20 Individual Recognition using positions of facial parts 1115081 2009 3 5
,,,,,,,,,,,,,,,,,,, 976%, i
Abstract Individual Recognition using positions of facial parts YOSHIHIRO Arisawa A facial recognition system is one of the biometric authentication of person As for this technique, the characteristics of not needing cooperation of a recognitized object are attracted attention The number of dimension of the facial image s feature is huge Therefore, computational complexity becomes huge when face recognition is performed Now, the mainstream technique is reducing the number of dimensions of the feature using the statistical method However, the feature that human is originally paying attention may be deleted by this technique In this research, the feature data which identify an individual was made based on the face cognitive processing of human When performing individual recognition of a face picture, human is setting arrangement patterns of parts of face such as eyes, noses, and mouths as one of the important features which individual identifies The useful individual feature data for performing personal identification was created using this arrangement pattern This research shows the result of having studied the useful features for performing individual identification using only the arrangement pattern of the partial feature of a face And the edperiment of individual identification was performed using this technique As a result,the recognition rate became 976% key words parts of face,arrangement patterns ii
1 1 11 1 12 2 13 3 14 4 2 5 21 5 22 7 221 7 222, 8 223 9 224 11 3 13 31 13 32 15 33 16 331 16 4 17 41 17 42 18 43 DB 19 44 20 iii
45 21 451 21 452 24 453 25 454 25 455 26 456 26 457 27 46 28 47 1 28 48 2 29 49 3 29 5 30 51 42 30 52 31 6 32 7 34 35 38 iv
21 5 22 5 23 8 24 9 31 14 32 15 33 A 16 34 B 16 35 C 16 36 D 16 41 18 42 A 20 43 B 20 44 21 45 d=2 22 46 d=3 22 47 d=4 22 48 d=5 22 49 23 410 23 411 A 24 412 24 v
413 25 414 26 415 26 vi
21 6 22 A 9 23 B 10 24 C 10 25 A 11 26 B 12 27 C 12 31 13 32 14 33 14 41 DB 19 42 A 27 43 B 27 51 30 52 21 31 vii
1 11,,,,,,DNA,, DNA 100% [1] [2],,,,,,,,,,,,,,,,,,,, 1
12,,,,,,, [5],,,,,,, 12,,,,,,,,,, [6],,,,,,,,, 2
13,,,,,,,,, [6],,,,,, [7],,,,,, 13, 1 2 3 4 3
14 2,,,,,,, 14,,,,,,, 4
2,,,, 21,,,,,,,,,, 22 14 21 21 22 5
21 21,,,,,, 6
22 22, 221,,, 12,,,,,, 1 2 3 3, DB 7
22 222,,221,,,,, 24 23 8
22 223,, - ( 14 ),,, - 22 23 24 24 22 A - ( ) 4303 - ( ) 7380 - ( ) 9399 - ( ) 7439 - ( ) 4490 - ( ) 7380 - ( ) 8119 - ( ) 12223 - ( ) 3408 - ( ) 12231 - ( ) 8045 - ( ) 5002 - ( ) 7439 - ( ) 5002 9
22 23 B - ( ) 3253 - ( ) 8169 - ( ) 9131 - ( ) 6821 - ( ) 4248 - ( ) 7178 - ( ) 9228 - ( ) 11897 - ( ) 3044 - ( ) 11890 - ( ) 8362 - ( ) 5150 - ( ) 3070 - ( ) 5150 24 C - ( ) 3471 - ( ) 8429 - ( ) 9436 - ( ) 7669 - ( ) 3607 - ( ) 7739 - ( ) 9753 - ( ) 12366 - ( ) 2912 - ( ) 12460 - ( ) 8230 - ( ) 5301 - ( ) 3298 - ( ) 5301 10
22 224, -,,,,DB,,, ( ), - 21 F 1 = (xi x c ) 2 + (y i x c ) 2 d (21) x i, y i : x, y, x c, y c : x, y, n :, d : 25 A - ( ) 0430 - ( ) 0340 - ( ) 0939 - ( ) 0804 - ( ) 0449 - ( ) 0743 - ( ) 0918 - ( ) 0738 - ( ) 0305 - ( ) 1222 - ( ) 0812 - ( ) 1223 11
22 26 B - ( ) 0316 - ( ) 0298 - ( ) 0887 - ( ) 0793 - ( ) 0412 - ( ) 0662 - ( ) 0860 - ( ) 0697 - ( ) 0296 - ( ) 1155 - ( ) 0811 - ( ) 1154 27 C - ( ) 0328 - ( ) 0311 - ( ) 0890 - ( ) 0795 - ( ) 0340 - ( ) 0723 - ( ) 0920 - ( ) 0730 - ( ) 0275 - ( ) 1166 - ( ) 0776 - ( ) 1175 12
