Masterarbeit DigitSense

Titel

DigitSense: Grasp Recognition Using Fingerprint Sensors

Aufgabenstellung

Testen der Möglichkeit, mittels Fingerabdruckscannern die Handhaltung an einem Handheld(?) zu erkennen
Entwickeln eines passenden Algorithmus
Bauen eines Prototyps
Echtzeitanforderung!

Zeitplan

Rest vom April: Literaturrecherche, AuthenTec AES2501 verstehen
Mai: Literaturrecherche, Treiber
Juni:
  • Algorithmensuche
  • SW-Entwicklung
  • Demoentwicklung
  • Datenschutz klären

Juli:
  • SURF/SIFT zum Laufen bringen (1 Woche)
  • Vergleich Performance BruteForce mit/ohne directional field, SURF
  • ganze Hände einscannen (sinnvoll? Uniqueness, klare Bilder)
  • Scannen im Magnetfeld? => keinerlei sichtbare Beeinflussung

August:
  • Studie: wie gut lassen sich verschiedene Handbereiche identifizieren?
  • Einfluss der Größe des Scanbereichs auf Erkennungsgenauigkeit ermitteln (Ausschnitte aus Ganzhandscans)
  • Flachprototyp zur Verifizierung der Erkenntnisse mit den Ganzhandscans

September:
  • Algorithmus zur Kombination von Scanslots zur gegenseitigen Verifizierung der Scans (Ganzhandscans theoretisch testen)
  • 3D-Modell der Hand falls nötig
  • 3D-Prototyp

Oktober: User Study + Schreiben

Misc. Links

http://home.gna.org/aes2501/index_en.html
http://www.reactivated.net/fprint/wiki/Aes2501
http://gkall.hobby.nl/authentec.html
http://opencv.willowgarage.com/documentation/python/index.html
http://www.suttoncourtenay.org.uk/duncan/accu/integratingpython.html

  • Lyx - WYSIWYM for LaTex
  • LaTeX-Vorlage auf Website

Notes

AES = Advanced Electronic Signature
AES2501 resolution = 500 dpi (http://www.griaulebiometrics.com/page/fingerprint_sdk/supported_readers)
Minutiae density = 0.215 - 0.230 minutiae/mm² (http://www3.interscience.wiley.com/cgi-bin/fulltext/110482926/PDFSTART)
=> number of minutiae on one aes2501 slot in [1.7; 1.82]

BF = Brute Force
DF = Directional Field (OM = Orientation Map)

Fingerprint classification systems:
  • Roscher (Germany and Japan)
  • Juan Vucetich (South America)
  • Henry (most English-speaking countries)

Level of Detail:
  • 1) general structure: left/right loop, (tented) arch, whorl...
  • 2) minutiae: ridge endings, ridge bifurcations...
  • 3) intra-ridge details: sweat pores...

Literature:
  • "Handbook of Fingerprint Recognition" By Davide Maltoni, Dario Maio, Anil K. Jain, Salil Prabhakar

aligning the fingerprint images by using the Iterative Closest Point (ICP) algorithm (Besland McKay, 1992)

To compensates for the amount of plastic distortion between two partial images, Lee et al. (2003) and Choi et al. (2007a) use non-rigid alignment transformations such as Chamfer matching (Borgefors, 1988) and Thin-Plate Spline (TPS) model (Ross, Dass, and Jain, 2006).

Morguet et al. (2004) proposed matching two fingerprints directly at slice level without any image reconstruction: a pair of slices are matched through normalized correlation and the optimal matching between two complete slice sequences is found by Viterbi search (Viterbi, 1967).

Compilation des modularen Versuchs:
gcc -Wall -c aes2501.c
gcc -Wall -c aes2501_ext.c
gcc usb.o aes2501.o aes2501_ext.o -o digitSense
=> executable "digitSense"

SIFT: http://blogs.oregonstate.edu/hess/code/sift/

~1,8 Minutiae/Slot, eher weniger => viele Slots ohne Minutiae
5-6 Scanners => 9-11 Minutiae (oder schätzungsweise 4-5 in low quality images), ausreichend für das Gesamtsystem?

