RU2745032C1 - Method of additional authentication of valuable document - Google Patents

Abstract

FIELD: computing.

SUBSTANCE: invention relates to the field of automated authentication of protected documents, and concerns a method of additional verification of a valuable document authenticity. When implementing the method, a digital image of the document being checked is obtained and the location of the control zone of the digital image of the document is determined, where there is a moire image formation property. When checking, a digital image is used consisting of sequentially scanned lines, as well as the skew angle of the document being checked in relation to the scanning direction. When scanning, the length of the registration area on the document measured in the scanning direction, from which the sensor receives at least 80 percent of the radiation recorded in the course of a single line scanning, is less than the interline spacing of the checked document relative to the sensor. The check consists in determining the presence in the control zone in the digital image of the checked document of the controlled features characteristic of the moiré image formed in the digital image of the authentic document for the specified type and orientation in the same control zone, with the skew angle of the original document equal to the skew angle of the checked document.

EFFECT: increased reliability of the confirmation of the protected document authenticity.

14 cl, 10 dwg

Description

The invention relates to the field of automated authentication of secure documents, such as banknotes. To protect these documents from counterfeiting, special printing technologies are used that are not widely available. Special printing technologies allow you to print an image with special characteristics, which are very difficult to reproduce using publicly available technologies. Such special characteristics of the printing process are, in particular, a very small width of continuous lines reproduced on a banknote (can reach values less than 25 microns), a small size of minimally reproducible white space elements (up to 5 microns), and an extremely high spatial discreteness of image synthesis, for example, the resolution of a digital image can be 10,000 dpi and higher. Hereinafter, the term "sampling rate" refers to the frequency of the location of pixels in the rows and columns of a digital image and is similar in meaning to the previously used expression "scan lineature". Resolution is measured in dots per inch (dpi) or dots per millimeter. Discreteness should not be confused with the term "resolution", which characterizes the transmission of alternating black and white lines in a digital image and depends not only on the number of pixels per unit length, but also on the quality of the optical system used. Resolution is measured in lines per inch (lpi) or line pairs per inch (lppi) or per millimeter that can be distinguished.

The printed design on the banknote is designed in such a way that its correct reproduction would essentially rely on the listed special characteristics of the printing process. When trying to reproduce this pattern on publicly available equipment, such as copiers or high-quality printers, the pattern of the banknote is reproduced with distortions that are clearly visible to the eye and attract the attention of even an unskilled person. These distortions look like the disappearance of certain areas of the image, the replacement of the image with a solid dark tone, as well as the appearance of new image elements that are not observed on genuine banknotes. For example, on the copied banknote the inscription “COPY” appears.

Another type of distortion that is highly visible is the so-called moire. Moire appears as alternating light and dark stripes superimposed on the original image. On copies of banknotes, it occurs as a result of the interference of the banknote pattern with a periodic line-column structure of a digital image that is scanned by a scanner and / or printed by a printer. This type of moiré is sometimes called sampling moire because it occurs when a continuous pattern is converted into a discrete digital image. When banknotes are printed on special equipment, their pattern is synthesized with such small discreteness that interference with the line-column structure does not occur. However, when trying to scan and / or print on publicly available equipment with a much coarser resolution, interference appears well on parts of the pattern with a periodic or quasi-periodic structure. In order to enhance the manifestation of moirés during copying, the pattern of protected documents is deliberately saturated with special elements that contain thin lines that form quasiperiodic structures. Such elements are found on most types of modern banknotes. They are considered a special type of printing security and effectively protect banknotes from copying. Thus, security documents having the property of forming a moire image during scanning form an extensive class of security documents.

Unfortunately, counterfeiters have found a way to bypass the scan moiré security. They scan the original banknote with a high-quality general-purpose scanner and then process the image on a computer. This processing is aimed both at eliminating the developed moiré images in the area of special elements that have arisen during scanning, and at replacing special elements with their simplified versions, which can then be printed on a conventional printer without the danger of creating a moiré image. In particular, with such processing, a quasiperiodic structure consisting of thin lines is replaced by points and segments of thicker lines, in the arrangement of which considerable randomness has been introduced. In the most carefully executed forgeries, the directionality of the line segments is also retained. When observed with the naked eye, such a substitution remains unnoticed, since the level of the average optical density, which the eye confidently fixes, remains the same as on a genuine banknote. The human eye does not see the differences of a fake in the structure of fine lines, since their confident observation is impossible without magnifying devices. However, the observer hesitantly determines the very presence of lines on a fake and their approximately correct direction. In order to reliably identify such a forgery, it is necessary to examine the microscopic structure of the pattern on the document, which already belongs to an expert study and cannot be performed at all points of acceptance of banknotes.

It would seem that automated banknote processing equipment should confidently detect such counterfeits. However, this does not happen, since the overwhelming majority of machines for processing banknotes scans the image on banknotes with a discreteness of the obtained digital image of 100-200 dpi or even less, which does not exceed the resolution of the human eye. Therefore, the task of verifying the authenticity of the thin-linear structure of the banknote pattern on automated equipment is very urgent.

A simple increase in the resolution of a digital image to several thousand dpi would solve this problem, but it conflicts with the required processing speeds of banknotes. If an expert examination of a document is carried out, then high-resolution scanning and processing of scan results are not limited in time and can be successfully performed on modern equipment. However, when it comes to the productivity of automated processing, reaching 10-25 banknotes per second, the state of the art can not provide the required data transfer rate from the photodetector matrix of the scanner, nor the computing power to process this data. The lag in these parameters from the level required for confident registration of fine lines of a protected document is more than one order of magnitude. Such a lag makes it possible to consider such a solution technically unfeasible at the moment. Therefore, it is necessary to look for alternative technical solutions that could be implemented at the state of the art.

Known patent RU 2668856 (publ. 03.10.2018, IPC G07D 7/20), in which to increase the discreteness of the digital image, a photodetector matrix with low resolution is used together with a laser lighting system that provides spatial modulation of document illumination using a sequence of random images. Using a linear photodetector array, document responses to a set of many random images are recorded. Then, as a result of computational processing of these responses, a part of the banknote image is formed, containing several lines, in each of which the number of pixels significantly exceeds the number of elements of the photodetector matrix. This approach makes it possible to control microprinting elements, including both thin lines and micro-texts. The disadvantage of the known device is the complexity of implementation and low speed of operation. A laser lighting system includes many optical elements, some of which, such as a fiber laser and a modulator, are very complex and expensive. The optical design needs careful alignment. Computational processing of the obtained data requires a noticeable increase in processor power in comparison with those processors that are now used in counting and sorting machines. The need to interrogate the photodetector matrix many times using various random images reduces the scanning speed of the document by a factor.

If the document has the property of forming a moiré image when scanned, then this moiré image can be used not only to make an attempt to copy the document obvious, but also to directly verify the authenticity of the document. In other words, it is possible to analyze the moiré image generated during the scanning of the document in the banknote processing equipment. If the moiré image on the checked document has the same characteristics as the moiré image that appears on the original document, then this indicates a high probability of coincidence of the thin-line pattern on the checked document with the thin-line pattern of the original document.

As such, the manifestation of moiré sampling on scanning in banknote processing equipment is well known and is generally considered to interfere with reliable banknote authentication. This is due to the fact that the moiré image cannot be regarded as a stationary image, since as a result of each scan, even of the same document, the moiré image is obtained completely different than in the previous scan. This is due to the influence of a number of random factors, such as the skew of the document in relation to the direction of the scanner line and the shift of the banknote pattern relative to the moment of the start of scanning. As a result, an unpredictably changing moire image is superimposed on the stationary banknote pattern. This introduces a significant noise component into such characteristics of the banknote image as the average level of optical density and dispersion of optical density, and also distorts the shape of the pattern elements. All these distortions worsen the accuracy of the authentication algorithms based on statistical parameters of optical density and on the analysis of the shape of the pattern elements.

In the prior art, examples are known when special measures are used to combat the harmful effects of sampling moirés. So, in the patent RU 2631161 (published on September 19, 2017, IPC G07D 7/12) it is proposed, during scanning, to randomly change the line spacing to a small extent. Thanks to this accident, it is possible to break the strict periodicity of the line-column structure of the digital image, which blocks the interference with it of the quasi-periodic structure of the document drawing and the formation of continuous moiré fringes associated with the interference.

On the other hand, the use of moire images as a security effect for authenticating security documents is well known and widely used in practice. For the development of moiré, as a rule, a so-called developing image is used on a separate medium, which is superimposed on the pattern of the security document at the time of verification. For example, a transparent polymer film is used as a carrier, and the developing image is a specially designed thin-line pattern applied to the film using black ink. When the developing image is overlaid, only certain, precisely calculated areas of the design on the surface of the document remain visible to the viewer through the transparent areas between the black lines. The interference between the developing image and the thin line pattern of the security document results in a highly visible high-contrast moire pattern.

A similar solution is described, in particular, in patent RU 2328036 (publ. 27.06.2008, IPC G07D 7/12). In the simplest case, the developing image has the form of parallel straight lines of equal thickness, following with a constant shift relative to each other. The pattern of the security document is formed in such a way that when the developing image is overlaid, moiré appears on it in the form of an additional, previously hidden pattern. Special techniques for designing a picture of a secured document can achieve a dynamic visual effect. The dynamic effect is that as the developing image moves over the surface of the security document, the shape of the moiré image undergoes continuous complex transformations. The same patent describes a device for automatic authentication, in which a document covered with a film with a developing image is photographed using a digital camera. After that, the moiré image is analyzed according to the digital photograph received from the camera, and a conclusion is made about the authenticity of the document.

The described solution has a high reliability of authentication, since to reproduce the planned moire image, it is required to accurately reproduce the pattern of a genuine banknote, with thin continuous lines embedded in it and fine sampling of the position of these lines. A counterfeiting attempt will not successfully reproduce the intended moire pattern unless counterfeiters are able to gain access to highly protected secure printing technologies. However, this solution is only suitable for expert research and is not suitable for automated processing of banknotes at high speed. This is because the imposition and positioning of the developing image implies that the document being checked is stationary. The observed moiré depends very much on the relative position and orientation of the developing image and the document; therefore, even the slightest mutual displacements are unacceptable during the registration of a digital image. In counting and sorting machines, banknotes move along complex paths at speeds of up to 10 kilometers per hour, so it would be very difficult to superimpose and continuously, without any relative displacements, hold the developing film on the document.

Thus, we can talk about two mechanisms of moiré formation. When trying to copy protected documents, sampling moire occurs. Moire sampling is, in fact, determined by the residual errors in the translation of an analog image of a document into a discrete digital image, and its contrast cannot be very high. Since the residual sampling errors refer to both the line and column directions, the sampling moire does not have a clearly defined direction of the highest sensitivity. When examining a security document using a developing film, a moire of interaction between two images that are on tightly aligned carriers occurs.

Another type of moire is generated by the interaction of the document image with the developing image. The developing image provides the highest sensitivity to the position of the elements of the drawing of the document in the direction perpendicular to its lines. In principle, both the one and the other type of moire can be used for authentication, but there are significant differences between these moirés. Since the developing image itself has a very high contrast, the moiré that occurs when interacting with it has a significantly higher contrast than the sampling moire that occurs when scanning the same document. In addition, the border of the black ink in the developing image has a high sharpness, and the tight alignment of the media provides little blur in the overlapping area of the two images. This makes the moiré generated by the developing image highly sensitive to the microscopic characteristics of fine lines on the document, such as thickness, continuity and uniformity. Experience shows that it is possible to control these characteristics with a sensitivity of 10-20 microns or even better, which is helped by the special sensitivity of this type of moiré to displacements along the normal to the line. The highest sensitivity to defects in fine lines is achieved when the width of the transparent areas between the black lines of the developing image is very small. This can be explained by the fact that the sensitivity turns out to be the higher, the larger the relative fraction of the width of the transparent area can cover the defect of the thin line of the drawing on the document. As will be shown below using the mathematical apparatus of two-dimensional Fourier transform, with a small relative width of transparent areas, the ratio between the thickness of the lines on the document and the period of their repetition is approximately equal to the ratio between the width of the lines of the moiré image and the period of their repetition.

