Strategize a Smooth Tenant-to-tenant Migration and Copilot Takeoff
Freni - Ph.D. Defense Slides
1. Biometric System
Template Editing
Template Replacement
Template Editing & Replacement: novel methods
for Biometric Template Selection & Update
Biagio Freni
Advisor: Prof. Fabio Roli
Pattern Recognition and Application Group
Dept. Electrical Electronic Engineering - University of Cagliari
05 March 2010
Biagio Freni Template Editing & Replacement in Biometric
3. Biometric System
Template Editing
Template Replacement
Overture
20 January 2010 . . . A man has been founded dead in a Dubai’s
hotel.
. . . couple of days later . . . Local Police discovered that 11 main
suspects got into the country illegally using forged passports of
European citizen. Police found out that pictures in the documents
were different from legitimate owner’s pictures.
. . . 14 January 2010 just a week before the Dubai affair, EU
delegates approved — 594 vs 51, while 37 abstained — the launch
of Biometric Passport including owner’s fingerprint and face.
Biagio Freni Template Editing & Replacement in Biometric
4. Biometric System Overview
Template Editing Template Representativeness
Template Replacement State-of-the-Art: Template Selection & Update
What’s Biometric?
Biometric refers to the use of physiological or behavioural
characteristics, “unique” for each person, with the aim of
established people’s identity.
Core of Biometric System is represented by Templates.
Biagio Freni Template Editing & Replacement in Biometric
5. Biometric System Overview
Template Editing Template Representativeness
Template Replacement State-of-the-Art: Template Selection & Update
Template Selection & Update
The issue of template selection and update, in biometric
recognition systems, is twofold and is related to:
Selection during Enrollment regarding the effective creation
of representative template gallery of client populations,
keeping the number of templates as small as possible at the
same time.
Update during Authentication regarding the need of adapt
over time templates, in order to capture the variations, in the
biometric traits not presented in the time of enrollment.
Selection & Update are different problems that share the common
notion of “best representative” templates.
Biagio Freni Template Editing & Replacement in Biometric
6. Biometric System Overview
Template Editing Template Representativeness
Template Replacement State-of-the-Art: Template Selection & Update
State-of-the-Art: summary
State-of-the-Art can be summurized by following modalities 1 :
Supervised: requires human intervention to labeling data.
Semi-Supervised 2 : queries labelled by the system are used
for the task.
Offline: a bunch of semi-labelled data are stored during the
system authentication, later, they are used to update system’s
templates when the system itself is not operative.
Online: each coming query is evaluated by the system during
authentication phase, template adaptation is performed online.
1
A. Rattani, B. Freni, G.L. Marcialis, F. Roli, Template Update Methods in
Adaptive Biometric Systems: A Critical Review, ICB09, pp 847-856.
2
B. Freni, G.L. Marcialis, and F. Roli, Online and offline fingerprint
template update using minutiae: an experimental comparison, AMDO08, July,
9-11, 2008, Eds., Springer LNCS 5098, pp. 441-448.
Biagio Freni Template Editing & Replacement in Biometric
7. Biometric System Overview
Template Editing Template Representativeness
Template Replacement State-of-the-Art: Template Selection & Update
PhD work
This PhD work explored the whole S-o-A and new methods have
been proposed and published:
S-o-A: Template Update Methods in Adaptive Biometric
Systems: A Critical Review, al. et Freni, ICB09.
Supervised: Template Selection by Editing Algorithms: a
case of Study in Face Recognition, Freni et al., S+SSPR08.
Semi-Supervised
Offline: Online and offline fingerprint template update using
minutiae: an experimental comparison, Freni et al., AMDO08.
Online: Replacement algorithms for fingerprint template
update, Freni et al., ICIAR08.
For sake of time just two works are addressed in this talk Editing
methods for Template Selection and Replacement algorithms for
Template Update.
Biagio Freni Template Editing & Replacement in Biometric
8. Biometric System Clustering Algorithms
Template Editing Editing Algorithms
Template Replacement Experimental Comparison
Template selection in Biometric
Problem statement Given a set of N templates for a given
person, select K templates that “best” represent the owner’s
identity.
State-of-the-Art Derived from the clustering theory,
consisting in exploring each template gallery according with
two criteria: maximum similarity among templates (MDIST),
maximum variation among them (DEND).
Main Cons
1. The procedure is not fully automatic since it requires the
manual insertion of parameter K .
2. All the template gallery are resized to the same dimension K ,
without taking into account “intrinsic” difficulty of each client.
