CA2100477A1 - Method and apparatus for detecting counterfeit documents - Google Patents
Method and apparatus for detecting counterfeit documentsInfo
- Publication number
- CA2100477A1 CA2100477A1 CA002100477A CA2100477A CA2100477A1 CA 2100477 A1 CA2100477 A1 CA 2100477A1 CA 002100477 A CA002100477 A CA 002100477A CA 2100477 A CA2100477 A CA 2100477A CA 2100477 A1 CA2100477 A1 CA 2100477A1
- Authority
- CA
- Canada
- Prior art keywords
- specimen
- light beam
- data
- printed
- lens
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Abandoned
Links
- 238000000034 method Methods 0.000 title claims abstract description 78
- 230000008569 process Effects 0.000 claims abstract description 32
- 238000007639 printing Methods 0.000 claims abstract description 31
- 238000006073 displacement reaction Methods 0.000 claims description 14
- 238000013528 artificial neural network Methods 0.000 claims description 13
- 238000005286 illumination Methods 0.000 claims 2
- 239000000976 ink Substances 0.000 description 11
- 239000013598 vector Substances 0.000 description 10
- 238000012545 processing Methods 0.000 description 7
- 230000008859 change Effects 0.000 description 6
- 238000010586 diagram Methods 0.000 description 6
- 230000007246 mechanism Effects 0.000 description 5
- 239000003086 colorant Substances 0.000 description 4
- 230000033458 reproduction Effects 0.000 description 4
- 238000001514 detection method Methods 0.000 description 3
- 239000011888 foil Substances 0.000 description 3
- 238000001228 spectrum Methods 0.000 description 3
- 239000000126 substance Substances 0.000 description 3
- 230000007723 transport mechanism Effects 0.000 description 3
- 230000000007 visual effect Effects 0.000 description 3
- 229920000914 Metallic fiber Polymers 0.000 description 2
- 239000000872 buffer Substances 0.000 description 2
- 230000003247 decreasing effect Effects 0.000 description 2
- 238000009792 diffusion process Methods 0.000 description 2
- 230000006870 function Effects 0.000 description 2
- 238000005259 measurement Methods 0.000 description 2
- 230000002441 reversible effect Effects 0.000 description 2
- RYGMFSIKBFXOCR-UHFFFAOYSA-N Copper Chemical compound [Cu] RYGMFSIKBFXOCR-UHFFFAOYSA-N 0.000 description 1
- 229910000831 Steel Inorganic materials 0.000 description 1
- 230000006978 adaptation Effects 0.000 description 1
- 230000004075 alteration Effects 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 230000006399 behavior Effects 0.000 description 1
- 230000008901 benefit Effects 0.000 description 1
- 230000002860 competitive effect Effects 0.000 description 1
- 239000002131 composite material Substances 0.000 description 1
- 229910052802 copper Inorganic materials 0.000 description 1
- 239000010949 copper Substances 0.000 description 1
- 238000012937 correction Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000002329 infrared spectrum Methods 0.000 description 1
- 239000007788 liquid Substances 0.000 description 1
- 239000004973 liquid crystal related substance Substances 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 239000002184 metal Substances 0.000 description 1
- 229910052751 metal Inorganic materials 0.000 description 1
- 230000007935 neutral effect Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 239000004033 plastic Substances 0.000 description 1
- 239000000843 powder Substances 0.000 description 1
- 230000002265 prevention Effects 0.000 description 1
- 238000011084 recovery Methods 0.000 description 1
- 230000004044 response Effects 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 239000004065 semiconductor Substances 0.000 description 1
- 125000006850 spacer group Chemical group 0.000 description 1
- 239000010959 steel Substances 0.000 description 1
- 238000012549 training Methods 0.000 description 1
- 230000037303 wrinkles Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G07—CHECKING-DEVICES
- G07D—HANDLING OF COINS OR VALUABLE PAPERS, e.g. TESTING, SORTING BY DENOMINATIONS, COUNTING, DISPENSING, CHANGING OR DEPOSITING
- G07D7/00—Testing specially adapted to determine the identity or genuineness of valuable papers or for segregating those which are unacceptable, e.g. banknotes that are alien to a currency
- G07D7/06—Testing specially adapted to determine the identity or genuineness of valuable papers or for segregating those which are unacceptable, e.g. banknotes that are alien to a currency using wave or particle radiation
- G07D7/12—Visible light, infrared or ultraviolet radiation
- G07D7/121—Apparatus characterised by sensor details
Abstract
ABSTRACT OF THE DISCLOSURE
A method is described for detecting counterfeit or forged documents where the genuine documents contain matter printed in relief by a security printing process, such as intaglio. The method comprises storing data representing a characteristic relief profile of matter printed by said process, scanning a specimen document to generate data representing a relief profile of the printed matter thereon, comparing the generated data with the stored data, and determining the validity of the specimen document on the basis of said comparison.
A method is described for detecting counterfeit or forged documents where the genuine documents contain matter printed in relief by a security printing process, such as intaglio. The method comprises storing data representing a characteristic relief profile of matter printed by said process, scanning a specimen document to generate data representing a relief profile of the printed matter thereon, comparing the generated data with the stored data, and determining the validity of the specimen document on the basis of said comparison.
Description
2l0a477 BACKGROUND OF THE INVENTION
This invention relates to a method of discriminating counterfeit or forged documents, where the genuine documents are printed in relief by a security printing process, such as intaglio.
Substantial losses occur to the economy each year due to counterfeit or forged documents, æuch as traveler's cheques and paper currency. For example, it is estimated that VisaTM, MastercardTM and Thomas CookeTM
together lose about forty-five million dollars per year due to fraud. No estimate is available for American Express, but a reasonable guess is about forty million dollars. The total loss to be addressed arising from counterfeit or forged traveler's cheques alone i8 in the region of ninety-five million dollars per year.
Losses due to counterfeit and forged paper money are estimated to be about two billion dollars per year.
Es~imates for other types of legal tender (bearer bonds, etc.) have not been obtained. Also, legal documents such as driver's licenses etc. do not cause a major cash loss but are of major concern to government~ and security agencies.
Attempts to prevent cheques from being counterfeit or forged in the first place have not met with great success because all prevention techniques implemented so far are easily and cheaply defeated with technologies that are widely available. Some techniques, which have been implemented on some currencies and other legal documents, include metal foils, holograms, and embedded metallic threads. These techniques are also expensive~to implement and require sophisticated techniques and expertise to discriminate forgeries from the originals.
211D~t7 Currently, counterfeit or forgeries of travelers' cheques are detectable in two ways. The first is by phoning the issuing company to cross-check the serial numbers. This is only effective if the original cheque has already been cashed and cleared through the issuing company, and the method is only available during normal business hours. It is also a costly technique for the merchant cashing the cheque. The second method takes place at the issuing company by automated methods that cross-check the serial numbers. At this time, the loss has already been incurred. Also, once the cheque has been passed successfully to the issuing company, the opportunity to identify the forger is lost.
One current method of counterfeit and forgery detection involves a visual examination for color incorrectness, watermarks, absence of appropriate foils etc., aberrations in expected patterns (visual comparison with a known legitimate document), or a microscopic examination for the absence of micro-lettering. The weakness of this approach is that not all persons have sufficient expertise to judge the authenticity of a document.
Another method involves feeling the document. This method requires that special attention be paid by the merchant to the texture of the item. This is very error-prone due to the variability of expertise.
Metallic ink detectors can be used. The weakness with this method of detection is that metallic ink detectors are easily defeated by counterfeits and forgeries that contain metallic inks.
A pen containing a special type of liquid that reacts with the chemicals in the ink to produce a distinctive color can be used. The weakness of the pen technique is that the appropriate chemicals can easily be . .
added to counterfeit or forged documents to defeat this method.
Also employed are luminescent powder, camera and microscope. The weaknesses of this method is that it is very expensive to use, requires special expertise to operate and diagnose the results, and is time-consuming.
- It is a solution that is more appropriate to a laboratory setting and is not appropriate as a point-of-sale detection device.
Unique document identification numbers can be cross-checked against an information database containing valid numbers or known forged numbers. This method requires a phone call to a central database of information which is not always available for access (for example on holidays). It is time-con~uming and only is reliable if the document number has been listed as previously cleared or is a known counterfeit or forged number.
A device that reads a hidden number or code on some paper money is used in change machines. This device is easily defeated by counterfeits that contain the hidden number or code.
Chemical analysis of the paper or inks can be performed and compared to legitimate parameters. These methods are not appropriate for point-of-sale devices because they require special expertise to execute and diagnose the results, as well as being slow and expensive. They are more suited to forensic laboratory use.
The specimen document can be photographed and the 3~ colors and patterns compared with legitimate parameters.
Within a reasonable price range, this method does not have sufficient resolution to detect adequately the fine details of a well designed intaglio printed document and,
This invention relates to a method of discriminating counterfeit or forged documents, where the genuine documents are printed in relief by a security printing process, such as intaglio.
