21__ISSN_1392-1215_Study of Finger Vein Authentication Algorithms for Physical Access Control
Content
ELECTRO ONICS AND ND ELECTR RICAL ENG GINEERIN NG ISSN 13 392 – 1215 ELEK KTRONIKA A IR ELEKT TROTECH HNIKA
SYSTEM M ENGINEE ERING, COM MPUTER TECHNOLOGY Y
2012. No. N 5(121)
T 120 0
SISTEMŲ INŽINERIJA A, KOMPIUT TERINĖS TECHNOLOGIJ IJOS
Study S of f Finger Vein Au uthentica ation Alg gorithms s for Phy ysical Ac ccess Control C
A. A Vencka auskas, N. N Morkev vicius
Computer C Dep partment, Kau unas Universit ty of Technolo ogy, Student S ų str. 5 50, LT-51368, Kaunas, Lithuania, phone: : +370 37 300 0386, e-mails: e algim [email protected], nerijus.morke [email protected] lt
K. K Kulika auskas
Synthesis S Inc., 318 3 72nd St., B Brooklyn, New w York, USA, e-mail: kristij [email protected] http:/ //dx.doi.org/10 0.5755/j01.eee e.121.5.1660 Introduction I The purp pose of any physical secu urity system is to permit p only authorized users to access restr ricted There are various objects/areas. o v exam mples of phy ysical security: s territ tory, room, au utomobile, etc. Physical sec curity rain, appliances a o operate in complex conditions: c temperature t fl luctuations, dust, d dirt, etc. Users of sec curity systems s encom mpass a vast range r of peopl le; they can b be old or o young and d have varying g degrees of ability to op perate such s systems. Security sy ystems must be able to both control c access to objects/are eas and also to o save and co ontain data d relevant t to the ac ccess of the ese objects/a areas. Therefore, T th hese systems have to be smart, rel liable (unbreakable), ( , simple to use, u resistant to environm mental impacts and ch heap (not requ uiring a lot of resources). Biometric B aut thentication methods m for access a contro ol Various user authent tication meth hods are use ed in contemporary c electronic se ecurity system ms [1, 2]: ce ertain codes c – passw word, persona al identificatio on number (P PIN); trical electronic e car rds – magn netic, smart-c cards; biomet methods m – phy ysical user par rameters: fing gerprint, face, hand iting. shape, s eye ir ris, vein image; behaviou ur: voice, wri Systems S that o operate by us sing password ds or PIN num mbers require r the in ndividual to memorize m the e code. How wever, such s systems cannot determ mine the pers son who ente ers it. Systems S that u use cards for r access contr rol only deter rmine card c authentic city, but not the authentic city of the pe erson who w presents it. Therefor re, systems ba ased on code e and card c usage ca annot ensure high object security. Va arious biotronic b mea asures are used in contem mporary electr tronic systems s [3]. B Biometric dev vices verify who w a person is by what w they are, , whether it is s their hand, eye, fingerpri int or 101 voic ce [4, 21]. Biometrics can also eliminat te the need fo or card ds. Growth of the biometric c industry also o highlights th he pros spect of biom metric systems s: the market t for biometri ic core e technology will w increase f from $2,584 billion b in 2009 9, to $10,882 billion n in 2017, a forecasted ye early growth of o 19.6 69% [5]. Biom metrics will b be a fundame ental embedde ed com mponent of th he digital wo orld, as it becomes a ke ey enab bler of trusted transaction control – da ata access an nd flow w - for persona al, commercia al and governm ment use. Distribution of biometric cal technologies is shown in i Fig. 1 [6].
Other O Mod dalities; 1,6% 1 AFI IS/LiveScan n; 38,3% Iris n; Recognition 5,1% Hand Geometry y; 1,8% Vein V reco ognition; 2,4% 2 Voice on; recognitio 3,0% Middleware; 8,0% Fingerprint; 28,4%
Face n Recognition ; 11,4%
Fig. 1. Biometric Revenues by y Technology, 2009 (© 200 08 Inter rnational Biome etric Group)
Finger vein authenticatio on thus offer rs considerabl ly more advantages compared to o other forms of biometric cs. Thes se comparativ ve advantages s are collectively shown in i Tabl le 1 [7].