Thus, the moiré generated by the developing image is highly effective for authentication. In contrast to this type of moiré, sampling moire is based on the values of individual pixels of a digital image, which are about 100 microns in size (with a resolution of 200-300 dots per inch) and convey the average value of the optical density of the document. In reality, the averaging zone in scanners is 2 or more times larger than the pixel size and, moreover, has blurry edges. This leads to a low contrast of the moire image and the difficulty of detecting it against the background of high-contrast elements of the printed image of the document. Moire sampling reflects the presence, direction and repetition period of fine lines. However, due to the formation of moiré with both columns and rows of a digital image, the direction and period of repetition of thin lines can be determined ambiguously. This type of moiré does not allow one to confidently determine the thickness of thin lines, their continuity and equal thickness. Therefore, sampling moiré is not very efficient as an authentication tool.

Thus, sampling moire is a parasitic effect with low potential efficiency for document authentication, while moiré interaction between document and developing image can be considered a highly effective authentication tool. On the other hand, sampling moiré occurs naturally in high-speed automated document processing, while interaction moire is poorly applicable for this kind of processing. It would be desirable to create a technical solution that combines the ease of obtaining sampling moiré in automated processing machines with a high accuracy of the moiré interaction between the document and the developing image, but without the aforementioned disadvantages of each of the named types of moiré.

In US patent US 7561308 (publ. 14.07.2009, IPC G06K 15/00), a developing image in the form of a digital image is superimposed on the scanned digital image of the document being checked in the course of computational processing. Moire in this method is generated by the developing image and has all the positive properties of this type of moire. However, for the formation of a moiré image, it is required that the image of the document being checked be scanned at a high resolution, in which the moiré-forming elements are clearly reproduced. This limits the application of this method both from the point of view of the required scanning equipment and from the point of view of possible performance. This method cannot be applied in high-speed machines for automated processing of security documents, such as banknote sorters.

Known Chinese patent CN 107730707 (publ. 02/23/2018, IPC G07D 7/20), in which the checked security document is scanned with a low resolution of 200 dpi, which leads to the appearance of sampling moirés in the image. Moiré locations in images of verifiable and known to be genuine documents are analyzed using the co-occurrence matrix method and compared with each other. The well-known patent does not specify the conditions for the formation of moiré and relies on sampling moire as a natural property of the document image. More precisely, moire is treated as an additional texture in the document. The co-occurrence matrix method is a proven method for statistical analysis and comparison of textures, therefore, it allows you to approximately confirm the similarity of moiré images of an original and a verified document. However, the moiré image is subject to major changes depending on the skew angle of the document when scanned. This variability is not taken into account in the known patent, which limits the accuracy of checking the similarity of moire images. The low contrast of the sampling moire compared to the contrast image of the document itself increases the errors and further reduces the verification accuracy. The device and method according to the known patent, from the point of view of speed and complexity of implementation, may well be applied in high-speed machines for automated processing of secure documents. However, this solution does not allow for reliable rejection of well-made fake documents.

Another disadvantage of the solution, according to the known patent, is the possibility of easily deceiving the check with the help of moiré by those counterfeiters who know about the essence of the check being carried out. These counterfeiters can add an additional halftone structure to the printed image of the document, which mimics the appearance of the moiré expected on the original document. Since the known patent does not analyze the process of moiré formation and does not take into account the influence of the skew angle on it, it is not possible to distinguish imitation from real moiré on the basis of only a statistical analysis of the image. When using moiré to verify the authenticity of a security document in a machine for automated processing, it should be borne in mind that at some angles of skewing of the document in the machine, moire may not be recorded at all during scanning, or it may be unsuitable for analysis. Because of this, moire verification cannot be relied upon as the primary means of verifying the authenticity of a document. Moire verification is a very powerful means of authentication, but it should only be used in addition to other verification methods. Such methods include, for example, checking a printed image of a banknote using radiation with different wavelengths, checking for the presence of magnetic marks, checking marks containing a phosphor, and others. As a prototype of the claimed invention, the aforementioned patent CN 107730707 is selected.

Observing the results of scanning documents with low resolution, in the range from 100 to 300 dots per inch, allowed the authors of the claimed invention to conclude that part of the printed image on the surface of the document is sometimes absent in the resulting digital image. Moreover, this is not associated with the smearing of small elements, described in detail in the literature on optics, namely, with the disappearance of these elements. A more in-depth analysis of the device of modern linear image sensors, the so-called contact type, showed that, with a certain combination of design parameters of the sensor, the width of the area on the document from which the sensor receives radiation during the scanning of one line turns out to be very small. This is manifested, in particular, in the fact that when such sensors are used in high-resolution scanners (more than 300 dots per inch), the resolution of the scanner in the direction of movement of the sensor turns out to be much higher than in the direction along the sensor. When a sensor with the same design is used for scanning with a low resolution, the sensor's field of view does not overlap when scanning adjacent lines, which is why some of the small elements of the printed image on the document surface do not fall into the sensor's field of view or partially fall into the sensor's field of view. This, in turn, leads to the partial or complete disappearance of these elements. This effect gives rise to certain difficulties when scanning macro objects on the surface of a document, but, at the same time, forms the basis of the claimed invention.

The technical result of the claimed invention is to increase the reliability of authenticity confirmation of a security document having the property of forming a moire image during scanning.

This result is achieved due to the fact that in the method of additional verification of the authenticity of a valuable document, performed after the basic verification of authenticity, during which a digital image of the document to be checked is obtained and the type of document and its orientation during filing for verification are determined, on the basis of which the location of the control zone is determined a digital image of a document, where there is a property of forming a moire image, while for additional verification of the authenticity of the document, the mentioned digital image of the document being checked is used, consisting of sequentially scanned lines and obtained under the following condition: during scanning, the length of the registration area measured in the scanning direction by document, from which the sensor receives at least 80 percent of the radiation recorded during the scanning of a single line, is less than the line spacing of the offset of the document being checked relative to the sensor, as well as the angle of intersection the axis of the checked document in relation to the scanning direction, moreover, the specified additional check consists in determining the presence in the mentioned control zone in the digital image of the checked document of controlled features, which are predetermined as characteristic of the moire image formed in the digital image of an authentic document for the established type and orientation in the same control zone, when the skew angle of the original document during scanning is equal to the skew angle of the checked document, and according to a predetermined rule, a decision is made to confirm the authenticity of the checked document.

The claimed invention relates to authentication in addition to basic document authentication. Basic authentication, as is known in the art, typically involves scanning a document submitted for verification line by line to obtain a digital image thereof. Based on the digital image, the type of document and its orientation are determined when it is submitted for verification, and then the authentication is carried out, provided for a certain type of document and its orientation. According to the claimed invention, an additional check is carried out for a predetermined type and orientation of the document, that is, when it has already been confirmed during the main authentication that the document being checked is of a predetermined type and orientation. To carry out the claimed method, a scanned digital image of the checked document is used, which is usually obtained during the main authentication.

If the basic authentication can be carried out for different types and orientations of the document, then for at least some combinations of these types and orientations, separate implementations of the claimed method of additional authentication can be provided.

In accordance with the claimed method, a specific requirement is imposed on the conditions for obtaining a scanned digital image of the document being checked. This requirement consists in the fact that during scanning, the length of the registration area on the document, measured in the scanning direction, from which the sensor receives at least 80 percent of the radiation recorded during the scanning of a separate line used to obtain a digital image of the document being checked, should be less than line-to-line. step of displacement of the checked document relative to the sensor. As mentioned earlier, this requirement can sometimes be fulfilled in known scanning devices due to a certain combination of their design parameters. However, such a requirement for scanning is not described in the known sources and it is not known about its application to achieve the claimed technical result. This determines the novelty of the claimed invention. In the claimed invention, a film with a developing image is not used, nor is it used in the prototype. However, it can be said that, during scanning, a virtual developing image is superimposed on the banknote pattern. Indeed, due to the fact that the interline pitch of the document displacement relative to the sensor exceeds the length of the registration area on the document, from which the sensor receives at least 80 percent of the radiation recorded during the scanning of a single line, a part of the pattern on the image in the form of narrow stripes corresponding to each line offset is practically not registered by the sensor. That is, it can be considered that, when scanning, a developing image is superimposed on the document pattern in the form of uniformly spaced black stripes of the same width, parallel to the direction of placement of the linear image sensor. The pitch of these stripes is equal to the scanning line spacing, and the gap between them is equal to the size of the area on the document from which the sensor receives radiation during the scanning of one line. The term "virtual" in relation to the developing image described herein reflects the fact that there is no physical medium for the developing image.

Thus, the moire that appears in the digital image obtained in accordance with the inventive method is by its nature not the sampling moire, as in the prototype, but the moire of interaction with the developing image. It has all the positive properties of this type of moire: high contrast and high sensitivity to the characteristics of fine lines in the document drawing. It is also important that the orientation and spatial frequency of the lines of the virtual developing image coincide with the orientation and spatial frequency of the line-column scanning grid. Therefore, those security documents, the pattern of which is specially created for the formation of sampling moiré during normal scanning, with a high probability will create a high-contrast moiré image when scanned in accordance with the claimed method. Accordingly, the characteristics of the thin-line elements of the design of such a security document can be checked in many parameters and compared with the characteristics of the original document. Therefore, the claimed invention provides a higher degree of validation confidence than the prototype.

The influence of the requirement in the claimed method for a scanned digital image of a document on improving the characteristics of the resulting moire image is not obvious to a specialist.

A fraction of radiation that is less than 20 percent of the total radiation detected by the sensor during the scanning of a single line can fall on the sensor from other places in the document outside the registration area, directly from the sensor's lighting system, as well as from the external space around the sensor. This share is due to the imperfection of the sensor as a technical object, as well as the conditions of its operation in a device for processing documents. The reasons for this can be mainly the residual energy outside the scattering spot of the optical system, as well as the influence of dust and microscopic defects on optical surfaces, re-reflection of illumination from the sensor elements, and external stray illumination. The choice of the limit of less than 20 percent for the specified proportion is due to the fact that at this value there is a slight decrease in the contrast of the virtual image. However, this decrease in contrast does not lead to a significant decrease in the contrast of the moiré interaction with the developing image. The contrast ratio of the virtual image remains high enough, in the worst case, close to 5 times. The brightness at each specific point of the moiré is the product of the brightness of the image to be checked and the light transmission of the developing image at the same point. Therefore, the contrast of the moiré, in the worst case, is reduced by no more than about 20 percent compared to an absolutely contrast developing image.

The claimed invention uses a skew angle with respect to the scanning direction that occurs when scanning a document. The skew angle changes the orientation of the virtual developing image and thereby significantly changes the characteristic features of the resulting moiré, such as the direction and length of the spatial moiré frequency vector. In the claimed method, as in the prototype, the check is carried out in the control zone, the location of which is chosen in such a way as to cover at least part of the moiré-forming elements of the document. The presence in the control zone in the digital image of the checked document of controlled features is checked, which are characteristic of the moiré image that appears in the digital image of an authentic document of a given type and orientation in the same control zone, obtained when the skew angle of the original document during scanning is equal to the skew angle of the checked document ... This provides a high degree of confidence in the authentication by taking into account the dependence of the moiré signature on the skew angle. In the prototype, on the contrary, the dependence of the moiré on the skew angle is not taken into account. Thus, the claimed invention provides significantly more reliable confirmation of authenticity than the prototype.

