ‬ͷ΋ͱɼମ‫ܕݧ‬ͷσϞΛ࣮ࢪதͰ͋Γɼ͜Ε·Ͱ. தͰ࠷େͷඃ‫਺ऀݧ‬͸ 4,007 ਓ ( [5]) Ͱ͋ΔͨΊɼ͢Ͱʹ. 31,090 ਓͷาߦσʔλΛऩूࡁΈͰ͋Δɽ. େ෯ʹ௒͍͑ͯΔɽ͜ͷలࣔ͸߹‫ܭ‬໿ 11 ͔݄ؒߦ͏༧ఆ. ௕‫ظ‬ͷৗઃల͕ࣔऴྃ‫ޙ‬ɼେ‫ن‬໛ͳาߦσʔλϕʔεΛ. Ͱ͋Γɼ࠷ऴతʹ͸ 60,000 ਓΛ௒͑ΔσʔλϕʔεΛߏங. ߏங͠ɼ͍͔ͭ͘ͷา༰ղੳͷΞϧΰϦζϜΛ༻͍ͯධՁ. ՄೳͰ͋Δͱ༧ଌ͍ͯ͠Δɽ. Λߦ͏༧ఆͰ͋Δɽ·ͨɼσʔλϕʔεΛެ։͢Δ͜ͱͰ. ‫ࡏݱ‬ऩूࡁΈͷମ‫ऀݧ‬ͷੑผɼ೥ྸΛਤ 8 ʹࣔ͢ɽ60. า༰ೝূͷ‫ڀݚ‬ΛΑΓଅਐͤ͞Δ͜ͱ͕Ͱ͖Δͱߟ͑ͯ. ࡀҎ্ͷߴྸऀΛআ͍ͯɼ෯޿͍೥ྸͱੑผͷσʔλ͕ू. ͍Δɽ. ·͍ͬͯΔ͜ͱ͕෼͔ΔɽͦͷͨΊɼߴྸऀͷσʔλ͕૿. ँࣙ. ͑ΔͱΑΓਫ਼౓ͷߴ͍೥ྸਪఆ͕ՄೳʹͳΔͱࢥΘΕΔɽ. ຊ‫ڀݚ‬͸ɼՊֶٕज़ৼ‫( ߏػڵ‬JST) ઓུత૑଄‫ڀݚ‬ਪਐ ࣄ‫( ۀ‬CREST)ɼ‫ ͼٴ‬JSPS ‫ج‬൫‫( ڀݚ‬A)15H01693 ͷॿ੒. 4.4 าߦσʔλऩूͷ੍‫ݶ‬. Λड͚ͨ΋ͷͰ͋Δɽ·ͨɼ‫ࡏݱ‬೔ຊՊֶະདྷ‫ؗ‬ͷϝσΟ. ‫ݸ‬ਓೝূͷධՁʹ༻͍Δσʔλϕʔε͸ɼ௨ৗ͸ొ࿥. ΞϥϘͷୈ 15 ‫ظ‬లࣔͷ৔Λ͓आΓ࣮ͯ͠ࢪ͓ͯ͠Γɼ௕. σʔλͱೖྗσʔλͷ 2 छྨΛɼ࣌ؒͷִؒΛۭ͚ͯσʔ. ‫ظ‬ͷৗઃలࣔʹΑΔσʔλऩूͷ‫ػ‬ձΛ͍͖ͨͩɼ೔ຊՊ. λऩूΛߦ͏͜ͱ͕๬·͍͕͠ɼࠓճͷσϞͰ͸΄ͱΜͲ. ֶະདྷ‫[ ؗ‬31] ͷελοϑͷํʹ͸‫ँײ‬க͠·͢ɽ. ࣌ؒͷִ͕ؒͳ͍ɽ͜ͷσϞ͸Պֶ‫΁ؗ‬ͷདྷ٬Λର৅ͱ͠ ͍ͯΔͨΊɼ࣌ؒΛۭ͚ͯ΋͏Ұ౓ମ‫ͯ͠ݧ‬΋Β͏͜ͱ͸. ࢀߟจ‫ݙ‬. ೉͍͠ɽ࣌ؒΛۭ͚ͨσʔλΛऩू͢ΔͨΊʹ͸ɼ2 ౓Ҏ্. [1]. ͷମ‫ݧ‬Λͯ͠΋Β͏ඞཁ͕͋ΓɼͦͷͨΊʹ͸ QR ίʔυ ͷ࠶ར༻Λ͠΍ͨ͘͢͠Γɼาߦө૾Λ༻͍݈ͨ߁νΣο. [2]. ΫͳͲɼ࠶౓ମ‫ͳͨ͘͠ݧ‬ΔΑ͏ͳ޻෉Λ͢Δඞཁ͕͋Δɽ ͞Βʹɼࣗಈ‫ܭ‬ଌγεςϜΛ૿ங͢Δ͜ͱ͸ɼखಈͰ σʔλऩूΛߦ͏ํ๏ͱൺ΂ͯॳ‫ظ‬ίετ͕͔͔Δɽ͔͠. c 2016 Information Processing Society of Japan. [3]. J. Wambaugh, The Blooding. HarperCollins Publishers, 1989. D. Maltoni, D. Maio, A. K. Jain, and S. Prabhakar, Handbook of Fingerprint Recognition, 2nd ed. Springer Publishing Company, Incorporated, 2009. D. Zhang, Palmprint Authentication, ser. International Series on Biometrics. Springer Publishing Company,. 7.

(8) Vol.2016-CVIM-200 No.18 2016/1/22. ৘ใॲཧֶձ‫ڀݚ‬ใࠂ IPSJ SIG Technical Report. [4]. [5]. [6]. [7]. [8]. [9]. [10]. [11]. [12]. [13]. [14]. [15]. [16] [17]. [18]. Incorporated, 2004, vol. 3. M. J. Burge and K. W. Bowyer, Handbook of Iris Recognition. Springer Publishing Company, Incorporated, 2013. H. Iwama, M. Okumura, Y. Makihara, and Y. Yagi, “The ou-isir gait database comprising the large population dataset and performance evaluation of gait recognition,” IEEE Transactions on Information Forensics and Security, vol. 7, no. 5, pp. 1511–1521, Oct. 2012. Y. Makihara, H. Mannami, A. Tsuji, M. Hossain, K. Sugiura, A. Mori, and Y. Yagi, “The ou-isir gait database comprising the treadmill dataset,” IPSJ Transactions on Computer Vision and Applications, vol. 4, pp. 53– 62, Apr. 2012. S. Samangooei, J. D. Bustard, R. D. S. M. S. Nixon, and J. N. Carter, “On acquisition and analysis of a dataset comprising of gait, ear and semantic data,” pp. 277–301, 2011. M. Hofmann, J. Geiger, S. Bachmann, B. Schuller, and G. Rigoll, “The tum gait from audio, image and depth (gaid) database: Multimodal recognition of