Table T 1. Compa arison of major r biometrics met thods
SECURITY BIOMETRICS B
AntiAccuracy Forgery Speed
CONV VENIENCE
e ResisEnrollme nt rates tance Cost Size
Fingerprint F O O Ir ris O O Face F O O Voice V O O Vein V Pattern Note: N : good, O: normal, : insu ufficient
O O O O
O
O O
O O
finger vein recog gnition. We w will study thr ree algorithm ms. Abra aham et al. pr roposed a meth hod for finger rprint matchin ng usin ng a hybrid shape and orien ntation descrip ptor (Algorithm m A1) [16]. Jain et al. proposed d filter-based algorithm, tha at uses s a bank of Gabor filters to capture both b local an nd glob bal details in a fingerprint t as a compac ct fixed lengt th Fing gerCode (Alg gorithm A2) [17]. Ng et al. analyzed a nove el fingerprint feature named d adjacent ori ientation vecto or for fingerprint f ma atching (Algor rithm A3) [18]. Exp periments During the experiment we explor red fingerprin nt reco ognition metho od suitability y for finger ve ein recognitio on as well w as how image resolu ution and no oise influence es resu ults. We used images i from t the experimen nt of Intelligen nt Biom metric Group [19]. In orde er to imitate different d finge er vein n scans, we al ltered the hue e, contrast and d brightness of o imag ges with imag ge editing soft ftware, so that t we obtained 5 imag ges of each finger. f To im mitate environm ment influenc ce such h as dirt or the illumination n of surroundi ings, we adde ed diffe erent levels of f noise with im mage editing software, s usin ng Perlin noise funct tion [20], resp pectively: leve el 1 - 0%, level 2 - 5%, level 3 - 10% and level 4 - 20 0%. To imitat te diffe erent resolutio ons of a CCD D camera, we used series of o diffe erent resolutio on images fo or the experim ment: level 1 180 x 60 pixels, level l 2 - 150 x 50 pixels, level l 3 - 120 x 40 pixels, p level 4 – 90 x 30 pix xels. By using g these method ds we had h 16 series, each made u up of 20 imag ges. Researche ed finger vein image examples are e presented in Fig. 2–Fig. 4.
The Rile ey‘s et al. evaluation [8] suggests s that vein technology t is s more acces ssible for an older popul lation using u than fin ngerprint techn nology. Using g fingerprint b based technologies t f for physical access is problematic for r the following f rea asons: environ nmental cond ditions (dust, dirt, temperature t fl luctuation), fi ingerprint ima age quality is low, they t can be fo orged and it can be more co omplicated fo or the elderly. e ger vein patter rn based auth hentication me ethod The fing is highly reliable, the vei ins are hidde en underneath h the skin's s surface so forgery is s extremely difficult; d it is noninvasive and e easy to use, of ffering a balan nce of advant tages. These T unique aspects of fin nger vein patte ern recognitio on set it apart from p previous forms s of biometrics [7]. Let’s exa amine finger vein v pattern usage u problem ms. A finger f vein im mage is obtaine ed using an in nfrared scanne er, so its quality, de epending on environment e can c be erratic c and sometimes s po oor. Quality y parameters of authentic cation system, s false acceptance and a false reje ection [9], hi highly depend d on fin nger vein ima age processin ng algorithms s and their t resistance e to noise i.e. image quality y. Importan nt issues for bi iometric syste ems are resear rch of efficiency e [10 0]. NIST perf formed compr rehensive ana alysis of o fingerprint t solutions (technologies s) [11]. Eigh hteen different d com mpanies comp peted, and 34 3 systems were evaluated. e Th here is a lack of o finger vein n based techno ology analysis. a Finger F vein re ecognition alg gorithm Finger v vein recognition – a rela atively embry yonic field, f new me ethods are de eveloped and existing one s are examined. e Mi iura et al. pro oposed a met thod of extra acting finger vein pa atterns by usi ing repeated line l tracking from various v startin ng positions [12]. [ Zhang et e al. propose ed an extraction e method based on o curve let information o of the pro p file of fing ger vein imag ges and locally interconne ected structured s neu ural networks [13]. Mahri et al. propose ed an algorithm a for finger vein re ecognition wit th less compl lexity in the image e preprocessing phase, where w finger vein pattern p extraction is not in ncluded at all l. In the prop posed algorithm, a th hey impleme ent a phase e-only correl lation function f at t the matching g stage with h a very si imple preprocessing p technique [14]. All of the previo ously described d meth hods use the features f from the extracted d vein patterns p for re ecognition. However, finge er vein image es are not n always cl n show irregu ular shadings s and lear and can highly h saturat ted regions. Therefore, T detection errors s can occur o when ex xtracting accur rate vein patte erns. Well-dev veloped finger rprint recognition methods have r, we been b used for many years [15, [ 16, 17 an nd 18]. Further will w research f fingerprint rec cognition met thod suitabilit ty for 102
Fig. 2. Unmodified image with ori iginal resolution n
olution image w with added noise Fig. 3. Original reso
lution image wi with added noise e Fig. 4. Lowest resol
Simulation was perfor rmed in Matlab M 2007 7b envi ironment, usin ng three algori ithms – A1, A2 A and A3. Experiment- we compared d each image from all serie es with h remaining same series images and d recorded th he resu ults: true ac cceptance, f false rejectio on and fals se acce eptance. Resul lts are present ted in Table 2. The true acceptance rate e’s dependenc ce on differen nt reso olution, using different algo orithms witho out Perlin nois se are presented in Fig. 5. As w we can see, we get the most stable results from m Algorithm A A1.