Substantial losses occur to the economy each year due to counterfeit or forged documents, æuch as traveler's cheques and paper currency. For example, it is estimated that VisaTM, MastercardTM and Thomas CookeTM
together lose about forty-five million dollars per year due to fraud. No estimate is available for American Express, but a reasonable guess is about forty million dollars. The total loss to be addressed arising from counterfeit or forged traveler's cheques alone i8 in the region of ninety-five million dollars per year.
Losses due to counterfeit and forged paper money are estimated to be about two billion dollars per year.
Es~imates for other types of legal tender (bearer bonds, etc.) have not been obtained. Also, legal documents such as driver's licenses etc. do not cause a major cash loss but are of major concern to government~ and security agencies.
Attempts to prevent cheques from being counterfeit or forged in the first place have not met with great success because all prevention techniques implemented so far are easily and cheaply defeated with technologies that are widely available. Some techniques, which have been implemented on some currencies and other legal documents, include metal foils, holograms, and embedded metallic threads. These techniques are also expensive~to implement and require sophisticated techniques and expertise to discriminate forgeries from the originals.
211D~t7 Currently, counterfeit or forgeries of travelers' cheques are detectable in two ways. The first is by phoning the issuing company to cross-check the serial numbers. This is only effective if the original cheque has already been cashed and cleared through the issuing company, and the method is only available during normal business hours. It is also a costly technique for the merchant cashing the cheque. The second method takes place at the issuing company by automated methods that cross-check the serial numbers. At this time, the loss has already been incurred. Also, once the cheque has been passed successfully to the issuing company, the opportunity to identify the forger is lost.
One current method of counterfeit and forgery detection involves a visual examination for color incorrectness, watermarks, absence of appropriate foils etc., aberrations in expected patterns (visual comparison with a known legitimate document), or a microscopic examination for the absence of micro-lettering. The weakness of this approach is that not all persons have sufficient expertise to judge the authenticity of a document.
Another method involves feeling the document. This method requires that special attention be paid by the merchant to the texture of the item. This is very error-prone due to the variability of expertise.
Metallic ink detectors can be used. The weakness with this method of detection is that metallic ink detectors are easily defeated by counterfeits and forgeries that contain metallic inks.
A pen containing a special type of liquid that reacts with the chemicals in the ink to produce a distinctive color can be used. The weakness of the pen technique is that the appropriate chemicals can easily be . .
added to counterfeit or forged documents to defeat this method.
Also employed are luminescent powder, camera and microscope. The weaknesses of this method is that it is very expensive to use, requires special expertise to operate and diagnose the results, and is time-consuming.
- It is a solution that is more appropriate to a laboratory setting and is not appropriate as a point-of-sale detection device.
Unique document identification numbers can be cross-checked against an information database containing valid numbers or known forged numbers. This method requires a phone call to a central database of information which is not always available for access (for example on holidays). It is time-con~uming and only is reliable if the document number has been listed as previously cleared or is a known counterfeit or forged number.
A device that reads a hidden number or code on some paper money is used in change machines. This device is easily defeated by counterfeits that contain the hidden number or code.
Chemical analysis of the paper or inks can be performed and compared to legitimate parameters. These methods are not appropriate for point-of-sale devices because they require special expertise to execute and diagnose the results, as well as being slow and expensive. They are more suited to forensic laboratory use.
The specimen document can be photographed and the 3~ colors and patterns compared with legitimate parameters.
Within a reasonable price range, this method does not have sufficient resolution to detect adequately the fine details of a well designed intaglio printed document and,
2~0~77 therefore, is not able to discriminate properly between copies and originals.
Also, when a photographed reproduction is examined (regardless of resolution capabilities and cost), it is not possible to tell if the original was printed with intaglio or lithograph or other techniques because a photograph does not pick up the three-dimensional characteristics of intaglio. Therefore, although this technique can establish that certain patterns and colors match the legitimate original, it cannot tell whether or not the target is printed with intaglio.
The travelers' cheque companies and banks and other financial institutions have sophisticated equipment on-site at central locations to analyze the propriety of the legal tender. These methods include checking for magnetic ink, cross-checking for bad serial numbers etc. This equipment is expensive and is suitable for institutions but not as a point-of-sale device.
An object of the invention is to provide a low-cost device suitable for point-of-sale use capable of discriminating counterfeit or forged documents.
SUMMARY OF THE INVENTION
Accordingly the invention provides a method of detecting counterfeit or forged documents where the genuine documents contain matter printed in relief by a security printing process, comprising storing data representing a characteristic relief profile of matter printed by said process, scanning a specimen document to generate data representing a relief profile of said printed matter thereon, comparing said generated data with said stored data, and determining the validity of said specimen on the basis of said comparison.
,, In a preferred embodiment, the invention comprises a scanner having a scanning bed for receiving said specimen, means for producing a collimated light beam, a lens for focusing said light beam onto said specimen on said scanning bed, means for determining when said light beam is in focus on the surface of said specimen, means for effecting relative vertical displacement of said specimen and said lens, means for effecting relative horizontal displacement of said specimen and said lens, and means responsive to the relative vertical displacement of said specimen and said lens to generate said relief profile data.
In this specification the expression "relief profilel' refers to any characteristic of the printed matter along the scanning line that is measurable on the Z-axis, where the X-Y axes define the plane of the paper.
This can be, for example, the complete contour in the Z
dimension of the printed matter along the scanning line.
Alternatively, it can be simply the stepwise pattern of segments that are sufficiently horizontal to reflect light back along the incident beam. To these can be added the slopes from which insufficient light is reflected back to be detected. In one embodiment these slopes can all be given the same value, normally zero, 80 that the pattern of zeros and data values representing the height of the horizontal surfaces makes up the relief profile.
In a still further embodiment the actual slope angles are measured and these constitute additional data making up the relief profile.
The invention provides a system that will detect counterfeits and forgeries on the basis of their relief characteristics. It can be used to discriminate illicit reproductions or forgeries of legal documents from the legitimate ones, provided the legitimate ones are printed (at least in partl in relief with a security printing process, such as an intaglio printing technique. Intaglio ~l~O~
is a printing technique whereby ~he print is etched into the printing plate. Very consistent and reproducible results can be obtained in relief. Intaglio, which has been applied using the sophisticated methods of engra~ing and printing of major printers of legal tender, has unique characteristics that the system uses to discriminate fakes. The engraving, printing and other equipment and supplies that are required to print intaglio of this quality are expensive and usually protected with security measures. The expertise to print using these techniques is also in short supply. Knowledge of the techniques is also secured. For these reasons, the "feel of steel" of the intaglio document is a very reliable indicator of authenticity.
By legal documents, are meant paper money, travelers' cheques, bearer bonds, or anything printed using the intaglio printing technique. By counterfeits and forgeries are meant unauthorized reproductions including, but not necessarily limited to photocopies, lithograph, thermal print, and embossed reproductions.
The apparatus, in essence, i8 able to discriminate genuine intaglio print from other types of print.
The apparatus can be used as a point-of-sale deterrent to the passing of counterfeit and forged documents.
In more detail, in its preferred embodiment, the apparatus comprises a high-resolution laser scanner, a motorized scanner carriage, a scanning bed with a flattening mechanism, an automatic insert and ejection mechanism, an LCD output screen, a CPU with operating control programs and EEPROM application software chips, enclosed in a secure box.
The method and apparatus according to the invention take advantage of the fact that intaglio printing has a ~ , . . .
:
,, , 2~0~V77 three-dimensional signature, which is not possible to reproduce by other printing techniques. The apparatus places no reliance on any other aspect of the specimen, thereby obviating or reducing the necessity for expensive anti-counterfeiting techniques such as special foils, embedded metallic threads, holograms, etc., for the specified applications. This could result in reduced printing costs for the targeted applications as well as reduced costs of identifying fakes.
Artificial intelligence is used to analyze the scanned information to determine if the target is legitimate or not. The technique that is employed in the preferred embodiment, namely a neural network embodying a self-organizing map.
A neural network iæ a computing system made up of a number of simple, highly interconnected processing elements, which processes information by its dynamic state response to external inputs. A neural network i5 neither sequential nor even necessarily deterministic.
The neural network does not execute a series of instructions; it responds to the input presented to it.
The result is not stored in a specific memory location, but consists of the overall state of the network after it has reached an equilibrium condition. In the present application the external input is the number of data points scanned in one sample line.
The self-organizing map (SOM) can be viewed as a two dimensional map that can, from a random starting point, find the natural relationships amongst various patterns without any external guidance. The SOM looks for regularities or trends and makes adaptations according to the function of the network while providing a graphical representation of its findings. The object of this paradigm is to compress information by forming reduced representations of the most relevant facts, without loss : ,," ,: .
.' ~ , ~ ., ~ , , 2100~77 of knowledge about their interrelationships. Each SOM
node is a composite profile or a typical representative of all cases that fit that profile.
