Analysis and design of computer information processing methods applied to bank-note sorting machines and validators

Dissertation abstract

Preface

1. Research field overview

1.1 Bank note protection methods

Technology - contains the techniques based on the calculation of certain object properties , such as

• paper composition 
• the preset composition and characteristics of colour substances
• special paper surface treatment
• water mark
• inclusion of colour and luminescent fibers in paper mass composition
• preset composition of protective inclusions
• protective thread (synthetic or metallic)
• relief printing 
• Kipp-effect - special kind of printing detectable under specific lighting only

Graphic level - a set of methods based on use of special forms, size characteristics, means of location and combination of graphic image elements:

• pseudo water mark
• giilloshire frames
• microtext 
• protective grid (e.g. antiscan grid)
• diverse form and combination of fonts
• special micro-deffects of graphic elements
• asymmetric graphic image location 
• image superposition – front and reverse side printing insuring their exact matching or supplement.

Physical level - a set of methods based on physical properties of objects and substances:

• hologram and methogram label elements
• luminescent substances combination
• use of substances that turn luminous in IR or UV rays
• use of OVI paints changing color depending on the angle of vision
• use of protective magnetic code thread - metallic thread including magnetic elements, the combination of which encodes bank note denomination

1.2 Protection degrees

      The combination of diverse bank note protection methods insures the reliability of protection and makes for the following classification of currency protection degrees[1]:

• Simple or visible: water marks, relief printing
• Complicated, needing special equipment to detect it: magnetic, UV and IR protection
• Special protection - the combination of protective features known by currency developers and manufacturers only, which is a State secret

1.3 Technical means of authentification

      The great number of protection techniques and a set of requirements as for the depth of authentification make for a wide range of currency counting and validation equipment that can be classified by 3 groups  [1]:

• not automated (simple manual visualisation devices)
• automated (semi-automatic bank note detectors, equiped with sensors)
• expert complexes (currency sorters able to detect the elements of special protection)

1.4 Leading researchers and manufacturers

      Today the leding manufacturers of the bank note counting, sorting and validation devices include such producers as  DeLaRuе (England)[4], SPEED (Japan)[5], PRO (Japan)[6], SCAN COIN (Swiss)[7] . The devices of the first and third group of technical means classification are also produced by such home producers as «Vildis» (Moscow), «Regula» (Minsk), «Systema» (Moscow),«Spectr»(Nikolaev)[8].

2. Existing methods analysis

      The analysis of the great range of counting, sorting and validation devices makes it possible to classify basic authentification methods:

• geometric size control - insured by optic sensors that detect the beginning and the end of the bank note passing through the transportation mechanism. The length and the width of each bank note is computed with use of data saved by encoder connected to transportation shaft of the mechanism.
• IR label presence and location control. Control asserted by IR sensors. The value of light reflected from the bank note surface is estimated by the sensors and determines the presence of IR labels
• paper quality control with use of UV light source and photoreceivers in visible spectrum part. The controlled bank note passes under the UV rays (wave length= 365 nm or 246 nm). Special paper used for bank note manufacturing is not luminous in visible spectrum part. The only luminous parts of the bank note are those that include special protective elements luminous in the visible part of light spectrum.
• paper quality check with use of tri-colour optic sensors
• magnetic thread and elements presence and location control with use of magnetic heads. Magnetic heads' sensitive areas detect magnetic marks included in protective magnetic thread. The magnetic heads digital response processing with standard comparison makes for a conclusion authenticity conclusion.

3. Research highlights

3.1 Problem actuality

3.2 Practical value of research

3.3 Research steps

3.3.1. Code standards identification

      Open sources do not contain any information about magnetic protection code sequences. That makes it necessary to identify standard code sequences using magnetic fields visualisation means  [9],[10]. It is proposed to take 5,10,20,100,200,500 euro bank notes as reasearch samples. Figure 3.1 represents the results of standard identification step.

                                  Fig.3.1 - Standard euro codes

   Magnetized parts are indicated with use of bold black lines. The length of magnetic code spaces is in mm .

3.3.2. Magnetic code reading and computer information input

  Fig. 3.2 -  Digital magnetic code sequence of 20 hrvn. bank note

3.3.3 Data processing

      The last, but not the least step of this reasearch consists in magnetic code sygnal processing. Here the problem turns into discrete nonstationary sygnal processing. Data analysis includes the following steps described further.    

3.3.3.1 Spectral analysis

• Discrete Fourier transform (DFT)[14 p. 34]
  For a discrete sygnal represented by a sequence of N values DFT computation looks like:
  
  , where s - discrete sygnal, С - Fourier transform coefficients
• Fast Fourier transform (FFT). This method represents a fast way of Discrete Fourier transform computation. The use of this method speeds up the process of Fourier transform calculations.

3.3.3.2 Filtering

• Digital filter design. The output sequence of any filter represents the initial sygnal and filter impulse response convolution. The task of filter calculation consists of filter type (linear - nonlinear; recursive - nonrecursive) and filter impulse response selection.
• Wavelet filtering. Wavelets represent a useful mechanism of sygnal filtering and preliminary processing of sygnal data[15],[16]. The essence of sygnal wavelet transform is in multiscale analysis which means a sygnal decomposition with use of basis of functions exposed to multiple dilations and shifts which assures sygnal representation both in time and frequency space. The base functions are called wavelets if they are defined in  L²space, oscillate at absciss axis and meet zero value with argument increment. The sygnal and wavelet convolution allows to distinguish sygnal singularities at the point of wavelet localization. Wavelet transform as well as Fourier transform represents a sygnal decomposition in a basis of determined functions  [15].
  Straight continuous wavelet transform is calculated using the following formula:
  ,where a and b are scale and shift factors of ψ function or wavelet, Cψ – rate setting factor. Wavelet filtration algorithm includes the following steps:
  
