METHOD AND DEVICE FOR CHECKING VALUE DOCUMENTS AND METHOD AND DEVICE FOR GENERATING CHECKING PARAMETERS FOR THE CHECKING METHOD

Abstract

A method is provided for generating templates for checking value documents of a predefined value document type having at least two predefined production elements that optionally partially overlap, as well as digital training images of training value documents of the predefined value document type. The method involves: for each of the training images, determining position data sets having position coordinates in a coordinate space each describing the positions of the production elements on the value document at least relative to one another; forming at least two position sub-regions of the coordinate space, each comprising at least one predefined number of position data sets. The position subregions contain no common position data sets; for each of the position sub-regions, determining a template using training images of the value documents; storing the template and position sub-region data describing the position and extent of each position sub-region.

Claims

1.-17. (canceled)

18. A method for generating templates for checking value documents of a predefined value document type, in which method value documents of the predefined value document type have at least two predefined production elements, which optionally partially overlap, wherein digital training images of training value documents of the predefined value document type are used, each of which have pixels each as-signed pixel data; the method comprising the following steps: for each of the training images, determining position data sets having position coordinates in a coordinate space each describing the positions of the production elements on the value document at least relative to one an-other, and forming at least two position sub-regions of the coordinate space, each comprising at least one predefined number of position data sets, the position sub-regions containing no common position data sets, for each of the position sub-regions, determining a template using training images of the value documents whose position coordinates lie in the position sub-region, and storing the template and position sub-region data describing the position and extent of each position sub-region.

19. The method according to claim 18, wherein forming the position subregions comprises a plurality of division steps, wherein in each of the division steps a current position sub-region from among those present which contains double the predefined number or more than double the predefined number of position data sets is in each case divided into a predefined division number of newly formed position sub-regions each comprising at least the predefined number of position data sets, and the respective current position sub-region from among those present is replaced by the newly formed position sub-regions, wherein a region containing the position coordinates of all the training value documents is used as position sub-region in the first division step, and wherein division steps are carried out as often as until all the position sub-regions formed satisfy at least one predefined termination criterion.

20. The method according to claim 19, wherein the coordinate space is n-dimensional where n>1, and in each of the division steps a division is carried out in only one of the dimensions, and successive divisions of a position sub-region and of resultant position sub-regions are carried out in different spatial dimensions in each case.

21. The method according to claim 19, wherein the termination criterion contains the criterion that the position sub-regions contain at most a pre-defined multiple of the predefined number of position data sets.

22. The method according to claim 18, wherein the termination criterion contains the criterion that the numbers of position data sets in the position sub-regions differ by less than a tolerance portion relative to a reference value.

23. The method according to claim 18, wherein the termination criterion contains the criterion that an extent of the position sub-regions in at least one direction of the coordinate space in each case falls below a predefined maximum extent for the at least one direction.

24. The method according to claim 18, wherein the termination criterion contains the criterion that the number of position sub-regions is less than or equal to a predefined number of position sub-regions.

25. The method according to claim 18, wherein determining the templates for the position sub-regions involves using in each case the training images of those training value documents whose position data sets lie in the respective position sub-region and training images of those training value documents whose position data sets lie within a predefined distance from the boundary of the respective position sub-region but outside the position sub-region.

26. A device for generating templates for checking value documents of a predefined value document type; wherein value documents of the predefined value document type have at least two predefined production elements, which optionally partially overlap, wherein digital training images of training value documents of the predefined value document type are used, each of which have pixels each assigned pixel data, the device comprising a storage unit for storing digital training images of value documents of the predefined value document type, wherein the device is configured to carry out a method according to the invention according to claim using the training images, wherein the device has an interface via which the generated templates and position subregion data can be transmitted to another device, and the device is configured to transmit the generated templates and the position sub-region data to the other device, and/or wherein the device is configured to store the generated templates and the position subregion data in a storage unit and/or the storage unit.

27. A computer program comprising program code for carrying out the method according to claim 18 when the program is executed on a computer.

28. A computer-readable data carrier comprising program code which is executable by a computer so that the computer carries out a method according to claim 18.