The basic mechanism of unsupervised or competitive learning as seen in a SOM will now be described. Since each SOM node can be rep~esentative of all input vectors, the problem is to decide which of the SOM nodes should be that representative. The SOM ~odes are allowed to compete for the right to represent a particular class of input vectors, and the winner (this is relaxed by having the neighbors of the winner included) will get the chance to adjust its weights so that it becomes even more similar to the input vector. Thus, the closest SOM node is, in a measurable way, the closet to the input vector and thus represents the input vector. In the present application different print relief signatures excite strongly a SOM
node.
Each SOM node measures the Euclidean distance of its weights to the incoming input vector. The Euclidean distance is the difference between the two vectors. The SOM node with the smallest difference is declared the winner. The difference is between the input vector and the weights from the input vector to each individual SOM
node.
The operatipn of the neural network can be broken down into two phases: the learning phase and the recall phase.
In the learning phase a number of simple scanned lines are taken from a variety of types of print (intaglio, photocopied, embossed, lithograph).
This is the training set. These are submitted in turn to the SOM until the weights or connections from the input sample to the SOM nodes have reached an equilibrium J ~ ' , ~: :'~ ' ' ' , ' ' ~' - '' "
i':
'' ' ' . ',~, .' ' ' ' , ' ' .
~1o~Y~7 condition (learning as stopped and weights or connection do not change).
The knowledge of the neural network is embedded in the weights or connections. It is these weights or connections that are downloaded onto a chip.
In the recall phase the apparatus scans one data point at a time. This data point is multiplied by all of the weights or connections leading into the SOM and added to each SOM node. This is done for each data point until the apparatus has finished scanning the sample line. This means that each SOM node keeps a running total by having each data point multiplied by each weight or connection until the end of the sample. At this time each SOM node will have a summation value. The SOM node with the highest value is declared the winner since it mostly represents the sample scanned line. This in turn determines whether the sample is genuine or a forgery based on the fact that the learning phase has determined which SOM node responds to which type of print.
More particularly, in the preferred embodiment the apparatus relies in its operation on the specific sequence of ze~oes (generated as a result of the inability to focus the light beam on a steep slope), and relative height measures. These comprise the so-called slope pattern which is unique for the genuine document printed using the intaglio type of security printing. Any other type of printing, including modified intaglio (forged, for example) will have different slope pattern thus allowing the device to identify them successfully.
For the present purposes the value of the slope pattern at any given moment is equal to FV.
In case of flat types of printing, or photocopies, both black and white and color, the specific slopes are absent (no zeroes are generated), even though they may give a correct measurement of the relative ink heights.
On a micro level, even the flat prints show certain high-low areas, different for different types of print, thus enabling the particular type of print used on the scanned document specimen to be determined.
Raised prints, other than intaglio lack the fine resolution, which will reflect on the width of the clusters of data with height measures, effectively decreasing the frequency of the slope pattern. This is enough for the neural network to classify such examples as "bad~. The same refer to the tampered genuine intaglio print. Once destroyed, the fine line structure of the genuine document cannot be restored 100~ and this also effects the frequency of the slope pattern.
The embodiment described so far responds to the presence or absence of certain slopes and the heights of the horizontal or nearly horizontal print surfaces. In a still further embodiment the actual anglefi of these slopes can be measured, thereby enabling even greater precision to be achieved, creating microtopographic maps of the scanned surface.
BRIEF DESGRIPTION OF THE DRAWINGS
The invention will now be described in more detail, by way of example only, with reference to the accompanying drawings, in which:-Figure 1 is a diagram of the scanning assembly;
Figures 2a to 2c show the reflected beam spot asseen by the photodetector;
Figure 3 is a diagram illustrating the method of determining when the beam is in focusi ,~ ~,. ..
, .
., Figure 4 shows the horizontal scanner displacement mechanism;
Figure 5 is a side elevation of the scanning apparatus;
Figure 6 is an end elevation of the scanning apparatus;
Figure 7 is block diagram of the display driver circuitry;
Figure 8 is blo~k diagram of the main part of the apparatus;
Figure 9 is block diagram of the power supply Figure 10 shows the output relief profile waveforms;
Figure 11 illustrates the mode of operation of the neural network; and Figure 12 is a diagram showing the degrees of freedom of the sy~tem.
DESCRIPTION OF THE PREFERRED EMBODIMENTS
The apparatus in accordance with the invention comprises a high-resolution scanner assembly 15 (Figure 1), which provides raw 3D information about the micro surface of the specimen under question, a scanning bed assembly (Figs. 5 and 6), a liquid crystal display assembly (Fig. 7), a main printed circuit board assembly (Fig. 8), and a power supply assembly (Fig.9).
Referring to Figure 1, the high-resolution scanner assembly 15 consists of a semiconductor laser light emitting diode (LBD) 1, a half mirror prism 2, a .
~.: , ... : , , , '' " ~ ' '': ' '' '. ' ` .
. .
'~ao477 collimating lens assembly 3, a mirror prism 4, a two-axis carriage 5, a focusing lens 6, a cylindrical lens 7, and a photodiode sensor array 8. All these components are mounted on a scanner carriage, driven linearly by way of a worm gear and a reversible DC micromotor.
The LED 1 emits continuous light with a typical wavelength of 755nm, i.e. just beyond the red part of the visible light spectrum. The light is divergent so lenses are needed to converge and focus the beam.
The half mirror prism 2 is so positioned that the incident laser light beam is essentially split into two equal parts: the reflected light is lost and only 50~ of the initial light energy is passed on to the collimating lens 3. Here the light is converged so that after this point it forms a strictly parallel light beam 9. The mirror prism 4 then changes the beam direction by 90, without further distortion. Now the collimated light travels in a vertical plane. The two-axis carriage 5 consists of the focusing lens mounted with four degrees of freedom (see Fig. 12), two in the vertical plane of the light, and two in a plane normal to the vertical (horizontal). The focal point of lens 6 is 3 mm from its geometric center. The carriage has two sets of coils, immersed in a stationary constant magnetic field. By controlling the amount of DC current through the coils, the lens 6 can be moved both vertically and horizontally within a range of 2 mm. Further to every particular DC
current value corresponds only one absolute vertical or horizontal coordinate of the lens. This is the FV.
When focused on the specimen, the light spot has a diameter of 1.7 mierons. Some of the light energy is dissipated in the specimen, the amount depending on its texture, color and surface unevenness (on a micro level).
-, :
, ' , ' .
~100~77 If properly focused, the reflected light follows thesame path bac~wards to the half mirror prism 2, where 50 of it is reflected towards the sensors (the other 50~
passed towards the LED, where it is dissipated). At this point we have lost 3/4 of the emitted light energy, providing the specimen's albedo factor is 1 (100~ of the incident light i8 reflected back, best case scenario).
Since this is practically impossible, a feedback system is employed to maintain the level of light energy received by the sensors steady by changing the emission power of the LED 1.
The light is further collimated by the collimating lens assembly 3a, and passed on to the cylindrical lens 7. Because of its geometry, the cylindrical lens 3a acts as a spot-converter, converging non-linearly the off-centered light beams. This feature is used to change the shape of the light spot on the sensors array to an ellipse, should the beam be not properly focused at the specimen. The ellipse appears tilted to the left or to the right, depending on if the lens is too close (closer then the focal point) or too far (beyond the focal point) from the specimen's surface. Photo diodes are used as light-sensitive sensors, arranged in a 2x2 square matrix array. The sensor array arrangement is shown in Fig. 2a below for the situation where the beam is properly focused.
There are three distinct situations from the sensor's point of view, depending on the position of the 2-axis carriage rel~tive to the specimen's surface, as shown in Figs. 2a to 2c.
The signal returned from the photo diodes is linearly proportional to the light energy received.
Optically the main beam is aimed at the center of the array of photo diodes A, B, C and D. When the beam is properly focused on the specimen (its spot is at its `, ' ' ' , ~ ,.
~, ' `. ` ' . ' . ~ .
/ , . .
smallest), a perfect circle will be projected onto the sensor array (Fig. 2b). Then each of the sensors will produce the same signal (since their inputs are equal).
Should the specimen be out of focus however, either left-or right-tilted ellipse will be observed, the amount of elliptic distortion proportional to the degree of disfocus. This theory is applicable only in a certain vicinity of the focal point, about O.lmm.
The focus error signal FE is derived as the difference between the diagonal sums of light intensities, returned from the four photo diodes:
FE=(A+D)-(B+C). Thus FE=0 when the lens is positioned in such a way, that the specimen is in the focal point; FE~0 when the lens is too far from the specimen and FE<0 when the lens is too close to the specimen. FE is derived strictly for the servo focusing system and is not transmitted as useful signal for the purposes of any further processing.
The 3D scanning ability is based on pure optical reflection of the laser light off the surface of the specimen. When the surface is flat, most of the light is reflected back towards the source, as explained above.