In particular, the claimed invention, in contrast to the prototype, allows, in the overwhelming majority of cases, to detect an imitation of a moire image made by counterfeiters using halftone printing. Taking into account the dependence of the controlled features of the moiré on the skew angle, the probability that counterfeiters accurately reproduced the moiré image, in its features corresponding to the skew angle, which was randomly obtained during the scanning of the document being checked, turns out to be very low.

As follows from what has been said here, the technical result of the claimed invention is achieved both by the very fact of using a virtual developing image based on a given requirement for the scanned digital image, and by taking into account the influence of the skew angle of the checked document on the formation of moiré. As mentioned earlier, it must be borne in mind that at some angles of skewing of the document in the machine, moire may not be recorded at all during scanning, or it may turn out to be unsuitable for analysis. The second possibility arises, for example, when the moiré itself is recorded, but the period of the moiré image exceeds the size of the moiré-forming structure on the document and cannot be measured reliably. This limitation is associated with the essence of the physical phenomenon of moiré itself, is common to all methods based on the formation of moiré during scanning, and manifests itself both for the prototype and for the claimed invention. Due to this limitation, during the application of the claimed method, it may sometimes turn out that additional authentication is not possible. Therefore, the claimed method, like any method based on the control of the moiré image, should be used as an addition to the main check. This allows you to ensure the validity of the decision on the authenticity of the document.

In the case when additional verification is not possible, according to a predetermined rule, a decision can be made to confirm or not to confirm the authenticity of the document. For example, if the machine is configured to work with the deepest authentication, a predefined rule for this case should provide for the absence of confirmation of the authenticity of the document, which leads to its rejection. This type of rejection is usually called technical rejection, since it is not directly related to the authenticity of the banknote. In machines for processing security documents, technical rejection occurs in many cases that impede processing, for example, with a very large skew angle. Conversely, if the machine is tuned to a medium level of verification depth, which sometimes allows for additional authentication to be omitted, a predefined rule may include an authentication decision in this case.

The use of an array of single magnification lenses producing an un-inverted image is a typical solution for linear contact-type sensors used in document scanning. Several lenses are simultaneously involved in transferring the image from the document to the receiving area of the photodetecting elements of the sensor, each of which transmits only a certain part of the radiation beam emitted by a small area on the surface of the document and directed to the receiving area.

The design of a contact linear image sensor suitable for creating a virtual developing image in accordance with the claimed invention may have an optical system designed to form an image of a document on a linear array of photodetectors arranged along a straight line, and containing an array of lenses that form a non-inverted image with a single magnification , and placed in one row or in two parallel rows, and the linear size of the field of view of each lens is more than 3 times greater than the distance between its optical axis and the optical axis of any of the adjacent lenses, while both the line of location of the photodetector elements of the sensor and and the row of lenses in the array are oriented in a direction practically perpendicular to the scanning direction, and the size of the receiving area of each photodetector element, measured in the scanning direction, is less than the line pitch of the document being checked from relative to the sensor.

This design achieves a very small width of the area on the document from which the sensor receives radiation during the scanning of one line, even when the document is displaced from the best focusing position. Such random displacements are almost constantly observed in counting and sorting machines, since the document moves past the sensor in a wide channel and can occupy different positions in this channel with respect to the sensor. The small width of the specified area is due to the fact that the previously mentioned beam, emitted by a small area on the surface of the document and directed to the receiving area, has an asymmetric shape. Namely, it is compressed in the scanning direction and stretched in the perpendicular direction because it consists of several beams passing through adjacent objectives. When the document surface is in the beam waist, that is, in the best focusing position, the size of the small area associated with the receiving area is very small and is determined by diffraction constraints. Due to the asymmetry of the beam, when the document is displaced from the position of the best focusing, the size of the specified small area increases significantly in the direction of the line of location of the photodetector elements of the sensor, but remains small in the direction of scanning. The sensor's field of view consists of individual small areas, each of which is associated with a corresponding receiving area.

The size of the receiving area of the photodetector element in the scanning direction, which is less than the line-to-line offset of the document being checked relative to the sensor, is a necessary condition for the line-to-line offset of the sensor's field of view along the document to exceed the size of the area on the document from which the sensor receives radiation during scanning one line. This is due to the single magnification of the lenses that make up the optical system. The influence of other design parameters of the sensor on the specified width will be discussed in detail below.

When describing the moiré image generated by the developing image, it was stated that its effectiveness for authentication is the higher, the narrower the areas of transparency between the black lines. In terms of the claimed invention, the width of the transparent regions of the virtual developing image is determined by the width of the area on the document from which the sensor receives radiation during the scanning of one line. This width, in turn, is the sum of the size of the sensor's field of view, measured in the direction of the displacement of the document, and the displacement of the document itself over the time interval for which the sensor registers the incident radiation when scanning one line. Hence it follows that in order to increase the efficiency of the check, both the size of the field of view and the offset when registering a line should be minimized.

An additional reduction in the size of the field of view in the direction of displacement of the document can be achieved if the optical system additionally uses elements that limit the path of the rays of each of the lenses. The use of such elements, for example, aperture diaphragms, is well known to those skilled in the art and can further reduce the cross-sectional dimension of the radiation beam measured in the scanning direction when the document surface is positioned away from the waist. In this case, however, the influence of diffraction effects, which leads to an increase in the size of the beam cross section near the waist, must be taken into account.

To reduce the displacement of the document for the time interval during which the sensor registers the incident radiation when scanning one line, you can reduce the duration of exposure to radiation. This makes it possible to reduce the size of the registration area on the document, measured in the scanning direction. For this, the image sensor can be equipped with a backlight system that pulses irradiates the location of the document in the sensor's field of view; moreover, an irradiation pulse is supplied during scanning of each line of the digital image, and it is ensured that the sum of the magnitude of the displacement of the sensor's field of view along the document during this pulse and the size of the field of view of the sensor in the scanning direction was equal to a value less than the interline pitch of the displacement of the field of view of the sensor along the document used to obtain a digital image of the document being checked.

Linear image sensors are often equipped with multiple illumination sources in order to scan different regions of the spectrum. Typically, red, blue and green light and infrared light are used. With pulsed illumination, the required level of photoelectric sensitivity of the sensor is provided by a certain pulse energy. In order to further reduce the duration of the backlight pulse without affecting the pulse energy, it is possible to use sources of several wavelengths, for which the response of a drawing on a document in the test area differs slightly. For example, for a black-and-white drawing, it is possible to simultaneously turn on the red, blue and green backlight source during a backlight pulse for a certain period of time, due to which the required pulse energy will be achieved in a much shorter time than when using a single source. The inventive method contains the step of determining the skew angle, and the subsequent use of the skew angle value when checking the area of the digital image for the presence of features characteristic of the moiré image that occurs in this area of the digital image of the original document. As already mentioned, the moiré image on an original document, when the skew angle is changed, can change beyond recognition. On the other hand, knowledge of the skew angle makes it possible to predict with sufficient accuracy what the moiré image should be on an original document, and, on the basis of this, form a standard for comparison with the moiré image on the analyzed document. Determination of the skew angle can be performed, for example, by locating the leading edge of the document on the digital image. Methods for detecting rectilinear boundaries of objects and determining their angle of inclination are well known to specialists in the field of technical vision.

Hereinafter, we will use the term "document drawing" to represent a high-resolution image of a document, which does not distort the transmission of microscopic elements, such as thin lines or micro-texts. A digital image of a document is obtained when scanning with low resolution and low resolution, therefore, microscopic elements in it are either distorted, or lost, or appear indirectly in the form of moiré images.

A moire image obtained on a digital image of an original document, even at the same skew angle, can look different. Since a moiré image is an interference pattern, the phase of this pattern depends on the phase shift of the two interfering images. In the claimed invention, the moiré phase depends on the shift of the document pattern with respect to the scanning phase of the lines of the digital image. Since the moiré image is formed by thin-linear periodic or quasiperiodic elements of the pattern, the phase shift is a relative shift in the position of the pattern within the distance between adjacent moiré-forming lines. The moiré phase is manifested in which part of the moire image will have the minimum brightness, and in which - the maximum. Changing the phase of the moiré by 180 degrees leads to the fact that the sections of the minimum and maximum brightness of the moiré image are reversed. The initial shift in the position of the document within the distance between adjacent moiré lines is a very small amount, which is determined by how much the edge of the document has offset with respect to the position of the nearest line of the virtual developing image. As each document passes through the scanner, the initial shift is a random value, therefore, from one scan to another, the moiré image changes its phase at random. Therefore, in the course of checking the image for the presence of features characteristic of the moiré image of an original document, it is necessary to exclude the influence of the moiré phase.

Various approaches can be used to eliminate the effect of the moiré phase. In one possible way, a two-dimensional Fourier transform is applied. The convenience of using the Fourier transform is due to the fact that it allows one to judge the presence of certain spatial frequencies in the image, excluding the phase from consideration. Let's consider in terms of two-dimensional Fourier transform, what happens when scanning a document. The drawing of a document that has the property of moiré formation during scanning contains periodic or quasiperiodic structures. In the Fourier transform of the figure, these structures are represented as a set of points and segments of straight lines, or compact regions of small size. The skew is manifested in the rotation of the Fourier image by the skew angle. When a virtual developing image is superimposed, which occurs during scanning in accordance with the inventive method, the brightness of the document image is multiplied by the brightness of the developing image.

The virtual developing image can be represented as the sum of the constant component corresponding to the average level of the developing image, the fundamental harmonic of the fundamental spatial frequency corresponding to the scanning line spacing, as well as the higher harmonics of the fundamental frequency. The multiplication of the document image by the fundamental frequency of the developing image leads to simultaneous shifts of its Fourier image both to the low frequency region and to the high frequency region, by the value of the fundamental frequency. This phenomenon is known as the formation of combination frequencies. The same happens when the figure is multiplied by the harmonic of the fundamental frequency, while the magnitude of the shift is equal to the product of the fundamental frequency and the number of the harmonic. It should be borne in mind that the original drawing also partially passes through the developing image unchanged, which corresponds to multiplication by a constant component. Therefore, the resulting Fourier transform of the overlay result is a linear combination of the original and shifted Fourier transforms of the document drawing. The linear combination coefficients for each shifted image are determined by the amplitude of the corresponding harmonic of the fundamental frequency in the virtual developing image.

When the Fourier transform of the digital image obtained during scanning, a limited rectangular section of the final Fourier image is obtained, the size of which in the scanning direction is equal to the fundamental spatial frequency, and in the direction perpendicular to it is equal to the spatial frequency of the location of the receiving areas in the linear photodetector matrix of the sensor. This limitation is a direct consequence of the sampling theorem, known as the Nyquist-Shannon theorem or the Kotelnikov theorem. Frequency coordinates in the specified rectangular section of the final Fourier image do not exceed half the fundamental spatial frequency and half the spatial frequency of the location of the receiving areas in the linear photodetector matrix of the sensor.

In the claimed invention, the Fourier transform of the digital image needs to be performed only for a predetermined control zone, since it contains the controlled characteristics of the moiré. This limitation significantly reduces computational costs. The result of the Fourier transform of the digital image of the document in the control zone will be called the controlled Fourier transform. The characteristic features of the moiré image are manifested in the controlled Fourier image in the form of the corresponding features, expressed as a certain pattern of the distribution of the amplitude of the spectral components of the Fourier image on the frequency plane.

In accordance with the above, in order to check for the presence of features characteristic of the moiré image of an original document, in the claimed method, using a discrete Fourier transform, a controlled Fourier image of the contents of a digital image of the document being checked in the control zone is calculated, the presence of signs in the controlled Fourier image is checked, corresponding to the controlled features, and by the presence of these signs, a conclusion is made about the presence of controlled features.