Log file format:
id, finger_target, x_target, y_target, angle_target, finger_nearest, x_nearest, y_nearest, angle_nearest, rank_correct_finger, rank_position_on_correct_finger, timestamp

Code

Sämtlicher Code verfügbar über
svn checkout https://svn.cip.ifi.lmu.de/~rademacb/svn/digitSense

Compilation:
make
=> executable "aes2501"

Swig:
swig -python -Wall digitSense.i
python setup.py build_ext --inplace

in Python:
import digitSense
digitSense.findScanners()
string = digitSense.scanSingleRow(0)
digitSense.closeAll()
from opencv import cv, highgui
image = cv.cvCreateMat(16, 192, cv.CV_8UC1)
for y in range(0, image.height):
    for x in range(0, image.width):
        image[y,x] = ord(string[y*image.width + x])
highgui.cvSaveImage("test.png", image)
import os
os.system("display test.png")

Zeigen im Fenster:
in Konsole: export LD_LIBRARY_PATH=.
in Python: highgui.cvNamedWindow("name"), highgui.cvShowImage("name", image), highgui.cvStartWindowThread()

Auftrag in Grid Engine geben:
qsub -cwd -q luna.q "statistics 2.sh"

Rekursives Finden:
find . -name \*1827

Statistics

  • Statistics 1: How many random cutout slots does BF with DF find at the correct location on one finger?
352/1100 = 32% linker Daumen
311/1100 = 28,3% linker Zeigefinger
332/1100 = 30,2% linker Mittelfinger
Parameters: Thresholds to hit for statistic: x < 12, y < 12, angle < 6, SPACING for cutout = 2
COH_THRESHOLD = 0.4 in BF, ANGLE_STEP = 5 in BF, SAMPLE_DIST = 5 in DF

No Penalty, value outside picture = 15 (no artificial edge):
315/1100 = 28,6%

Penalty = 90° for slot structure outside the finger image:
214/1100 = 19,5% linker Daumen
With fill values = 15: 248/1100 = 22,5% linker Daumen

Penalty = 90° for slot structure outside the finger image and low finger coherence:
171/1100 = 15,6% linker Daumen

  • Statistics 2: How many random cutout slots from 10 fingers does BF with DF find at the correct finger and location? => 295/1260 = 23,4% on correct finger, 186/295 = 63,1% on correct location
Parameters: Thresholds to hit for statistic: x < 12, y < 12, angle < 6, SPACING for cutout = 2
COH_THRESHOLD = 0.4 in BF, ANGLE_STEP = 5 in BF, SAMPLE_DIST = 5 in DF

  • Statistics 3: At which position in the ranking is the first good hit for random cutout slots on 1 finger?
ranks: [649, 131, 66, 52, 36, 25, 28, 25, 13, 18, 15, 16, 11, 16, 12, 7, 8, 5, 14, 5, 13, 7, 6, 9, 6, 4, 7, 6, 5, 5, 12, 7, 2, 8, 6, 5, 5, 2, 1, 9, 1, 7, 2, 3, 4, 5, 3, 0, 4, 3, 6, 2, 3, 5, 1, 7, 4, 5, 0, 3, 4, 1, 0, 1, 4, 2, 0, 2, 4, 1, 2, 4, 3, 6, 2, 5, 2, 2, 1, 4, 0, 1, 3, 1, 1, 3, 0, 3, 5, 3, 3, 0, 1, 3, 5, 0, 1, 4, 2, 0, 566] out of 2000 ranks.
Parameters: "links Daumen.png"; Thresholds to hit for statistic: x < 12, y < 12, angle < 6, SPACING for cutout = 2
COH_THRESHOLD = 0.4 in BF, ANGLE_STEP = 5 in BF, SAMPLE_DIST = 5 in DF

With penalty = 90° for slot structure outside the finger image:
[410, 98, 47, 44, 31, 22, 31, 17, 14, 15, 9, 26, 14, 12, 10, 9, 8, 7, 10, 10, 11, 4, 9, 5, 6, 12, 10, 5, 5, 4, 1, 9, 3, 0, 7, 6, 8, 3, 2, 3, 6, 5, 5, 4, 1, 6, 5, 2, 9, 4, 3, 1, 7, 4, 2, 3, 3, 8, 4, 2, 3, 1, 1, 0, 6, 3, 2, 0, 3, 1, 2, 5, 4, 2, 3, 0, 3, 1, 1, 4, 3, 1, 1, 2, 2, 3, 4, 1, 5, 0, 1, 4, 2, 3, 1, 2, 2, 2, 1, 0, 874]