Biagio Freni Template Editing & Replacement in Biometric
9. Biometric System Clustering Algorithms
Template Editing Editing Algorithms
Template Replacement Experimental Comparison
SoA MDIST: maximum similarity among templates
apply to all client’s gallery
1. Compute distance between
N templates
2. For each template
compute the average
distance with the other
(N − 1)
3. Choose K templates with
smallest average distance
as new selected gallery
Biagio Freni Template Editing & Replacement in Biometric
10. Biometric System Clustering Algorithms
Template Editing Editing Algorithms
Template Replacement Experimental Comparison
SoA DEND: maximum variation among templates
apply to all client’s gallery
1. Generate a NxN dissimilarity
matrix DM
2. Apply Complete Link Clustering
to DM in order to generate a
Dendrogram D, using D to
identify K clusters
3. For each K cluster select the
center
4. The set of templates selected in 3.
represent a new selected gallery
Biagio Freni Template Editing & Replacement in Biometric
11. Biometric System Clustering Algorithms
Template Editing Editing Algorithms
Template Replacement Experimental Comparison
Novel Proposal: Template Editing for Biometric
Editing Algorithms
Editing algorithms belong to the K − NN classifier theory. K − NN
use a set of prototype to perfom classification. A pattern is
classified according to the majority of “K ” prototypes close to it.
Biometric could be seen as a “1 − NN” classifier where templates
are prototypes.
Editing consist in generating from a given Template Set T a subset
E able to maintain the same classification accuracy on T itself.
Characteristics of Editing Algorithms:
1. the procedure is completly automatic
2. build up variable length galleries accordingly with the
“difficult” of each client
3. a superior generalization ability is expected
Biagio Freni Template Editing & Replacement in Biometric
12. Biometric System Clustering Algorithms
Template Editing Editing Algorithms
Template Replacement Experimental Comparison
3
CNN: Condensed Nearest Neighbour
1. E ← x1, ..., xC , C number of clients, T template set, E
edited set and x1..xC are templates randomly selected from T
2. T ← T − E
3. classify T using E
4. Y set of misclassified templates in T
5. if Y = φ then
5.1 E ← E ∪ Y
5.2 T ← T − Y
5.3 repeat from point 4
6. Stop
3
P.E. Hart, The Condensed Nearest Neighbor Rule, IEEE Transactions on
Information Theory, 14, 515-516.
Biagio Freni Template Editing & Replacement in Biometric
13. Biometric System Clustering Algorithms
Template Editing Editing Algorithms
Template Replacement Experimental Comparison
4
RNN: Reduced Nearest Neighbour
1. E ← T
2. for each x ∈ E
2.1 E ←E −x
2.2 classify T using E
2.3 Y set of misclassified templates in T
2.4 if Y = φ then
2.4.1 E ← E ∪ x
3. Stop
4
G.W. Gates, The Reduced Nearest Neighbor Rule, IEEE Transactions on
Information Theory, 18 (3) 431-433.
Biagio Freni Template Editing & Replacement in Biometric
14. Biometric System Clustering Algorithms
Template Editing Editing Algorithms
Template Replacement Experimental Comparison
Data sets, Protocol and Perfomance
Data sets
Results are carried out over Equinox, public Faces Dataset.
50 clients have been randomly choosen from the dataset. Each one
made up of 100 samples. A total of 5000 faces images.
Protocol
All the images have been grouped in two equal size sets, T and t.
T has been used as Template Set and t as a complete separated
test set to assess performance.
Performance
System’s performance has been evaluated as identification
accuracy : number of correct identified queries over total number
of submitted queries.
Results are showed as mean and (std) over six runs.
Biagio Freni Template Editing & Replacement in Biometric
15. Biometric System Clustering Algorithms
Template Editing Editing Algorithms
Template Replacement Experimental Comparison
5
Results: Accuracy
Accuracy over a test set obtained by different selection methods.
Gallery #instances×class TEST
TRAIN 50 (0) 99.62 (0.14)
CNN 7 (3) 97.6 (0.45)
SNN 4 (3) 73.66 (3.31)
RNN 17 (9) 98.43 (0.53)
ENN 49 (1) 99.35 (0.27)
MDIST 6 (0) 94.15 (0.68)
MDIST 9 (0) 96.56 (0.58)
DEND 6 (0) 89.11 (1.39)
DEND 9 (0) 94.03 (0.70)
5
B. Freni, G.L. Marcialis, and F. Roli, Template Selection by Editing
Algorithms: a case of Study in Face Recognition, S+SSPR08, Springer
LNCS5342, 755-764.
Biagio Freni Template Editing & Replacement in Biometric
17. Biometric System Clustering Algorithms
Template Editing Editing Algorithms
Template Replacement Experimental Comparison
Summary
Editing algorithms have been showed as a good alternative to the
State-of-the-Art Template Selection techniques.
Results pointed out main characteristics of Editing algorithms:
1. Completly automatic procedures, no futher intervention is
needed by supervisor.
2. Capability to build up variable length galleries, according to
client intrinsic difficulty.
3. Superior identification accuracy.
As a step futher a combined use of both techniques could be
investigated.
Biagio Freni Template Editing & Replacement in Biometric
18. Biometric System Semi-Supervised Template Update
Template Editing Template Update with Replacement
Template Replacement Results
Template Update in Biometric
Problem statement The problem is quite intuitive and
consists in making adaptive the biometric recognition systems
over the time.
Templates collected during enrollment tend to be non
representative by the time, due by the large intra-class
variation.