29. A method for checking value documents of a predefined value document type, each having two predefined production elements, which optionally partially overlap, using templates, produced by a method according to claim 18, which are predefined for predefined position sub-regions for positions of the production elements, the method comprising the following steps: providing a digital value document image of a value document to be checked of the predefined value document type, comprising pixels each assigned pixel data, determining a position of the production elements in the provided value document at least relative to one another, determining a template for the digital value document image depending on the determined position of the production elements and the predefined position sub-regions, and checking the digital value document image using the determined template.

30. A device for checking value documents, in particular banknotes, each having at least two predefined production elements, in particular print layers and/or security elements, which optionally partially overlap, using templates, and position subregion data assigned to the templates, the device comprising an evaluation unit having at least one memory in which the templates and the position sub-region data respectively assigned thereto are stored, and an interface for providing a digital value document image, wherein the evaluation unit is designed to carry out a method according to claim 29.

31. The device according to claim 30, which furthermore has an image capture unit for capturing a digital value document image of a value document to be checked, said image capture unit being connected to the interface for providing a digital value document image via a signal connection.

32. A computer program comprising program code for carrying out a method according to claim 29 when the program is executed on a computer.

33. A computer-readable data carrier comprising program code which, when executed by a computer, causes a method according to claim 29 to be carried out.

34. A device for processing, in particular checking and/or counting and/or sorting and/or destroying, value documents of a predefined value document type, in particular banknotes, each having at least two predefined production elements, in particular print layers and/or security elements, which optionally partially overlap, the device comprising: a feed unit for feeding individual or separated value documents to be processed, a dispensing unit having at least one dispensing section for receiving processed value documents, a transport unit for transporting individual or separated value documents from the feed unit to the dispensing unit, and a checking device according to claim 30, wherein the image capture unit of the checking device is arranged on the transport path and designed such that digital value document images of value documents to be checked that are transported past the image capture unit are captured while being transported past and are provided for the use in the checking device.

Description

[0057] The invention is explained even further below by way of example with reference to the drawings, in which:

[0058] FIG. 1 shows a schematic illustration of a value document processing device, a banknote sorting device in the example,

[0059] FIG. 2 shows a roughly schematic illustration of a device for generating templates,

[0060] FIGS. 3A, B show roughly schematic illustrations of value documents which have different positions of production elements,

[0061] FIG. 3C shows a schematic illustration of a digital image captured from the value document in FIG. 3B,

[0062] FIG. 4 shows a roughly schematic flowchart for one example of an adaptation method for generating templates,

[0063] FIG. 5 shows a schematic frequency chart for positions of production elements in predefined training images,

[0064] FIG. 6 shows a flowchart with sub-steps of a step S12 in FIG. 4,

[0065] FIGS. 7A-D show illustrations of position data sets in a coordinate space in various division cycles in accordance with FIG. 6,

[0066] FIG. 8 shows a roughly schematic flowchart for a further example of an adaptation method for generating templates, and

[0067] FIG. 9 shows a roughly schematic flowchart for one example of a checking method.

[0068] A value document processing device 10 in FIG. 1, in the example a device for processing value documents 12 of a predefined value document type in the form of bank notes, is configured for sorting value documents 12 depending on the state determined by means of the value document processing device 10, and depending on the authenticity of processed value documents, which is checked by means of the value document processing device 10.

[0069] The device has a feed unit 14 for feeding value documents 12, a dispensing unit 16 for dispensing or receiving processed, i.e. sorted, value documents, and a transport unit 18 for transporting separated value documents from the feed unit 14 to the dispensing unit 16.

[0070] In the example, the feed unit 14 comprises an introduction compartment 20 for a value document stack and a separator 22 for separating value documents 12 from the value document stack in the introduction compartment 20 and feeding separated value documents to the transport unit 18.

[0071] In the example, the dispensing unit 16 comprises three dispensing sections 24, 25 and 26, into which processed value documents can be sorted depending on the interim result of the processing, checking in the example. In the example, each of the sections comprises a stacking compartment and a stacking wheel, not shown, by means of which value documents fed can be placed in the stacking compartment. In other exemplary embodiments, a dispensing section can be replaced by a unit for destroying banknotes.