By Parameters

All times inclusive cutout times for slot retrieval: 4783.53816915s for 2000 slots => 2.391769084575s per slot

slot size (mean, DF,
DF resolution 8x8)
ranks correct finger correct position time to calculate
192x16 [512, 183, 131, 112, 78, 85, 54, 46, 47, 29, 32, 28, 38, 35, 30, 28, 30, 33, 14, 14, 441]
out of 2000
512/2000 = 25.6% 286/512 = 55.86% 399174.9s
96x32 [581, 185, 125, 98, 83, 55, 63, 45, 43, 40, 45, 31, 30, 22, 34, 34, 29, 26, 28, 21, 382]
out of 2000
581/2000 = 29.05% 351/581 = 60.41% 419837.68s
56x56 [623, 202, 128, 95, 84, 60, 63, 51, 34, 35, 35, 35, 23, 32, 37, 23, 26, 17, 18, 23, 356]
out of 2000
623/2000 = 31.15% 458/623 = 73.52% 443247.74s
288x24 [827, 138, 81, 70, 45, 46, 37, 29, 27, 27, 23, 25, 22, 27, 22, 14, 15, 18, 10, 24, 473]
out of 2000
827/2000 = 41.35% 618/827 = 74.73% 813587.4s

slot size (median, DF,
DF resolution 8x8)
ranks correct finger correct position time to calculate
192x16 [356, 231, 197, 131, 122, 85, 85, 80, 73, 61, 48, 43, 46, 22, 33, 28, 30, 22, 17, 28, 262]
out of 2000
356/2000 = 17.8% 141/356 = 39.61% 416482.27s
96x32 [359, 236, 172, 126, 127, 113, 95, 96, 63, 53, 45, 59, 35, 35, 30, 28, 30, 23, 9, 23, 243]
out of 2000
359/2000 = 17.95% 131/359 = 36.49% 453547.49s
56x56 [406, 265, 191, 143, 129, 115, 75, 64, 56, 41, 48, 48, 50, 26, 22, 32, 27, 20, 14, 14, 214]
out of 2000
406/2000 = 20.3% 222/406 = 54.68% 473136.6s
288x24 [1072, 210, 115, 74, 70, 48, 33, 28, 20, 28, 24, 17, 16, 10, 10, 12, 8, 7, 13, 8, 177]
out of 2000
1072/2000 = 53.6% 902/1072 = 84.14% 848461.48s

DF resolution
(192x16, mean)
ranks correct finger correct position time to calculate
2x2 [80, 3, 5, 4, 3, 0, 4, 2, 0, 1, 5, 1, 1, 1, 1, 1, 0, 1, 0, 0, 40]
out of 153
80/153 = 52.29% 72/80 = 90.0% 5263087.37s
3x3 [387, 20, 21, 17, 16, 8, 16, 6, 5, 10, 10, 8, 6, 11, 4, 5, 3, 6, 7, 3, 177]
out of 746
387/746 = 51.88% 345/387 = 89.15% 5488749.63s
4x4 [887, 106, 86, 61, 61, 43, 22, 27, 27, 26, 25, 23, 28, 27, 17, 15, 23, 14, 26, 22, 434]
out of 2000
887/2000 = 44.35% 753/887 = 84.89% 4307668.82s
5x5 [699, 146, 90, 54, 53, 44, 41, 40, 33, 24, 42, 31, 29, 28, 18, 26, 21, 21, 22, 19, 519]
out of 2000
699/2000 = 34.95% 540/699 = 77.25% 2522856.34s
6x6 [718, 149, 112, 74, 54, 42, 40, 35, 31, 28, 39, 33, 31, 29, 19, 23, 20, 19, 18, 21, 465]
out of 2000
718/2000 = 35.9% 574/718 = 79.94% 1323708.05s
7x7 [554, 160, 100, 88, 68, 58, 55, 42, 40, 46, 37, 42, 37, 25, 31, 26, 23, 23, 21, 27, 497]
out of 2000
554/2000 = 27.7% 334/554 = 60.29% 882518.13s
8x8 [512, 183, 131, 112, 78, 85, 54, 46, 47, 29, 32, 28, 38, 35, 30, 28, 30, 33, 14, 14, 441]
out of 2000
512/2000 = 25.6% 286/512 = 55.86% 399174.9s
9x9 [417, 209, 123, 95, 106, 71, 66, 53, 52, 41, 36, 31, 46, 30, 35, 29, 27, 21, 25, 25, 462]
out of 2000
417/2000 = 20.85% 184/417 = 44.12% 295831.79s
10x10 [391, 187, 141, 96, 87, 80, 74, 49, 50, 49, 34, 40, 31, 26, 25, 27, 28, 22, 26, 26, 511]
out of 2000
391/2000 = 19.55% 154/391 = 39.39% 223951.87s

