ImageBio

  1. What is positive recognition?
    • To je specifičen scenarij:
    • Vzorec, ki gre v sistem je že znan sistemu.

    Ali si nekdo, ki je že znan sistemu.
  2. What is negative recognition?
    • To je specifičen scenarij:
    • Vzorec, ki gre v sistem še ni znan sistemu.
    • Ali si nekdo ki ni znan sistemu pol te passa.
  3. What is pair matching for images?
    Sistem poskuša določiti, ali se 2 sliki ujemata.
  4. What is verification?
    Sistem poskuša določiti ali se specifična oseba ujema z eno v sistemu.
  5. What is identification?
    Sistem poskuša določiti ali oseba obstaja v našem sistemu.
  6. What is negative authentication?
    • Sistem poskuša določiti ali oseba ne pripada dooločeni skupini.
    • (Black list)
  7. What is deduplication?
    Sistem poskuša določiti ali oseba že obstaja v sistemu, da je ne kreira še enkrat.
  8. What is false match rate?
    Verjetnost, da sistem vhodni vzorec klasificira napačnemu (non-match) vzorcu v bazi.

    • Procenti napačnih vhodov ki so napačno sprejeti(accepted).
    • Pomeni da ga je sistem sprejel čeprav ga nebi smel.
  9. What is false non-match rate?
    Verjetnost, da sistem ne uspe klasificirati klasificirati vhodnega vzorca.

    Procenti ustreznih vhodnih vzorcev, ki so zavrženi.
  10. What ROC stands for?
    Receiver Operating Characteristics
  11. What does CMC stands for?
    • Cumulative Match Curve.
    • A plot of rank measures for different n-s.
  12. What is a rank measure?
    Rank nam pove verjetnost da matchamo pravo osebo v n best matches produced by out system.

    Rank nam pove verjetnost da se vzorec (closest match) v bazi za katerega sistem meni, da se najbolje ujema, res pravi in da se ujema z vhodno osebo.
  13. How many ridges per centimeter on their fingerprints do people have?
    • Male: 20.7 ridges/cm
    • Female: 23.4 ridges/cm
  14. What do we observe at level 1 fingerprint features?
    Ridge orientation, coarse features at pixels.
  15. What is singular point?
    Lokacija kjer se orientacija of ridge spremeni (zavoji).
  16. What are the two basic types of singular point?
    • Loops


    • Delta
  17. What is the Core point of the fingerprint?
    • Core point corresponds to the north most loop-type singular point.
  18. What are the most common types of finger print patterns?


    • Left and right loops 65%
    • Whorls 24%
    • Twin loop 4%
    • Tended arch 3%
    • Plain arch 4%
  19. What is the resolution needed to observe L1 fingerprint features?
    250 ppi
  20. What is the resolution needed to observe L2 fingerprint features?
    500 ppi
  21. What is the resolution needed to observe L3 fingerprint features?
    1000 ppi
  22. What do we observe at level 2 fingerprint features?
    Minutiae points, location of ridge emersions and splits
  23. What are three common type of fingerprints based on how we acquire them?
    • Rolled
    • Plain
    • Latent fingerprints

  24. What do we observe at level 3 fingerprint features?
    Inner holes (sweat pores) and outer contours (edges)
  25. Which 5 sensors for obtaining fingerprints do we
    most commonly use?
    • Optical
    • Capacitance
    • Ultrasound reflection
    • Piezo electric
    • Swipe
  26. What is L2 skeletonisation?
    Proces kjer fingerprint ridges pretvorimo v 1 pixel široke črte.
  27. Which 3 techniques may we use to evaluate the quality of fingerprints?
    • Features counting:
    • Slabe slike ponavadi dajo zelo veliko ali malo število of feature points, takrat zavržemo.

    • NFIQ method:
    • Za vsak feature point (minutae) izračunaj ridge density.
    • Potem vsak blok evaluiram kvaliteto (4 = perfect quality ... 0-background)

    Pixel Intensity:
  28. How may we use feature counting to evaluate quality of the fingerprint?
    • Slabe slike ponavadi dajo zelo veliko ali malo minutae points.
    • Obdržimo le tiste odtise, ki imajo število minutae points v nekem treshold območju od do.
  29. How may we use pixel intensity method to evaluate the quality of the fingerprint?
    Use histogram
  30. Which 3 methods may we use to improve the quality of the fingerprints?
    • Hong's method
    • 1D smoothing
    • Short Time Fourier Transform approach
  31. Which level features we use in commercial systems for fingerprints?
    L1 and L2 features.
  32. Which 2 types of minutae points do we know?
    • Bifurcations
    • Takrat, ko ima ridge piksel 3 sosednje piksle

    • Endings
    • Takrat, ko ima ridge piksel 1 sosedni piksel
  33. How do we detect both types of minutae points?
    • Bifurcations:
    • Ridge neighbours == 3

    • Endings:
    • Ridge neighbours == 1
  34. How do we trace minutae direction?
    • Endings:
    • Gremo po poti nazaj recimo 10 pikslov in potegnemo premico.

    • Bifurcations:
    • obtain three points and trace direction of two points whose difference is the smallest
  35. What are the three steps of matching two fingerprints based on minutae points.
    • Aligment 
    • Transformiramo eno sliko tako, da se prilagaja drugi.

    • Correspondence:
    • Najdem pare tako da združim točke ki so v območju recimo 15 pikslov, to določimo s treshold.

    • Score generation:
    • Število paired minutae points/ Vsi minutae points iz obeh slik.
  36. What is a plain fingerprint?
    • Fingerprint obtained when a finger directly touches the sensor.
  37. What is a rolled fingerprint?
    • Fingerprint obtained when a finger is rolled from one side to the other on the sensor.
  38. What is latent fingerprint?
    • Fingerprint obtained from uncontrolled environment such as crime scene.
Author
wolf
ID
354388
Card Set
ImageBio
Description
fdsfds
Updated