Due to the diffusion (because of the paper's micro structure, as well as ink surface micro craters), a reflection factor cannot be achieved (See Fig. 3), but the circuitry is calibrated to recognize this as a normal condition. However, when the beam meets a steep slope of ink (either positive or negative), the only light returned to the sensors will be due to diffusion, since the angle of the slope steepness is greater than the angle of possible reflected beam recovery. This allows the pattern of certain slopes on the specimen's surface to be used as a validity criterion since the patterns provided by intaglio printing are uniquely different from other methods of printing.
~i~0~77 In a further embodiment a secondary sensor array 8a (Fig. 3. 3) is provided to detect the obliquely reflected light and determine the slope angles and these cannot be used as additional variables forming part of the relief profile.
Figure 3 shows in more detail the focusing lens 6, the specimen 10 with the relief 11 formed by the intaglio printing process on its upper surface. As can be seen in this figure, when the incident beam 11 is on a steep slope, the reflected beam 12 does not return along the path of the incident beam and cannot be focused.
Since the laser light is outside the visible light spectrum, it is by default insensitive to visual colors.
At the same time, it is sensitive (to a degree) to hues at the bottom of the IR spectrum, which may or may not be accounted for in the process of intaglio printing. The scanner is sensitive also to the albedo of the scanned surface, less at the blue end of the spectrum and more at the red one.
As shown in Figure 4, the scanner carriage assembly, generally referenced 15, moves along a trajectory normal to the one of feeding the specimen in the device (and parallel to the one of scanning). The mechanism for moving scanner carriage assembly comprises a worm gear 13 driven by a DC motor 14 through a drive belt lS. A limit switch at the end of free movement produces a signal, which is used by the processor to determine the position of the scanner carriage assembly.
Figures 5 and 6 show the main scanning platform 20, a passive flattener 21, a transport mechanism 22, sensors 23 and guide 24.
The main scanning platform 20 is a polished copper plate to which are attached the flattener 21, the ~1~0477 transport mechanism 22, the sensors 23, and the scanner assembly 15. The plate's polish ensures flawless movement of the specimen. The scanner is mounted on spacers above the plate 20. The direction of movement of the scanner carriage is parallel to its shorter side. The specimen is loaded along the longer side of the platform. This type of arrangement allows for scanning at any given point of the specimen.
The scanner assembly lays parallel (maximum allowed offset 10 microns) to the plane of the scanning platform, to ensure precision and consistency of the measurements.
The passive flattener is made out of PlexiglasTM and mounted on top of the platform 20, allowing clearance of one and a half bill thickness between them. The flattener has an air gap along the trajectory of scanning, to allow direct light beam on the specimen's surface. Thus, the latter is kept reasonably flat while being scanned, to avoid errors due to wrinkles for example.
The transport mechanism consists of two rubber rollers 24, the top one on a fixed axis, while the bottom one can move vertically about 0.5 mm. A coil spring ensures that both rollers are brought together, when the specimen is tightly caught between them. The springs allow for specimens with different thickness to be loaded for scanning.
The top roller is motor-driven by way of a plastic gear assembly 25. The motor 26 is a DC reversible micromotor and is driven by the driver circuitry on the printed circuit board.
The sensors 23 are two infrared LED-photo diode couples. One of them is so positioned, that signals the printed circuit board when to start processing a ,'~ ' , .
', : .
.: '.
~V~4'~7 specimen, the other, accordingly, when the specimen is positioned for scanning.
As shown in Fig~re 7, The LCD serves as a machine-human interface, providing the user with instructions, information and the results of scanning in natural language (different and multiple-language displays can be supported). It con ists of a LCD and a control processor board. The LCD displays the information passed by the control processor board. The control processor board communicates with the printed circuit board (Fig. 8)) and interprets the received machine language code for the purposes of the LCD. The control processor board consists of I/0 buffers 31, character generation circuit 32, timing generation circuit 33, display data Random Access Memory (RAM) and LCD drivers 34. There are 132 characters available at 5x7 display resolution (2-line, 16 characters per line display capability).
As shown in Figure 8, The printed circuit board contains all electronic components necessary for the operation of the device. These include the main processor 40, clock circuitry 41, reset circuitry, security circuitry, D/A converters 42, data buffers 43, drivers 44, various interfaces 45-48.
me power supply, shown in Figure 10, the necessary power to the printed circuit board and the LCD. It is of the switching variety and is capable of operating in the range of 110 - 240V AC, 50/60Hz. It consists of an input rectifier and input filter 50, a flyback transformer driver transformer 51, an output rectifier and filter 52, a output voltage watchdog and control circuitry 53.
me power supply delivers energy to the printed circuit board, necessary for the normal work of the device. It starts working as soon as it is plugged in (in a battery version, at the moment of connecting the battery). All other blocks receive their energy from the printed circuit board, according to their needs. At first, the main processor executes reset/initialization routine, which sets the device in a known state. During this time (2.8 microseconds) no useful function can be performed. Immediately after this, the processor switches in a low-power wait mode, which can be indefinitely long.
At this point the laser LED 1 is off, and the LCD is on and displays a status message and the circuitry is in a WAIT state. It will stay in this state until either a specimen is fed in~o the inlet of the device, or power is disconnected (battery dead in case of a battery version).
When a specimen is fed into the device, the microprocessor wakes up and starts executing the scanning program. First, the laser is turned on at full power, then the specimen is positioned inside the device and the scanner assembly carriage is positioned on top of the area of interest. This is accomplished by the motor and worm gear.
Next, at the LCD the status message is changed. Now, the device will not accept any other jobs, until the current one is completed. Scanning begins with the establishment of the average Focus Value (FV). To establish the FV, the fosusing lens is fully retracted and mo~ed down at a certain rate. The scanner assembly is so mounted on top of the scanning bed that the surface of the latter is within the focusing range of the lens. As the lens moves downwards, the point reached where the specimen's surface appears close to the focal point of the lens. This causes some light to be reflected back to the sensors. The laser emission feedback system makes the necessary corrections in the emission power of the laser, to compensate for different albedo factors (of different specimens). Essentially, the laser power is reduced by certain amount so that different colors look alike to the sensors. As the lens moves further down, it reaches the - ,. .
~, ,~ ,. ~ ,.. f .
~100~77 point when the beam is maximally focused on the sur~ace of the specimen. The processor takes the measure of the current required to deflect vertically the focusing lens by this amount and assigns to it the label FV(1). The FV
is generated as a result of the condition code FE
supplied by the main beam sensors' array. In both cases FV is a true description of the vertical position of the lens relative to the specimen. Now the routine splits into two parallel branches:
The processor makes an adjustment to the FV, preparing it for the next step scanning. Appropriate changes are made in the vertical positioning of the focusing lens. Next, the processor moves the focusing lens horizontally a step. Some delay is also necessary to compensate for the mechanical inertia of the 2-axis Carriage Assembly.
After scaling down the FV, the processor executes the neural network computations on it using the pre-recorded weight map. The results of these computations 2C are cumulatively stored in a virtual two dimensional array for further processing.
Upon finishing these tasks, the processor proceeds with finding FV(2), much in same manner, except for the initial vertical position of the focusing lens. Instead of 0, the adjusted FV is used for determining the initial state.
This process repeats a predetermined number of steps and after the last one is executed, the focusing lens is positioned in neutral position, the specimen is transported out of the device, while simultaneously the processor analyzes the results stored in the array.
According to the result, the processor passes a message to be displayed: either affirmative, negative or percentage probability of the specimen being an original.
., ,,.j;
.
~' "' ' ' ,,~ , ' -, ,~ , .
2~0~77 After some delay, providing no other specimen is fed into the device, the processor sets the circuitry back to the low-power WAIT mode.
The described scanning process does not mark or mutilate the specimens in any way. However, some results may be stored, depending on their importance, for the purposes of the system self-enhancement. These results are not associated in any way with a particular bill, but are rather generalization of a field expertise.
The sampling actually consists of subsequent focusing sessions performed at a given rate. In other words, the lens is made to closely follow the bumps on the specimen's surface.
Provisions have been made in the device described not to take into consideration the steepness of slopes.
Thus, a specific value is assigned to FV every time a measure is taken at a steep slope on the specimen's surface. To determine 3-dimensionality, it is enough to have a secure measure of having a slope, not what exactly slope it is. In another application of the novel process, the slope characteristics become important and are accounted for (e.g., micro topographical maps).
Figure 10 shows the projected time-amplitude relationships of the FE, FV, and AV. As shown in Fig. 10 the FV signal follows exactly the envelope of the specimen's surface, although it may be shifted vertically relative to the abscissa. The FV is accepted as a measure of the relative height of an element of the specimen. The FE signal is different from zero only when a change of the focus value is required, i.e., when there is a slope on the specimen. The direction of change of FE depends on what direction we want to move the lens in. Again, FE
amplitude is not related to the steepness of the scanned slope, but only to its change, it is just a servo signal 2~a~rl7 for the 2-axis device to move. AV depends both on the steepness of the slope and the reflecting ability of the surface, so the flatter and more reflective the specimen, the higher the value of AV.