finger part / feature
(192x16, mean, DF, 8x8)
left
thumb
left
index finger
left
middle finger
left
ring finger
left
little finger
right
thumb
right
index finger
right
middle finger
right
ring finger
right
little finger
all
fingertip 56/168
= 33.33%
48/153
= 31.37%
29/105
= 27.62%
24/118
= 20.34%
29/134
= 21.64%
47/173
= 27.17%
57/150
= 38.0%
52/159
= 32.70%
36/172
= 20.93%
30/119
= 25.21%
408/1451
= 28.12%
phalanx 7/32
= 21.88%
3/47
= 6.38%
18/95
= 18.95%
19/82
= 23.17%
14/66
= 21.21%
9/27
= 33.33%
6/50
= 12.0%
9/41
= 21.95%
6/28
= 21.43%
13/81
= 16.05%
104/549
= 18.94%
0.907 7.907 1.298 0.148 0.003 0.239 6.847 0.996 0.002 1.573 10.795

penalty (mean, DF,
DF resolution 8x8)
ranks correct finger correct position time to calculate
no penalty [512, 183, 131, 112, 78, 85, 54, 46, 47, 29, 32, 28, 38, 35, 30, 28, 30, 33, 14, 14, 441]
out of 2000
512/2000 = 25.6% 286/512 = 55.86% 399174.9s
penalty = 90° [470, 159, 126, 97, 75, 64, 47, 43, 41, 41, 43, 41, 39, 38, 36, 28, 25, 25, 21, 23, 518]
out of 2000
470/2000 = 23.5% 208/470 = 44.26% 401708.17s
penalty = 180° [431, 158, 101, 81, 82, 62, 67, 44, 46, 41, 34, 36, 37, 39, 29, 26, 30, 22, 17, 29, 588]
out of 2000
431/2000 = 21.55% 173/431 = 40.14% 394353.69s

erosion (mean, DF,
DF resolution 8x8)
ranks correct finger correct position time to calculate
without erosion [512, 183, 131, 112, 78, 85, 54, 46, 47, 29, 32, 28, 38, 35, 30, 28, 30, 33, 14, 14, 441]
out of 2000
512/2000 = 25.6% 286/512 = 55.86% 399174.9s
with erosion [523, 182, 123, 111, 87, 64, 66, 65, 44, 47, 45, 31, 21, 36, 38, 19, 22, 30, 21, 28, 397]
out of 2000
523/2000 = 26.15% 283/523 = 54.11% 395993.5s

weighting (stat2, standard, left middle finger) ranks correct position
Threshold = 40% (standard) [66, 18, 3, 5, 2, 3, 4, 0, 1, 1, 1, 2, 2, 3, 1, 1, 1, 1, 1, 1, 93] 66/210 = 31.43%
coh_finger [62, 9, 7, 7, 3, 3, 3, 1, 2, 2, 2, 1, 1, 1, 1, 0, 0, 0, 2, 0, 103] 62/210 = 29.52%
coh_slot [56, 11, 7, 3, 6, 0, 4, 2, 2, 2, 0, 1, 3, 0, 2, 0, 2, 3, 1, 1, 104] 56/210 = 26.67%
(coh_finger+coh_slot)/2 [63, 11, 9, 3, 5, 0, 3, 1, 0, 2, 0, 1, 1, 1, 0, 1, 2, 0, 1, 1, 105] 63/210 = 30.0%
min(coh_finger, coh_slot) [58, 11, 9, 3, 6, 2, 1, 2, 2, 2, 0, 1, 0, 2, 1, 1, 2, 0, 0, 2, 105] 58/210 = 27.62%
max(coh_finger, coh_slot) [63, 13, 5, 4, 6, 2, 2, 2, 0, 1, 0, 2, 1, 1, 1, 2, 1, 0, 0, 0, 104] 63/210 = 30.0%
coh_finger * coh_slot [61, 8, 6, 4, 6, 3, 5, 0, 1, 2, 2, 2, 1, 0, 1, 2, 2, 2, 5, 3, 94] 61/210 = 29.05%