Table 2. Results of finger vein recognition algorithm research
Algorithm A1 Resolution Perlin Noise True Level Level Acceptance Rate, % False Rejection Rate, % False True Acceptance Acceptance Rate, % Rate, % Algorithm A2 False Rejection Rate, % False True Acceptance Acceptance Rate, % Rate, % Algorithm A3 False Rejection Rate, % False Acceptance Rate, %
1 1 1 1 2 2 2 2 3 3 3 3 4 4 4 4 100 80 60 40 20 0 1
1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4
98 91 58 56 91 84 81 79 92 91 87 86 93 95 86 94
0 7 40 44 9 16 16 21 7 7 12 14 7 0 9 5
2 2 2 0 0 0 3 0 1 2 1 0 0 5 5 1
90 55 5 0 90 70 35 5 87 76 69 62 85 95 90 95
10 40 90 100 10 25 55 90 13 18 23 38 15 5 10 5
0 5 5 0 0 5 10 5 0 6 8 0 0 0 0 0
100 100 100 100 95 95 85 75 83 82 70 53 60 60 60 40
0 0 0 0 5 5 15 15 17 18 30 35 40 40 40 40
0 0 0 0 0 0 0 10 0 0 0 12 0 0 0 20
Table 3. Overall algorithm assessment A1 Resolution True Acceptance False Rejection False Acceptance Amount Perlin Noise True Acceptance False Rejection False Acceptance Amount Total 3 2 2 7 3 2 3 8 15 Algorithm A2 3 1 3 7 1 1 1 3 10 A3 2 3 1 6 3 2 2 7 13
2
Algorithm A1 Algorithm A3
3
4
Algorithm A2
Fig. 5. True acceptance rate’s dependence on different resolution using different algorithms
The true acceptance rate’s dependence on Perlin noise level, using different algorithms with resolution level 1 are presented in Fig. 6. As we can seem we get the most stable results from algorithm A3. 100 80 60 40 20 0 1 2
Algorithm A1 Algorithm A3
We summarized our results in a table, assessing results of algorithms: 3 – good, 2 – average, 1 – poor (Table 3), using level 4 resolution and level 4 Perlin noise. It is clear that in the conditions of our experiment we get the best results from algorithm A1. As we can see from Table 3, when image resolution is decreasing and noise is increasing, we get the best results from algorithm A1 - method for fingerprint matching using a hybrid shape and orientation descriptor. Using primary image resolution and all noise levels, we get best results from algorithm A3. Algorithm A2 is very sensitive to noise and as it increases, false acceptance rate highly increases. Conclusions
3
Algorithm A2
4
Fig. 6. True acceptance rate’s dependence on Perlin noise, using different algorithms
Biometrical access control methods are increasingly being used in physical security systems. Finger vein authentication methods appear promising for the future when assessing reliability and user-friendliness. In this study, we researched the possibilities of using fingerprint recognition methods for finger vein recognition.
103
Experiments showed that the method for fingerprint matching using a hybrid shape and orientation descriptor from these researched algorithms for physical access systems suits most. This method is least sensitive to finger vein image quality – the true acceptance rate is 98% with primary quality and it only drops to 56% with the worst quality. Such indicators satisfy practical extreme conditions. References
1. Anderson Ross J. Security Engineering: A Guide to Building Dependable Distributed Systems. – John Wiley and Sons, 2010. – 1080 p. 2. O’Gorman L. Comparing passwords, tokens, and biometrics for user authentication // Proceedings of the IEEE, 2003. – Vol. 91. – No. 12. – P. 2019–2040. 3. Balaišis P., Eidukas D., Keras E., Valinevičius A. The Selection of Biotronics Measures // Electronics and Electrical Engineering. – Kaunas: Technologija, 2010. – No. 1(97). – P. 9–14. 4. Ivanovas E., Navakauskas D.. Development of Biometric Systems for Person Recognition: Biometric Feature Systems, Traits and Acquisition // Electronics and Electrical Engineering. – Kaunas: Technologija, 2010. – No. 5(101). – P. 87–90. 5. Acuity Market Intelligence. The Future of Biometrics. Market Analysis, Segmentation & Forecasts. Insights into the Trends, Drivers & Opportunities that will Shape the Industry through 2020. – 2009. Online: http://www.acuity– mi.com/Future_of_Biometrics.html. 6. International Biometric Group. Biometrics Market and Industry Report 2009–2014, December 2009. // http://www.ibgweb.com/products/reports/bmir–2009–2014. 7. Hashimoto J. Finger Vein Authentication Technology and Its Future // VLSI Circuits, 2006. – Digest of Technical Papers. – P. 5–8. 8. Riley C., McCracken H., Buckner K. Fingers, veins and the grey pound: accessibility of biometric technology // Proceedings of the 14th European conference on Cognitive ergonomics (ECCE’07). – New York, NY, USA, 2007. – P. 149–152. 9. Jain A. K., Pankanti S., Prabhakar S., Lin Hong, Ross A. Biometrics: a grand challenge // Proceedings of the 17th
10.