As shown in Fig. 10, and as explained above, the system responds to the presence or absence of reflected light along the path of the incident light beam to determine. Where insufficient light is reflected back, - i.e. in the presence of a steep slope, a zero is returned, otherwise one of 255 values representing the height of the ink surface is returned. This creates a relief pattern, which is a unique signature of the printing process and consists of the presence of slopes and heights of the ink surface across the specimen.
Figure 11 shows a scanning line consisting of N
pixels 61, each capable of having one of 255 values, zero representing the absence of sufficient reflected light as described above. These can represent the elements of a vector of a elements, each of which is connected to one of 16 nodes 60 of the self-organizing map.
What happens in the learning process is that after each scan of say legitimate intaglio the value of each pixel is multiplied by the weight associated with its connection to each node and the weighted values for each pixel are accumulated in each node. Initially the weightings for each connection are random. After this process one of the nodes will have a higher accumulated value than the others, and this is taken to be the node corresponding to genuine intaglio print. The weightings are then adjusted to still further favor this node.
The same process is repeated for various types of counterfeit or forged documents, with a different node being favored in each case until an equilibrium condition , . .
C~100~77 is reached where each node represents a certain type of printing process.
The system software is written in assembly language.
Residing on-chip, and the user having no access to the program, it is classified as firmware. It comprises several routines to satisfy the control, computation and security needs of the device and a weights' map, to implement the AI elements of the process.
From software point of view, both sequential and parallel processing are employed, and the system demonstrates both linear and nonlinear behavior, as required. The sequential part of the program is executed sequentially, requiring 100% of the processor time, one step at a time. The parallel processing is used when computing the equations for the neural network, when the processor is used to crunch more than one piece of data at a time. This yields a saving of processing time, thus decreasing the total process time.
. - .- .. ,.,,, ~ . , , '
Also, when a photographed reproduction is examined (regardless of resolution capabilities and cost), it is not possible to tell if the original was printed with intaglio or lithograph or other techniques because a photograph does not pick up the three-dimensional characteristics of intaglio. Therefore, although this technique can establish that certain patterns and colors match the legitimate original, it cannot tell whether or not the target is printed with intaglio.
The travelers' cheque companies and banks and other financial institutions have sophisticated equipment on-site at central locations to analyze the propriety of the legal tender. These methods include checking for magnetic ink, cross-checking for bad serial numbers etc. This equipment is expensive and is suitable for institutions but not as a point-of-sale device.
An object of the invention is to provide a low-cost device suitable for point-of-sale use capable of discriminating counterfeit or forged documents.
SUMMARY OF THE INVENTION
Accordingly the invention provides a method of detecting counterfeit or forged documents where the genuine documents contain matter printed in relief by a security printing process, comprising storing data representing a characteristic relief profile of matter printed by said process, scanning a specimen document to generate data representing a relief profile of said printed matter thereon, comparing said generated data with said stored data, and determining the validity of said specimen on the basis of said comparison.
,, In a preferred embodiment, the invention comprises a scanner having a scanning bed for receiving said specimen, means for producing a collimated light beam, a lens for focusing said light beam onto said specimen on said scanning bed, means for determining when said light beam is in focus on the surface of said specimen, means for effecting relative vertical displacement of said specimen and said lens, means for effecting relative horizontal displacement of said specimen and said lens, and means responsive to the relative vertical displacement of said specimen and said lens to generate said relief profile data.
In this specification the expression "relief profilel' refers to any characteristic of the printed matter along the scanning line that is measurable on the Z-axis, where the X-Y axes define the plane of the paper.
This can be, for example, the complete contour in the Z
dimension of the printed matter along the scanning line.
Alternatively, it can be simply the stepwise pattern of segments that are sufficiently horizontal to reflect light back along the incident beam. To these can be added the slopes from which insufficient light is reflected back to be detected. In one embodiment these slopes can all be given the same value, normally zero, 80 that the pattern of zeros and data values representing the height of the horizontal surfaces makes up the relief profile.
In a still further embodiment the actual slope angles are measured and these constitute additional data making up the relief profile.
The invention provides a system that will detect counterfeits and forgeries on the basis of their relief characteristics. It can be used to discriminate illicit reproductions or forgeries of legal documents from the legitimate ones, provided the legitimate ones are printed (at least in partl in relief with a security printing process, such as an intaglio printing technique. Intaglio ~l~O~
is a printing technique whereby ~he print is etched into the printing plate. Very consistent and reproducible results can be obtained in relief. Intaglio, which has been applied using the sophisticated methods of engra~ing and printing of major printers of legal tender, has unique characteristics that the system uses to discriminate fakes. The engraving, printing and other equipment and supplies that are required to print intaglio of this quality are expensive and usually protected with security measures. The expertise to print using these techniques is also in short supply. Knowledge of the techniques is also secured. For these reasons, the "feel of steel" of the intaglio document is a very reliable indicator of authenticity.
By legal documents, are meant paper money, travelers' cheques, bearer bonds, or anything printed using the intaglio printing technique. By counterfeits and forgeries are meant unauthorized reproductions including, but not necessarily limited to photocopies, lithograph, thermal print, and embossed reproductions.
The apparatus, in essence, i8 able to discriminate genuine intaglio print from other types of print.
The apparatus can be used as a point-of-sale deterrent to the passing of counterfeit and forged documents.
In more detail, in its preferred embodiment, the apparatus comprises a high-resolution laser scanner, a motorized scanner carriage, a scanning bed with a flattening mechanism, an automatic insert and ejection mechanism, an LCD output screen, a CPU with operating control programs and EEPROM application software chips, enclosed in a secure box.
The method and apparatus according to the invention take advantage of the fact that intaglio printing has a ~ , . . .
:
,, , 2~0~V77 three-dimensional signature, which is not possible to reproduce by other printing techniques. The apparatus places no reliance on any other aspect of the specimen, thereby obviating or reducing the necessity for expensive anti-counterfeiting techniques such as special foils, embedded metallic threads, holograms, etc., for the specified applications. This could result in reduced printing costs for the targeted applications as well as reduced costs of identifying fakes.
Artificial intelligence is used to analyze the scanned information to determine if the target is legitimate or not. The technique that is employed in the preferred embodiment, namely a neural network embodying a self-organizing map.
A neural network iæ a computing system made up of a number of simple, highly interconnected processing elements, which processes information by its dynamic state response to external inputs. A neural network i5 neither sequential nor even necessarily deterministic.
The neural network does not execute a series of instructions; it responds to the input presented to it.
The result is not stored in a specific memory location, but consists of the overall state of the network after it has reached an equilibrium condition. In the present application the external input is the number of data points scanned in one sample line.
The self-organizing map (SOM) can be viewed as a two dimensional map that can, from a random starting point, find the natural relationships amongst various patterns without any external guidance. The SOM looks for regularities or trends and makes adaptations according to the function of the network while providing a graphical representation of its findings. The object of this paradigm is to compress information by forming reduced representations of the most relevant facts, without loss : ,," ,: .
.' ~ , ~ ., ~ , , 2100~77 of knowledge about their interrelationships. Each SOM
node is a composite profile or a typical representative of all cases that fit that profile.
The basic mechanism of unsupervised or competitive learning as seen in a SOM will now be described. Since each SOM node can be rep~esentative of all input vectors, the problem is to decide which of the SOM nodes should be that representative. The SOM ~odes are allowed to compete for the right to represent a particular class of input vectors, and the winner (this is relaxed by having the neighbors of the winner included) will get the chance to adjust its weights so that it becomes even more similar to the input vector. Thus, the closest SOM node is, in a measurable way, the closet to the input vector and thus represents the input vector. In the present application different print relief signatures excite strongly a SOM
node.
Each SOM node measures the Euclidean distance of its weights to the incoming input vector. The Euclidean distance is the difference between the two vectors. The SOM node with the smallest difference is declared the winner. The difference is between the input vector and the weights from the input vector to each individual SOM
node.
The operatipn of the neural network can be broken down into two phases: the learning phase and the recall phase.
In the learning phase a number of simple scanned lines are taken from a variety of types of print (intaglio, photocopied, embossed, lithograph).
This is the training set. These are submitted in turn to the SOM until the weights or connections from the input sample to the SOM nodes have reached an equilibrium J ~ ' , ~: :'~ ' ' ' , ' ' ~' - '' "
i':
'' ' ' . ',~, .' ' ' ' , ' ' .
~1o~Y~7 condition (learning as stopped and weights or connection do not change).
The knowledge of the neural network is embedded in the weights or connections. It is these weights or connections that are downloaded onto a chip.