subjects and traits,” in Journal of Visual Communication and Image Representation, Special Issue on Visual Understanding and Applications with RGB-D Cameras, vol. 25, no. 1, 2014, pp. 195–206. J. Shutler, M. Grant, M. Nixon, and J. Carter, “On a large sequence-based human gait database,” in Proc. of the 4th Int. Conf. on Recent Advances in Soft Computing, Nottingham, UK, Dec. 2002, pp. 66–71. S. Sarkar, J. Phillips, Z. Liu, I. Vega, P. G. ther, and K. Bowyer, “The humanid gait challenge problem: Data sets, performance, and analysis,” IEEE Transactions of Pattern Analysis and Machine Intelligence, vol. 27, no. 2, pp. 162–177, 2005. S. Yu, D. Tan, and T. Tan, “A framework for evaluating the effect of view angle, clothing and carrying condition on gait recognition,” in Proc. of the 18th Int. Conf. on Pattern Recognition, vol. 4, Hong Kong, China, Aug. 2006, pp. 441–444. Y. Makihara, A. Tsuji, and Y. Yagi, “Silhouette transformation based on walking speed for gait identification,” in Proc. of the 23rd IEEE Conf. on Computer Vision and Pattern Recognition, San Francisco, CA, USA, Jun 2010. M. A. Hossain, Y. Makihara, J. Wang, and Y. Yagi, “Clothing-invariant gait identification using part-based clothing categorization and adaptive weight control,” Pattern Recognition, vol. 43, no. 6, pp. 2281–2291, Jun. 2010. B. DeCann, A. Ross, , and J. Dawson, “Investigating gait recognition in the short-wave infrared (swir) spectrum: Dataset and challenges,” in Proc. of the SPIE Conference on Biometric Technology for Human Identification X, 2013, pp. 1–16. H. Mannami, Y. Makihara, and Y. Yagi, “Gait analysis of gender and age using a large-scale multi-view gait database,” in Proc. of the 10th Asian Conf. on Computer Vision, Queenstown, New Zealand, Nov. 2010, pp. 975–986. R. Gross and J. Shi, “The cmu motion of body (mobo) database,” CMT, Tech. Rep., Jun. 2001. R. Tanawongsuwan, “Impact of speed variations in gait recognition,” Ph.D. dissertation, Atlanta, GA, USA, 2003, aAI3110453. T. Chalidabhongse, V. Kruger, and R. Chellappa, “The. c 2016 Information Processing Society of Japan. [19]. [20]. [21]. [22]. [23]. [24]. [25]. [26]. [27]. [28]. [29]. [30]. [31]. umd database for human identification at a distance,” University of Meryland, Tech. Rep., 2001. M. Nixon, J. Carter, J. Shutler, and M. Grant, “Experimental plan for automatic gait recognition,” Southampton, Tech. Rep., 2001. D. Matovski, M. Nixon, S. Mahmoodi, and J. Carter, “The effect