11. 12.
13.
14.
15.
16. 17. 18.
19. 20. 21.
International Conference (ICPR’2004), 2004. – Vol. 2. – P. 935–942. Balaišis P, Gužauskas R., Eidukas D., Valinevičius A. Directions of Efficiency Research for Biotronics Technologies // Electronics and Electrical Engineering. – Kaunas: Technologija, 2009. – No. 1(89). – P. 49–54. Wilson C., Hicklin R. A., Korves H., et al. Finger–print Vendor Technology Evaluation 2003. – Summary of Results and Analysis Report, NIST Research Report: 7123, 2004. Miura N., Nagasaka A., Miyatake T. Feature extraction of finger–vein patterns based on repeated line tracking and its application to personal identification // Machine Vision and Applications, 2004. – Vol. 15. – No. 4. – P. 194–203. Zhang Z., Ma S., Han X. Multiscale Feature Extraction of Finger–Vein Patterns Based on Curvelets and Local Interconnection Structure Neural Network // Pattern Recognition, 2006. – 18th International Conference ICPR’2006. – Vol. 4. – P. 145–148. Mahri N., Suandi S. A. S., Rosdi B. A. Finger Vein Recognition Algorithm Using Phase Only Correlation // Emerging Techniques and Challenges for Hand–Based Biometrics (ETCHB), 2010. – P. 1–6. Kisel A., Kochetkov A., Kranauskas J. Fingerprint Minutiae Matching without Global Alignment Using Local Structures // Informatica, 2008. – Vol. 19. – No. 1. – P. 31– 44. Abraham J., Kwan P., Junbin Gao. Fingerprint Matching using A Hybrid Shape and Orientation Descriptor // InTech, 2011. – P. 25–56. Jain A. K., Prabhakar S., Hong L., Pankanti S. Filterbank– based fingerprint matching // Image Processing, IEEE Transactions, 2000. – Vol. 9. – No. 5. – P. 846–859. Ng G. S., Tong X., Tang X., Shi D. Adjacent orientation vector based fingerprint minutiae matching system // Pattern Recognition, 2004. – Proceedings of the 17th International Conference ICPR’2004. – Vol. 1. – P. 528–531. USM–IBG. Finger vein biometric system. Online: http://ibg– usm.org. Perlin K. Improving noise // ACM Transactions on Graphics (TOG), 2002. – Proceedings of ACM SIGGRAPH’2002. – Vol. 21. – Iss. 3. – San Antonio, Texas, 2002. – P. 681–682. Wang F., Han J. Robust Multimodal Biometric Authentication Integrating Iris, Face and Palmprint // Information Technology And Control. – Kaunas: Technologija, 2008. – Vol. 37. – No. 4. – P. 326–332. Received 2012 01 21 Accepted after revision 2012 03 19
A. Venckauskas, N. Morkevicius, K. Kulikauskas. Study of Finger Vein Authentication Algorithms for Physical Access Control // Electronics and Electrical Engineering. – Kaunas: Technologija, 2012. – No. 5(121). – P. 101–104. In this study problems of using biometric authentication methods for physical access systems are analyzed. Possibilities for applying biometric authentication methods are also examined. It is suggested that fingerprint recognition methods could be applied to finger vein recognition. The experiment shows that method for fingerprint matching using hybrid shape and orientation descriptor finger vein recognition could be used in finger vein based physical access systems. Finger vein image quality has low impact for this method. Ill. 6, bibl. 21, tabl. 3 (in English; abstracts in English and Lithuanian). A. Venčkauskas, N. Morkevičius, K. Kulikauskas. Pirštų kraujagyslių tapatumo nustatymo algoritmų taikymo fizinės apsaugos sistemose tyrimas // Elektronika ir elektrotechnika. – Kaunas: Technologija, 2012. – Nr. 5(121). – P. 101–104. Darbe sprendžiamos biometrinių tapatumo nustatymo metodų taikymo fizinės apsaugos sistemose problemos. Išnagrinėtos biometrinių tapatumo nustatymo metodų taikymo galimybės. Pasiūlyta pirštų atspaudų atpažinimo metodus pritaikyti pirštų kraujagyslėms atpažinti. Eksperimentiškai parodyta, kad pirštų kraujagyslėms atpažinti fizinės saugos sistemose galima taikyti figūros konteksto ir orientacijos deskriptoriais atvaizdų atpažinimo principu veikiantį metodą. Šis metodas nelabai jautrus kraujagyslių atspaudų kokybei. Il. 6, bibl. 21, lent. 3 (anglų kalba; santraukos anglų ir lietuvių k.).