With a complex character of moiré, which differs from straight stripes of equal slope located with a constant period, the entire distribution of the amplitude of the spectral components of the Fourier image on the frequency plane, which should be present in the controlled Fourier transform, is considered as a set of features for verification. In accordance with this, a digital image of an authentic document for each type and orientation is pre-obtained by scanning the specified original document without skewing, providing a discreteness in the scanning direction, three or more times greater than the discreteness of the monitored digital image in the scanning direction, and performing a discrete Fourier transform the contents of this digital image in the control zone to obtain a reference Fourier image of the control zone, and to check for the presence of signs corresponding to the controlled features, in the controlled Fourier image of the control zone, taking into account the skew angle, the reference Fourier image of the control zone is converted to obtain the target Fourier transform, the area of spatial frequencies of which coincides with that of the controlled Fourier image, and, according to a predetermined similarity criterion, the distribution of the amplitude of the spectral components of the Fourier image on the frequency plane of the target Fourier image is compared with that distribution of the control zone controlled by the Fourier image, and when a predetermined degree of similarity is reached, a conclusion is made about the presence of features corresponding to the controlled features, if a predetermined degree of similarity is not achieved, then a conclusion is made about the absence of features corresponding to the controlled features.

Since the discrete Fourier transform is used, the distribution of the amplitude of the spectral components of the Fourier image on the frequency plane is simply a two-dimensional image consisting of pixels. At the same time, the distributions for the controlled and target Fourier images have the same dimension, since their spatial frequency regions coincide. To find the degree of similarity, any of the known methods for finding the degree of similarity of images can be used, for example, calculating the correlation coefficient. Exceeding a certain predetermined threshold value of the correlation coefficient can be considered as achieving a given degree of similarity.

In Fourier transform theory, the amplitude of the spectral component corresponding to a particular spatial frequency is calculated as the square root of the sum of the squares of the real and imaginary values of the Fourier transform for that spatial frequency. Therefore, the amplitude of a spectral component is sometimes called the modulus of the complex amplitude of that component. The amplitude of the spectral component does not depend on the moiré phase. Comparison of the distribution of the amplitude of the spectral components of the Fourier image on the frequency plane of the target Fourier image with the distribution of the amplitude of the spectral components of the control zone controlled by the Fourier image thus excludes the moiré phase from consideration. Due to this, the dependence of the comparison result on the small initial displacement of the document with respect to the position of the lines of the virtual developing image is eliminated.

A digital image of an original document for each type and orientation is obtained by scanning with a resolution in the scanning direction, three or more times the resolution of the controlled digital image in the same direction. Note that, in accordance with the sampling theorem, the discreteness of the monitored digital image sets the highest spatial frequency in the scanning direction, equal to 1/2 of the spatial frequency of scanning the lines of the monitored digital image. Increasing the discreteness by three or more times increases the highest spatial frequency of the reference Fourier image to 3/2 of the spatial frequency of scanning the lines of the controlled digital image. Such an increase in the scanning resolution is necessary in order to convey the moiré-forming structure of the pattern of the original document. It allows performing a frequency shift of the reference Fourier image towards lower frequencies by the value of the spatial scanning frequency of the lines of the controlled digital image, which corresponds to the interaction with the first harmonic of the fundamental frequency of the virtual developing image. As a result of the shift, a part of the shifted Fourier transform covers the frequency interval of the target Fourier transform. Such coverage is necessary to fill the target Fourier transform with spectral components characterizing moiré.

The higher the resolution in the scanning direction, the more accurate representation of the original document in the target Fourier transform can be obtained. So, with an increase in the discreteness five times in relation to the discreteness of the controlled digital image, the results of interaction not only with the first, but also the second harmonics of the fundamental frequency can be used to form the target Fourier image. The discreteness in the direction perpendicular to the scanning direction must be provided with approximately equal resolution in the scanning direction in order to stably transmit the structure of thin moiré-forming lines and the spatial frequencies corresponding to them. Scanners of originals used in publishing and printing activities can be used for scanning with high resolution. Since it is necessary to obtain a high resolution image only for the inspection zone, a micrograph of the inspection zone obtained using a microscope with a low distortion level can be used as a digital image of a genuine document.

In one of the possible implementations, to obtain the target Fourier transform, a rotated Fourier transform is obtained by rotating the reference Fourier transform by a skew angle, at least one shifted Fourier transform is obtained by shifting the rotated Fourier transform along the axis corresponding to the direction scanning, by an amount that is a multiple of the spatial scanning frequency determined by the pitch of the lines of the digital image, and a linear combination of the rotated Fourier transform and at least one shifted Fourier transform is calculated according to the predetermined linear combination coefficients, after which a region is extracted from the specified linear combination spatial frequencies, the size of which coincides with the area of spatial frequencies of the controlled Fourier transform, and use it as a target Fourier transform.

This implementation repeats the previously described process of forming a Fourier image of a moiré by superimposing a virtual developing image on a document drawing. The resulting target Fourier transform, thus, should be close to the controlled Fourier image, if, of course, the drawings of the obviously genuine document and the checked document within the control zone turn out to be close. The linear combination coefficients can be found on the basis of an experimental study of the apparatus function of a linear sensor, which determines the spatial distribution of the brightness of the virtual developing image and the harmonic composition of its Fourier image. Methods for studying the hardware function of an image sensor are well known to those skilled in the art. For small coefficients of the linear combination that do not significantly affect the result of comparing the degree of similarity, the corresponding shifted Fourier transforms should be excluded from the linear combination in order to simplify the calculations. Also, you should exclude those shifted Fourier transforms that are shifted by a large amount and, because of this, do not contribute to the selection of the target Fourier transform.

To provide the fastest computational speed, the obvious solution for obtaining Fourier transforms is to use the Fast Discrete Fourier Transform algorithm, which is well known to specialists in the field of image processing. There is an additional option to reduce the computational overhead performed during authentication. Computational manipulations with the Fourier transform of a genuine banknote, including rotation, shift, linear combination and target Fourier transform extraction, can be performed in advance for a wide range of possible angles, and the set of obtained results can be saved as blanks. During authentication, from this set, it will be necessary to extract the blank corresponding to the angle closest to the measured skew angle of the document being checked, and use it as a target Fourier transform for comparison.

A fairly common case is when the moiré-forming elements in the control zone have a periodic or almost periodic arrangement. The reference Fourier transform of such a control zone is characterized by increased values of the amplitude of the spectral components in one or several compact regions and small values of the amplitude outside the named regions. We will refer to such compact areas as power concentration areas. The location of these regions on the frequency plane is determined by the period of the moiré-forming elements, and the size of the regions corresponds to deviations from the periodicity. As a rule, there is one dominant region, the spatial frequency of the center of which corresponds to the period of the moiré-forming elements, and many secondary regions corresponding to the harmonics of the indicated frequency. For such a reference Fourier transform, a target Fourier transform using rotation, displacement and linear combination can be used as previously described. However, the complexity of computational processing can be significantly reduced if, instead of operations with Fourier transforms, operations of rotation and displacement of the known coordinates of the power concentration regions are performed. For this, as a feature corresponding to the monitored feature, the excess of the specified amplitude level of the monitored Fourier image in at least one characteristic region on the frequency plane is used, and the size and location of the characteristic region on the frequency plane are calculated based on the skew angle according to a predetermined algorithm.

Exceeding the specified amplitude level can be determined in various ways. For comparison, the peak, average or rms value of the amplitude in the characteristic area can be used. The algorithm for calculating the position and size of the characteristic area relies on the previously known location of the center of the power concentration area in the reference Fourier transform and performs a rotation sequence in accordance with the skew and shift angle by an amount multiple of the scan frequency of the lines of the controlled digital image.

The position of the power concentration regions can be obtained directly from the reference Fourier transform. However, in many cases, they can be calculated without obtaining the reference Fourier image itself, only on the basis of manual measurement of the period of moiré-forming elements in the drawing of a genuine document, observed under a microscope. This avoids the need for sophisticated equipment to scan an original document at high resolution.

In one of the implementations of the claimed invention, the line spacing, as scanned, is changed by a value limited to predetermined limits, and this change is taken into account during the check for the presence of monitored features. A change in the interline pitch in one or more consecutive lines causes a change in the moiré phase, and a change in the interline pitch in adjacent groups of lines causes a change in the moiré frequency. Both types of changes can be detected by analyzing the digital image of the checked document and correlated with the change in the line spacing made. As a result, it can be determined to what extent the response of moiré parameters, such as phase and spatial frequency, to a change in line pitch is as expected for an authentic document. This makes it possible to additionally detect the imitation of the moiré image made by counterfeiters using halftone printing.

As it was found out by the authors in the course of experiments, the moiré image obtained in accordance with the claimed method carries information about the thickness and continuity of thin lines of the document drawing. When a thin-line structure contains a group of parallel segments located at an equal distance, then a moiré structure is also a group of segments located at an equal distance. The angle at which the segments in the moiré pattern are directed and the spatial frequency of the moiré are determined by the direction and spatial frequency of the segments in the pattern and the skew angle of the document. It is important that between the thickness and the spatial period of the segments in the moiré image, approximately the same relationship is observed as between the thickness and the spatial period of the segments in the drawing of the document. The narrower the transparent stripes in the virtual developing image, the higher the accuracy of this relationship. This allows you to measure the thickness of the lines in the drawing based on the moiré parameters. In addition, moiré reflects the continuity of the lines in the drawing. For continuous lines, the transverse profile of the brightness of the moiré, measured in the direction perpendicular to the moiré lines, has a characteristic shape, which is determined by the line pitch of the sensor's field of view along the document and the size of the area on the document from which the sensor receives radiation during scanning one line. However, if the lines of the drawing on the document have discontinuities and / or significant modulation of thickness, then the transverse profile of the moiré luminance becomes significantly flatter, and the longitudinal profile of the luminance along the moire line becomes uneven. By these differences, one can judge the continuity of the lines and the constancy of their thickness.

In a possible implementation of the claimed method, in the course of checking for the presence of controlled features, the fulfillment of a predetermined criterion is checked, indicating the presence of a controlled feature and based on an assessment of the thickness of the lines of the printed image on the surface of the document. Verification can be implemented, in particular, due to the fact that the inspection zone is predetermined in such a way that it contains a plurality of parallel segments spaced equal to each other in the digital image of a genuine document of a given type and a given orientation, so that they define the vector of the spatial frequency of the segments, and during the check for the presence of controlled features, the vector of the spatial frequency of the moiré is found based on the vector of the spatial frequency of the segments and the skew angle, using a discrete Fourier transform, the controlled Fourier image of the contents of the digital image of the document being checked in the control zone is calculated, and fulfillment of a predetermined criterion using the distribution of the values of the controlled Fourier image along a straight line passing through the starting point on the frequency plane in the direction of the vector of the spatial moiré frequency.

This check is based on the fact that the moiré generated by a set of parallel segments spaced at equal distances from each other itself consists of periodically spaced segments. The Fourier transform of such a moiré is a set of characteristic regions with increased amplitude located along a straight line passing through the starting point on the frequency plane in the direction of the spatial moiré frequency vector. The characteristic regions are located along a straight line with a step equal to the spatial frequency of the moiré. The order of the characteristic regions is here numbered from zero from the origin, so that the zero region corresponds to the constant component, the first region corresponds to the spatial moiré frequency, and the second corresponds to twice the spatial moiré frequency, and so on.

The spatial moire frequency vector can be found based on the position of the first characteristic region. A method for calculating the position of the representative area based on the known location of the power concentration area of the reference Fourier transform has been described previously. As applied to the spatial frequency of moiré, in the reference Fourier transform, a power concentration region is preselected, which, when interacting with a virtual developing image, passes into the first characteristic region. To calculate the position of the characteristic region, the coordinates of the compact region are rotated by a skew angle and shifted towards lower frequencies by a vector that is a multiple of the fundamental spatial frequency.