[0072] The transport unit 18 has at least two, in the example three, branches 28, 29 and 30, at each of the ends of which one of the dispensing sections 24 and 25 and 26, respectively, is arranged, and, at the branch junctions, diverters 32 and 34, which are controllable by actuating signals and by means of which value documents are able to be fed to the branches 28 to 30 and thus to the dispensing sections 24 to 26 depending on actuating signals.

[0073] On a transport path 36defined by the transport unit 18between the feed unit 14, more precisely in the example the separator 22, and the first diverter 32 in the transport direction T downstream of the separator 22, there is arranged a sensor unit 38, which, while value documents are being transported past, captures physical properties of the value documents and forms sensor signals which represent the capture results and which constitute sensor data. In this example, the sensor unit 38 has an image capture unit 40 with an optical reflection sensor, which captures a reflection color image of the value document, and also other sensors 42merely symbolized by boxesfor other physical properties of a value document.

[0074] A control and evaluation unit 46 is connected via signal connections to the sensor unit 38 and the transport unit 18, in particular the diverters 32 and 34. In conjunction with the sensor unit 38, it classifies a value document into one of predefined sorting classes depending on the signals or sensor data of the sensor unit 38 for the value document. These sorting classes can be predefined for example depending on a state value determined by means of the sensor data, and likewise depending on an authenticity value determined by means of the sensor data. It is possible to use for example the values capable of circulation or not capable of circulation as state values, and the values counterfeit, suspected counterfeit or authentic as authenticity values. Depending on the sorting class determined, by way of the outputting of actuating signals, it controls the transport unit 18, more precisely here the diverters 32 and 34, such that the value document, according to its sorting class determined during the classification, is dispensed into a dispensing section of the dispensing unit 16 that is assigned to the class. In this case, the assignment to one of the predefined sorting classes or the classification is effected depending on criteria which are predefined for the assessment of the state and the assessment of the authenticity and which depend on at least one portion of the sensor data.

[0075] For this purpose, in particular besides at least one corresponding interface 44 for the sensor unit 38 or the sensors thereof, in particular of the image capture unit 40, the control and evaluation unit 46 has a processor 48 and a memory 50 connected to the processor 48, in which memory there is stored at least one computer program comprising program code which, when executed, causes the processor 48 to control the device and to evaluate the sensor signals of the sensor unit 38, in particular in order to determine a sorting class of a processed value document. Furthermore, program code is stored therein which, when executed, causes the processor 48 to control the device, and to control the transport unit 18 according to the evaluation.

[0076] The interface 44, the processor 50 and the memory 48 or a section of the memory 48 in which a corresponding computer program and method parameters are stored are part of a computer and form an evaluation unit 47 within the meaning of the present disclosure. In this example, the evaluation unit 47 evaluates the signals of the reflection sensor 40 separately from those of the other sensors. In addition, the processor 50 and other sections of the memory 48 can also fulfil other functions, in the example the control of the value document processing device 10.

[0077] The reflection sensor 40 is configured to capture an RGB reflection image of a value document while it is being transported past the reflection sensor 40 by means of the transport unit 18, and to generate a digital image therefrom, which the evaluation unit 47 evaluates.

[0078] Depending on the value document properties, the control and evaluation unit 46, more precisely the evaluation unit 47, using the sensor data of the various sensors, in sub-evaluations, determines in each case whether or not the value document properties determined represent an indication of the state and/or the authenticity of the value document. Subsequently, corresponding data can be stored in the control and evaluation unit 46, for example the memory 50, for later use. Depending on the sub-evaluations, the control and evaluation unit 46 then determines a sorting class as the overall result for the checking in accordance with a predefined overall criterion and forms the sorting or actuating signal for the transport unit 18 depending on the sorting class determined.

[0079] For the processing of value documents 12, value documents 12 that have been inserted into the introduction compartment 20 as a stack or individually are separated by the separator 22 and fed in separated form to the transport unit 18, which transports the separated value documents 12 past the sensor unit 38. The latter captures the properties of the value documents 12, with sensor signals that represent the properties of the respective value document being formed. The control and evaluation unit 46 captures the sensor signals or sensor data, and depending on these determines a sorting class, in the example a combination of an authenticity class and a state class, of the respective value document and, depending on the result, controls the diverters such that the value documents, according to the sorting class determined, are transported into a dispensing section assigned to the respective sorting class.