104
SYSTEM M ENGINEE ERING, COM MPUTER TECHNOLOGY Y
2012. No. N 5(121)
T 120 0
SISTEMŲ INŽINERIJA A, KOMPIUT TERINĖS TECHNOLOGIJ IJOS
Study S of f Finger Vein Au uthentica ation Alg gorithms s for Phy ysical Ac ccess Control C
A. A Vencka auskas, N. N Morkev vicius
Computer C Dep partment, Kau unas Universit ty of Technolo ogy, Student S ų str. 5 50, LT-51368, Kaunas, Lithuania, phone: : +370 37 300 0386, e-mails: e algim [email protected], nerijus.morke [email protected] lt
K. K Kulika auskas
Synthesis S Inc., 318 3 72nd St., B Brooklyn, New w York, USA, e-mail: kristij [email protected] http:/ //dx.doi.org/10 0.5755/j01.eee e.121.5.1660 Introduction I The purp pose of any physical secu urity system is to permit p only authorized users to access restr ricted There are various objects/areas. o v exam mples of phy ysical security: s territ tory, room, au utomobile, etc. Physical sec curity rain, appliances a o operate in complex conditions: c temperature t fl luctuations, dust, d dirt, etc. Users of sec curity systems s encom mpass a vast range r of peopl le; they can b be old or o young and d have varying g degrees of ability to op perate such s systems. Security sy ystems must be able to both control c access to objects/are eas and also to o save and co ontain data d relevant t to the ac ccess of the ese objects/a areas. Therefore, T th hese systems have to be smart, rel liable (unbreakable), ( , simple to use, u resistant to environm mental impacts and ch heap (not requ uiring a lot of resources). Biometric B aut thentication methods m for access a contro ol Various user authent tication meth hods are use ed in contemporary c electronic se ecurity system ms [1, 2]: ce ertain codes c – passw word, persona al identificatio on number (P PIN); trical electronic e car rds – magn netic, smart-c cards; biomet methods m – phy ysical user par rameters: fing gerprint, face, hand iting. shape, s eye ir ris, vein image; behaviou ur: voice, wri Systems S that o operate by us sing password ds or PIN num mbers require r the in ndividual to memorize m the e code. How wever, such s systems cannot determ mine the pers son who ente ers it. Systems S that u use cards for r access contr rol only deter rmine card c authentic city, but not the authentic city of the pe erson who w presents it. Therefor re, systems ba ased on code e and card c usage ca annot ensure high object security. Va arious biotronic b mea asures are used in contem mporary electr tronic systems s [3]. B Biometric dev vices verify who w a person is by what w they are, , whether it is s their hand, eye, fingerpri int or 101 voic ce [4, 21]. Biometrics can also eliminat te the need fo or card ds. Growth of the biometric c industry also o highlights th he pros spect of biom metric systems s: the market t for biometri ic core e technology will w increase f from $2,584 billion b in 2009 9, to $10,882 billion n in 2017, a forecasted ye early growth of o 19.6 69% [5]. Biom metrics will b be a fundame ental embedde ed com mponent of th he digital wo orld, as it becomes a ke ey enab bler of trusted transaction control – da ata access an nd flow w - for persona al, commercia al and governm ment use. Distribution of biometric cal technologies is shown in i Fig. 1 [6].
Other O Mod dalities; 1,6% 1 AFI IS/LiveScan n; 38,3% Iris n; Recognition 5,1% Hand Geometry y; 1,8% Vein V reco ognition; 2,4% 2 Voice on; recognitio 3,0% Middleware; 8,0% Fingerprint; 28,4%
Face n Recognition ; 11,4%
Fig. 1. Biometric Revenues by y Technology, 2009 (© 200 08 Inter rnational Biome etric Group)
Finger vein authenticatio on thus offer rs considerabl ly more advantages compared to o other forms of biometric cs. Thes se comparativ ve advantages s are collectively shown in i Tabl le 1 [7].