The values of the monitored Fourier transform along the specified straight line characterize the thickness of the moiré-forming segments, as will be described in detail below. In particular, the thinner the moiré-forming segment with respect to the period of the location of these segments, the higher the amplitudes of the Fourier transform in the second and subsequent characteristic regions. As a criterion, the excess of the amplitude of the controlled Fourier image in the second characteristic region over some predetermined threshold can be used. As an alternative criterion, one can consider the excess of the ratio of amplitudes in the second and in the first characteristic regions over a predetermined threshold value. The fulfillment of the criterion indicates that the thickness of the moiré-forming segments in the control zone of the checked document turns out to be no more than that characteristic of the original document.

FIG. 1 shows a diagram of transferring an image from a document surface to a receiving area of a linear image sensor.

FIG. 2 shows the formation of a virtual developing image during line scanning.

FIG. 3 illustrates the formation of moiré between a group of parallel equally spaced segments and a virtual developing image.

FIG. 4, the same process of moiré formation is shown using a two-dimensional Fourier transform.

FIG. 5 shows the effect of skew angle on moiré formation.

FIG. 6 illustrates the skew angle limitations of moiré formation.

FIG. 7 shows the process of formation of a moiré between a group of parallel segments spaced at the same distance from each other and a virtual developing image, in which the thickness of the segments can be judged by the controlled Fourier image.

FIG. 8 shows a flowchart of a banknote verification process.

FIG. 9 and FIG. 10 shows a block diagram of an additional check of a banknote in accordance with the first and second embodiments, respectively.

The practical implementation of the claimed invention is intended for use in a counting and sorting machine for processing banknotes. This machine has a feeding pocket, a mechanism with a banknote path and a system of elements for redirecting banknotes, a set of sensors located in the path, a controller and several receiving pockets. The controller is used to control the mechanism, collect information from sensors, and analyze this information to confirm the authenticity and make a decision on the direction of the banknote in one or another pocket. The banknotes are placed by the user in a receiving pocket. At the command of the user, the controller starts the mechanism, as a result of which the banknotes are folded from the feeding pocket and pass by the sensors along the path. The redirecting elements, controlled by the controller according to the results of the validation of the banknotes, route each banknote to a particular acceptance pocket.

In practical implementation, a linear contact image sensor is used, similar to the commercially available sensors that are commonly used in scanners and counting and sorting machines. The contact sensor contains a pulsed LED lighting system that provides uniform illumination of the scanned area of the document with red, blue and green light. The transfer of the image from the surface of the document to the receiving area 5 of the linear image sensor is shown in FIG. 1. The optical system of the sensor contains a linear array of objectives 1, located in one row close to each other. Objective 1 provides the formation of a non-inverted image with a magnification of Г = 1. It is implemented in the form of a gradient lens with a diameter of D = 300 μm (the diameter of the lens coincides with the aperture), having a radiation reception angle θ = 12 ° and a working distance X ₀ = 2.8 mm. These parameters are based on commonly used lens arrays from various manufacturers. For example, this standard size includes an array of gradient lenses of the Selfoc Lens Array, model EG 12. The angle θ is large enough for the radiation to hit the receiving area 5, passing simultaneously through four adjacent lenses 1. In practice, when the parameters of these lenses change, their the number varies from 3 to 5 depending on the combination of their aperture D, the working distance X ₀ and the angle of radiation reception θ.

A group of four objectives 1 provides the transfer of radiation energy from an elementary spot at positions 2-4 on the document surface to the receiving area 5. We call the spot elementary to show that it is coupled with the receiving area 5 of the photodetector element of the sensor. The spot at position 2 is shown to indicate the position of the document corresponding to the best focusing, and the spot at positions 3 and 4 corresponds to the extreme permissible positions of the document with respect to the contact image sensor within the depth of field D _f = 1 mm. The value of the required depth of field is determined by the possibility of displacement of the document surface in the directions towards the sensor and away from the sensor during machine operation. Since the energy from the area of an elementary spot in positions 2-4 is collected on the receiving platform 5 of a given size, it is convenient to consider the operation of a group of lenses in the reverse path of the rays, that is, from the receiving platform to the document.

The size A of the receiving area 5, measured in the scanning direction (perpendicular to the axis of the location of the receiving areas), is chosen as small as possible and close to the size of the minimum scattering spot of the objective 1. Minimization of the size A is necessary to obtain the smallest spot size in the scanning direction in positions 2- 4, which ensures the minimum width of the transparent stripes of the virtual developing image. In this case, the level of photosensitivity required for the sensor to operate at the highest scanning speed must be ensured. For modern technology of photodetector elements and gradient lenses, the optimal choice of the size A of the receiving area 5 is 40-55 microns.

Radiation beam 6 has a waist in the best focusing position corresponding to spot 2. Beam 6 far from the waist has an asymmetric cross-section with an aspect ratio of 1: 4. However, in the area of constriction, i.e. in position 2, the spot is, as experience shows, practically round, in spite of the large asymmetry of the beam 6. The narrowing of the spot in the direction perpendicular to the scanning direction, which could be expected from considerations of diffractive optics, does not occur. This is most likely due to the fact that the lengths of different lenses have significant differences that exceed the radiation wavelength, which does not allow considering a group of lenses as a coordinated working single diffraction-limited lens with a highly elongated asymmetric aperture. The spot size Wmin at position 2 in the scanning direction is determined by the size of the receiving area 5 and the diameter of the minimum scattering spot. This size is influenced by both frame diffraction and aberration. Experience shows that for the considered standard size of the lens array, the W _min value does not exceed approximately 65-80 μm.

In the prior art, the asymmetry of the beam of an array of gradient lenses and the associated dependence of the spot size on the distance to the best focusing position is discussed in US Pat. No. 5,450,157 (publ. 09/12/1995, IPC G02B 3/00). However, in this patent asymmetry is considered as a disadvantage in providing the radiant illumination of the receiver, and in order to eliminate it, it is proposed to increase the number of parallel rows of gradient lenses. An increase in the number of parallel rows of gradient lenses leads to an increase in the spot size far from the waist, measured in the scanning direction. In the context of the claimed invention, this is the opposite of the desired spot narrowing effect. The specified patent has informative value, since it contains a well-founded method for calculating the beam parameters.

Таким образом, при указанном выше интервале оптимальных размеров приемной площадки величина W_max находится в интервале 94-109 мкм.Positions 3 and 4 of the spot are in the far zone with respect to the beam waist and their size in the scanning direction is determined, for the most part, by the geometric path of the beams. Based on the geometric path of the rays and the increase in Г = 1, this size can be approximately represented as

Thus, with the above range of optimal sizes of the receiving area, the value of W _max is in the range of 94-109 microns.

In the direction perpendicular to the scanning direction, the size L _{max of the} elementary spot at positions 3 and 4 is significantly greater than the size W _max in the scanning direction. This is due to the asymmetry of the cross section of the beam 6 with an aspect ratio of 1: 4.

For scanning, a power pulse of the lighting system is applied simultaneously to all red, blue and green LEDs. Due to the high illumination in the pulse, its required duration turns out to be sufficiently short, so that during the pulse the document is displaced by no more than a = 10 μm. As a result, the length of the registration area W on the document, from which the sensor receives the radiation recorded during the scanning of a single line, does not exceed the value of a + W _max , that is, 104-119 μm.

The resolution in the scanning direction is determined by the synchronization of the scanning of individual lines with the movement of the protected document, and is provided by the machine controller in accordance with the signals received from the movement sensor of the mechanism. This discreteness can be selected by setting the moment of issuing a signal to scan a separate line in the controller in relation to the signals from the motion sensor. Discreteness in the perpendicular direction is determined by the step of the receiving areas of the sensor's photodetector matrix. The discreteness values in each of the indicated directions can be selected independently in the range of 100-200 dpi. However, for simplicity of description, in the described implementation, the resolution in both directions is equal to 150 dpi. It corresponds to the scanning line spacing S and the step of the receiving areas of the photodetector matrix, each of which is equal to 169 μm.

Moiré-forming fields are often introduced into secure documents, which are specially calculated to create moiré when scanned on office scanners. Such scanners usually have a resolution from a range of values of 1200dpi, 600 dpi, 300 dpi, where adjacent values of resolution differ by 2 times. With a resolution of 150 dpi, which is a continuation of this series, the likelihood of moiré formation when scanning a security document is also quite high.

Many manufacturers of linear photodetector arrays make them adaptable to different discreteness values. To do this, the receiving sites are made corresponding to the highest applied discreteness value, and to obtain smaller discreteness values, neighboring receiving sites are electrically connected into groups and each group is used to register one pixel of a row. So, for example, for a resolution of 600 dpi, the size of the square receiving area of the matrix is made equal to about 40 microns and they are placed with a pitch of 42.33 microns along a straight line. To obtain a resolution of 300 dpi, adjacent receiving areas are combined into groups of two; to obtain a resolution of 150 dpi, the receiving areas are combined into groups of four. In the described implementation, just such a matrix is used, used in the mode of combining four sites in a group.

The line-to-line offset of the checked document S = 169 microns relative to the sensor exceeds the length of the registration area on the document W = 104 microns, from which the sensor receives radiation recorded during scanning of a separate line used to obtain a digital image of the checked document, where the specified length is measured in the scanning direction ... As shown in FIG. 2, between the areas of registration 7 when scanning sequentially located lines 10 there are areas of insensitivity 9 with a width of S-W = 65 μm. The image of the document in them is not registered at all by the sensor during scanning. Within the registration area 7, the amount of radiation energy received by the sensor during the scanning of the line is unevenly distributed in the scanning direction 8. As always happens when the optical system is imaging for light objects close in size to the minimum scattering spot, a higher energy level is provided in the center of the object than on its periphery. Because of this, when viewed in the scanning direction 8, more energy is collected from the central portion of the registration area 7 than from the edge portions of this area. Thus, in terms of the virtual developing image, between the registration region 7 and the dead region 9, there is a blurred transparency border located within the registration region 7.

The scanned line consists of individual pixels P1, P2, ..., PN, where the brightness of each pixel is formed in proportion to the average response of the four receiving areas 5. Thus, the pixel has a size of 169 μm in the direction perpendicular to the scanning direction. In this direction, the length of the sensor sensitivity region, from which the sensor receives radiation when registering one pixel of a row, significantly exceeds the pixel size due to the large size L _{max of the} elementary spot.

пространственной частоты группы отрезков, который перпендикулярен отрезкам 11 (вектор на рисунке не показан).FIG. 3 shows how a group of segments 11 is scanned on the surface of the document, located at equal intervals 1 / F and slightly inclined at an angle ϕ with respect to the scanning direction 8. Here F denotes the length of the vector

the spatial frequency of a group of segments, which is perpendicular to segments 11 (the vector is not shown in the figure).

фундаментальной пространственной частоты виртуального проявляющего изображения, то есть частоты сканирования строк, ориентированного в направлении сканирования 8.The virtual image 12 consists of registration areas 7 and dead areas 9, arranged in an alternating order with a repeating interval S = 1 / f. Here f denotes the length of the vector

the fundamental spatial frequency of the virtual developing image, that is, the scanning frequency of the lines, oriented in the scanning direction 8.