[0080] The evaluation unit 47 together with the image capture unit 40 form one example of a checking device for checking value documents of a predefined value document type, each having two predefined production elements, in particular print layers and/or security elements. Accordingly, the computer program contains instructions for carrying out a method for checking value documents of a predefined value document type, in particular banknotes, each having two predefined production elements, in particular print layers and/or security elements, using templates, produced in particular by means of an adaptation method outlined below, which are predefined for predefined positions of the production elements. In the checking method, by means of the reflection sensor 40, a digital value document image of a value document to be checked is captured and is provided in a corresponding section of the memory 50 in the evaluation unit 47. For the value document image provided, the relative position of the production elements with respect to one another is determined. In real time, a template for the digital value document image or the checking thereof is then stipulated depending on the determined position of the production elements using the position sub-region data. Afterward, the digital value document image is checked using the determined template.

[0081] For providing templates, use is made of an adaptation device shown roughly schematically in FIG. 2, that is to say a device 70 for generating templates for checking value documents of a predefined value document type, wherein value documents of the predefined value document type have at least two predefined production elements, in particular print layers and/or security elements, which optionally partially overlap. The device 70 is a data processing unit comprising a storage unit 72 for storing digital training images of the predefined value document type and preferably generated templates. The adaptation device 70 is configured to carry out an adaptation method described below, using the training images, and to store the generated templates and the assigned position sub-region data in the storage unit 72. For this purpose, the device can have at least one processor 74 connected to the storage unit 72 via a data connection, and a program memory 76 connected to the processor 74 via a data connection, in which program memory there is stored program code which, when executed, causes the device to carry out, by means of the processor 74, the adaptation method outlined below using the training images stored in the storage unit 72. In other exemplary embodiments, the program memory 76 can also be formed by a section of the storage unit 72. The adaptation device 70 can furthermore have a data interface, not shown in the figure, for example a network card or LAN card, via which generated element templates stored in the storage unit 72 can be transmitted to another device.

[0082] One example of a value document 12 of a predefined value document type having two predefined production elements 62 and 64 in the form of print layers is shown roughly schematically in FIG. 3A.

[0083] FIG. 3B shows another value document of the same value document type as in FIG. 3A, in which the relative position of the production elements 62 and 64corresponding to the production elements 62 and 64with respect to one another differs from that for the value document in FIG. 3A. The position of the production element 62 in FIG. 3A is represented by a dashed line in FIG. 3B. The difference in the relative positions is illustrated in FIGS. 3A and 3B by a vector V and V, respectively, which points from one predefined characteristic element 62A and 62A, respectively, of the production element 62 and 62, respectively, to another predefined characteristic section or element 64A and 64A, respectively, of the production element 64 and 64, respectively. This vector will in turn be represented by two corresponding components in a two-dimensional Cartesian coordinate system, with an x-axis and a y-axis in the example.

[0084] The digital image 60 of the value document in FIG. 3B is illustrated schematically again in FIG. 3C. It contains pixels 66, which in the example are arranged on a square grid and correspond to locations in the digital image and thus on the value document imaged. In this case, the image is preprocessed such that it represents only the value document 12 over the whole area, that is to say that the edges of the value document in the image run along the corresponding edge pixels. In this example, this preprocessing is carried out for all of the images, such that the images include corresponding representations. The vector V and V, respectively, can then be represented by corresponding pixel coordinates, which can preferably be non-negative integers. In this case, the x-axis and the y-axis run parallel to the long and short edges, respectively, of the value document in the image.

[0085] In the adaptation method, digital training images of training value documents of the predefined value document type are used. These are provided for the adaptation method in a first step. In the example, finished and clean, preferably freshly printed, value documents of the predefined value document type are used for this purpose, which preferably have variations in the position of the production elements. Preferably, the value documents also encompass those having large differences in the position of the production elements. Preferably, these are chosen such that they include corresponding value documents having different relative positions of the production elements 62 and 64, where with particular preference the frequency of value documents having given relative positions corresponds at least approximately to the frequencies of actual value documents of the predefined value document type.

[0086] The following exemplary adaptation method illustrated roughly schematically in FIG. 4 is used for generating templates.