method (192x16, mean, DF,
DF resolution 8x8)
ranks correct finger correct position time to calculate
rotation on DF level via offsets
(statistics 2.py)
[512, 183, 131, 112, 78, 85, 54, 46, 47, 29, 32, 28, 38, 35, 30, 28, 30, 33, 14, 14, 441]
out of 2000
512/2000 = 25.6% 286/512 = 55.86% 399174.9s
rotation on pixel level via image rotation
(statistics 4.py)
[814, 220, 106, 80, 70, 58, 58, 41, 45, 40, 32, 29, 26, 26, 16, 23, 17, 24, 13, 10, 252]
out of 2000
814/2000 = 40.7% 653/814 = 80.22% 887427.41s

rotation on pixel level via image rotation (statistics 4)

DF resolution
(192x16, mean)
ranks correct finger correct position time to calculate
2x2 [34, 0, 0, 1, 1, 2, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 6]
out of 46
34/46 = 73.91% 33/34 = 97.06% 3528763.91s
3x3 [139, 9, 4, 5, 7, 5, 1, 0, 0, 0, 1, 0, 1, 2, 3, 0, 3, 3, 0, 0, 26]
out of 209
139/209 = 66.51% 132/139 = 94.96% 3558083.56s
4x4 [339, 35, 9, 11, 8, 5, 6, 8, 6, 4, 4, 5, 2, 2, 3, 3, 3, 3, 3, 4, 63]
out of 526
339/526 = 64.45% 312/339 = 92.04% 2470044.16s
5x5 [728, 84, 58, 41, 31, 25, 18, 18, 19, 13, 24, 18, 20, 18, 19, 11, 5, 11, 8, 10, 237]
out of 1416
728/1416 = 51.41% 649/728 = 89.15% 3552371.32s
6x6 [979, 154, 93, 80, 54, 27, 32, 40, 41, 28, 28, 22, 15, 17, 18, 16, 13, 16, 17, 13, 297]
out of 2000
979/2000 = 48.95% 856/979 = 87.44% 2189592.92s
7x7 [772, 168, 111, 80, 61, 57, 51, 46, 43, 40, 34, 27, 20, 11, 28, 15, 26, 22, 19, 16, 353]
out of 2000
772/2000 = 38.6% 629/772 = 81.48% 1616745.98s
8x8 [814, 220, 106, 80, 70, 58, 58, 41, 45, 40, 32, 29, 26, 26, 16, 23, 17, 24, 13, 10, 252]
out of 2000
814/2000 = 40.7% 653/814 = 80.22% 887427.41s

slot size (mean, DF,
DF resolution 8x8)
ranks correct finger correct position time to calculate
192x16 [814, 220, 106, 80, 70, 58, 58, 41, 45, 40, 32, 29, 26, 26, 16, 23, 17, 24, 13, 10, 252]
out of 2000
814/2000 = 40.7% 653/814 = 80.22% 887427.41s
96x32 [1112, 189, 83, 77, 53, 49, 32, 29, 27, 22, 17, 27, 21, 19, 11, 17, 19, 11, 8, 8, 169]
out of 2000
1112/2000 = 55.6% 990/1112 = 89.03% 689535.61s
56x56 [1229, 180, 100, 59, 39, 34, 30, 26, 22, 20, 22, 18, 14, 12, 14, 14, 12, 9, 7, 3, 136]
out of 2000
1229/2000 = 61.45% 896/1229 = 72.9% 717551.22s
288x24 [1362, 118, 64, 52, 31, 30, 26, 9, 14, 12, 16, 12, 10, 9, 10, 14, 12, 3, 10, 6, 180]
out of 2000
1362/2000 = 68.1% 1145/1362 = 84.07% 1438755.87s

penalty (mean, DF,
DF resolution 8x8)
ranks correct finger correct position time to calculate
no penalty [814, 220, 106, 80, 70, 58, 58, 41, 45, 40, 32, 29, 26, 26, 16, 23, 17, 24, 13, 10, 252]
out of 2000
814/2000 = 40.7% 653/814 = 80.22% 887427.41s
penalty = 90° [791, 215, 129, 96, 71, 63, 53, 52, 33, 33, 31, 30, 30, 23, 19, 13, 18, 16, 13, 17, 254]
out of 2000
791/2000 = 39.55% 643/791 = 81.29% 641748.42s
penalty = 180° [792, 210, 107, 112, 76, 47, 48, 53, 36, 34, 36, 27, 25, 23, 20, 14, 24, 16, 15, 11, 274]
out of 2000
792/2000 = 39.6% 633/792 = 79.92% 640631.25s