пространственной частоты муара отличаются от ориентации и длины векторов

, и определяются углом наклона ϕ. Можно продемонстрировать, что даже малые изменения угла ϕ приводят к значительным изменениям вектора

Этот эффект сильного влияния угла наклона на пространственную частоту муара хорошо известен специалистам. Математическое рассмотрение процесса образования муара может проводиться разными способами. Например, можно вычислить положение муаровых полос между группой отрезков 11 и виртуальным проявляющим изображением 12 путем геометрического расчета точек пересечения этих структур. Однако наиболее полное представление в отношении явления муарообразования дает двумерное преобразование Фурье. Использование преобразования Фурье позволяет анализировать муар, образованный не только группой отрезков, но также любым иным рисунком на поверхности документа, обладающим свойством муарообразования.As a result of the superposition of the group of segments 11 and the virtual developing image 12, a moiré image is formed in the form of straight stripes 13. This image has a spatial frequency M and a repetition period of the stripes 1 / M. Orientation and vector length

the spatial frequency of moiré differs from the orientation and length of the vectors

, and are determined by the angle of inclination ϕ. It can be demonstrated that even small changes in the angle ϕ lead to significant changes in the vector

This effect of the strong influence of the tilt angle on the spatial moiré frequency is well known to those skilled in the art. The mathematical consideration of the moire formation process can be carried out in different ways. For example, the position of the moiré fringes between the group of line segments 11 and the virtual developing image 12 can be calculated by geometric calculation of the intersection points of these structures. However, the most complete understanding of the phenomenon of moiré formation is given by the two-dimensional Fourier transform. The use of the Fourier transform makes it possible to analyze moiré formed not only by a group of segments, but also by any other pattern on the surface of a document that has the property of moiré formation.

In this description, we will graphically represent the individual spectral components of the Fourier transform using small circles, where the center of the circle corresponds to the spatial frequency of the component, and the size of the circle corresponds to its amplitude. The Fourier transform of the image is centrally symmetric because the image is only represented by real values. Although the theoretical size of the Fourier transform is not limited and can occupy the entire frequency plane, in reality, very high spatial frequencies cannot be recorded in images due to physical limitations of the imaging process. Therefore, in this description, the Fourier transform is limited to a rectangle in the frequency domain corresponding to a spatial frequency interval of practical value.

и разделены интервалами F.FIG. 4A shows the Fourier transform of a group of segments 11. It consists of the spectral components of the first harmonic 16 of the spatial frequency F, the second harmonic of this frequency 17, as well as its higher harmonics (not shown in the figure). Component 15 with zero frequency is the zero harmonic and corresponds to the average brightness level of the image of the group of segments 11. These spectral components are located on the straight line 18, which coincides in direction with the vector

and separated by intervals F.

FIG. 4B shows the Fourier transform of the virtual developing image 12. It consists of the spectral components of the first, second and third harmonics of the spatial scanning frequency f, located at the points ± f, ± 2f, and ± 3f on the ordinate axis, as well as higher harmonics (not shown in the figure. ). The component with zero frequency is the zero harmonic and corresponds to the average level of light transmission of the developing image 12. The superposition of the developing image 12 on the pattern on the document surface in the form of a group of segments 11 is represented as multiplying the brightness of the original image by the light transmission of the developing image at each point to obtain the brightness of the resulting image in that same point. The Fourier transform represents each of the original images as a sum of harmonic functions. With pointwise multiplication of images, there is a pairwise multiplication of the harmonic functions of one image with the harmonic functions of another image, in all possible combinations. The result of multiplying a pair of harmonic functions is the sum of a harmonic function with a frequency equal to the difference between the frequencies of the original functions, and a function with a frequency equal to the sum of the frequencies of the original functions. This process is known as obtaining combination difference and sum components.

For clarity of understanding, it should be pointed out that the virtual developing image is a physical effect associated not with the light transmission of the film of a conventional developing image, but with the concentration of radiation perceived by the sensor in the registration area 7. However, for describing the process of moiré formation, this difference is not essential. , since it does not affect the conclusions drawn. Therefore, in this description we use the concept of light transmission. In addition, when obtaining a two-dimensional Fourier image, it is customary to talk about the brightness of the original image. In accordance with this approach, in order to obtain a Fourier image of a virtual developing image, its light transmission should be formally considered as a normalized brightness.

With this multiplication, the resulting Fourier transform of the moiré structure is obtained as follows. Each spectral component of the Fourier transform of the document drawing (Fig. 4A) interacts with all Fourier transform components of the virtual developing image (Fig. 4B) to obtain the difference and total components from each interaction. The difference component is obtained by shifting the spectral component of the Fourier image of the pattern by -nf, where n is the number of the harmonic of the fundamental spatial frequency of the virtual developing image. Likewise, the total component is obtained by shifting by nf. The Fourier transform of the moiré pattern generated by the group of line segments 11 is shown in FIG. 4C. For the original frequency component 16, it is possible to trace the sum and difference components obtained by shifting by ± nf in the direction of the ordinate. The amplitude of the resulting sum and difference components is proportional to the product of the amplitudes of the interacting spectral components. Combination components that go beyond the range of practically used spatial frequencies have low amplitudes and are not shown in the figure. In vector form, the spatial frequency of the combinational component can be represented as

In general, to obtain the Fourier image of a moiré of an arbitrary pattern on the surface of a document with a virtual developing image, it is necessary to construct a linear combination of a set of copies of the Fourier image of this arbitrary pattern, each of which corresponds to the nth harmonic of the virtual developing image and is shifted by ± nf in direction of the ordinate axis. The linear combination coefficients are proportional to the amplitude of the corresponding harmonic of the virtual developing image. In practice, the number of terms in a linear combination that affect the reproduction of moiré turns out to be small due to a rapid decrease in the amplitude of the combination components with increasing harmonic number, and because these components go beyond a practically significant interval of spatial frequencies. Therefore, one should restrict oneself to a linear combination containing several practically significant members.

The discreteness of the scanned digital image limits the transmission of high spatial frequencies of the moiré image in it. Rectangle 19 denotes the border of this limitation. The height of the rectangle is equal to the spatial frequency of scanning the lines f, and its width is equal to the spatial frequency of the pixels P0, P1, ..., PN in the line of pixels corresponding to the registration area 7. Since in the described implementation the discreteness of the digital image is the same in both directions and is equal to 150 dpi, then rectangle 19 is a square with side f. In a digital form, in accordance with the sampling theorem, it is impossible to transmit a spatial frequency exceeding f / 2 in each direction. In this description, we will refer to rectangle 19 as the scope.

Этот вектор направлен параллельно прямой 21. Рисунки Фиг. 3 и Фиг. 4 выполнены в одинаковом масштабе по обеим координатным осям и с соблюдением отношения размеров на чертеже для значений F и f. Поэтому, вектор

вычисленный графическим образом на Фиг. 4, по величине и направлению соответствует муаровым полосам 13 на Фиг. 3. Полосы 13 перпендикулярны прямой 21.Of the entire set of combination frequency components obtained by superimposing a virtual developing image 12 on a group of segments 11, only symmetrically located components 20, corresponding to a single vector of spatial frequency, fall into the field of view

This vector is directed parallel to straight line 21. Figures FIG. 3 and FIG. 4 are made in the same scale along both coordinate axes and in compliance with the ratio of dimensions in the drawing for the values of F and f. Therefore, the vector

calculated graphically in FIG. 4 corresponds in magnitude and direction to the moire fringes 13 in FIG. 3. Stripes 13 are perpendicular to straight line 21.

Let us now consider the influence of the skew angle α of the document. The Fourier transform of the document pattern in the control zone is also rotated through the angle α when skewed. With one spectral component 25 as an example, this is shown in FIG. 5A. As a result of the skewing, the spectral component 25 undergoes a rotation through the angle α and turns into the spectral component 26. Further, when a virtual developing image is superimposed, the spectral component turns into a series of combination components by displacement by ± nf parallel to the ordinate axis. The combinational component 27, which is the result of interaction with the second harmonic of the Fourier image of the virtual developing image and is shifted by -2f relative to the component 26, appears in the field of view 19. It appears as a moiré in the form of parallel straight stripes observed on the digital image of the document being checked in the control zone ... The skew angle a corresponds to the rotation matrix A, which allows you to perform the rotation as a vector operation. The spatial frequency transform in FIG. 5A can be written in vector form as

The skew angle α during the operation of counting and sorting machines is due to the imperfection of the folding mechanism, varies from sheet to sheet according to a random law and is usually within ± 15 °. FIG. 5B shows the transformation of the spectral component 31, the location of which corresponds to the Fourier transform of the document pattern in the inspection zone, without skewing. For different skew angles, the positions 30-32 of the same spectral component are located on the arc of a circle 33 centered at the origin. The corresponding positions 34-36 of the combinational component, obtained after a shift by -f from the interaction with the first harmonic, lie on the arc of a circle 38.

The described actions, consisting of rotation and displacement, can be used to obtain a vector of the spatial moiré frequency for each individual spectral component of the Fourier transform of the document pattern in the control zone. For the Fourier transform of the document drawing in the control zone, which contains many spectral components capable of creating moiré, it is necessary to first rotate the specified Fourier transform by the angle α. Next, you should shift several copies of the rotated Fourier transform, each by its own value ± nf parallel to the ordinate axis. Then you need to form a linear combination from the displacement results, and select that part of the linear combination that falls into the scope 19.

Very often moiré occurs as a result of interaction with only one specific harmonic of the spatial scanning frequency. In this case, the linear combination degenerates to a single term. When considering this term, it is enough to rotate and shift for a part of the Fourier-image of the document drawing in the control zone. FIG. 5C shows an example of interaction with the first harmonic of the spatial scanning frequency. Shown here are three different square areas 37 of the Fourier image of the document pattern in the control zone, which at different skew angles within ± 15 ° pass into the field of view 19 as a result of rotation and displacement by -f when interacting with the first harmonic of the spatial scanning frequency. Rotation and displacement should be performed not for the entire Fourier image of the document drawing, but only for the rectangle 36 enclosing squares 37. This can significantly reduce the computational processing time.

In some cases, moiré in the control zone of the digital image of the document can be formed at some skew angles and not formed at others. The origin of this phenomenon is shown in FIG. 6. For different possible spatial frequencies of the moiré-forming structure of the document pattern in the inspection zone, different positions of the circles 38, 40, 41 are possible, along which, in accordance with the skew angle, the position of the combination spectral component corresponding to the moiré moves.

At some angles of skew, the combinational spectral component on the circle goes beyond the field of view 19, which leads to the disappearance of the moiré. For circles 38 and 41, this occurs at too large and too small skew angles, and for circle 40 additionally also at skew angles close to zero.

In addition, sometimes moiré is observed, but at some skew angles it has such a low spatial frequency that it is impossible to measure its value and direction. This is shown by the example of circle 41, which runs near the origin. Square 48 with side 2z defines the limits of the spatial frequency, in which it is impossible to measure its value and direction. The impossibility of measurement is caused by uncertainty when the moiré period is larger than the size of the moiré-forming structure in the banknote pattern. Due to uncertainty, the position of the Raman spectral component 39 merges with the spectral component 15 having a zero spatial frequency.

Using thick lines in FIG. 6 shows the segments of circles, when located on which the combination spectral components, moiré is formed and is suitable for measuring the magnitude and direction of its spatial frequency. At angles where moiré is not formed or is not suitable for measuring the magnitude and direction of its spatial frequency, additional verification of authenticity using moiré is not possible.

In the case of moiré-forming structures in the form of a group of 11 parallel segments spaced at equal distances, the corresponding areas of power concentration 16, 17 have a small size on the frequency plane. In the case of a discrete Fourier transform, this size is close to the value of the discrete frequency step. When the moiré structure contains wavy or curved lines that are equidistantly spaced, this leads to the fact that the power concentration areas increase slightly in size. The same dimensions are for the characteristic regions corresponding to the combination components generated from the interaction of the virtual developing image with the said regions of power concentration. When rotated and displaced, the size on the frequency plane of the region of concentration of spectral components does not change, and therefore, without change, passes into the size of the resulting characteristic region. When the concentration region of the enlarged size passes into the characteristic region of the enlarged size, the positions of the regions are transformed into each other in accordance with the rotation and displacement of the spectral component corresponding to the center of the region, similarly to FIG. 4 and FIG. five.