Table T 1. Compa arison of major r biometrics met thods
SECURITY BIOMETRICS B
AntiAccuracy Forgery Speed
CONV VENIENCE
e ResisEnrollme nt rates tance Cost Size
Fingerprint F O O Ir ris O O Face F O O Voice V O O Vein V Pattern Note: N : good, O: normal, : insu ufficient
O O O O
O
O O
O O
finger vein recog gnition. We w will study thr ree algorithm ms. Abra aham et al. pr roposed a meth hod for finger rprint matchin ng usin ng a hybrid shape and orien ntation descrip ptor (Algorithm m A1) [16]. Jain et al. proposed d filter-based algorithm, tha at uses s a bank of Gabor filters to capture both b local an nd glob bal details in a fingerprint t as a compac ct fixed lengt th Fing gerCode (Alg gorithm A2) [17]. Ng et al. analyzed a nove el fingerprint feature named d adjacent ori ientation vecto or for fingerprint f ma atching (Algor rithm A3) [18]. Exp periments During the experiment we explor red fingerprin nt reco ognition metho od suitability y for finger ve ein recognitio on as well w as how image resolu ution and no oise influence es resu ults. We used images i from t the experimen nt of Intelligen nt Biom metric Group [19]. In orde er to imitate different d finge er vein n scans, we al ltered the hue e, contrast and d brightness of o imag ges with imag ge editing soft ftware, so that t we obtained 5 imag ges of each finger. f To im mitate environm ment influenc ce such h as dirt or the illumination n of surroundi ings, we adde ed diffe erent levels of f noise with im mage editing software, s usin ng Perlin noise funct tion [20], resp pectively: leve el 1 - 0%, level 2 - 5%, level 3 - 10% and level 4 - 20 0%. To imitat te diffe erent resolutio ons of a CCD D camera, we used series of o diffe erent resolutio on images fo or the experim ment: level 1 180 x 60 pixels, level l 2 - 150 x 50 pixels, level l 3 - 120 x 40 pixels, p level 4 – 90 x 30 pix xels. By using g these method ds we had h 16 series, each made u up of 20 imag ges. Researche ed finger vein image examples are e presented in Fig. 2–Fig. 4.
The Rile ey‘s et al. evaluation [8] suggests s that vein technology t is s more acces ssible for an older popul lation using u than fin ngerprint techn nology. Using g fingerprint b based technologies t f for physical access is problematic for r the following f rea asons: environ nmental cond ditions (dust, dirt, temperature t fl luctuation), fi ingerprint ima age quality is low, they t can be fo orged and it can be more co omplicated fo or the elderly. e ger vein patter rn based auth hentication me ethod The fing is highly reliable, the vei ins are hidde en underneath h the skin's s surface so forgery is s extremely difficult; d it is noninvasive and e easy to use, of ffering a balan nce of advant tages. These T unique aspects of fin nger vein patte ern recognitio on set it apart from p previous forms s of biometrics [7]. Let’s exa amine finger vein v pattern usage u problem ms. A finger f vein im mage is obtaine ed using an in nfrared scanne er, so its quality, de epending on environment e can c be erratic c and sometimes s po oor. Quality y parameters of authentic cation system, s false acceptance and a false reje ection [9], hi highly depend d on fin nger vein ima age processin ng algorithms s and their t resistance e to noise i.e. image quality y. Importan nt issues for bi iometric syste ems are resear rch of efficiency e [10 0]. NIST perf formed compr rehensive ana alysis of o fingerprint t solutions (technologies s) [11]. Eigh hteen different d com mpanies comp peted, and 34 3 systems were evaluated. e Th here is a lack of o finger vein n based techno ology analysis. a Finger F vein re ecognition alg gorithm Finger v vein recognition – a rela atively embry yonic field, f new me ethods are de eveloped and existing one s are examined. e Mi iura et al. pro oposed a met thod of extra acting finger vein pa atterns by usi ing repeated line l tracking from various v startin ng positions [12]. [ Zhang et e al. propose ed an extraction e method based on o curve let information o of the pro p file of fing ger vein imag ges and locally interconne ected structured s neu ural networks [13]. Mahri et al. propose ed an algorithm a for finger vein re ecognition wit th less compl lexity in the image e preprocessing phase, where w finger vein pattern p extraction is not in ncluded at all l. In the prop posed algorithm, a th hey impleme ent a phase e-only correl lation function f at t the matching g stage with h a very si imple preprocessing p technique [14]. All of the previo ously described d meth hods use the features f from the extracted d vein patterns p for re ecognition. However, finge er vein image es are not n always cl n show irregu ular shadings s and lear and can highly h saturat ted regions. Therefore, T detection errors s can occur o when ex xtracting accur rate vein patte erns. Well-dev veloped finger rprint recognition methods have r, we been b used for many years [15, [ 16, 17 an nd 18]. Further will w research f fingerprint rec cognition met thod suitabilit ty for 102
Fig. 2. Unmodified image with ori iginal resolution n
olution image w with added noise Fig. 3. Original reso
lution image wi with added noise e Fig. 4. Lowest resol
Simulation was perfor rmed in Matlab M 2007 7b envi ironment, usin ng three algori ithms – A1, A2 A and A3. Experiment- we compared d each image from all serie es with h remaining same series images and d recorded th he resu ults: true ac cceptance, f false rejectio on and fals se acce eptance. Resul lts are present ted in Table 2. The true acceptance rate e’s dependenc ce on differen nt reso olution, using different algo orithms witho out Perlin nois se are presented in Fig. 5. As w we can see, we get the most stable results from m Algorithm A A1.