[0087] The training images can be captured for example by the processing device 10 described, in particular the reflection sensor 40, for which purpose the digital images fed to the evaluation unit 47 are stored. These can be transmitted by means of a storage medium or via a data connection (not shown) to the adaptation device 70, where they are stored in the storage unit 72 and thus provided. The digital images each have identical numbers and arrangements of pixels and show the entire value document.

[0088] In step S10, a position data set having position coordinates in a coordinate space is determined for each of the training images. The position data set or the position coordinates therein describes in each case the position of the production elements on the value document relative to one another. In the example, two production elements are used; the coordinate space is therefore two-dimensional.

[0089] The positions are each determined in relation to the same position reference system, given by the edges of the value document in the image or, since the image shows only the entire value document, by the edges of the image or corresponding axes.

[0090] For determining the positions, anchor regions 62A and 64A are used in the present example, which were stipulated beforehand for value documents of the predefined value document type and are characteristic of the production element and in particular always visible. In order to simplify the illustration, only one anchor region in each case is used in this exemplary embodiment; in other exemplary embodiments, preferably at least two or more anchor regions are used for each of the production elements. Using methods known per se, the anchor regions can be sought in each case in the digital images. In the present example, the average value over the locations of the specimen can be used as the position of the anchor region. For each of the training value documents or training images, the relative position in the form of a vector as mentioned above is stored in a manner assigned to the training image, for which purpose corresponding vector components are used, which are also regarded as position coordinates. In the example, each of the vectors leads from the anchor region 62A to the anchor region 64A. The result of this step is therefore a set of training images andrespectively assigned theretovectors or vector components ?x.sub.AB and ?y.sub.AB in the mutually orthogonal directions or dimensions of the coordinate space: a vector is thus assigned to each training image.

[0091] An exemplary distribution of vectors or vector components is illustrated in the representation in FIG. 5. Since the coordinates and thus the vector components ?x.sub.AB and ?y.sub.AB are integers, the frequencies or numbers ng with which specific relative positions or vectors have occurred in the training images can be represented as bars in a two-dimensional bar chart, the height of which corresponds to the number of training value documents. Typically, relative positions in the vicinity of the relative position predefined by a specification are particularly frequent, while relative positions with larger deviations from the predefined relative position are less common.

[0092] Step S12 involves forming at least two position sub-regions of the coordinate space, each comprising at least one predefined number of position data sets, the position sub-regions, that is to say in each case different position sub-regions from among the position sub-regions, containing no common position data sets.

[0093] The position sub-regions can have any desired predefined shapes, in principle, but the position sub-regions preferably all have the same shape, a rectangular shape in the example. The position sub-regions can differ, however, at least in terms of the position and optionally also the dimensions, the side lengths in the example. The position, shape and dimensions or size of a position sub-region are described by suitable position sub-region data, which preferably, as in this exemplary embodiment, are given by coordinates of suitable points in the coordinate space which describe the position and at least dimensions. In the case of the example of rectangles, assuming for example that the shape of a rectangle is chosen as shape in the method, the coordinates of diagonally opposite corner points of a respective rectangular position sub-region can be used.

[0094] As illustrated roughly schematically in FIG. 6, in step S12 for forming the position subregions, sub-steps are carried out, if appropriate repeatedly, until a predefined termination criterion is satisfied. In the steps, each current position sub-region, if possible, is divided into N newly formed position sub-regions, each containing however at least one predefined number of position data sets. In the present example, N=2.

[0095] When the steps are carried out for the first time, the position sub-region used for the first division step is a region which contains the position coordinates of all the training value documents, in the example the smallest rectangular region in which all the position data sets lie.

[0096] In this exemplary embodiment, during a division of a respective position sub-region, a division is carried out in only one of the dimensions, preferably one of the dimensions or directions of the coordinate space.

[0097] In the example, successive divisions of a position sub-region and of resultant position subregions are carried out in each case in different spatial dimensions or spatial directions of the coordinate space.

[0098] More precisely, in the adaptation method of the example, a division is effected in each case along one of two mutually orthogonal division directions. These are the coordinate axes of the coordinate space in the example.