of time on gait recognition performance,” Information Forensics and Security, IEEE Trans. on, vol. 7, no. 2, pp. 543 –552, april 2012. L. Wang, H. Ning, T. Tan, and W. Hu, “Fusion of static and dynamic body biometrics for gait recognition,” in Proc. of the 9th International Conference on Computer Vision, vol. 2, 2003, pp. 1449–1454. D. Tan, K. Huang, S. Yu, and T. Tan, “Efficient night gait recognition based on template matching,” in Proc. of the 18th International Conference on Pattern Recognition, vol. 3, Hong Kong, China, Aug. 2006, pp. 1000– 1003. A. Mori, Y. Makihara, and Y. Yagi, “Gait recognition using period-based phase synchronization for low framerate videos,” in Proc. of the 20th International Conference on Pattern Recognition, Istanbul, Turkey, Aug. 2010, pp. 2194–2197. M. Hofmann, S. Sural, and G. Rigoll, “Gait recognition in the presence of occlusion: A new dataset and baseline algorithms,” in In: 19th International Conferences on Computer Graphics, Visualization and Computer Vision (WSCG), 2011. Y. Makihara and Y. Yagi, “Silhouette extraction based on iterative spatio-temporal local color transformation and graph-cut segmentation,” in Proc. of the 19th International Conference on Pattern Recognition, Tampa, Florida USA, Dec. 2008. Y. Makihara, M. Okumura, Y. Yagi, and S. Morishima, “The online gait measurement for characteristic gait animation synthesis,” in Proc. of Human Computer Interaction Int. 2011, Virtual and Mixed Reality - New Trends, ser. Lecture Notes in Computer Science, R. Shumaker, Ed., vol. 6773. Orlando, FL, USA: Springer, 2011, pp. 325–334. Y. Makihara, R. Sagawa, Y. Mukaigawa, T. Echigo, and Y. Yagi, “Gait recognition using a view transformation model in the frequency domain,” in Proc. of the 9th European Conference on Computer Vision, Graz, Austria, May 2006, pp. 151–163. Y. Makihara, M. Okumura, H. Iwama, and Y. Yagi, “Gait-based age estimation using a whole-generation gait database,” in Proc. of the Int. Joint Conf. on Biometrics (IJCB2011), Washington D.C., USA, Oct. 2011, pp. 1–6. Z. Liu and S. Sarkar, “Simplest representation yet for gait recognition: Averaged silhouette,” in Proc. of the 17th International Conference on Pattern Recognition, vol. 1, Aug. 2004, pp. 211–214. J. Han and B. Bhanu, “Individual recognition using gait energy image,” IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 28, no. 2, pp. 316– 322, 2006. http://www.miraikan.jst.go.jp/info/1506151518361. html.. 8.

(9)