stat4 192x16, mean, matchTemplate pixels:
ranks correct finger correct position time to calculate
[395, 22, 22, 14, 16, 13, 8, 5, 11, 5, 125, 13, 18, 21, 13, 12, 6, 16, 12, 8, 1245]
out of 2000
393/2000 = 19.65% 0/393 = 0.0% 769188.97s

DF resolution
(192x16, mean, speed optimised per "cut")
check all x pixel angle threshold ranks correct finger correct position time to calculate
2x2 5 5 [8, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 10]
out of 18
8/18 = 44.44% 6/8 = 75.0% 1097046.49s
8x8 5 5 [585, 158, 87, 56, 55, 28, 31, 19, 20, 24, 11, 7, 16, 20, 12, 13, 6, 4, 6, 8, 834]
out of 2000
585/2000 = 29.25% 423/585 = 72.31% 676824.44s
2x2 10 10 [15, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 3]
out of 20
15/20 = 75.0% 15/15 = 100.0% 1322482.97s
8x8 10 10 [786, 172, 125, 99, 56, 50, 50, 50, 40, 30, 27, 21, 20, 22, 21, 20, 17, 6, 16, 13, 359]
out of 2000
786/2000 = 39.3% 621/786 = 79.01% 768913.68s

Test of Independence

observed method 1 method 2
success o1s o2s
failure o1f o2f

Calculation of the expectation values (http://en.wikipedia.org/wiki/Pearson%27s_chi-square_test#Test_of_independence):
expected method 1 method 2
success e1s e2s
failure e1f e2f

e1s = (o1s + o2s)*(o1s + o1f) / (o1s + o1f + o2s + o2f)
e2s = (o2s + o1s)*(o2s + o2f) / (o1s + o1f + o2s + o2f)
e1f = (o1f + o2f)*(o1f + o1s) / (o1s + o1f + o2s + o2f)
e2f = (o2f + o2s)*(o2f + o1f) / (o1s + o1f + o2s + o2f)

Calculation of the X² statistics:
X² = (o1s - e1s)² / e1s + (o1f - e1f)² / e1f + (o2s - e2s)² / e2s + (o2f - e2f)² / e2f

degree of freedom = (2-1)*(2-1) = 1

Comparison with critical X² distribution value (http://www.itl.nist.gov/div898/handbook/eda/section3/eda3674.htm): 5% corresponds to X² = 3,841
X² > 3,841: with 95% probability one method is better than the other

Reverse Engineering

U1: unnötig (=> R1 unnötig)
Q1: nötig. Ohne Q1 wird USB-Gerät zwar erkannt, aber Kontakte auf dem Sensor werden nicht mehr registriert.
U2: Spannungswandler (5 V -> 3.32 V)
U4: unnötig (=> R5 unnötig)

Ideen für U1 und/oder U4:
  • Sicherungen, die Fehler in der Massenproduktion auffangen
  • Schalter zum Stromsparen, da Fingerprintsensor normalerweise nur kurz zum Anmelden genutzt wird, aber ständig am USB-Anschluss hängt. Besonders nützlich für mobile Geräte ohne ständigen Netzzugang

TODO

Ränder des DF?

Wie gut wird gescannter Slot an richtigem Finger gefunden? BF mit DF (BruteForce mit Directional Field), 10 Finger | 1. haufenweise Scans als Bilder speichern; 2. Bilder einlesen und auswerten
Welche Bereiche der Hand/Finger lassen sich wie gut wiedererkennen?
R-Baum mit Feature-Vektoren

SURF geht nicht? Baudisch/Holz interessiert?