Table 2. Results of finger vein recognition algorithm research
Algorithm A1 Resolution Perlin Noise True Level Level Acceptance Rate, % False Rejection Rate, % False True Acceptance Acceptance Rate, % Rate, % Algorithm A2 False Rejection Rate, % False True Acceptance Acceptance Rate, % Rate, % Algorithm A3 False Rejection Rate, % False Acceptance Rate, %
1 1 1 1 2 2 2 2 3 3 3 3 4 4 4 4 100 80 60 40 20 0 1
1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4
98 91 58 56 91 84 81 79 92 91 87 86 93 95 86 94
0 7 40 44 9 16 16 21 7 7 12 14 7 0 9 5
2 2 2 0 0 0 3 0 1 2 1 0 0 5 5 1
90 55 5 0 90 70 35 5 87 76 69 62 85 95 90 95
10 40 90 100 10 25 55 90 13 18 23 38 15 5 10 5
0 5 5 0 0 5 10 5 0 6 8 0 0 0 0 0
100 100 100 100 95 95 85 75 83 82 70 53 60 60 60 40
0 0 0 0 5 5 15 15 17 18 30 35 40 40 40 40
0 0 0 0 0 0 0 10 0 0 0 12 0 0 0 20
Table 3. Overall algorithm assessment A1 Resolution True Acceptance False Rejection False Acceptance Amount Perlin Noise True Acceptance False Rejection False Acceptance Amount Total 3 2 2 7 3 2 3 8 15 Algorithm A2 3 1 3 7 1 1 1 3 10 A3 2 3 1 6 3 2 2 7 13
2
Algorithm A1 Algorithm A3
3
4
Algorithm A2
Fig. 5. True acceptance rate’s dependence on different resolution using different algorithms
The true acceptance rate’s dependence on Perlin noise level, using different algorithms with resolution level 1 are presented in Fig. 6. As we can seem we get the most stable results from algorithm A3. 100 80 60 40 20 0 1 2
Algorithm A1 Algorithm A3
We summarized our results in a table, assessing results of algorithms: 3 – good, 2 – average, 1 – poor (Table 3), using level 4 resolution and level 4 Perlin noise. It is clear that in the conditions of our experiment we get the best results from algorithm A1. As we can see from Table 3, when image resolution is decreasing and noise is increasing, we get the best results from algorithm A1 - method for fingerprint matching using a hybrid shape and orientation descriptor. Using primary image resolution and all noise levels, we get best results from algorithm A3. Algorithm A2 is very sensitive to noise and as it increases, false acceptance rate highly increases. Conclusions
3
Algorithm A2
4
Fig. 6. True acceptance rate’s dependence on Perlin noise, using different algorithms
Biometrical access control methods are increasingly being used in physical security systems. Finger vein authentication methods appear promising for the future when assessing reliability and user-friendliness. In this study, we researched the possibilities of using fingerprint recognition methods for finger vein recognition.
103
Experiments showed that the method for fingerprint matching using a hybrid shape and orientation descriptor from these researched algorithms for physical access systems suits most. This method is least sensitive to finger vein image quality – the true acceptance rate is 98% with primary quality and it only drops to 56% with the worst quality. Such indicators satisfy practical extreme conditions. References
1. Anderson Ross J. Security Engineering: A Guide to Building Dependable Distributed Systems. – John Wiley and Sons, 2010. – 1080 p. 2. O’Gorman L. Comparing passwords, tokens, and biometrics for user authentication // Proceedings of the IEEE, 2003. – Vol. 91. – No. 12. – P. 2019–2040. 3. Balaišis P., Eidukas D., Keras E., Valinevičius A. The Selection of Biotronics Measures // Electronics and Electrical Engineering. – Kaunas: Technologija, 2010. – No. 1(97). – P. 9–14. 4. Ivanovas E., Navakauskas D.. Development of Biometric Systems for Person Recognition: Biometric Feature Systems, Traits and Acquisition // Electronics and Electrical Engineering. – Kaunas: Technologija, 2010. – No. 5(101). – P. 87–90. 5. Acuity Market Intelligence. The Future of Biometrics. Market Analysis, Segmentation & Forecasts. Insights into the Trends, Drivers & Opportunities that will Shape the Industry through 2020. – 2009. Online: http://www.acuity– mi.com/Future_of_Biometrics.html. 6. International Biometric Group. Biometrics Market and Industry Report 2009–2014, December 2009. // http://www.ibgweb.com/products/reports/bmir–2009–2014. 7. Hashimoto J. Finger Vein Authentication Technology and Its Future // VLSI Circuits, 2006. – Digest of Technical Papers. – P. 5–8. 8. Riley C., McCracken H., Buckner K. Fingers, veins and the grey pound: accessibility of biometric technology // Proceedings of the 14th European conference on Cognitive ergonomics (ECCE’07). – New York, NY, USA, 2007. – P. 149–152. 9. Jain A. K., Pankanti S., Prabhakar S., Lin Hong, Ross A. Biometrics: a grand challenge // Proceedings of the 17th
10.