[0099] If a position sub-region was divided by division in a first division direction, in the case of this the division is effected in a second division direction orthogonal to the first. This is done in such a way that, for the resultant position sub-regions in the event of a division, the direction along which the division was effected during their formation is stored or the direction along which the next division has to be effected is stored. During the very first division, one of the division directions is predefined and used. During the next division, the division direction to be correspondingly used can then be determined.

[0100] More precisely, in step S12.1, a next position sub-region to be processed is selected from the current position sub-regions present.

[0101] The subsequent step S12.2 involves checking whether the number of position data sets in the current position sub-region selected is greater than N times the minimum number. If that is not the case, a division into position sub-regions each having at least the minimum number of position data sets is not possible. The method is then continued with step S12.5, in which a termination criterion is checked. This is described in even greater detail below.

[0102] By contrast, if the number of position data sets in the current position sub-region selected is greater than N times the minimum number, the current position sub-region is divided into N position sub-regions in step S12.3. If the current position sub-region resulted from division along a first of the division directions, the second of the division directions, which runs orthogonally to the first division direction, is used in step S12.3. The division is effected such that at least the minimum number of position data sets lies in each of the resulting position sub-regions.

[0103] There are usually a number of possibilities for this. In the example, the division is effected such that for a rectangular position sub-region, the number of position data sets therein is determined. If this number is less than three times the predefined number, the position sub-region is divided along the predefined direction into two newly formed position sub-regions containing approximately the same number of position data sets. Otherwise, division along the predefined direction results in the formation of one position sub-region containing the predefined number of position data sets, and another position sub-region having at least double the predefined number.

[0104] In step S12.4, the current position sub-region or the position sub-region data stipulating the latter is replaced by the newly formed position sub-regions or the position sub-region data stipulating the latter. For the newly formed position sub-regions, position sub-region data specifying in each case the position and shape of the position sub-regions are stored, in a manner respectively assigned to said position sub-regions. Afterward, step S12.5 is carried out, in which the termination criterion already mentioned above is checked.

[0105] The division steps S12.1 to S12.4 are carried out as often as until all the position sub-regions formed satisfy at least one predefined termination criterion, which in the example is checked in step S12.5 already mentioned. If said criterion is satisfied, no further division is performed, and the method is continued with step S14. Otherwise a renewed division is attempted, for which purpose step S12.1 is carried out anew on the basis of the position sub-regions then present.

[0106] The termination criterion can be a single criterion or can comprise a plurality of sub-criteria which have to be satisfied either cumulatively or alternatively in order that the termination criterion is deemed to be satisfied.

[0107] In the present exemplary embodied, what is checked as criterion is merely whether at least one position sub-region present at this point in time contains less than N times or N times the minimum number of position data sets. If this is the case, the termination criterion is satisfied, step S12 is ended and the method is continued with step S14. Otherwise the method is continued with step S12.1.

[0108] Some of the division steps are illustrated in a highly simplified manner for a concrete example in FIGS. 7A to 7D. FIG. 7A shows the initial situation before the first division: in the coordinate space, here a Cartesian coordinate system with coordinate axes ?x.sub.AB and ?x.sub.AB, the relative positions of the production elements of all the training images used are represented by small crosses. They lie within a position sub-region T.sub.0 in the form of a rectangle, chosen as first position sub-region in step S12.1. The position and size of the position sub-region are described by two diagonally opposite corner points A.sub.0 and B.sub.0 (top left corner and bottom right corner) or the coordinates thereof as position sub-region data. To simplify the illustration, 5 is the number chosen as the minimum number of position data sets.

[0109] After checking the number of position data sets in the region T.sub.0 in a step corresponding to step S12.2, the ?x.sub.AB direction is chosen as division direction for the chosen position sub-region in a step corresponding to step S12.3. A corresponding division line is illustrated in a dashed manner in FIG. 7B. Division results in the position sub-regions T.sub.1 and T.sub.2, each having at least the minimum number of position data sets.

[0110] In a step corresponding to step S12.4, the position sub-region T.sub.0 is replaced by the newly formed position sub-regions T.sub.1 and T.sub.2. Furthermore, the position sub-region data A.sub.0 and B.sub.0 are replaced by corresponding position sub-region data A and B.sub.1, and A.sub.2 and B.sub.2.