In the recall phase the apparatus scans one data point at a time. This data point is multiplied by all of the weights or connections leading into the SOM and added to each SOM node. This is done for each data point until the apparatus has finished scanning the sample line. This means that each SOM node keeps a running total by having each data point multiplied by each weight or connection until the end of the sample. At this time each SOM node will have a summation value. The SOM node with the highest value is declared the winner since it mostly represents the sample scanned line. This in turn determines whether the sample is genuine or a forgery based on the fact that the learning phase has determined which SOM node responds to which type of print.
More particularly, in the preferred embodiment the apparatus relies in its operation on the specific sequence of ze~oes (generated as a result of the inability to focus the light beam on a steep slope), and relative height measures. These comprise the so-called slope pattern which is unique for the genuine document printed using the intaglio type of security printing. Any other type of printing, including modified intaglio (forged, for example) will have different slope pattern thus allowing the device to identify them successfully.
For the present purposes the value of the slope pattern at any given moment is equal to FV.
In case of flat types of printing, or photocopies, both black and white and color, the specific slopes are absent (no zeroes are generated), even though they may give a correct measurement of the relative ink heights.
On a micro level, even the flat prints show certain high-low areas, different for different types of print, thus enabling the particular type of print used on the scanned document specimen to be determined.
Raised prints, other than intaglio lack the fine resolution, which will reflect on the width of the clusters of data with height measures, effectively decreasing the frequency of the slope pattern. This is enough for the neural network to classify such examples as "bad~. The same refer to the tampered genuine intaglio print. Once destroyed, the fine line structure of the genuine document cannot be restored 100~ and this also effects the frequency of the slope pattern.
The embodiment described so far responds to the presence or absence of certain slopes and the heights of the horizontal or nearly horizontal print surfaces. In a still further embodiment the actual anglefi of these slopes can be measured, thereby enabling even greater precision to be achieved, creating microtopographic maps of the scanned surface.
BRIEF DESGRIPTION OF THE DRAWINGS
The invention will now be described in more detail, by way of example only, with reference to the accompanying drawings, in which:-Figure 1 is a diagram of the scanning assembly;
Figures 2a to 2c show the reflected beam spot asseen by the photodetector;
Figure 3 is a diagram illustrating the method of determining when the beam is in focusi ,~ ~,. ..
, .
., Figure 4 shows the horizontal scanner displacement mechanism;
Figure 5 is a side elevation of the scanning apparatus;
Figure 6 is an end elevation of the scanning apparatus;
Figure 7 is block diagram of the display driver circuitry;
Figure 8 is blo~k diagram of the main part of the apparatus;
Figure 9 is block diagram of the power supply Figure 10 shows the output relief profile waveforms;
Figure 11 illustrates the mode of operation of the neural network; and Figure 12 is a diagram showing the degrees of freedom of the sy~tem.
DESCRIPTION OF THE PREFERRED EMBODIMENTS
The apparatus in accordance with the invention comprises a high-resolution scanner assembly 15 (Figure 1), which provides raw 3D information about the micro surface of the specimen under question, a scanning bed assembly (Figs. 5 and 6), a liquid crystal display assembly (Fig. 7), a main printed circuit board assembly (Fig. 8), and a power supply assembly (Fig.9).
Referring to Figure 1, the high-resolution scanner assembly 15 consists of a semiconductor laser light emitting diode (LBD) 1, a half mirror prism 2, a .
~.: , ... : , , , '' " ~ ' '': ' '' '. ' ` .
. .
'~ao477 collimating lens assembly 3, a mirror prism 4, a two-axis carriage 5, a focusing lens 6, a cylindrical lens 7, and a photodiode sensor array 8. All these components are mounted on a scanner carriage, driven linearly by way of a worm gear and a reversible DC micromotor.
The LED 1 emits continuous light with a typical wavelength of 755nm, i.e. just beyond the red part of the visible light spectrum. The light is divergent so lenses are needed to converge and focus the beam.
The half mirror prism 2 is so positioned that the incident laser light beam is essentially split into two equal parts: the reflected light is lost and only 50~ of the initial light energy is passed on to the collimating lens 3. Here the light is converged so that after this point it forms a strictly parallel light beam 9. The mirror prism 4 then changes the beam direction by 90, without further distortion. Now the collimated light travels in a vertical plane. The two-axis carriage 5 consists of the focusing lens mounted with four degrees of freedom (see Fig. 12), two in the vertical plane of the light, and two in a plane normal to the vertical (horizontal). The focal point of lens 6 is 3 mm from its geometric center. The carriage has two sets of coils, immersed in a stationary constant magnetic field. By controlling the amount of DC current through the coils, the lens 6 can be moved both vertically and horizontally within a range of 2 mm. Further to every particular DC
current value corresponds only one absolute vertical or horizontal coordinate of the lens. This is the FV.
When focused on the specimen, the light spot has a diameter of 1.7 mierons. Some of the light energy is dissipated in the specimen, the amount depending on its texture, color and surface unevenness (on a micro level).
-, :
, ' , ' .
~100~77 If properly focused, the reflected light follows thesame path bac~wards to the half mirror prism 2, where 50 of it is reflected towards the sensors (the other 50~
passed towards the LED, where it is dissipated). At this point we have lost 3/4 of the emitted light energy, providing the specimen's albedo factor is 1 (100~ of the incident light i8 reflected back, best case scenario).
Since this is practically impossible, a feedback system is employed to maintain the level of light energy received by the sensors steady by changing the emission power of the LED 1.
The light is further collimated by the collimating lens assembly 3a, and passed on to the cylindrical lens 7. Because of its geometry, the cylindrical lens 3a acts as a spot-converter, converging non-linearly the off-centered light beams. This feature is used to change the shape of the light spot on the sensors array to an ellipse, should the beam be not properly focused at the specimen. The ellipse appears tilted to the left or to the right, depending on if the lens is too close (closer then the focal point) or too far (beyond the focal point) from the specimen's surface. Photo diodes are used as light-sensitive sensors, arranged in a 2x2 square matrix array. The sensor array arrangement is shown in Fig. 2a below for the situation where the beam is properly focused.
There are three distinct situations from the sensor's point of view, depending on the position of the 2-axis carriage rel~tive to the specimen's surface, as shown in Figs. 2a to 2c.
The signal returned from the photo diodes is linearly proportional to the light energy received.
Optically the main beam is aimed at the center of the array of photo diodes A, B, C and D. When the beam is properly focused on the specimen (its spot is at its `, ' ' ' , ~ ,.
~, ' `. ` ' . ' . ~ .
/ , . .
smallest), a perfect circle will be projected onto the sensor array (Fig. 2b). Then each of the sensors will produce the same signal (since their inputs are equal).
Should the specimen be out of focus however, either left-or right-tilted ellipse will be observed, the amount of elliptic distortion proportional to the degree of disfocus. This theory is applicable only in a certain vicinity of the focal point, about O.lmm.
The focus error signal FE is derived as the difference between the diagonal sums of light intensities, returned from the four photo diodes:
FE=(A+D)-(B+C). Thus FE=0 when the lens is positioned in such a way, that the specimen is in the focal point; FE~0 when the lens is too far from the specimen and FE<0 when the lens is too close to the specimen. FE is derived strictly for the servo focusing system and is not transmitted as useful signal for the purposes of any further processing.
The 3D scanning ability is based on pure optical reflection of the laser light off the surface of the specimen. When the surface is flat, most of the light is reflected back towards the source, as explained above.
Due to the diffusion (because of the paper's micro structure, as well as ink surface micro craters), a reflection factor cannot be achieved (See Fig. 3), but the circuitry is calibrated to recognize this as a normal condition. However, when the beam meets a steep slope of ink (either positive or negative), the only light returned to the sensors will be due to diffusion, since the angle of the slope steepness is greater than the angle of possible reflected beam recovery. This allows the pattern of certain slopes on the specimen's surface to be used as a validity criterion since the patterns provided by intaglio printing are uniquely different from other methods of printing.
~i~0~77 In a further embodiment a secondary sensor array 8a (Fig. 3. 3) is provided to detect the obliquely reflected light and determine the slope angles and these cannot be used as additional variables forming part of the relief profile.
Figure 3 shows in more detail the focusing lens 6, the specimen 10 with the relief 11 formed by the intaglio printing process on its upper surface. As can be seen in this figure, when the incident beam 11 is on a steep slope, the reflected beam 12 does not return along the path of the incident beam and cannot be focused.
Since the laser light is outside the visible light spectrum, it is by default insensitive to visual colors.
At the same time, it is sensitive (to a degree) to hues at the bottom of the IR spectrum, which may or may not be accounted for in the process of intaglio printing. The scanner is sensitive also to the albedo of the scanned surface, less at the blue end of the spectrum and more at the red one.
As shown in Figure 4, the scanner carriage assembly, generally referenced 15, moves along a trajectory normal to the one of feeding the specimen in the device (and parallel to the one of scanning). The mechanism for moving scanner carriage assembly comprises a worm gear 13 driven by a DC motor 14 through a drive belt lS. A limit switch at the end of free movement produces a signal, which is used by the processor to determine the position of the scanner carriage assembly.