Performing several enrollment sessions is expensive.
State-of-the-Art
Semi-supervised paradigms exploit unlabelled samples
submitted to the system in order to find out “highly genuine”
to adapt system’s templates.
Biagio Freni Template Editing & Replacement in Biometric
19. Biometric System Semi-Supervised Template Update
Template Editing Template Update with Replacement
Template Replacement Results
S-o-A summary: Semi-Supervised Template Update
Semi-Supervised methods can be summarized by basic operations:
1. Insertion. A “highly genuine” is added into template gallery.
2. Condensing. A template gallery is “fused” in a
“super-template”.
Main Cons:
1. Sistematic use of Insertion made up long galleries. For real
systems Memory and Time of Matching are constrains.
2. Condensing absolves constrains but is less representative of
the original template galleries.
Replacement is a novel basic operation. Able to:
1. Absolve constrains of Memory and Time of Matching.
2. Assure high level of perfomance.
Biagio Freni Template Editing & Replacement in Biometric
20. Biometric System Semi-Supervised Template Update
Template Editing Template Update with Replacement
Template Replacement Results
Novel Proposal: Replacement Algorithm
T c indicates the template gallery of client c.
M is the maximum number of template slots allowed.
|T c | is the length of client’s gallery.
Replacement algorithm consists in the following steps:
for each client c = 1..C
1. x ← i, i as novel input
2. s = ms(x, T c ), matching score
3. if s > threshold, “highly genuine”
3.1 if |T c | < M then T c = T c ∪ x
3.2 else T c = replace(T c , x)
Function replace is made up according to some criteria.
Biagio Freni Template Editing & Replacement in Biometric
21. Biometric System Semi-Supervised Template Update
Template Editing Template Update with Replacement
Template Replacement Results
Replacement criteria
Random Novel template replaces one randomly chosen.
Naive Novel template replaces the one nearest to it.
FIFO Template galleries are managed as a First In First Out
queue. The new element supersede the oldest one.
LFU Template galleries are seen as a priority queue Least
Frequently Used. Less used template is substituted by novel
one.
MDIST applied to semi-supervised scenario. A new gallery is
created adding by a novel template. MDIST is applied to
pruned one element from the gallery.
DEND applied in semi-supervised scenario. A new gallery is
created adding by a novel template, then, a Dendrogram is
made up. Based on Dendrogram an element is removed from
the gallery.
Biagio Freni Template Editing & Replacement in Biometric
22. Biometric System Semi-Supervised Template Update
Template Editing Template Update with Replacement
Template Replacement Results
Data sets, Protocol and Perfomance
Data sets
Results are carried out over 12 public Fingerprint datasets. Each
one made up of 100 clients, 8 samples per client, a total of 800 of
fingerprint images for dataset.
Protocol
50 clients have been selected as system’s users. Other 50 as
impostors. For each user 3 sets have been created L, U and T. L
refers to user’s template gallery, U as unlabelled coming inputs and
T as separeted test. U contains genuine and impostors.
Performance
Equal Error Rate has been calculated over seven runs. EER
represents the error of the verification system when a number of
false acceptances is equal to a number of false rejections.
Biagio Freni Template Editing & Replacement in Biometric
24. Biometric System Semi-Supervised Template Update
Template Editing Template Update with Replacement
Template Replacement Results
6
Results: EER vs gallery dimension M
6
B. Freni, G.L. Marcialis, F. Roli, Replacement algorithms for fingerprint
template update, 5th Int. Conf. On Image Analysis and Recognition ICIAR08,
June, 25-27, 2008, Povoa de Varzim (Portugal), A. Campihlo and M. Kamel
Eds., Springer LNCS 5112, pp. 884-893.
Biagio Freni Template Editing & Replacement in Biometric
25. Biometric System Semi-Supervised Template Update
Template Editing Template Update with Replacement
Template Replacement Results
Summary
Results pointed out:
1. Less EER respect to update without replacement has been
showed.
2. Perfomance differences among replacement criteria are strong
with small “M”. Which means when strong requirements of
storing memory is a constrain.
3. MDIST outperfom other criteria, due to the fact that it
performs replacement only if it is necessary.
Biagio Freni Template Editing & Replacement in Biometric
26. Biometric System Semi-Supervised Template Update
Template Editing Template Update with Replacement
Template Replacement Results
Conclusions
Biometric plays a central role in the problem of security and
its importance is going to grow.
Template representativeness is the key for the success of a
Biometric system. Templates that “best” represent people’s
identity must be choosen during “enrollment”, as well during
the “authentication”, “highly genuine” must be detected in
the coming input queries.
Among the whole S-o-A explored in this investigation:
1. the employ of Editing for template selection during enrollment
2. the use of Replacement for template update during
authentication
Template representativeness is crucial for other important
issues in Biometric as Sensor-Interoparability in fingerprint.
This problem has been addressed too, but for sake of room
this talk was dedicated just to Editing and Replacement.
Biagio Freni Template Editing & Replacement in Biometric