Checking banknotes is performed as follows (Fig. 8). During the operation of the machine, the banknote is folded from the feeding pocket and guided along the path of the machine past the linear image sensor (S1). In this case, every 169 µm, the image line corresponding to the registration area 7 is scanned. As a result, a digital image of the checked banknote is formed, consisting of lines, where each line consists of pixels P0-PN (S2). During scanning, the position of the leading edge of the banknote and its displacement along the sensor are monitored with each newly scanned line. On the basis of the obtained information about the displacement, find the skew angle of the banknote α and form the corresponding rotation matrix A (S3).

When the banknote has completely passed the image sensor, a digital image with a reduced resolution equal to 50 dpi is created on the basis of the digital image of the checked banknote (S4). This is done by averaging the brightness levels in square groups of 3x3 adjacent pixels, and then assigning the average value to the corresponding pixel in the downsampled image. Further, on the basis of the reduced-resolution image, the type and orientation of the banknote are recognized (S5). The type of banknote is specified by its currency, denomination, year of issue, that is, classification characteristics. If the type and orientation were successfully determined, then basic authentication is performed. The methods used for determining the type and orientation, as well as verification of authenticity, are well known in the art and are widely described in the patent literature. Data obtained from other sensors installed in the path of the machine can also be used to determine the authenticity. Such sensors include magnetic property sensors, luminescence sensors, infrared transmission sensors, and others well known to those skilled in the art. If the type of the banknote could not be established, or if the banknote was found to be inauthentic, no further check is carried out and the banknote is marked for sending to the reject pocket (S9).

If the banknote is recognized as authentic, but the additional authentication method according to the claimed invention is not implemented for its type and orientation, then no further verification is carried out, and the processing of the banknote in the machine is carried out on the basis of the previously made conclusion about its authenticity (S8). Otherwise, an additional check (S7) is performed in accordance with the claimed invention, in its implementation, where the predetermined type and orientation coincide with the type and orientation of the checked banknote. The result of this check can be either authentication or rejection.

Taking into account the above description of the moiré formation process, let us turn to the presentation of the first practical implementation of the method.

The preliminary work in the method is carried out for a predetermined type, as well as the orientation of banknotes when passing through the machine. On the basis of the analysis of the pattern of banknotes and their test scanning, a control zone is assigned, in which the moiré appears especially noticeably and has a character close to parallel stripes. Such moiré gives straight or slightly wavy or curved lines of the banknote pattern, located equidistantly and having a high spatial frequency. The control area of a genuine banknote of a given type is scanned without skewing, with a high resolution exceeding 2000 dpi, using a publishing and printing scanner, and a digital image of a genuine banknote is obtained in the control area.

и диаметр ее охватывающего круга d, а также номер N гармоники фундаментальной пространственной частоты, при взаимодействии с которой комбинационные компоненты попадают в область видимости 19 и наблюдаются как муар. Эти параметры далее используют при выполнении способа в машине.A discrete two-dimensional Fourier transform of this digital image is performed using scientific image processing software widely used in practice. The field of view 19 is marked on the obtained reference Fourier transform. Next, the power concentration region is found in which there is a high amplitude of the spectral components, and which, when interacting with the virtual developing image, gives the combination spectral components in the field of view 19 corresponding to the moiré. For this, interaction with a virtual developing image and obtaining a moiré image are simulated by shifting the position of the power concentration region by ± nf parallel to the ordinate axis, as shown in FIG. 4A-C and as previously described. As a result, the spatial frequency of the center of the power concentration region is fixed

and the diameter of its enclosing circle d, as well as the number N of the harmonic of the fundamental spatial frequency, upon interaction with which the combination components fall into the field of view 19 and are observed as moiré. These parameters are further used when executing the method in the machine.

в которую, в результате взаимодействия рисунка проверяемой банкноты и виртуального проявляющего изображения, переходит пространственная частота центра области концентрации мощности

эталонного Фурье-образа. После этого, проводят проверку допустимости значения

(S22). Если координаты вектора

выходят за интервал

то есть за пределы области видимости 19, то делают вывод о невозможности проведения дополнительной проверки подлинности по причине отсутствия муара. Если обе координаты

оказываются в пределах заранее заданного интервала [-z, z] от начала координат на частотной плоскости, что соответствует квадрату 48, то делают вывод о невозможности проведения дополнительной проверки подлинности из-за невозможности анализа значения и направления вектора пространственной частоты муара вследствие очень низкого значения пространственной частоты. В остальных случаях считают

допустимым и продолжают проверку.In the first implementation of the method (Fig. 9), additional authentication in the machine begins with a validation of the spatial frequency of the center of the characteristic region. To do this, calculate (S21) the spatial frequency corresponding to the center of the characteristic region

into which, as a result of the interaction of the pattern of the checked banknote and the virtual developing image, the spatial frequency of the center of the power concentration region

reference Fourier transform. After that, the validity of the value is checked.

(S22). If the coordinates of the vector

go beyond the interval

that is, outside the scope 19, it is concluded that additional authentication is impossible due to the absence of moiré. If both coordinates

are within a predetermined interval [-z, z] from the origin on the frequency plane, which corresponds to square 48, then it is concluded that it is impossible to carry out additional authentication due to the impossibility of analyzing the value and direction of the moiré spatial frequency vector due to the very low value of the spatial frequency. In other cases, consider

valid and continue checking.

и точно соответствующие интервалам области видимости 19. По кругу диаметром d с центром в центре характерной области

вычисляют среднеквадратичную амплитуду спектра в контролируемом Фурье-образе (S24). Если среднеквадратичная амплитуда оказывается менее заранее определенного предела L, то делают вывод о неподлинности банкноты и ее браковке (S31). В противном случае, принимают решение о подтверждении подлинности банкноты (S32). Среднеквадратичный метод усреднения дает амплитуду, соответствующую среднему значению спектральной мощности в охватывающем квадрате характерной области. Это позволяет исключить влияние неравномерности распределения мощности в характерной области, возникающей из-за дискретности разбиения частотной плоскости.From the digital image of the checked banknote, a section corresponding to the inspection zone is selected, and for this section a fast two-dimensional Fourier transform is performed to obtain a controlled Fourier transform (S23). The resulting Fourier transform has frequency intervals along the abscissa and ordinate axes equal to

and exactly corresponding to the intervals of the field of view 19. In a circle of diameter d centered in the center of the characteristic area

calculating the rms amplitude of the spectrum in the controlled Fourier transform (S24). If the rms amplitude is less than the predetermined limit L, it is judged that the note is not authentic and is rejected (S31). Otherwise, a decision is made to confirm the authenticity of the banknote (S32). The root-mean-square averaging method gives the amplitude corresponding to the average value of the spectral power in the spanning square of the characteristic region. This makes it possible to exclude the influence of uneven power distribution in the characteristic region arising from the discreteness of the division of the frequency plane.

The value of L is determined in advance, experimentally, by performing the previous steps of the described implementation of the method and using a sample of genuine banknotes of a given type. At the same time, it is achieved that on genuine banknotes the root-mean-square amplitude of the spectrum in a controlled Fourier image in a circle with a diameter d centered in the center of the characteristic region would always be slightly higher than L. the image will not have a concentration of spectral power in the indicated circle, or the concentrated power will be small. Accordingly, the rms amplitude of the spectrum in the specified circle will be less than L, which makes it possible to distinguish a genuine banknote from a counterfeit one. The decision on authenticity, which is provided in this implementation, confirms the presence of moiré-forming elements in the control zone and their spatial frequency.

The decision taken if additional authentication is not possible depends on the depth setting previously made by the user of the machine. If the machine is set to operate with the deepest authentication, the note is marked for sending to the reject pocket as shown in FIG. 9. Thus, the technical rejection of the banknote is realized. On the contrary, if the machine is tuned to a medium level of verification depth, at which it is sometimes allowed not to carry out additional verification, then a conclusion is made about the authenticity of the banknote in accordance with the result of the previously performed authentication.

The test described here allows you to check the spatial frequency of moiré-forming elements such as lines. However, it does not allow you to control how the thickness of the lines in the control zone of the checked banknote corresponds to the thickness of the lines of the pattern of the original banknote. The following is a description of the theoretical basis for verifying line thickness compliance.

The thickness of the lines directly affects the amplitude of the second and higher harmonics in the Fourier transform. When the thickness of the lines is equal to the period of their repetition, in the Fourier transform the amplitude of the first harmonic is large, and the second and subsequent harmonics have small amplitudes. With decreasing line thickness, the amplitude of the first harmonic decreases, while the amplitude of the higher harmonics increases. With a very small thickness of the lines, in comparison with the period of their repetition, the amplitudes of all harmonics become approximately the same. This conclusion is directly obtained from the analysis of the spectrum of a periodic sequence of extremely narrow pulses in the form of the Dirac delta function, which is well known in mathematics.

Thus, control of the amplitude of the second and higher harmonics in the spectrum allows you to control the thickness of the lines. For the amplitude of the harmonics of moiré-forming lines, direct control is usually not possible, since they are outside the field of view 19. However, these amplitudes can be controlled indirectly through the amplitudes of the harmonics of the moiré structure falling into the field of view 19.

FIG. 7, using circles without filling, the spectral components of the Fourier transform of a group of thin moiré-forming segments are shown. Component 42 corresponds to the first harmonic and component 45 corresponds to the second harmonic. None of these components fall into scope 19 and cannot be directly inspected. However, the combinational component 43 from the interaction of the component 42 with the first harmonic of the fundamental frequency of the virtual developing image falls into the field of view 19. Similarly, the combinational component 46 from the interaction of the second harmonic component 45 with the second harmonic of the fundamental frequency of the virtual developing image also falls into the field of view 19. These combinational components the components are on straight line 47 due to the proportional relationship between the coordinates of the vectors of their spatial frequencies. They are harmonics of the spatial frequency of the moiré observed in the test area in the digital image being tested. Their amplitudes are proportional to both the amplitudes of the spectral components of a group of thin moiré-forming segments and the amplitudes of the corresponding harmonics of the fundamental frequency of the virtual developing image. Straight line 47 is perpendicular to the moiré stripes.

In general, the combination components corresponding to the spatial moiré frequency harmonics result from the interaction of harmonics with the same number, corresponding to moiré-forming segments, on the one hand, and a virtual developing image, on the other. For example, the first harmonic 42 interacts with the first harmonic of the fundamental frequency and gives the first harmonic of the moiré 43, the second harmonic 45 interacts with the second harmonic of the fundamental frequency and gives the second harmonic of the moiré 46, and so on. For the second harmonic of the moiré 46 to be in the field of view 19, the first harmonic 43 must have a sufficiently low spatial frequency.

Thus, control of the second harmonic amplitude is not possible for all moirés.

гармоники фундаментальной пространственной частоты виртуального проявляющего изображения. Для улучшения проверки толщины линий желательно увеличить обе эти амплитуды, чтобы снизить влияние шумов. С этой целью следует, по возможности, уменьшить ширину W области 7 регистрации. Для этого, нужно уменьшить размер элементарного пятна датчика в направлении сканирования для интервала возможных положений пятна 2-4 на банкноте, путем оптимизации параметров массива объективов 1 и уменьшения размера приемной площадки 5. Кроме того, уменьшению ширины W способствует уменьшение длительности импульса освещения. Эти методы понятны специалистам.The amplitude of the second harmonic of the moiré 46 is the greater, the greater the amplitude of the second

harmonics of the fundamental spatial frequency of the virtual developing image. To improve the line thickness checking, it is desirable to increase both of these amplitudes in order to reduce the influence of noise. For this purpose, the width W of the registration area 7 should be reduced as much as possible. For this, it is necessary to reduce the size of the elementary sensor spot in the scanning direction for the interval of possible spot positions 2-4 on the banknote, by optimizing the parameters of the lens array 1 and reducing the size of the receiving area 5. In addition, reducing the width W is facilitated by a decrease in the duration of the illumination pulse. These methods are understandable to specialists.