Figures 5 and 6 show the main scanning platform 20, a passive flattener 21, a transport mechanism 22, sensors 23 and guide 24.
The main scanning platform 20 is a polished copper plate to which are attached the flattener 21, the ~1~0477 transport mechanism 22, the sensors 23, and the scanner assembly 15. The plate's polish ensures flawless movement of the specimen. The scanner is mounted on spacers above the plate 20. The direction of movement of the scanner carriage is parallel to its shorter side. The specimen is loaded along the longer side of the platform. This type of arrangement allows for scanning at any given point of the specimen.
The scanner assembly lays parallel (maximum allowed offset 10 microns) to the plane of the scanning platform, to ensure precision and consistency of the measurements.
The passive flattener is made out of PlexiglasTM and mounted on top of the platform 20, allowing clearance of one and a half bill thickness between them. The flattener has an air gap along the trajectory of scanning, to allow direct light beam on the specimen's surface. Thus, the latter is kept reasonably flat while being scanned, to avoid errors due to wrinkles for example.
The transport mechanism consists of two rubber rollers 24, the top one on a fixed axis, while the bottom one can move vertically about 0.5 mm. A coil spring ensures that both rollers are brought together, when the specimen is tightly caught between them. The springs allow for specimens with different thickness to be loaded for scanning.
The top roller is motor-driven by way of a plastic gear assembly 25. The motor 26 is a DC reversible micromotor and is driven by the driver circuitry on the printed circuit board.
The sensors 23 are two infrared LED-photo diode couples. One of them is so positioned, that signals the printed circuit board when to start processing a ,'~ ' , .
', : .
.: '.
~V~4'~7 specimen, the other, accordingly, when the specimen is positioned for scanning.
As shown in Fig~re 7, The LCD serves as a machine-human interface, providing the user with instructions, information and the results of scanning in natural language (different and multiple-language displays can be supported). It con ists of a LCD and a control processor board. The LCD displays the information passed by the control processor board. The control processor board communicates with the printed circuit board (Fig. 8)) and interprets the received machine language code for the purposes of the LCD. The control processor board consists of I/0 buffers 31, character generation circuit 32, timing generation circuit 33, display data Random Access Memory (RAM) and LCD drivers 34. There are 132 characters available at 5x7 display resolution (2-line, 16 characters per line display capability).
As shown in Figure 8, The printed circuit board contains all electronic components necessary for the operation of the device. These include the main processor 40, clock circuitry 41, reset circuitry, security circuitry, D/A converters 42, data buffers 43, drivers 44, various interfaces 45-48.
me power supply, shown in Figure 10, the necessary power to the printed circuit board and the LCD. It is of the switching variety and is capable of operating in the range of 110 - 240V AC, 50/60Hz. It consists of an input rectifier and input filter 50, a flyback transformer driver transformer 51, an output rectifier and filter 52, a output voltage watchdog and control circuitry 53.
me power supply delivers energy to the printed circuit board, necessary for the normal work of the device. It starts working as soon as it is plugged in (in a battery version, at the moment of connecting the battery). All other blocks receive their energy from the printed circuit board, according to their needs. At first, the main processor executes reset/initialization routine, which sets the device in a known state. During this time (2.8 microseconds) no useful function can be performed. Immediately after this, the processor switches in a low-power wait mode, which can be indefinitely long.
At this point the laser LED 1 is off, and the LCD is on and displays a status message and the circuitry is in a WAIT state. It will stay in this state until either a specimen is fed in~o the inlet of the device, or power is disconnected (battery dead in case of a battery version).
When a specimen is fed into the device, the microprocessor wakes up and starts executing the scanning program. First, the laser is turned on at full power, then the specimen is positioned inside the device and the scanner assembly carriage is positioned on top of the area of interest. This is accomplished by the motor and worm gear.
Next, at the LCD the status message is changed. Now, the device will not accept any other jobs, until the current one is completed. Scanning begins with the establishment of the average Focus Value (FV). To establish the FV, the fosusing lens is fully retracted and mo~ed down at a certain rate. The scanner assembly is so mounted on top of the scanning bed that the surface of the latter is within the focusing range of the lens. As the lens moves downwards, the point reached where the specimen's surface appears close to the focal point of the lens. This causes some light to be reflected back to the sensors. The laser emission feedback system makes the necessary corrections in the emission power of the laser, to compensate for different albedo factors (of different specimens). Essentially, the laser power is reduced by certain amount so that different colors look alike to the sensors. As the lens moves further down, it reaches the - ,. .
~, ,~ ,. ~ ,.. f .
~100~77 point when the beam is maximally focused on the sur~ace of the specimen. The processor takes the measure of the current required to deflect vertically the focusing lens by this amount and assigns to it the label FV(1). The FV
is generated as a result of the condition code FE
supplied by the main beam sensors' array. In both cases FV is a true description of the vertical position of the lens relative to the specimen. Now the routine splits into two parallel branches:
The processor makes an adjustment to the FV, preparing it for the next step scanning. Appropriate changes are made in the vertical positioning of the focusing lens. Next, the processor moves the focusing lens horizontally a step. Some delay is also necessary to compensate for the mechanical inertia of the 2-axis Carriage Assembly.
After scaling down the FV, the processor executes the neural network computations on it using the pre-recorded weight map. The results of these computations 2C are cumulatively stored in a virtual two dimensional array for further processing.
Upon finishing these tasks, the processor proceeds with finding FV(2), much in same manner, except for the initial vertical position of the focusing lens. Instead of 0, the adjusted FV is used for determining the initial state.
This process repeats a predetermined number of steps and after the last one is executed, the focusing lens is positioned in neutral position, the specimen is transported out of the device, while simultaneously the processor analyzes the results stored in the array.
According to the result, the processor passes a message to be displayed: either affirmative, negative or percentage probability of the specimen being an original.
., ,,.j;
.
~' "' ' ' ,,~ , ' -, ,~ , .
2~0~77 After some delay, providing no other specimen is fed into the device, the processor sets the circuitry back to the low-power WAIT mode.
The described scanning process does not mark or mutilate the specimens in any way. However, some results may be stored, depending on their importance, for the purposes of the system self-enhancement. These results are not associated in any way with a particular bill, but are rather generalization of a field expertise.
The sampling actually consists of subsequent focusing sessions performed at a given rate. In other words, the lens is made to closely follow the bumps on the specimen's surface.
Provisions have been made in the device described not to take into consideration the steepness of slopes.
Thus, a specific value is assigned to FV every time a measure is taken at a steep slope on the specimen's surface. To determine 3-dimensionality, it is enough to have a secure measure of having a slope, not what exactly slope it is. In another application of the novel process, the slope characteristics become important and are accounted for (e.g., micro topographical maps).
Figure 10 shows the projected time-amplitude relationships of the FE, FV, and AV. As shown in Fig. 10 the FV signal follows exactly the envelope of the specimen's surface, although it may be shifted vertically relative to the abscissa. The FV is accepted as a measure of the relative height of an element of the specimen. The FE signal is different from zero only when a change of the focus value is required, i.e., when there is a slope on the specimen. The direction of change of FE depends on what direction we want to move the lens in. Again, FE
amplitude is not related to the steepness of the scanned slope, but only to its change, it is just a servo signal 2~a~rl7 for the 2-axis device to move. AV depends both on the steepness of the slope and the reflecting ability of the surface, so the flatter and more reflective the specimen, the higher the value of AV.
As shown in Fig. 10, and as explained above, the system responds to the presence or absence of reflected light along the path of the incident light beam to determine. Where insufficient light is reflected back, - i.e. in the presence of a steep slope, a zero is returned, otherwise one of 255 values representing the height of the ink surface is returned. This creates a relief pattern, which is a unique signature of the printing process and consists of the presence of slopes and heights of the ink surface across the specimen.
Figure 11 shows a scanning line consisting of N
pixels 61, each capable of having one of 255 values, zero representing the absence of sufficient reflected light as described above. These can represent the elements of a vector of a elements, each of which is connected to one of 16 nodes 60 of the self-organizing map.
What happens in the learning process is that after each scan of say legitimate intaglio the value of each pixel is multiplied by the weight associated with its connection to each node and the weighted values for each pixel are accumulated in each node. Initially the weightings for each connection are random. After this process one of the nodes will have a higher accumulated value than the others, and this is taken to be the node corresponding to genuine intaglio print. The weightings are then adjusted to still further favor this node.
The same process is repeated for various types of counterfeit or forged documents, with a different node being favored in each case until an equilibrium condition , . .
C~100~77 is reached where each node represents a certain type of printing process.
The system software is written in assembly language.
Residing on-chip, and the user having no access to the program, it is classified as firmware. It comprises several routines to satisfy the control, computation and security needs of the device and a weights' map, to implement the AI elements of the process.
From software point of view, both sequential and parallel processing are employed, and the system demonstrates both linear and nonlinear behavior, as required. The sequential part of the program is executed sequentially, requiring 100% of the processor time, one step at a time. The parallel processing is used when computing the equations for the neural network, when the processor is used to crunch more than one piece of data at a time. This yields a saving of processing time, thus decreasing the total process time.