11. 12.
13.
14.
15.
16. 17. 18.
19. 20. 21.
International Conference (ICPR’2004), 2004. – Vol. 2. – P. 935–942. Balaišis P, Gužauskas R., Eidukas D., Valinevičius A. Directions of Efficiency Research for Biotronics Technologies // Electronics and Electrical Engineering. – Kaunas: Technologija, 2009. – No. 1(89). – P. 49–54. Wilson C., Hicklin R. A., Korves H., et al. Finger–print Vendor Technology Evaluation 2003. – Summary of Results and Analysis Report, NIST Research Report: 7123, 2004. Miura N., Nagasaka A., Miyatake T. Feature extraction of finger–vein patterns based on repeated line tracking and its application to personal identification // Machine Vision and Applications, 2004. – Vol. 15. – No. 4. – P. 194–203. Zhang Z., Ma S., Han X. Multiscale Feature Extraction of Finger–Vein Patterns Based on Curvelets and Local Interconnection Structure Neural Network // Pattern Recognition, 2006. – 18th International Conference ICPR’2006. – Vol. 4. – P. 145–148. Mahri N., Suandi S. A. S., Rosdi B. A. Finger Vein Recognition Algorithm Using Phase Only Correlation // Emerging Techniques and Challenges for Hand–Based Biometrics (ETCHB), 2010. – P. 1–6. Kisel A., Kochetkov A., Kranauskas J. Fingerprint Minutiae Matching without Global Alignment Using Local Structures // Informatica, 2008. – Vol. 19. – No. 1. – P. 31– 44. Abraham J., Kwan P., Junbin Gao. Fingerprint Matching using A Hybrid Shape and Orientation Descriptor // InTech, 2011. – P. 25–56. Jain A. K., Prabhakar S., Hong L., Pankanti S. Filterbank– based fingerprint matching // Image Processing, IEEE Transactions, 2000. – Vol. 9. – No. 5. – P. 846–859. Ng G. S., Tong X., Tang X., Shi D. Adjacent orientation vector based fingerprint minutiae matching system // Pattern Recognition, 2004. – Proceedings of the 17th International Conference ICPR’2004. – Vol. 1. – P. 528–531. USM–IBG. Finger vein biometric system. Online: http://ibg– usm.org. Perlin K. Improving noise // ACM Transactions on Graphics (TOG), 2002. – Proceedings of ACM SIGGRAPH’2002. – Vol. 21. – Iss. 3. – San Antonio, Texas, 2002. – P. 681–682. Wang F., Han J. Robust Multimodal Biometric Authentication Integrating Iris, Face and Palmprint // Information Technology And Control. – Kaunas: Technologija, 2008. – Vol. 37. – No. 4. – P. 326–332. Received 2012 01 21 Accepted after revision 2012 03 19
A. Venckauskas, N. Morkevicius, K. Kulikauskas. Study of Finger Vein Authentication Algorithms for Physical Access Control // Electronics and Electrical Engineering. – Kaunas: Technologija, 2012. – No. 5(121). – P. 101–104. In this study problems of using biometric authentication methods for physical access systems are analyzed. Possibilities for applying biometric authentication methods are also examined. It is suggested that fingerprint recognition methods could be applied to finger vein recognition. The experiment shows that method for fingerprint matching using hybrid shape and orientation descriptor finger vein recognition could be used in finger vein based physical access systems. Finger vein image quality has low impact for this method. Ill. 6, bibl. 21, tabl. 3 (in English; abstracts in English and Lithuanian). A. Venčkauskas, N. Morkevičius, K. Kulikauskas. Pirštų kraujagyslių tapatumo nustatymo algoritmų taikymo fizinės apsaugos sistemose tyrimas // Elektronika ir elektrotechnika. – Kaunas: Technologija, 2012. – Nr. 5(121). – P. 101–104. Darbe sprendžiamos biometrinių tapatumo nustatymo metodų taikymo fizinės apsaugos sistemose problemos. Išnagrinėtos biometrinių tapatumo nustatymo metodų taikymo galimybės. Pasiūlyta pirštų atspaudų atpažinimo metodus pritaikyti pirštų kraujagyslėms atpažinti. Eksperimentiškai parodyta, kad pirštų kraujagyslėms atpažinti fizinės saugos sistemose galima taikyti figūros konteksto ir orientacijos deskriptoriais atvaizdų atpažinimo principu veikiantį metodą. Šis metodas nelabai jautrus kraujagyslių atspaudų kokybei. Il. 6, bibl. 21, lent. 3 (anglų kalba; santraukos anglų ir lietuvių k.).
104