Minutiae Detection -> Minutiae alignieren -> DFs vergleichen

PCA: http://www.cognotics.com/opencv/servo_2007_series/part_5/index.html
SURF: http://www.chrisevansdev.com/computer-vision-opensurf.html
Topic attachments
I Attachment Action Size Date Who Comment
Abschlussvortrag.odpodp Abschlussvortrag.odp manage 10843.2 K 02 Nov 2010 - 16:16 EricRademacher Folien des Abschlussvortrags 2.11.2010
Abschlussvortrag.pdfpdf Abschlussvortrag.pdf manage 6718.7 K 10 Nov 2010 - 19:10 EricRademacher Folien des Abschlussvortrags 2.11.2010
Antrittsvortrag.odpodp Antrittsvortrag.odp manage 700.0 K 29 Oct 2010 - 16:13 EricRademacher Folien des Antrittsvortrags 29.6.2010
MasterThesis.pdfpdf MasterThesis.pdf manage 26247.2 K 10 Nov 2010 - 19:11 EricRademacher Ausarbeitung Master Thesis Stand 10.11.2010
lD.pngpng lD.png manage 1.5 K 23 Jun 2010 - 13:48 EricRademacher Slot aus linkem Daumen (unrotiert)
lM.pngpng lM.png manage 1.9 K 23 Jun 2010 - 13:48 EricRademacher Slot aus linkem Mittelfinger (unrotiert)
lR.pngpng lR.png manage 1.7 K 23 Jun 2010 - 13:49 EricRademacher Slot aus linkem Ringfinger (unrotiert)
lZ.pngpng lZ.png manage 1.5 K 23 Jun 2010 - 13:49 EricRademacher Slot aus linkem Zeigefinger (unrotiert)
links_Daumen.pngpng links_Daumen.png manage 53.8 K 23 Jun 2010 - 13:46 EricRademacher Fingerscan linker Daumen
links_Mittelfinger.pngpng links_Mittelfinger.png manage 69.6 K 23 Jun 2010 - 13:45 EricRademacher Fingerscan linker Mittelfinger
links_Ringfinger.pngpng links_Ringfinger.png manage 55.2 K 23 Jun 2010 - 13:46 EricRademacher Fingerscan linker Ringfinger
links_Zeigefinger.pngpng links_Zeigefinger.png manage 59.1 K 23 Jun 2010 - 13:47 EricRademacher Fingerscan linker Zeigefinger
links_kleiner_Finger.pngpng links_kleiner_Finger.png manage 60.4 K 23 Jun 2010 - 13:43 EricRademacher Fingerscan linker kleiner Finger
lk.pngpng lk.png manage 1.8 K 23 Jun 2010 - 13:48 EricRademacher Slot aus linkem kleinen Finger (unrotiert)
rD.pngpng rD.png manage 1.0 K 23 Jun 2010 - 13:52 EricRademacher Slot aus rechtem Daumen (rotiert)
rM.pngpng rM.png manage 1.1 K 23 Jun 2010 - 13:53 EricRademacher Slot aus rechtem Mittelfinger (rotiert)
rR.pngpng rR.png manage 0.9 K 23 Jun 2010 - 13:54 EricRademacher Slot aus rechtem Ringfinger (rotiert)
rZ.pngpng rZ.png manage 0.9 K 23 Jun 2010 - 13:54 EricRademacher Slot aus rechtem Zeigefinger (rotiert)
rechts_Daumen.pngpng rechts_Daumen.png manage 64.2 K 23 Jun 2010 - 13:50 EricRademacher Fingerscan rechter Daumen
rechts_Mittelfinger.pngpng rechts_Mittelfinger.png manage 71.7 K 23 Jun 2010 - 13:51 EricRademacher Fingerscan rechter Mittelfinger
rechts_Ringfinger.pngpng rechts_Ringfinger.png manage 60.9 K 23 Jun 2010 - 13:51 EricRademacher Fingerscan rechter Ringfinger
rechts_Zeigefinger.pngpng rechts_Zeigefinger.png manage 61.7 K 23 Jun 2010 - 13:52 EricRademacher Fingerscan rechter Zeigefinger
rechts_kleiner_Finger.pngpng rechts_kleiner_Finger.png manage 79.6 K 23 Jun 2010 - 13:50 EricRademacher Fingerscan rechter kleiner Finger
rk.pngpng rk.png manage 1.1 K 23 Jun 2010 - 13:53 EricRademacher Slot aus rechtem kleinen Finger (rotiert)
Topic revision: r114 - 30 Nov 2010, EricRademacher
 
This site is powered by FoswikiCopyright © by the contributing authors. All material on this collaboration platform is the property of the contributing authors.
Ideas, requests, problems regarding Medieninformatik-Wiki? Send feedback