The second implementation of the invention is intended to check not only the presence and spatial frequency of moiré-forming lines, but also their thickness. In this implementation, the control zone is selected in such a way that the moiré arising in it would be sufficiently low-frequency, and its second harmonic would be within the field of view 19 for the majority of practically occurring skew angles of the banknote.

соответствующей первой гармонике, и диаметр ее охватывающего круга d. Пространственная частота

соответствует области концентрации мощности, образованной второй гармоникой.Previously, a reference Fourier transform is obtained, in the same way as in the first implementation, and the location of the power concentration regions corresponding to the first and second harmonics of the moiré-forming structure on the banknote is investigated. Based on this, the spatial frequency of the center of the power concentration region is fixed

corresponding to the first harmonic, and the diameter of its enclosing circle d. Spatial frequency

corresponds to the area of power concentration formed by the second harmonic.

характерной области, соответствующей первой гармонике (S25). Проверку допустимости (S22) проводят аналогично первой реализации для положения центра

характерной области, и принимают решение на основании проверки допустимости таким же образом.When performing the second implementation of the method in the machine (Fig. 10), as in the first implementation, additional authentication begins with checking the validity of the spatial frequency of the center of the characteristic region. To do this, calculate the center

the characteristic area corresponding to the first harmonic (S25). The admissibility check (S22) is carried out similarly to the first implementation for the center position

characteristic area, and make a decision based on validation in the same way.

Obtaining a controlled Fourier transform and checking (S26) the root-mean-square value of the amplitude in the characteristic region corresponding to the first harmonic is carried out similarly to the first implementation, using the value of the limit L ₁ , which is set in advance for the first harmonic. If the rms amplitude is less than the limit of L ₁ , then it is concluded that the banknote is rejected (S31), considering it not genuine.

Otherwise, the banknote is preliminarily considered authentic.

характерной области, соответствующей второй гармонике.Next, calculate (S27) the spatial frequency of the center

characteristic region corresponding to the second harmonic.

не выходят за интервал

то есть за пределы области видимости 19.The spatial frequency of the center of the characteristic region corresponding to the second harmonic is considered acceptable when the coordinates of the vector

do not go beyond the interval

that is, out of scope 19.

оказывается недопустимым, то завершают проверку и выносят решение о подлинности банкноты (S32). Оно основывается только на предварительном подтверждении наличия муарообразующих линий и их пространственной частоты, так как провести проверку толщины линий по второй гармонике невозможно.If a

is invalid, verification is completed and the note is authenticated (S32). It is based only on preliminary confirmation of the presence of moiré-forming lines and their spatial frequency, since it is impossible to check the thickness of the lines by the second harmonic.

вычисляют (S28) среднеквадратичную амплитуду второй гармоники муара в контролируемом Фурье-образе. Если эта амплитуда оказывается менее заранее определенного предела L₂, то делают вывод о браковке банкноты (S31), считая ее неподлинной. В противном случае, делают вывод о подтверждении подлинности банкноты (S32) на основании проверки не только наличия муарообразующих линий и их пространственной частоты, но также и их толщины.In the case of admissibility of the spatial frequency of the center of the characteristic region corresponding to the second harmonic, in a circle of diameter d with a center given by the spatial frequency

calculating (S28) the root-mean-square amplitude of the second harmonic of the moiré in the controlled Fourier transform. If this amplitude is less than the predetermined limit L ₂ , then the banknote is rejected (S31) as not genuine. Otherwise, it is concluded that the banknote is authenticated (S32) based on checking not only the presence of moiré-forming lines and their spatial frequency, but also their thickness.

The limits L ₁ and L _{2 are} found in advance, experimentally, similar to the limit L in the first implementation.

Each of the implementations described herein is intended to further validate the authenticity of a note relating to a single combination of note type and orientation. However, if the machine is designed to process many types of banknotes in different orientations, then the applicability of the claimed method can be extended to several types and orientations of banknotes. For this purpose, it is necessary to choose from the machine-processed combinations of type and orientation those combinations in which a noticeable moire is observed in the digital scanned image of the checked banknote, and for them also implement the claimed method. This can be done according to the first or second implementation. Then, depending on the type and orientation of the checked banknote, the machine will be able to carry out the corresponding implementation of the claimed method. In the course of checking a banknote, the selection of the implementation of a suitable method for several types and orientations of banknotes is provided in step S10, as shown in FIG. 8.

The use of the Fourier transform provides a convenient, but not the only, approach to the analysis of moiré. In accordance with the claimed method, other methods can also be used to check the monitored features, for example, based on direct measurement of the size, period and direction of moiré fringes in the monitored zone.

Claims (44)

1. A method of additional verification of the authenticity of a valuable document, carried out after the basic verification of authenticity, during which a digital image of the checked document is obtained and the type of the document and its orientation during filing for verification are established, on the basis of which the location of the control zone of the digital image of the document is determined, where there is the property of the formation of a moire image, while to carry out additional verification of the authenticity of the document, the mentioned digital image of the checked document is used, consisting of sequentially scanned lines and obtained when the following condition is met: during scanning, the length of the registration area on the document, measured in the scanning direction, from which the sensor receives at least 80 percent of the radiation recorded during the scanning of a single line, is less than the line spacing of the document being checked relative to the sensor, as well as the skew angle of the checked document in relation to the scanning direction, moreover, the specified additional check consists in determining the presence in the mentioned control zone in the digital image of the checked document of the controlled features that are predetermined as characteristic of the moiré image formed in the digital image of an authentic document for the specified type and orientation in the same control zone, at a skew angle the original document during scanning, equal to the skew angle of the checked document, and according to a predetermined rule, a decision is made to confirm the authenticity of the document being checked. 2. The method according to claim 1, in which a linear image sensor is used to obtain a digital image, which has an optical system designed to form a document image on a linear array of photodetecting elements located along a straight line, and containing an array of lenses that form a non-inverted image with a unit magnification , and placed in one row, or in two parallel rows, moreover, the linear size of the field of view of each lens is at least 3 times greater than the distance between its optical axis and the optical axis of any of the adjacent lenses, in this case, both the line of arrangement of the photodetector elements of the sensor and the row of placement of the lenses in the array are oriented in a direction practically perpendicular to the scanning direction, and the size of the receiving area of each photodetector element, measured in the scanning direction, is less than the line pitch of the document being checked relative to the sensor. 3. The method according to claim. 2, in which the linear image sensor is equipped with a backlight system made with the ability to pulse irradiate the location of the document in the field of view of the sensor, while an irradiation pulse is applied during scanning of each line of the digital image and ensure that the sum of the amount of displacement of the sensor's field of view along the document during this pulse and the size of the sensor's field of view in the scanning direction is equal to a value less than the value of the interline spacing of the sensor's field of view along the document used to obtain a digital image of the document being checked. 4. The method according to claim. 3, in which when obtaining a digital image of the checked document, a backlight system is used containing radiation sources of different wavelengths, which are simultaneously turned on during the generation of the backlight pulse for at least part of the duration of said pulse. 5. The method according to claim 4, in which an optical system is used, additionally equipped with structural elements that limit the path of the rays of each of the lenses. 6. The method according to any one of claims. 1-5, in which, in order to check for the presence of controlled features in the control area in the digital image of the checked document using a discrete Fourier transform, the controlled Fourier image of the contents of the digital image of the checked document in the control area is calculated, check for the presence in the controlled Fourier transform of the features corresponding to the controlled features, and by the presence of these signs, a conclusion is made about the presence of controlled features. 7. The method according to claim 6, wherein a digital image of an authentic document for each type and orientation is preliminarily obtained by scanning the specified original document without skewing, providing a discreteness in the scanning direction, at least three times the discreteness of the monitored digital image in the scanning direction, and performing a discrete Fourier transform of the contents of this digital image in the control zone with obtaining a reference Fourier-image of the control zone, and to check for the presence of signs corresponding to the controlled features, in the controlled Fourier image of the control zone, taking into account the skew angle, the reference Fourier image of the control zone is transformed to obtain the target Fourier image, the spatial frequency range of which coincides with that of the controlled Fourier image , and according to a predetermined similarity criterion, the distribution of the amplitude of the spectral components of the Fourier image on the frequency plane of the target Fourier image is compared with the distribution of the control zone controlled by the Fourier image, moreover, upon reaching a predetermined degree of similarity, a conclusion is made about the presence of features corresponding to the controlled features, if the predetermined degree of similarity is not achieved, then a conclusion is made about the absence of features corresponding to the controlled features. 8. The method according to claim 7, in which the transformation of the reference Fourier transform into the target Fourier transform is performed as follows: a rotated Fourier transform is obtained by rotating the reference Fourier transform by a skew angle, at least one shifted Fourier transform is obtained by shifting the rotated Fourier transform along the axis corresponding to the scanning direction by a multiple of the spatial scanning frequency determined by the pitch of the digital image lines, and calculating a linear combination of the rotated Fourier transform and at least measure of one shifted Fourier transform according to the predetermined coefficients of the linear combination, after that, from the specified linear combination, a region of spatial frequencies is selected, which coincides in size with the region of spatial frequencies of the controlled Fourier transform, and it is used as a target Fourier transform. 9. The method according to claim 8, wherein the reference Fourier transform is pre-transformed into a set of preforms of target Fourier transforms, each of which corresponds to a specific value of the skew angle from a predetermined set, and then, during the verification of the proximity criterion, one of the pre-obtained preforms is used as the target Fourier transform corresponding to which the skew angle used in its transformation is closest to the skew angle of the document being checked. 10. The method according to claim 6, wherein as a sign corresponding to the monitored feature, the excess of the specified amplitude level of the monitored Fourier image in at least one characteristic region on the frequency plane is used, moreover, the size and location of the characteristic region on the frequency plane are calculated based on the skew angle according to a predetermined algorithm. 11. The method according to claim 10, wherein before calculating the controlled Fourier transform, the admissibility of the skew angle is additionally evaluated using a predetermined condition, and if the skew angle is unacceptable, the check for the presence of controlled features in the control zone in the digital image of the checked document is completed ahead of schedule, after that, a decision, predetermined for this case, is made on the authenticity of the document being checked. 12. The method according to any one of claims. 1-5, in which the line pitch is changed by a value limited to predetermined limits as it is scanned, and this change is taken into account when checking for the presence of monitored features. 13. The method according to any one of claims. 1-5, in which, in the course of checking for the presence of controlled features, the fulfillment of a predetermined criterion corresponding to the controlled feature and based on the assessment of the thickness of the lines of the printed image on the surface of the document is checked. 14. The method according to claim 13, wherein the inspection zone is predetermined in such a way that it contains in a digital image of a genuine document of a given type and a given orientation a plurality of parallel segments spaced at equal distances from each other, so that they define a vector of the spatial frequency of the segments, and in the course of checking for the presence of controlled features, the vector of the spatial frequency of the moiré is found based on the vector of the spatial frequency of the segments and the skew angle, using a discrete Fourier transform, the controlled Fourier image of the contents of the digital image of the document being checked in the control zone is calculated, and checking the fulfillment of the predetermined criterion using the distribution of the values of the monitored Fourier transform along a straight line passing through the starting point on the frequency plane in the direction of the vector of the spatial moiré frequency.

Images