[0111] After checking the termination criterion in a step corresponding to step S12.5, in a further loop in a step corresponding to step S12.1 the position sub-region T.sub.1 is chosen as the next position sub-region to be divided (also cf. FIG. 7C).

[0112] As illustrated in FIG. 7C, a division is then effected in a direction orthogonal to the division direction in the preceding step S12.3, namely the second axis, the ?y.sub.AB axis, of the coordinate space or coordinate system. The division is effected in such a way that both newly formed position sub-regions T.sub.3 and T.sub.4 contain at least the minimum number of position data sets. In the example, the region T.sub.3 is chosen such that one of the position sub-regions formed contains the smallest possible number of position data sets, which however is greater than or equal to the minimum number. The division line is once again illustrated in a dashed manner.

[0113] In a step corresponding to step S12.4, the position sub-region T.sub.1 is replaced by the newly formed position sub-regions T.sub.3 and T.sub.4. Furthermore, the position sub-region data A.sub.1 and B.sub.1 are replaced by corresponding position sub-region data A.sub.3 and B.sub.3, and A.sub.4 and B.sub.4.

[0114] After checking the termination criterion in a step corresponding to step S12.5, in a further loop in a step corresponding to step S12.1, from the position sub-regions T.sub.2, T.sub.3 and T.sub.4 now present, a position sub-region, in the example the position sub-region T.sub.4, is chosen as the next position sub-region to be divided (also cf. FIG. 7D). The position sub-region T.sub.3 has less than double the predefined minimum number of position data sets, and therefore can no longer be divided further.

[0115] As illustrated in FIG. 7D, a division is then effected in a direction orthogonal to the division direction in the preceding step S12.3, namely the first axis, the ?x.sub.AB axis, of the coordinate space or coordinate system. The division is effected in such a way that both newly formed position sub-regions T.sub.5 and T.sub.6 contain at least the minimum number of position data sets. In the example, the region T.sub.5 is chosen such that one of the position sub-regions formed contains the smallest possible number of position data sets, which however is greater than or equal to the minimum number. The division line is once again illustrated in a dashed manner.

[0116] In a step corresponding to step S12.4, the position sub-region T.sub.4 is replaced by the newly formed position sub-regions T.sub.5 and T.sub.6. Furthermore, the position sub-region data A.sub.4 and B.sub.4 are replaced by corresponding position sub-region data A.sub.5 and B.sub.5, and A.sub.6 and B.sub.6.

[0117] After checking the termination criterion in a step corresponding to step S12.5, in a further loop in a step corresponding to step S12.1 the sole position sub-region T.sub.2 now present is chosen as the next position sub-region to be divided (also cf. FIG. 7D). The division is then continued analogously to the divisions described above until the termination criterion is satisfied.

[0118] After the end of step S12 and thus of the divisions, step S14 involves, for each of the position sub-regions formed, determining a template using training images of the value documents whose position coordinates lie in the respective position sub-region. Furthermore, for the position sub-regions formed, the template respectively determined and, assigned to the template, position sub-region data describing the position and extent of the respective position sub-region are stored.

[0119] The pixel data for the pixels of the template are determined depending on the corresponding pixel data of those training images whose position coordinates lie in the respective position sub-region. For checking contamination, for example, as pixel data it is possible to stipulate lower and upper limits for permissible pixel data values, these resulting from the minimum and maximum, respectively, of the corresponding pixel data of the corresponding pixels in the training images.

[0120] The position sub-region data preferably specify the position and extent for a given shape of the position sub-region or the position, extent and the shape of the respective position sub-region. In the present example, a rectangle is predefined as the shape, and the coordinates of diagonally opposite corners of the rectangle in the coordinate space used are used as position sub-region data. These represent both the position and the extent of the respective position sub-region.

[0121] A further exemplary embodiment of an adaptation method, illustrated roughly schematically in FIG. 8, differs from the preceding exemplary embodiment merely in that step S14 is replaced by a step S14.

[0122] Step S14 differs from step S14 in that determining the templates for the position sub-regions involves using not only the training images of those training value documents whose position data sets lie in the respective position sub-region but in addition also training images of those training value documents whose position data sets lie within a predefined maximum distance from the boundary of the respective position sub-region but outside the position sub-region. The actual determination is effected analogously to the determination in the first example.