. - .- .. ,.,,, ~ . , , '
Claims (30)
PROPERTY OR PRIVILEGE IS CLAIMED ARE DEFINED AS FOLLOWS:
1. A method of detecting counterfeit or forged documents where the genuine documents contain matter printed in relief by a security printing process, comprising storing data representing a characteristic relief profile of matter printed by said process, scanning a specimen document to generate data representing a relief profile of said printed matter thereon, comparing said generated data with said stored data, and determining the validity of said specimen on the basis of said comparison.
2. A method as claimed in claim 1, wherein said stored data are derived by scanning numerous legitimate samples printed by said security printing process and numerous counterfeit or forged samples printed by other printing processes.
3. A method as claimed in claim 2, which is implemented by a neural network.
4. A method as claimed in claim 1, wherein said printing process is intaglio.
5. A method as claimed in claim 1, wherein said relief profile comprises the pattern of slopes and heights of the ink surfaces of said matter printed in relief.
6. A method as claimed in claim 1, wherein said relief profile includes the angles of said slopes.
7. A method as claimed in claim 2, wherein said printing processes are classified by said stored data, and said validity determination is performed by identifying the printed process employed on said specimen with the aid of said stored data.
8. A method as claimed in claim 3, wherein said printing processes are classified by said stored data, each said printed process being associated with a node of a memory map of said neural network, and said validity determination is performed by identifying the node most closely matching the generated data.
9. An apparatus for detecting counterfeit or forged documents where the genuine documents contain matter printed in relief by a security printing process, comprising means for scanning a specimen document to generate data representing a relief profile of said printed matter thereon, means for storing data representing a characteristic relief profile of matter printed by said process, and means for comparing said generated data with said stored data to determine the validity of said specimen document on the basis of said comparison.
10. An apparatus as claimed in claim 9, wherein said scanning means comprises a scanning bed for receiving said specimen, means for positioning said scanning means over a desired area of said specimen, means for producing a collimated light beam, a lens for focusing said light beam onto said specimen on said scanning bed, means for determining when said light beam is in focus on the surface of said specimen, means for effecting relative vertical displacement of said specimen and said lens, means for effecting relative horizontal displacement of said specimen and said lens, and means responsive to the relative vertical displacement of said specimen and said lens to generate said at least part of said relief profile data.
11. An apparatus as claimed in claim 10, wherein said positioning means comprises a worm gear for effecting said relative displacement of said scanner means and said scanning bed.
12. An apparatus as claimed in claim 10, wherein said means for determining when said light beam is in focus on the surface of said specimen comprises means for separating out light reflected along the path of the incident light beam, and a photodetector responsive to the presence of light reflected back along the path of the incident light beam.
13. An apparatus as claimed in claim 12, wherein said photodetector comprises an sensor array responsive to eccentricity in the reflected light beam to determine an out-of-focus condition.
14. An apparatus as claimed in claim 13 wherein said sensor array is divided into quadrants, said sensor array being responsive to uneven illumination of said quadrants to determine said out-of-focus condition.
15. An apparatus as claimed in claim 10, further comprising means for detecting the presence or absence of slopes on said printed matter and producing corresponding slope data, said slope data being included as part of said relief profile data.
16. An apparatus as claimed in claim 12, wherein said photodetector also generates slope data indicating the presence or absence of slopes on said printed matter, said slope data being included as part of said relief profile data.
17. An apparatus as claimed in claim 15, further comprising a secondary sensor array for generating slope angle data representing the angles of slopes on said printed matter, said slope angle data being included as part of said relief profile data.
18. An apparatus as claimed in claim 12, wherein said separating means comprises a partially reflecting mirror.
19. An apparatus as claimed in claim 10, wherein said means for effecting relative vertical displacement of said specimen and said light beam comprises a coil immersed in a stationary constant magnetic field for displacing said lens relative to said scanning bed.
20. An apparatus as claimed in claim 10, wherein said means for effecting relative horizontal displacement of said specimen and said light beam comprises a coil immersed in a stationary constant magnetic field for displacing said lens relative to said scanning bed.
21. An apparatus as claimed in claim 9, wherein said comparing means comprise a microprocessor and a memory implementing a neural network embodying a self-organizing map.
22. An apparatus as claimed in claim 21, wherein said self-organizing map comprises a plurality of nodes, each associated with a respective printing process.
23. An apparatus as claimed in claim 9, wherein said security printing process is intaglio.
24. Relief profile sensing apparatus, comprising a scanning bed for receiving said specimen whose relief profile is to be sensed, means for producing a collimated light beam, a lens for focusing said light beam onto said specimen on said scanning bed, means for determining when said light beam is in focus on the surface of said specimen, means for effecting relative vertical displacement of said specimen and said light beam, means for effecting relative horizontal displacement of said specimen and said light beam, and means responsive to the relative vertical displacement of said specimen and said lens to generate said relief profile data.
25. An apparatus as claimed in claim 24, wherein said means for determining when said light beam is in focus on the surface of said specimen comprises means for separating out light reflected along the path of the incident light beam, and a photodetector responsive to the presence of light reflected back along the path of the incident light beam.
26. An apparatus as claimed in claim 25, wherein said photodetector comprises a sensor array responsive to eccentricity in the reflected light beam to determine an out-of-focus condition.
27. An apparatus as claimed in claim 26, wherein said sensor array is divided into quadrants, said sensor array being responsive to uneven illumination of said quadrants to determine said out-of-focus condition
28. An apparatus as claimed in claim 24, wherein said separating means comprises a partially reflecting mirror.
29. An apparatus as claimed in claim 24, wherein said means for effecting relative horizontal displacement of said specimen and said light beam comprises a worm gear for displacing said lens relative to said scanning bed.
30. An apparatus as claimed in claim 25, wherein said photodetector is divided into a plurality of equal segments, and means are provided to determine the coverage of said various segments by the reflected beam.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CA002100477A CA2100477A1 (en) | 1993-07-14 | 1993-07-14 | Method and apparatus for detecting counterfeit documents |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CA002100477A CA2100477A1 (en) | 1993-07-14 | 1993-07-14 | Method and apparatus for detecting counterfeit documents |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CA2100477A1 true CA2100477A1 (en) | 1995-01-15 |
Family
ID=4151928
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CA002100477A Abandoned CA2100477A1 (en) | 1993-07-14 | 1993-07-14 | Method and apparatus for detecting counterfeit documents |
Country Status (1)
| Country | Link |
|---|---|
| CA (1) | CA2100477A1 (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| RU2455690C2 (en) * | 2006-08-18 | 2012-07-10 | Де Ля Рю Интернэшнл Лимитед | Method and apparatus for identifying raised material |
-
1993
- 1993-07-14 CA CA002100477A patent/CA2100477A1/en not_active Abandoned
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| RU2455690C2 (en) * | 2006-08-18 | 2012-07-10 | Де Ля Рю Интернэшнл Лимитед | Method and apparatus for identifying raised material |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| JP5314419B2 (en) | Bill authenticity judging method and bill authenticity judging device | |
| EP2081130B1 (en) | Methods and apparatuses for creating authenticatable printed articles and subsequently verifying them | |
| EP1730675B1 (en) | Authenticity verification methods, products and apparatuses | |
| US6373965B1 (en) | Apparatus and methods for authentication using partially fluorescent graphic images and OCR characters | |
| US20070153078A1 (en) | Cartridges For Reprographics Devices | |
| WO2005076229A1 (en) | Document processing system using captured primary and secondary pictorial images which are compared to respective master images | |
| GB2411954A (en) | Determining a signature from an article to verify its authenticity | |
| GB2417707A (en) | Printer with integral scanner for authenticatable document creation and verification. | |
| JP2003141595A (en) | Paper sheet discriminating system | |
| KR101076322B1 (en) | Method for discriminating the authenticity of paper money by using a contact image sensor | |
| CA2100477A1 (en) | Method and apparatus for detecting counterfeit documents | |
| KR20010076926A (en) | Discrimination method for money and discriminator thereof | |
| National Research Council et al. | Counterfeit deterrent features for the next-generation currency design | |
| JPH04227365A (en) | Copying machine | |
| KR20080071028A (en) | Image forming apparatus to prevent counterfeiting of money and method thereof | |
| JP3385357B2 (en) | Reader for binarized data given by fine perforations | |
| CA3233218A1 (en) | Security solution with a digital image with integrated security feature, and image conversion method and image conversion device for the production thereof | |
| EP0936579A2 (en) | Sheet recognition system | |
| Chopra et al. | Portable Fake Currency Detector in Indian Scenario | |
| JPH04227372A (en) | Picture processing method | |
| JPH04227367A (en) | Picture processing unit | |
| KR19980062837U (en) | Counterfeit Bill Recognition Device |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| FZDE | Dead |