3 31,, 31 3 31 Victor GZ-MG77,,, 13
31 32 Victor GZ-MG77 1/39 218 CCD 123 200 F12 20 f = 38 38mm(35mm 358 457mm) MPEG-2 Dolby Digital NTSC HDD, SD 33 1mm/m=00527 525mm/m= 02999 31 14
32 32, ( ),,,,,,,, 32 15
33 環境設定 33 331 環境設定 カメラの位置設定 カメラを三脚に取り付ける際に, 水準器を用いてカメラが水平になるように雲台を調整す る 調整を行ったら, 雲台を固定する カメラの液晶モニターに背景を映し出し, 目標物をモニターの中央に捕らえるようにカメ ラの調整する 目標物との高さが合わない場合は, 三脚のエレベータ機能を使用して高さを合 わせるようにする 図 33 原画像 A 図 34 原画像 B 図 35 原画像 C 図 36 原画像 D 16
4 41, 21, 2,3 680 480 3,1 DB, 2 DB 21, 42, DB 17
42 42 41 18
43 DB 43 DB DB 21,, -, - ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) 41 DB - - ( ) - ( ) - ( ) - ( ) - ( ) - ( ) - ( ) - ( ) - ( ) - ( ) - ( ) - ( ) - ( ) - ( ) - ( ) - ( ) - ( ) - ( ) - ( ) - ( ) - ( ) - ( ) - ( ) - ( ) - ( ) - ( ) 19
44 画像の前処理 44 画像の前処理 DB 登録画像, 未知入力画像共に, 元画像から顔領域をプログラムを用いて検出した 顔 検出プログラムは,OpenCV で公開されている顔検出プログラムが基となっている 顔画像 データにより学習によってあらかじめ獲得された, 分類器のカスケードが記述された xml ファイルを読み込み, 顔領域を検出する [9][10] 検出された顔領域は 256 256 のサイズでリサイズした また, 撮影時の照明条件の若干 の違いを考慮し, 輝度の平坦化のためイコライズ処理を施した 今回顔の検出処理を施した 画像は全て顔領域の抽出に成功している (a) 原画像 (b) 正規化画像 図 42 被験者 A の顔画像の正規化 (a) 原画像 (b) 正規化画像 図 43 被験者 B の顔画像の正規化 20
45 45, 451,, Laplacian,Sobel,,, [8], 41,42, 44 d=2,3,4,5, 45, 46, 47, 48, 5,, d 5,, f xx = f(i + d, j) + 2f(i, j) f(i d, j) (41) f yy = f(i, j + d) + 2f(i, j) f(i, j d) (42) 44 21
45 45 d=2 46 d=3 47 d=4 48 d=5 22
45,,, 49, 410,,,,,,, f xx = f(i + d, j) 2f(i, j) + f(i d, j) (43) f yy = f(i, j + d) 2f(i, j) + f(i, j d) (44) 49 410 23
45 452,,, X Y, 411 A X,Y, 412,, X 2 X, Y, 4,, (a)x 411 A (b)y 412 24
45 453, X 2 X, Y 4,,,, 454 X 2 X, Y, 2,, 413 25
45 455 X 2 X, Y, 4, 414 456 X 2 X, Y 2,, 415 26
45 457, 42 43 42 A ( ) (101,72) ( ) (208,114) ( ) (37,82) ( ) (99,174) ( ) (160,74) ( ) (162,171) ( ) (218,87) ( ) (88,221) ( ) (98,111) ( ) (171,220) ( ) (47,111) ( ) (78,104) ( ) (161,114) ( ) (178,107) 43 B ( ) (101,79) ( ) (206,107) ( ) (36,79) ( ) (95,163) ( ) (156,73) ( ) (156,166) ( ) (214,79) ( ) (83,213) ( ) (95,109) ( ) (163,214) ( ) (156,106) ( ) (73,101) ( ) (83,110) ( ) (176,102),,,,, 27
46 46, 3, 47 1 DB ( ) F 1 = n i=1 (x i x i )2 + (y i y i )2 (45) x i, y i : DB x, y x i, y i : x, y n : DB, DB,, DB, DB, DB 28
48 2 48 2 ( ) F 2 = n (xi x c ) 2 + (y i y c ) 2 (46) i=1 x c, y c : x, y DB -, -, DB, DB 49 3 ( ) F 3 = n (xi x c ) 2 + (y i y c ) 2 i=1 d (47) d : DB, - DB, DB 29
5 3 51 42 51 643% 524% 952%, 42,DB 30
52 52 21, 2, 52 21 5 5 11 7 6 8 0 1 20 31
6 1 2,643%,524%,, 3,976%,,,,,,,,,,,,,,,,,,,,, 32
, -,,,, -,,,,, 33
7,,,,,,,,,DB,, 34
,,,, 10, 4,, 4, 4,,,,,,,,,,,,,,,,, 1,,,,, 35
NTT OB,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,, 6,, 4 36
,,,, 6,, 37
[1],,,, DNA-ID : STR, ISEC, Technical report of IEICE ISEC, Vol101, No214(20010718)pp1-7 [2],,, NC, IEICE technical report Neurocomputing Vol103, No732(20040310) pp 65-70 [3],,, 18 [4],,2004 [5] MTurk and APentland, Face Recognition Using Eigene faces, IEEE Computer Vision and Pattern Recognition,pp586-591, 1991 [6],,,2004 [7], 18,,2002 [8],,, 20 [9] Paul Viola and Michael J Jones, Rapid Object Detection using a Boosted Cascade of Simple Features, IEEE CVPR, 2001 [10] Rainer Lienhart and Jochen Maydt, An Extended Set of Haar-like Features for Rapid Object Detection, IEEE ICIP 2002, Vol 1, pp 900-903, Sep 2002 38