[0123] In the example, the maximum distance is 1 pixel.

[0124] One example of a checking method for checking value documents of the predefined value document type which involves using templates predefined for predefined positions of the production elements is illustrated roughly schematically in FIG. 9. Templates used can be, in particular, templates and assigned position sub-region data which were determined by one of the described examples of adaptation methods. The checking method can be carried out by means of the value document processing device 10. The templates and assigned position sub-region data can be stored in the evaluation unit 47, for example.

[0125] In the checking method, in step S20, during the transport of such a value document, by means of the sensor unit 38, in particular the image capture unit 40, a digital value document image of the value document to be checked that is being transported past or through the sensor unit 38 is captured. The digital value document image comprises pixels each assigned pixel data. The resolution of the value document image corresponds to that of the training images. That means that the value document images have substantially the same numbers of pixels and arrangements as the training images. A representation of such a value document image corresponds to that in FIG. 3C for training value documents. The same correspondingly applies to the templates.

[0126] In step S22, a position of the production elements is determined for the captured value document image, in the example use being made of the same method as in the adaptation method in the first example. More precisely, position coordinates are determined in the same coordinate space as was also used in the adaptation method.

[0127] Step S24 involves determining, depending on the determined position coordinates, a template for the digital value document image and thus a template for use during the further checking. This is done by checking in which of the stored position sub-regions the position coordinates determined for the current value document lie, with the position sub-region data respectively assigned to the templates being used. In the example, this involves determining more specifically that position sub-region, i.e. rectangle, stipulated by the position sub-region data in which the determined position coordinates lie. The template corresponding to the position sub-region data and thus to the position sub-region is used as a template for the subsequent step S26.

[0128] In step S26, the digital value document image is then checked using the determined template by means of a predefined image checking method. In the example of checking contamination, in the image checking method, in the simplest example, for each of the pixels it is possible to check whether the pixel data lie between the minimum and the maximum which are predefined for the pixel by the template. If the number of pixels for which this is not the case exceeds a predefined number, a contamination is recognized, otherwise a sufficiently good state.

[0129] Depending on the checking result, step S28 involves generating and outputting a corresponding signal representing the checking result. Depending on this signal, a sorting signal can then be formed and output, as described initially.

[0130] Other exemplary embodiments of adaptation methods differ from the first two exemplary embodiments in that, instead of the criterion in regard to the number of position data sets in a position sub-region, the termination criterion contains the criterion that the numbers of position data sets in the position sub-regions differ from a reference value by less than a tolerance portion. The arithmetic mean over the numbers of position data sets in the position sub-regions currently present is used as reference value in the present example. A value of 20%, for example, can be chosen as tolerance portion. A more uniform division can thus be achieved. The division steps may need to be adapted for this, however.

[0131] Still other exemplary embodiments can differ from the first two exemplary embodiments described in that, instead of the criterion in regard to the number of position data sets in a position sub-region, the termination criterion contains the criterion that an extent of the position subregions in at least one direction of the coordinate space in each case falls below a predefined maximum extent for the at least one direction. In the example, it is possible to predefine an extent in both coordinate directions. Excessively large position sub-regions can thus be avoided.

[0132] Still other exemplary embodiments can differ from the first two exemplary embodiments described in that the termination criterion contains the criterion that the number of position subregions is less than or equal to a predefined number of position sub-regions. This number can be chosen depending on the number of available training images and the necessary speed during checking. Preferably, the criterion used in the preceding paragraph can additionally be used. A termination takes place only when both criteria are satisfied.

[0133] Further exemplary embodiments of the adaptation method can differ from the exemplary embodiments outlined above in that N>2, for example N=3, is chosen. In a division step, a division into three new position sub-regions is then effected, but these sub-regions must contain at least the minimum number of position data sets.

[0134] Other exemplary embodiments can differ from the exemplary embodiments described above in that the templates are determined from the assigned training images in a different way. By way of example, an average value of the corresponding pixel data of the pixel in the training images can be determined as pixel data for a pixel of the template.

[0135] In other exemplary embodiments, the image checking method used can be a method as described in WO2008/058742A1. The template data then have the shape specified there. In the adaptation method, the determination of templates would take place analogously to the determination of the adaptation data in the cited document.