  
  1. Wavelet sygnal decomposition 
  2. Threshold selection for each level of wavelet decomposition
  3. Threshold filtering of detalization coefficients
  4. Sygnal reconstruction 

3.3.3.3 Feature detection 

      The features of magnetic code sygnals we are interested in, that represent the beginning and the end of magnetized parts of protective thread, include positive and negative piques. The exact localization of particularities mentioned above has a great value for subsequent data analysis. It is proposed to use the following methods to solve the problem of feature detection in magnetic code sygnals:

• Feature detection using wavelet analysis. The main particularity and the main advantage of  wavelet sygnal analysis is 2 domain (time and frequency) sygnal representation. The time-frequency representation of the sygnal in question makes for it's singularities localization [18],[19],[20]. The features in question may include piques, bursts and threshold points. The type of singularity we are interested in dictates the type of wavelet we are to chose for analysis. One of the classical instances of wavelet feature localization is a problem of R-piques detection in ECG processing  [21]. The coefficient amplitude is a main sign of local pique in use of wavelet feature detection. Therefore the local coefficients' extremum is the main indication of a local singularity. 
• Design of one's own feature localization algorithm. Defining exact indications of sygnal piques (such as their length in time, their amplitude and their number for a code length) makes for a creation of one's own simple feature localization algorithm.

3.3.3.4 Classification

      The fact that the magnetic code in question coinsides with one of the standards defined earlier serves an authenticity proof of the bank note in question. In case where the percentage of coinsidence is low (e.g. lower than 80%), a hypothesis of bank note authenticity should be denied, the bank note is either a counterfeit or defected. Therefore at the final stage of researchthe problem of magnetic code classification is to be solved.  

• Correlation analysis.
  A cross-cprrelation function is one of a special characteristics of two sygnals. This function is an indicator of both the differences in the form of sygnals and their relative location on the time axis. For discrete sygnals f_i  and g_i (i= 0,1...N), their cross-correlation function is calculated as:

 
      To evaluate the similarity of two discrete sequences the correlation coefficient should be calculated:

      The coefficient value always belongs to the range [-1;1]. If the value of correlation coefficient is close to 1, the forms of two evaluated sygnals are similar. The use of cross-correlation analysis is proposed to evaluate the similarity of the magnetic code sygnal in question to each standard in the set of standards defined at the earlier step.

4. Assumed scientific novelty

5. Conclusion

• Continue research on the methods of data processing mentioned above to specify the parameters of sygnal processing (wavelet types, filter characteristics) and introduce the acquired results into development of data processing algorithms.
• Acquire results of computer aided sygnal processing 
• Chose the best suited methods as a result of comparative analysis of the results
• Develop programming code implementing chosen data processing methods
• Use the results of research to develop and modernize computer controlled currency counters and validators

1. http://bankir.ru/analytics/antilegal/ ,
   Методы и оборудование для определения подлинности денежных знаков и ценных бумаг
2. http://www.bre.ru/security/payment/ ,
   Аппаратурные средства проверки подлинности документов на основе оптического метода неразрушающего контроля
3. http://money.dmd.ru/
   author:Шапошников Ю.И. , book title: Деньги.Информация.
4. http://www.bankob.ru/catalog/counters/det/delarue/
   Счетчики банкнот фирмы De La Rue (Англия)
5. http://www.bankob.ru/catalog/counters/det/pro/
   Счетчики банкнот фирмы PRO (Япония)
6. http://www.bankob.ru/catalog/counters/det/scancoin/
   Счетчики банкнот фирмы Scan Coin (Швейцария)
7. http://www.bankob.ru/catalog/counters/det/speed/
   Счетчики банкнот фирмы SPEED (Япония)
8. http://www.spectr.nikolaev.ua/
   Сайт компании-производителя средств пересчета и детекции валют НПО "Спектр", г. Николаев, Украина 
9. http://epos.kiev.ua/pubs/
   С.Коженевский, С.Левый, В. Вишневский, С. Прокопенко
   article: Методы визуализации магнитных полей носителей информации
10. http://www.lad.org.ua/index.html
    Лаборатория методов и средств риминалистических исследований радиотехнического факультета НТУУ КПИ.Сканер магнитной защиты документов "СЕЗАМ"
11. аuthor: Ефимов Е.Г.,книга: Магнитные головки, М., Энергия, 1967,350 с.370
12. журнал «CHIP News», №4 (24), май 2003
    статья «Использование магнитных датчиков фирмы NVE в детекторах фальшивых купюр».
13. author: Carl H. Smith
    article: The color of money: using magnetic media detection to identify currency.
14. author:А.Б. Сергиенко
    book title: Цифровая обработка сигналов, Питер, 2002г., 606 с.
15. author: C.Valens
    article: A really friendly guide to wavelets, 1999
16. author: Киселев А.
    article: Приложения вейвлет-анализа, BaseGroup Labs
17. http:// www.basegroup.ru/labs/
    article: Основы теории вейвлет-преобразования
18. Дремин И.М., Иванов О.В., Нечитайло В.А.,журнал «Успехи физических наук», том 171, №5, май 2001, статья «Вейвлеты и их использование»
19. Астафьева Н.М., журнал «Успехи физических наук», том 166, №11, ноябрь 1996, статья «Вейвлет-анализ: основы теории и примеры применения»
20. author: Naoki Saito
    PhD thesis: Local feature extraction and it's applications using a library of bases. ;186p.
    Yale University
21. author: Шитов А.Б.
    PhD thesis: Разработка численных методов и программ, связанных с применением вейвлет-анализа»
    Ивановский Государственный университет, специальность «Математическое моделирование, численные методы и комплексы программ»

Note: