EP3170154B1 - Method and device for fitness testing of value documents - Google Patents

Description

The invention relates to a method and a device for testing the fitness of documents of value, such as banknotes, checks, tickets, .... In the context of the present application, fitness testing is understood to mean, on the one hand, checking the fitness of used documents of value, e.g. banknotes in circulation, on the other hand also the quality check of new documents of value after their production before they go into circulation, for example the quality check of banknotes freshly printed.

To check the fitness of value documents, it is known to check the value documents with the help of sensors and to compare the recorded fitness measurement values with threshold values in order to distinguish between fit and unfit value documents and, if necessary, further fitness classes such as ATM-fit to distinguish. It is necessary that a user of the device who carries out the fitness test, e.g. an operator, an adapter or a service person of the device, selects and defines suitable threshold values for the sensors. These defined threshold values then serve to classify the value documents to be checked with the help of a value document processing device into fit and non-fit value documents and to sort the value documents, for example, into different output compartments of the value document processing device.

US2002 / 0043560A discloses an assessment system for banknotes using a neural network and fuzzy logic. Also the EP0706698B deals with the assessment of banknotes. With the help of measurement parameters, it is decided whether the banknote is suitable for reuse. the US2008 / 0034313A generally refers to the use of fuzzy logic to replace threshold values and Boolean logic. the US7978899B discloses a bank note processing system in which a fitness type and / or a fitness level is assigned to a bank note.

A disadvantage of the known methods is that it is complicated for the user to define suitable threshold values for the sensors. In this case, for example, by the manufacturer of a value document processing device already predetermined threshold values, which are rigidly predetermined, assumed. Problems arise B. by aging or soiling of the value document processing device or due to changes, for example aging, of the value documents to be processed over time. If one or more of the threshold values is set by the user even just slightly too high, value documents which are actually no longer suitable for circulation are classified as fit by the value document processing device. However, if one or more of the threshold values is set by the user even just slightly too low, value documents which are actually suitable for circulation are classified as unfit by the value document processing device. Thus, the value documents to be processed are not sorted into fit and unfit value documents in the manner desired by the user.

Furthermore, the fact that many fitness measurement values contribute to the fitness of the respective value document has not yet been sufficiently taken into account. This is because it can happen that several fitness measurement values are each just below their threshold value and the relevant document of value is classified as fit, although a human observer would classify it as unfit - based on the appearance. Many fitness measurement values also mean that the fitness test for a user of the device for fitness testing - due to the large number of threshold values required for this - is often difficult to survey. If the user wants to change the severity of the fitness test, e.g. tighten it, he has to change a large number of parameters.

The object of the present invention is therefore to improve the fitness check of documents of value.

This object is achieved by the subjects of the independent claims.

For the fitness test of the value documents, at least two different fitness criteria of the value documents are selected for the state the documents of value are characteristic. The value documents are checked by recording measurement data, a fitness measurement value for the respective value document being determined from the recorded measurement data for each of the selected fitness criteria. A degree of unfit of the respective value document is then determined from the respective measured fitness value for each of the selected fitness criteria. The unfit degree is determined with the help of an unfit function, which uniquely assigns an unfit degree to each fitness measurement value. Each unfit function is characterized by a first threshold value, a second threshold value and an uncertainty range lying between the first and the second threshold value, in which the respective unfit function either has a monotonically decreasing or a monotonically increasing profile. The unfit degrees of the various fitness criteria are then combined to form an unfit probability that is specific for the respective value document. Based on the unfit probability that was determined for the respective value document, a fitness classification of the respective value document is carried out, in which the respective value document is classified as fit or unfit. In addition to unfit and fit, one or more further fitness classes can also be provided, for example ATM-fit for documents of value with particularly great fitness.

If a fitness measured value lies in the uncertainty range according to the invention between the two threshold values, the respective value document is neither clearly classified as fit nor clearly classified as unfit with regard to the respective fitness criterion, but receives a corresponding degree of unfit between 0 and 1 for the respective fitness criterion Unfit degree of the respective fitness criterion is a quantitative measure for the fitness of the value document in relation to the respective fitness criterion. The same value document is assigned its own unfit degree (only valid for the respective fitness criterion) for each of the fitness criteria. The unfit function according to the invention introduces a fuzzy distinction between fit and unfit with regard to the respective fitness criterion. The fitness test is based on the perception of a human observer. This is because a human observer would also classify a value document - visually - as unfit if several fitness criteria are just below the limit of unfit (ie the corresponding fitness measured values are in the uncertainty range of the unfit function). The areas of the respective fitness measured value outside the uncertainty area are reached in those cases in which the viewer would classify a value document as clearly fit or clearly unfit with regard to the respective fitness criterion. The combination of the unfit degrees to form an unfit probability reflects the overall impression that a human observer makes of the fitness of a document of value.

By combining the unfit degrees, a common unfit probability is determined for several fitness criteria. The fitness test is therefore easier to oversee for the user of the device. If the user wants to change the severity of the fitness test, e.g. tighten it, he can do this simply by changing the one threshold value with which the unfit probability is compared. The unfit functions of the various fitness criteria do not need to be changed. In contrast, up to now, to tighten the fitness test, all threshold values have to be tightened individually, with the size of the shift in the threshold value also having to be specified for each fitness criterion.

In contrast to other methods for fitness testing, the invention allows an intuitively understandable procedure, since firstly the uncertainty areas are based on actual, comprehensible fitness measurement values can be established, and secondly, a single unfit probability is derived for the respective value document. The method according to the invention is also stable with respect to small changes in the fitness of the documents of value examined.

The results of the fitness classification can be used to observe and thus monitor the development of the fitness of a large number of value documents over time. If the fitness of the value documents on average no longer corresponds to the specifications made for them, measures can be taken to meet the specifications again in the future. For example, the fitness of used documents of value that are in circulation can thus be monitored and controlled, but also the fitness of new, freshly printed documents of value in the course of the quality check before the documents of value go into circulation. As a parameter for observation and comparison, for example, the number or the proportion of the value documents classified in a certain fitness class (e.g. as fit or unfit) (e.g. the unfit proportion, the fit proportion, etc.) can be used or an average value of the unfit Probability over the multitude of documents of value. For example, the fitness classification of the same or different types of value-document processing devices can be compared with one another or of value-document processing devices that are located at different installation locations or the value documents of different regions can be checked.

• mindestens zwei verschiedenen Fitnesskriterien der Wertdokumente, die für den Zustand der Wertdokumente charakteristisch sind, auszuwählen,
• aus den aufgenommenen Messdaten für jedes des ausgewählten Fitnesskriterien einen Fitnessmesswert für das jeweilige Wertdokument zu bestimmen,
• für jedes der ausgewählten Fitnesskriterien aus dem jeweiligen Fitnessmesswert des jeweiligen Wertdokuments mit Hilfe einer Unfit-Funktion jeweils einen Unfitgrad zu bestimmen, wobei die Unfit-Funktion jedem Fitnessmesswert eindeutig einen Unfitgrad zuordnet, und wobei jede Unfit-Funktion durch einen ersten Schwellenwert, einen zweiten Schwellenwert und einen zwischen dem ersten und dem zweiten Schwellenwert liegenden Unsicherheitsbereich charakterisiert ist, in dem die jeweilige Unfit-Funktion entweder einen monoton abfallenden oder einen monoton ansteigenden Verlauf hat, und
• die Unfitgrade der verschiedenen Fitnesskriterien des jeweiligen Wertdokuments zu einer Unfit-Wahrscheinlichkeit zu kombinieren, die für das jeweilige Wertdokument spezifisch ist, und
• eine Fitness-Klassifizierung des jeweiligen Wertdokuments durchzuführen, basierend auf der Unfit-Wahrscheinlichkeit, die für das jeweilige Wertdokument bestimmt wurde.

The invention also relates to a device for the fitness check of value documents, in particular a value document processing device designed for fitness check, with a measuring device for recording measurement data of the value documents, and an evaluation device for the fitness classification of the value documents on the basis of the recorded measurement data. The evaluation device is designed to

• to select at least two different fitness criteria of the value documents which are characteristic of the state of the value documents,
• to determine a fitness measurement value for the respective value document from the recorded measurement data for each of the selected fitness criteria,
• to determine a degree of unfit for each of the selected fitness criteria from the respective measured fitness value of the respective document of value with the aid of an unfit function, the unfit function uniquely assigning an unfit degree to each measured fitness value, and each unfit function by a first threshold value, a second threshold value and an uncertainty range lying between the first and the second threshold value is characterized, in which the respective unfit function has either a monotonically decreasing or a monotonically increasing profile, and
• to combine the unfit degrees of the various fitness criteria of the respective value document to form an unfit probability that is specific for the respective value document, and
• carry out a fitness classification of the respective value document, based on the unfit probability that was determined for the respective value document.

• ein Eingabefach zur Aufnahme der zu prüfenden Wertdokumente,
• eines oder mehrere Ausgabefächer,
• eine Transporteinrichtung zum Transportieren der Wertdokumente aus dem Eingabefach, an der Messeinrichtung vorbei, in das/ die Ausgabefächer,
• eine Benutzerschnittstelle zur Eingabe von Parametern für die Fitnessprüfung und ggf. zur Ausgabe der Ergebnisse der Fitness-Klassifizierung,
• eine Steuereinrichtung zur Steuerung der Vorrichtung, um die Wertdokumente in Abhängigkeit ihrer jeweiligen Fitness in unterschiedliche Ausgabefächer zu sortieren, wobei die Steuereinrichtung und die Auswerteeinrichtung zusammen oder voneinander getrennt ausgebildet sein können.

The value-document processing device designed for the fitness test also usually has the following devices:

• an input compartment for receiving the documents of value to be checked,
• one or more output trays,
• a transport device for transporting the documents of value from the input compartment, past the measuring device, into the output compartment (s),
• a user interface for entering parameters for the fitness test and, if necessary, for outputting the results of the fitness classification,
• a control device for controlling the device in order to place the documents of value in different output compartments depending on their respective fitness to sort, wherein the control device and the evaluation device can be formed together or separately from one another.

In order to ensure the secure separation of fit and unfit value documents, individual threshold values can be used for each type of value document, since each value document type has its own physical properties which can differ greatly from one another. In the case of banknotes, for example, individual threshold values can be used for each denomination and / or issue of the respective currency. However, the same threshold values can also be used for similarly created value documents, e.g. for bank notes of different denominations but the same currency.

The threshold values can be established for the respective value documents or optimized if necessary in advance of the fitness test, e.g. when adapting the respective value document type. The first and / or second threshold value of the unfit function of a fitness criterion can be derived from the hitherto customary (single) threshold value that was previously used for the fitness test on the basis of this fitness criterion. For example, the previous threshold value is used as the upper threshold value of the unfit function and the lower threshold value of the unfit function is selected below it. Alternatively, the first and second threshold values can be placed symmetrically around the previous threshold value.

The selection of the fitness criteria of the various fitness criteria can be carried out manually or automatically. When the fitness criteria are automatically selected, for example at least two predefined fitness criteria are selected which were established in advance of the fitness test for the respective value documents, for example individually for the respective value document type. However, the setting can also be carried out by an expert Based on empirical values. The fitness criteria selected are preferably those in which the respective frequency distribution of the fit and unfit value documents are separated from one another as far as possible or overlap as little as possible. For example, those fitness criteria are selected in which the respective frequency distributions of the fit and unfit value documents have a maximum of 30% overlap.

The value-document processing device can propose to the user such fitness criteria for selection (e.g. display on the user interface of the value-document processing device) which are particularly well suited for distinguishing between fit and unfit value documents. For example, those fitness criteria are suggested whose fit and unfit frequency distribution show the smallest overlap. The fitness criteria are displayed, for example, in the order of descending overlap of the frequency distributions on the user interface of the value document processing device. The results of the fitness classification can be displayed on the user interface, e.g. the development of the fitness of a large number of value documents over time or in comparison with the results of the fitness classification of other value documents.

The selected fitness criteria relate in particular to one or more of the following properties of the value documents: soiling, wear and tear, damage, foreign bodies (for example adhesive strips) or limpness of the respective value document. At least two different of these fitness criteria are preferably selected. If the method according to the invention is used to check the quality of new value documents, one or more of the following fitness criteria can also be selected - in addition to or instead of these fitness criteria: quality of the print (color, error), position of the print image in relation to the value document edges, manufacturing quality of authenticity features (e.g. based on their optical or magnetic properties), position of authenticity features on the value document, etc.

• Fleckengröße (Fläche) oder Fleckenanzahl,
• Größe (Fläche) oder Anzahl fehlender Teile, z.B. Eselsohren, Löcher,
• die Risslänge oder -fläche,
• die Länge oder Fläche eines Klebestreifens,
• der Verschmutzungsgrad des Wertdokuments in einer oder mehreren ROIs (Regions of Interest), z.B. in einem unbedruckten Bereich des Wertdokuments (Weißfeld),
• den Abnutzungsgrad (Abrieb oder Ausbleichen) der Druckfarbe in einem bedruckten Bereich des Wertdokuments,
• den Abnutzungsgrad von Echtheitsmerkmalen.

For example, one or more of the following fitness criteria can be selected:

• Spot size (area) or number of spots,
• Size (area) or number of missing parts, e.g. dog ears, holes,
• the crack length or area,
• the length or area of an adhesive strip,
• The degree of soiling of the document of value in one or more ROIs (Regions of Interest), e.g. in an unprinted area of the document of value (white field),
• the degree of wear (abrasion or fading) of the printing ink in a printed area of the document of value,
• the degree of wear of authenticity features.

The fitness measurement values in question can be determined quantitatively, for example, using the spatially resolved optical transmission, remission or luminescence intensity and, if necessary, suitable image processing. The degree of wear of magnetic authenticity features can be determined quantitatively with the help of a magnetic sensor. The dimensions of adhesive strips or the missing parts can also be determined using the ultrasonic transmission intensity. Furthermore, the limpness, folds or creasing of the value document can be determined quantitatively on the basis of the ultrasound transmission or remission intensity in one or more ROIs or the entire value document and selected as a fitness criterion.

The combination of the unfit degrees of the selected fitness criteria is carried out, for example, in such a way that, for each selected fitness criterion, up to one certain fitness measurement value (e.g. up to the first threshold value) the respective fitness criterion does not influence the fitness classification (unfit probability) of the respective value document at all, but that the respective fitness criterion from a certain fitness measurement value (e.g. from the second threshold value) decides the fitness classification of the respective value document, and that the respective fitness criterion for fitness measured values in the uncertainty range only partially influences the fitness classification in interaction with the other selected fitness criteria. This is achieved, for example, by the following formula (1).

When combining the unfit degrees G _{j of} the various fitness criteria to form the unfit probability P, the unfit degrees of the various fitness criteria can be multiplied - individually for each value document. The unfit probability P can be determined from the unfit degrees using the following formula, for example: $$\mathrm{P.}=1-{\displaystyle \prod _j{\left(1-G_j\right)}^{k_j}=1-{\left(1-G1\right)}^{{\mathrm{<figure-callout\; id="1"\; label="k"\; filenames="imgf0001.png"\; state="\{\{state\}\}">k</figure-callout>}}_1}\cdot {\left(1-G2\right)}^{k_2}\cdot \mathrm{...}}$$

When combining the unfit degrees G _j according to this formula, those fitness criteria with a high unfit degree G _{j dominate} over fitness criteria with a low unfit degree G _j . Fitness criteria with a very low unfit degree G _j ≈0 (ie almost fit) have a negligible influence on the unfit probability P. Already a single unfit degree of approximately 1 (ie almost unfit) causes the resulting unfit probability P of the value document to be approximately 100%, even if the other unfit degrees of this document of value are negligible (ie fit).

Since an exponent k _j is available for each fitness criterion, this simplifies the handling of the fitness test for different types of documents of value, since the fitness test can be carried out for all documents of value on the basis of this one generic formula (1) and, if necessary, the exponents k _j can be selected depending on the type of document of value. In the simplest case, the exponents are k _j = 1. The respective degree of unfit G _{j of} the relevant fitness criterion is then "normally" taken into account. The respective degree of unfit G _j can be masked out by an exponent k _j = 0, ie the respective fitness criterion can be ignored. In the case of a non-linear course of the unfit function in the uncertainty range, an approximately linear course of the unfit component as a function of the fitness threshold can be generated _{by exponents k j> 1.}

Combining the unfit degrees of the various fitness criteria to form the unfit probability P can, however, also be carried out by a linear combination of the unfit degrees G _{j of} the various fitness criteria, in particular by adding up the unfit degrees, possibly with different weighting of the unfit degrees of various fitness criteria.

• Bereitstellen einer ersten Gruppe von fit-Wertdokumenten und einer zweiten Gruppe von unfit-Wertdokumenten. Die fit- und unfit-Wertdokumente können zur selben Wertdokumentsorte gehören (gleiche Währung der Banknoten, ggf. auch gleiche Stückelung), können aber auch zu unterschiedlichen Sorten gehören. Die Einstufung als fit oder unfit kann z.B. durch manuelle Prüfung (auf Basis der menschlichen Wahrnehmung) oder durch Prüfung mittels eines Referenz-Messsystems durchgeführt worden sein.
• Prüfen der fit- und der unfit-Wertdokumente der ersten und zweiten Gruppe durch Aufnehmen von Messdaten dieser Wertdokumente mit Hilfe einer Messeinrichtung,
• Bestimmen mindestens eines Fitnessmesswerts, insbesondere mindestens zweier verschiedener Fitnessmesswerte für jedes Wertdokument aus den Messdaten des jeweiligen Wertdokuments,
• Bestimmen einer ersten Häufigkeitsverteilung des jeweiligen Fitnessmesswerts für die erste Gruppe der fit-Wertdokumente und einer zweiten Häufigkeitsverteilung des jeweiligen Fitnessmesswerts für die zweite Gruppe der unfit-Wertdokumente,
• Verwenden der ersten und zweiten Häufigkeitsverteilung des jeweiligen Fitnessmesswerts (der Häufigkeitsverteilung für die fit-Wertdokumente und der für unfit-Wertdokumente), um die bei der Fitnessprüfung der Wertdokumente zu verwendenden Fitnesskriterien (manuell oder automatisch) auszuwählen und/oder um die Unfit-Funktion des jeweiligen Fitnesskriteriums (manuell oder automatisch) festzulegen oder zu verändern.

The fitness test can be optimized in that the frequency distribution of the fitness measurement values is determined from a selection of fit and unfit value documents provided and this is used to select the fitness criteria or to optimize the unfit function. The unfit function of the respective fitness criterion is determined, for example, in advance of the fitness test, based on fit and unfit value documents, the following steps being carried out:

• Providing a first group of fit documents of value and a second group of unfit documents of value. The fit and unfit value documents can belong to the same value document type (same currency of the bank notes, possibly also the same denomination), but can also belong to different types. The classification as fit or unfit can have been carried out, for example, by manual testing (based on human perception) or by testing using a reference measuring system.
• Checking the fit and unfit value documents of the first and second group by recording measurement data of these value documents with the aid of a measuring device,
• Determining at least one fitness measurement value, in particular at least two different fitness measurement values for each value document from the measurement data of the respective value document,
• Determining a first frequency distribution of the respective measured fitness value for the first group of fit documents of value and a second frequency distribution of the respective measured fitness value for the second group of unfit documents of value,
• Using the first and second frequency distribution of the respective fitness measurement value (the frequency distribution for the fit value documents and that for unfit value documents) to select the fitness criteria to be used in the fitness check of the value documents (manually or automatically) and / or to use the unfit function of the set or change the respective fitness criterion (manually or automatically).

For example, the two frequency distributions of the respective fitness measurement value are used to determine the first and / or the second threshold value of the respective unfit function for the respective fitness criterion and / or to determine / optimize the course of the unfit function in the uncertainty range. This can be done manually, by a person, or automatically, by the device. For example, a value range of the respective fitness measurement value in which both fitness measurement values of fit value documents and fitness measurement values of unfit value documents are located is used as the uncertainty range of the unfit function of the respective fitness criterion. The uncertainty range can encompass the entire range of values in which both fitness measurement values of fit documents of value and of unfit documents of value lie, or also just a section of this range of values.

Alternatively or additionally, the two frequency distributions of the respective fitness measurement value are used to select those fitness criteria for the fitness classification in which the frequency distribution of the fit value documents and the frequency distribution of the unfit value documents overlap as little as possible (e.g. a maximum of 30% overlap of the two frequency distributions).

When determining the unfit functions, the respective unfit function can be determined, for example, in such a way that a first threshold value of the unfit function is placed on a fitness measurement value for which the fit frequency is much greater than the unfit frequency, in particular at least one specific one Has ratio (eg 5: 1), and the second threshold value is placed on a fitness measurement value, in which the fit frequency is much smaller than the unfit frequency (see, for example, the threshold values X1 and Y1 in the histogram of the Fig. 2a ). Alternatively, the accumulated frequency distribution (cumulative histogram) of the fitness measurement values can also be used to determine the first and second threshold values. For example, the first / second threshold value is placed on a fitness measurement value, in which the accumulated frequency of the fit value documents has a certain ratio to the accumulated frequency of the unfit value documents.

In the uncertainty area of the respective unfit function, a course of the unfit function can be selected that was defined in advance of the fitness test for the respective fitness criterion, in particular on the basis of empirical values. However, the two threshold values can also be determined manually by selecting the respective threshold value from a large number of predetermined threshold values.

• Ermitteln des Unfit-Anteils der einen oder mehreren Wertdokumentgruppe, der den Anteil an Wertdokumenten angibt, die bei der Fitness-Klassifizierung der jeweiligen Wertdokumentgruppe als unfit klassifiziert wurden,
• Prüfen des Unfit-Anteils auf mindestens eine für den Unfit-Anteil bestimmte Vorgabe (z.B. ob ein Höchstwert für den Unfit-Anteil und/ oder ein Mindestwert für den Unfit-Anteil erreicht wird oder ein bestimmter vorgegebener Unfit-Anteil zumindest näherungsweise erreicht wird),
• Verändern der Unfit-Funktion eines oder mehrerer der verwendeten Fitnesskriterien in Abhängigkeit des ermittelten Unfit-Anteils der Wertdokumentgruppe, um eine veränderte Fitness-Klassifizierung der Wertdokumente zu erreichen. Dabei wird, falls der ermittelte Unfit-Anteil die bestimmte Vorgabe erfüllt, die Unfit-Funktion unverändert gelassen. Falls der ermittelte Unfit-Anteil jedoch die bestimmte Vorgabe nicht erfüllt, wird die Unfit-Funktion verändert und die folgenden Schritte a)-f) im Rahmen der Simulation durchgeführt und ggf. einmal oder mehrmals wiederholt:
  1. a) erneutes Bestimmen der (i.a. veränderten) Unfitgrade des jeweiligen Wertdokuments für die mindestens zwei verschiedenen Fitnesskriterien unter Verwendung der veränderten Unfit-Funktion des jeweiligen Fitnesskriteriums,
  2. b) erneutes Kombinieren der Unfitgrade der verschiedenen Fitnesskriterien zu einer (i.a. veränderten) Unfit-Wahrscheinlichkeit des jeweiligen Wertdokuments und
  3. c) erneutes Fitness-Klassifizieren des jeweiligen Wertdokuments anhand der jeweiligen Unfit-Wahrscheinlichkeit,
  4. d) erneutes Ermitteln des (i.a. veränderten) Unfit-Anteils der einen oder mehreren Wertdokumentgruppe, der den Anteil an Wertdokumenten angibt, die bei der Fitness-Klassifizierung der jeweiligen Wertdokumentgruppe als unfit klassifiziert werden,
  5. e) erneutes Prüfen des erneut ermittelten Unfit-Anteils auf die bestimmte Vorgabe,
  6. f) erneutes Verändern der Unfit-Funktion eines oder mehrerer der verwendeten Fitnesskriterien in Abhängigkeit des erneut ermittelten Unfit-Anteils der Wertdokumentgruppe, wobei, falls der erneut ermittelte Unfit-Anteil die bestimmte Vorgabe erfüllt, die Unfit-Funktion unverändert gelassen wird, und, falls der erneut ermittelte Unfit-Anteil die bestimmte Vorgabe nicht erfüllt, die Unfit-Funktion verändert wird und die Schritte a)-f) im Rahmen der Simulation einmal oder ggf. mehrmals wiederholt werden. Sobald der erneut ermittelte Unfit-Anteil die bestimmte Vorgabe, die für den Unfit-Anteil definiert wurde, erfüllt, wird die Unfit-Funktion unverändert gelassen und die zuletzt durchgeführte Fitnessklassifizierung (Schritt c) als endgültige Fitness-Klassifizierung verwendet. Das Erfüllen der Vorgabe kann eine näherungsweises Erreichen eines bestimmten Unfit-Anteils sein. Durch das Wiederholen der Schritte a)-f) kann z.B. der Unterschied des automatisch ermittelten Unfit-Anteils zu einem vorgegebenen (z.B. manuell ermittelten) Unfit-Anteil minimiert werden. Durch das Anpassen des ermittelten Unfit-Anteils an den vorgegebenen wird erreicht, die Maßstäbe der manuellen Fitnessprüfung auf die automatische Fitnessprüfung zu übertragen.

After a first fitness classification, the unfit function can be optimized by simulating the fitness classification, for example in order to achieve a certain unfit proportion. For the value documents of one or more value document groups to be checked (for example a certain value document stack containing a mixture of fit and unfit value documents), the following steps are carried out in the context of a simulation:

• Determining the unfit portion of the one or more value document groups, which indicates the portion of value documents that were classified as unfit in the fitness classification of the respective value document group,
• Checking the unfit portion for at least one specification determined for the unfit portion (e.g. whether a maximum value for the unfit portion and / or a minimum value for the unfit portion is reached or a certain predetermined unfit portion is at least approximately reached),
• Changing the unfit function of one or more of the fitness criteria used as a function of the determined unfit portion of the value document group in order to achieve a changed fitness classification of the value documents. In this case, if the determined unfit component fulfills the specific specification, the unfit function is left unchanged. However, if the determined unfit component does not meet the specific requirement, the unfit function is changed and the following steps a) -f) are carried out as part of the simulation and, if necessary, repeated once or several times:
  1. a) renewed determination of the (generally changed) unfit degrees of the respective value document for the at least two different fitness criteria using the changed unfit function of the respective fitness criterion,
  2. b) recombining the unfit degrees of the various fitness criteria to form an (generally changed) unfit probability of the respective value document and
  3. c) renewed fitness classification of the respective value document based on the respective unfit probability,
  4. d) renewed determination of the (generally changed) unfit portion of the one or more value document groups, which specifies the portion of value documents that are classified as unfit in the fitness classification of the respective value document group,
  5. e) renewed checking of the re-determined unfit component for the specified specification,
  6. f) renewed changing of the unfit function of one or more of the fitness criteria used as a function of the re-determined unfit portion of the value document group, with the unfit function being left unchanged if the re-determined unfit portion meets the specified specification, and if so the re-determined unfit component does not meet the specified specification, the unfit function is changed and steps a) -f) are repeated once or, if necessary, several times within the scope of the simulation. As soon as the re-determined unfit component fulfills the specific requirement that was defined for the unfit component, the unfit function is left unchanged and the fitness classification carried out last (step c) is used as the final fitness classification. Fulfilling the requirement can be an approximate achievement of a certain unfit proportion. By repeating steps a) -f), for example, the difference between the automatically determined unfit component and a predetermined (for example manually determined) unfit component can be minimized. By adapting the determined unfit proportion to the specified one, the standards of the manual fitness test are transferred to the automatic fitness test.

The simulation has the advantage that the fitness classification can be optimized without renewed recording of measurement data from the value documents of the value documents to be checked. Thus, an additional mechanical stress on the value documents, which is repeated Recording of measurement data in a value document processing device would be avoided. For example, the simulation is carried out after the value document check on the basis of a large number of checked value documents (which may have been checked by several different value document processing devices), for example by the central bank, in order to control the quality of banknotes in circulation.

For the above-mentioned changing of the unfit function of the respective fitness criterion, in particular the position of the two threshold values (i.e. the position of the uncertainty area) and / or the course of the unfit function in the uncertainty area is changed for the respectively selected fitness criterion. When changing the unfit function of the fitness criterion, the unfit function of this fitness criterion, depending on the result of the fitness classification of the value documents of the value document group, can be changed in such a way that the unfit portion in the repeated fitness classification, as desired, is either increased or decreased. In particular, the unfit portion can be adapted to the unfit portion that was previously determined during manual presorting for this value document group.

The unfit portion can, however, also be adapted to a predetermined unfit portion that has resulted for the same value document group at one or more other value document processing devices. For this purpose, the same value document stack is brought to a plurality of value document processing devices and there (with the same or also with different sensors) is subjected to a fitness test. As a result, the fitness test on several value document processing devices can be equated. This is because the unfit portion achieved at different value document processing devices has not always matched up to now, for example when the sensors of the value document processing devices are different, use different measuring principles or are not calibrated to match, or if the value-document transport takes place at different speeds or by different transport devices.

If the unfit portion is less than a predetermined or desired unfit portion, the first and / or the second threshold value is changed (e.g. one or both of them reduced) so that the rigor of the fitness test is increased when the fitness is re-classified. And, if the unfit proportion is higher than a predetermined or desired unfit proportion, the first and / or second threshold values are changed (e.g. one or both of them increased) in such a way that the severity of the fitness test is reduced when the fitness is re-classified. In the case of fitness criteria in which the degree of unfit in the uncertainty area increases monotonically with increasing fitness measurement value (e.g. damaged area) and the percentage of unfit is too large, one or both threshold values are increased in order to reduce the percentage of unfit, and if the percentage of unfit If the proportion is too low, one or both threshold values are reduced in order to increase the unfit proportion. In the case of fitness criteria in which the degree of unfit in the uncertainty area decreases monotonically with increasing fitness measurement value (e.g. remission in the white field) and the unfit portion is too large, one or both threshold values are reduced in order to reduce the unfit portion, and if the unfit -Proportion is too low, one or both threshold values are increased in order to increase the unfit proportion.

For the first run of the simulation for fitness classification, an original unfit function can be used, which was determined, for example, in advance of the value document check or was automatically selected. Based on this original unfit function, the unfit function is changed when the simulation is repeated.

The unfit function clearly assigns a degree of unfit to each fitness measurement value. The degree of unfit of the respectively selected fitness criterion is determined by inserting the respective measured fitness value of the respective value document into the unfit function of the respectively selected fitness criterion. The respective unfit function is a rule by means of which a degree of unfit is assigned to the fitness measurement values that the value documents have in relation to the respective fitness criterion. However, an individual unfit function is used for each fitness criterion. The degree of unfit is therefore specific to the respective fitness criterion.

In the uncertainty range of the unfit function, the respective fitness measurement value is classified neither clearly as fit nor as unfit. The unfit function is therefore not a simple sorting threshold. The uncertainty range is limited by a first and a second threshold value. In the uncertainty range between the first and second threshold value, it assumes either a monotonically decreasing or a monotonically increasing, in particular linear or non-linear, course. The unfit function assigns a degree of unfit that is greater than 0 and less than 1 to the measured fitness values in the uncertainty range. It assigns a degree of unfit of 0 to all fitness measurements lying beyond the first threshold value (i.e. those lying on the side of the first threshold value facing away from the uncertainty region) and all fitness measurements values lying beyond the second threshold value (i.e. those lying on the side of the second threshold value facing away from the uncertainty region) an unfit degree of 1. In particular, the unfit function assigns a fitness criterion-specific unfit degree of 1 to all those fitness measured values that are above the second (upper) threshold value, and a fitness criterion-specific unfit degree of 1, and to all those fitness measured values that are below the first (lower) threshold value Fitness criterion-specific unfit degree from 0 to.

The unfit functions of the selected fitness criteria differ from one another, in particular with regard to the position of the first and / or the second threshold value. However, you can also differentiate between the first and second threshold values with regard to the course of the unfit functions. Preferably, the same unfit function is used for all documents of value of a stack of documents of value to be checked for fitness to determine the degree of unfit of the respective fitness criterion.

The unfit probability of the respective value document determined by combining the unfit degrees provides a quantitative measure of the overall condition of the respective value document. The unfit probability can also be determined by combining the unfit degrees of more than two different fitness criteria. In particular, the unfit probability of the respective value document can be determined on the basis of the unfit degrees of at least five, preferably of at least 10, different fitness criteria. At least one unfit grade can also flow into the combination of the unfit grade, which was obtained for a fitness criterion with the hitherto customary sharp distinction between fit and unfit (without uncertainty range), ie a fitness criterion for which either an unfit grade of 0 or a degree of unfit of 1, but no degrees of unfit between 0 and 1 are used, cf. Fig. 1 .

In particular, at least one of the fitness measurement values can be a combined fitness measurement value in which at least two different fitness measurement values are combined. For example, several ROIs can be defined on the bank note for a fitness criterion, the fitness measurement values of which are then combined into a single fitness measurement value. At least one of the unfit degrees that is included in the unfit probability can be a group unfit degree that defines the fitness of the value document in relation to at least two different fitness criteria, the group unfit degree being determined with the aid of an unfit function that was set up for the combined fitness measurement value. For example, a first group unfit degree is determined for a first group of (at least two) fitness criteria, each relating to the soiling of the document of value, and a second group unfit degree is determined for a second group of (at least two) fitness criteria, each with the Affect damage to the value document. Optionally, a third group unfit degree is also formed for a third group of fitness criteria, for example for the wear and tear of the document of value or the limpness.

The unfit probability of the respective document of value is then determined by combining the first group unfit degree, which relates to the damage, with the second group unfit degree, which relates to the contamination of the banknote, and possibly with further unfit degrees, in particular further group unfit degrees. The group unfit grades have the advantage that they reduce the number of fitness criteria and thus also reduce the complexity of the fitness test. The fine adjustment of the fitness test is therefore simplified for the user of the device.

When combining the unfit degrees, it is also possible to combine with one another those unfit degrees that are determined from fitness measurement values that were recorded at different positions on the value document, which in particular lie in different ROIs of the value document.

For the fitness classification of the document of value, the unfit probability determined for the document of value is compared, for example, with a single fitness threshold, the document of value being classified as unfit when the fitness threshold is exceeded and otherwise as fit. The fitness classes fit and / or unfit can each also be divided into other fitness classes be divided, for example, the fitness class fit can be divided into the two fitness classes fit and ATM-fit.

The fitness threshold can be changed in order to control the unfit portion of the stack of value documents to be checked. For example, the fitness threshold can be changed for a user of the value document processing device. In this way, the severity of the fitness test with regard to all fitness criteria can easily be controlled by selecting a single threshold without having to adjust further adjustments or further thresholds. In addition, the unfit portion of the stack of value documents to be checked can be changed in this way easily.

If necessary, after checking the value documents for the respective checked value document group, a pre-calculation can be carried out in which the expected unfit portion of the respective value document group is determined for different values of the fitness threshold and the dependency of the unfit portion on the value of the Fitness threshold is determined. This information can be communicated to the user, e.g. output at a user interface of the value document processing device. In particular, the dependency of the unfit portion on the value of the fitness threshold can be shown as a look-up table. The user can then select the fitness threshold with which the unfit portion he wants is achieved in the fitness test. Information about the general quality of the processed value documents can also be output at the user interface.

The control of the unfit portion can also be carried out from a central point which compares the unfit portions of several value document processing devices with one another and, in order to adjust them, accordingly changed fitness thresholds on these Adjusts value document processing devices. This can be carried out by remote access from the central point to the value document processing devices (for example connected in the network).

In the fitness classification, in addition to the distinction between fit and unfit value documents, further fitness classes can also be differentiated, for example for value documents that are suitable for use in a cash dispenser (further fitness class ATM-fit). Such documents of value have to meet higher demands with regard to their fitness than is necessary for the fit classification, since the frequency of faults in the machines depends on the fitness of the documents of value.

In the simplest case, only a further, lower fitness threshold is used for the ATM-fit fitness class, with which the unfit probability of the respective value document is compared. If the unfit probability is also below this further fitness threshold, the value document is classified as ATM-fit. If the unfit probability is below the (previously mentioned) fitness threshold but above this further fitness threshold, the value document is classified as fit. If the unfit probability is above the (previously mentioned) fitness threshold, the value document is classified as unfit.

Alternatively, the decision between the ATM-fit fitness class and the fit fitness class can also be proceeded in the same way as for the decision between the fit fitness class and the unfit fitness class described above. For example, an ATM-fit degree is used - analogously to the unfit degree - and for this - analogously to the unfit function - an ATM-fit function is set up, likewise with two threshold values and an uncertainty range in between, in which the ATM-fit -Function monotonically decreases or increases. For fitness metrics below of a first threshold value, the ATM-fit grade is 0, for fitness measurements above a second threshold value, the ATM-fit grade is 1, and in the uncertainty range the ATM-fit grade is between 0 and 1. For the decision fit to unfit and the differentiation fit to ATM-fit, the same or different fitness criteria can be selected. If one considers the same fitness criterion, the two threshold values for the decision fit to ATM-fit are different from those for the decision fit to unfit, namely in such a way that higher fitness requirements are placed for the ATM-fit fitness class than for the fitness class fit. Depending on the fitness criterion, higher fitness requirements are met either by higher threshold values or by lower threshold values. The ATM-fit grades of the fitness criteria selected for this decision are combined - analogously to the unfit probability - to an ATM-fit probability of the respective value document. In the fitness classification of the value document, it is then decided on the basis of the ATM-fit probability whether the respective value document is ATM-fit or not, for example by comparing it with an ATM-fit threshold.

Fig. 1a

Häufigkeitsverteilung des Fitnessmesswerts M1 für umlauffähige (fit) und nicht-umlauffähige (unfit) Wertdokumente,

Fig. 1b

bisher übliche Fitnessklassifizierung mittels einer Schwelle,

Fig. 2a

Häufigkeitsverteilung des Fitnessmesswerts M1 eines Fitnesskriteriums K1 für fit- und für unfit-Wertdokumente,

Fig. 2b-c

zwei Beispiele einer Unfit-Funktion für Fitnesskriterium K1,

Fig. 3a

Häufigkeitsverteilung des Fitnessmesswerts M2 eines Fitnesskriteriums K2 für fit- und für unfit-Wertdokumente,

Fig. 3b

Beispiel einer Unfit-Funktion für das Fitnesskriterium K2,

Fig. 4

prinzipieller Aufbau einer Banknotenbearbeitungsmaschine,

Fig. 5a-c

Unfit-Funktionen für drei verschiedene Fitnesskriterien,

Fig. 6a-b

Tabelle zur Fitnessauswertung (Fig. 6a) anhand dreier verschiedener Fitnesskriterien und daraus ermittelte Unfit-Wahrscheinlichkeit (Fig. 6b) für drei Wertdokumente A,B,C

Fig. 7a-b

Zusammenfassen von Fitnessmesswerten und Gruppen-Unfitgrad für den zusammengefassten Fitnessmesswert.

Further advantages of the present invention emerge from the dependent claims and the following description of the exemplary embodiments. Show it:

Fig. 1a

Frequency distribution of the fitness measurement value M1 for fit and unfit documents of value,

Figure 1b

Previously common fitness classification using a threshold,

Fig. 2a

Frequency distribution of the fitness measurement value M1 of a fitness criterion K1 for fit and for unfit value documents,

Figures 2b-c

two examples of an unfit function for fitness criterion K1,

Fig. 3a

Frequency distribution of the fitness measured value M2 of a fitness criterion K2 for fit and for unfit value documents,

Figure 3b

Example of an unfit function for the fitness criterion K2,

Fig. 4

basic structure of a banknote processing machine,

Figures 5a-c

Unfit functions for three different fitness criteria,

Figures 6a-b

Fitness evaluation table ( Figure 6a ) on the basis of three different fitness criteria and the unfit probability determined from them ( Figure 6b ) for three documents of value A, B, C

Figures 7a-b

Summarizing fitness measurements and group unfit degrees for the summarized fitness measurement.

In Figure 4 a banknote processing machine 1 is shown which has an input compartment 20 into which banknotes 10 to be processed can be inserted, for example banknotes that are to be separated into fit and unfit banknotes. One after the other, the bank notes 10 are transferred individually from a singler 25 to a transport system 30. The transport system 30 transports the individual banknotes through the banknote processing machine, past a measuring device 41 into one or more output compartments 32, 34. The banknotes of different fitness can be sorted into different output compartments.

The measuring device 41 contains one or more sensors, the measurement data of which allow conclusions to be drawn about the state of the respective bank note in order to be able to assess and classify the bank note as fit or unfit. The sensors of the measuring device 41 can be, for example, one or more optical sensors with suitable light sources, the sensors detecting light reflected by the respective bank note or light transmitted through the respective bank note, e.g. B. Light of a certain wavelength or a certain wavelength range. Further sensors can check, for example, acoustic (for example ultrasound) and / or mechanical (for example thickness measurement) and / or thermal and / or magnetic and / or electrical properties of the respective bank note. On the basis of the measurement data from the sensors mentioned, statements can be made as to whether the respective banknote is soiled or damaged, or whether it has foreign objects such as clips or adhesive strips that affect the fitness of the respective banknote.

On the basis of the measurement data made available by the measuring device 41, an evaluation device 40 determines the fitness of the respective banknote, e.g. whether the respective banknote is a fit or an unfit banknote. The evaluation device 40 has, for. B. on a microprocessor that executes software for fitness testing, which is stored in an associated memory. Depending on the fitness of the banknote determined by the evaluation device 40, gates 31 and 33 are activated in the transport system 30 in order, for example, to store fit banknotes in a first output compartment 32 and unfit banknotes in a second output compartment 34. Further switches or output compartments can be provided in the transport system 30 of the bank note processing machine 1 and are indicated by a continuation 35.

A user interface 45 connected to the evaluation device 40, which z. B. can consist of a keyboard and a display or a touch screen, is used for the operation of the bank note processing machine 1 by a user. Commands can be entered or processing modes selected via the user interface 45, and processing results can be displayed or the user can be prompted to perform certain actions by means of instructions. The user interface can be accessed directly or by remote control.

In order to check the fitness of banknotes with regard to a specific fitness criterion, a fitness measurement value M of the banknote has usually been included compared to a single threshold value X. This threshold value is selected such that, in the case of a fitness measurement value, it lies between the frequency distribution for fit banknotes and the frequency distribution for unfit banknotes, cf. Fig. 1a . If the fitness measured value of the banknote is above the respective threshold value X, the respective banknote is classified as unfit, otherwise as fit, cf. Figure 1b . Up to now, such a comparison has been carried out for each fitness criterion and, if one (or more) of the fitness measurement values M of the bank note exceeds its respective threshold value X, the respective bank note is classified as unfit.

In Fig. 2a the same two frequency distributions are shown for a fitness measurement value M1 of a fitness criterion K1 as in FIG Fig. 1a , but now an upper threshold value X1 and a lower threshold value Y1 are used, which limit an uncertainty range U1 in which the bank notes are neither clearly classified as fit nor clearly as unfit. In the case of the fitness criterion K1, a large fitness measurement value M1 speaks for the presence of an unfit bank note. If the measured fitness value M1 is above the upper threshold value Y1, the banknote is clearly classified as unfit (unfit grade 1) with regard to the respective fitness criterion K1, and below the lower threshold value X1 unequivocally as fit (unfit grade 0). For fitness measurements that lie in the uncertainty range U1 between X1 and Y1, the unfit degree is between 0 and 1. The value of this unfit degree depends on the course of the selected unfit function F1. In the example from Figure 2b a linear, monotonically increasing course of the unfit function was used for the fitness criterion K1. Alternatively, however, an unfit function F1 'can also be used, which has a non-linear, monotonically increasing curve in the uncertainty region U1, for example an S-shaped curve, cf. Figure 2c . The non-linearity can be advantageous, for example, if the frequency curves run non-linearly in the overlap area of the two frequency distributions.

In Figures 3a-b an example of another fitness criterion K2 is shown, in which a low fitness measurement value M2 speaks for the presence of an unfit banknote. Correspondingly, the frequency distribution of the unfit banknotes is with relatively low fitness measurement values M2 compared to the frequency distribution of the fit banknotes. Correspondingly, an unfit function with the opposite course is also used, that is to say it drops monotonically from 1 to 0 in the uncertainty range U2. If the measured fitness value M2 is above the upper threshold value Y2, the banknote is clearly classified as fit (unfit grade 0) with regard to the fitness criterion K2 - and below the lower threshold value X2 clearly as unfit (unfit grade 1). Here, too, the unfit function assumes a non-linear course in the uncertainty area.

In the Figures 5a-c three unfit functions F2, F2, F3 for three different fitness criteria are shown by way of example, which are characterized by the uncertainty ranges U1, U2, U3 and the threshold values X1, Y1, X2, Y2, X3, Y3. Figure 5a shows the unfit function F1 for a fitness criterion relating to the damage to the bank note, the damaged area of the bank note being used here as the fitness measurement value. Figure 5b shows the unfit function for a fitness criterion F2, which relates to the soiling of the bank note, the remission intensity of the bank note in one or more ROIs being used here as the fitness measurement value. In Figure 5c shows the unfit function F3 for a fitness criterion that relates to the limpness of the banknote, the ultrasound intensity transmitted through the banknote being used here as the fitness measurement value.

The damaged area is, for example, the sum of all damaged areas of the respective bank note (damage such as holes, cracks, dog-ears, etc.), as can be seen from an image recorded by an optical sensor Bank note produced with the help of known image processing methods. The remission is measured, for example, in one or more spectral channels in one or more ROIs on the bank note, in which the contamination of the respective bank note is checked. The limp is determined, for example, with the help of an ultrasonic transmission measurement.

In Figures 5a-c the fitness measurement values M for these three fitness criteria for three banknotes A, B and C are also given by way of example, the black circle being used as the symbol for banknote A, the white circle for banknote B and the cross for banknote C. For each individual bank note A, B, C from the respective unfit function F1, F2, F3, a respective unfit degree G1, G2, G3 results from the respective measured fitness value M.

In the table Figure 6a the respective unfit grades G1, G2 and G3 are entered for these three banknotes A, B and C. With regard to damage, the bank note A is assigned an unfit grade G1 of 0.80 due to its damaged surface, the bank note B a unfit grade G1 of 0.40 and the bank note C a unfit grade G1 of 0. With regard to the contamination, the bank note A Due to its remission, an unfit grade G2 of 0 is assigned to the banknote B a unfit grade G2 of 0.75 and the banknote C a unfit grade G2 of 1. With regard to the limpness, the banknote A has an unfit grade G3 of 0 due to its ultrasonic measured value, 7 assigned, the bank note B an unfit grade G3 of 0 and the bank note C a unfit grade G3 of 0. In the example of FIG Fig. 5 However, several fitness measured values can also flow into the unfit grades G1, G2 and G3, e.g. several ROIs can be defined on the banknote for the contamination check, whose fitness measured values are then combined into a single fitness measured value, e.g. by adding up, possibly with different weighting, or multiply, if necessary with exponent k ≠ 1.

The respective unfit degrees G1, G2, G3 are now combined to form an unfit probability P for each individual bank note. For this purpose, the degrees of unfit can be multiplied with one another using the following formula, for example, in which the exponents k ₁ = k ₂ = k ₃ = 1 were set: $$\begin{array}{l}\mathrm{P.}=1-{\displaystyle \prod _j{\left(1-G_j\right)}^{k_j}=1-{\left(1-G1\right)}^{{\mathrm{<figure-callout\; id="1"\; label="k"\; filenames="imgf0001.png"\; state="\{\{state\}\}">k</figure-callout>}}_1}\cdot {\left(1-G2\right)}^{k_2}\cdot {\left(1-G3\right)}^{k_3}=}\\ =1-\left(1-\mathrm{<figure-callout\; id="1"\; label="G"\; filenames="imgf0001.png"\; state="\{\{state\}\}">G</figure-callout>}1\right)\cdot \left(1-G2\right)\cdot \left(1-G3\right)\end{array}$$

This multiplication ensures that a bank note which has an unfit grade of 1 in at least one fitness criterion will receive an overall unfit probability of 100%, regardless of the unfit grade that this bank note has in the other fitness criteria. For example, the degree of soiling G2 = 1 for the bank note C leads to an unfit probability of the bank note C of P = 100%, regardless of how low the degree of unfit for the limpness and damage may be.

In Figure 6b the unfit probabilities P calculated in this way for the three banknotes A, B and C are shown, and a fitness threshold T that can be used for their fitness classification, for example T = 90%. Since the unfit probabilities P of the bank note B are below the fitness threshold T = 90%, the bank note B is classified as fit. Since the unfit probabilities P of the bank notes A and C are above the fitness threshold T, the bank notes A and C are classified as unfit and can be sorted out by the bank note processing machine. In addition, a further fitness threshold T 'can be used for the fitness class ATM-fit, which is below the fitness threshold T, ie for the classification as ATM-fit the banknotes require an even lower unfit probability. For example, the unfit probability P is compared with the further fitness threshold T '.

During the fitness check of the bank note stack 10 to be checked, an unfit probability P is determined for each of these documents of value. This unfit probability P is compared with a fitness threshold T that applies to the overall condition of the value documents. This fitness threshold T can be specified by the user or in advance of the value document check, e.g. during the adaptation or also by remote access from a central point. With a fixed fitness threshold, the number of banknotes whose unfit probability P exceeds this fitness threshold T then results in a corresponding unfit proportion, e.g. 20%.

However, it can also be provided that the user uses the user interface 45 to enter a desired unfit portion for the bank notes to be classified as unfit, e.g. B. in percent, indicates. If, for example, not 20%, but only 10% of the banknotes of the banknote stack 10 are to be classified as unfit, then the fitness threshold T is changed in such a way that only 10% of the banknotes exceed the fitness threshold. In order to achieve this, based on the fitness threshold T ₂₀ , which led to a 20% unfit share, the evaluation device would then set the fitness threshold T correspondingly higher, taking into account the frequency of the unfit probabilities in this banknote stack (T ₁₀ ). If necessary, the bank notes of the bank note _{stack 10 can} then be checked again - with the fitness threshold T 10 - and sorted according to their fitness.

In order to define the unfit functions, the following procedure can be used: In the run-up to the fitness check of a stack of banknotes to be checked, the user selects a first group of banknotes that he classifies as fit, ie these banknotes have e.g. B. at most a slight soiling and / or damage that is not perceived as annoying, and a second group of banknotes that he classifies as unfit, ie these banknotes have abnormalities such as soiling, damage, Clips, tape, etc. Using the user interface 45, the user selects a configuration operating mode of the banknote processing machine 1 in which parameters for the fitness test can be set, in particular which fitness criteria are to be used for the fitness test and / or in which the unfit functions and their threshold values are defined or changed can be.

In the configuration operating mode, the user is requested, for example, to first insert the banknotes which he has classified as unfit into the input compartment 20. The bank notes classified as unfit are individually detected by the singler 25 and transferred to the transport system 30. The measuring device 41, or the sensor or sensors contained therein, determines measurement data representative of the respective bank note, which are transmitted to the evaluation device 40. After all banknotes classified as unfit have been processed, the user is requested to insert the banknotes classified as fit into the input compartment 20, which are then processed analogously to the fit banknotes. Alternatively, the unfit and fit banknotes can also be placed together in the input compartment 20 in the operating mode for determining the threshold value (s) if they can be clearly separated from one another by the banknote processing machine 1, for example by means of a separation card placed between the unfit - and the fit banknotes is inserted. During processing, the separation card is recognized by the control device 40 on the basis of the measurement data from the measuring device 41, so that the control device 40 can perform the separation between unfit and fit banknotes.

The parameters for the fitness test are then set on the basis of the frequency distributions of the fitness measurement values of the fit and unfit banknotes. This can be done manually by the user (operator, adapter, service person), but also automatically by the evaluation device the value document processing device. For example, the first threshold value is placed on a fitness measurement value in which the fit frequency is much greater than the unfit frequency (e.g. have at least a certain ratio, e.g. 5: 1 or 10: 1) and the second threshold value is set on a fitness measurement value, in which the fit frequency is much smaller than the unfit frequency frequency, e.g. have at least a certain ratio (e.g. 1: 5 or 1:10). The uncertainty area then lies accordingly in the overlap area of the two frequency distributions.

In order to reduce the number of fitness criteria that have to be adapted by a user, several fitness criteria can be combined, for example several fitness criteria that relate to the damage to the bank note. For example, the damaged area can be used as the fitness criterion K3 and the length of the crack of the respective bank note can be used as the fitness criterion K4. The fitness measurement values M3 and M4 of the two fitness criteria are combined, for example, by a linear combination to form the fitness measurement value MK = a * M3 + b * M4, in which the fitness measurement values M3 and M4 can be included with different weightings a, b. The result of the linear combination provides the combined fitness measurement value MK. In Figure 7a the distributions of the two fitness measured values M3 and M4 can be seen for a group of unfit banknotes, each represented by a black circle, and for a group of fit banknotes, each represented by a white circle. In addition, a two-dimensional area “clear fit” is drawn in, in which the group unfit degree is 0, and a two-dimensional area “clear unfit”, in which the group unfit degree is 1. The two threshold values X and Y are formed in the two-dimensional case by the two straight lines a * M3 + b * M4 = X and a * M3 + b * M4 = Y. A bank note in which a * M3 + b * M4 <X (ie MK <X) is classified as clearly fit in relation to the combined fitness measurement value MK (group unfit degree = 0), a bank note in which a * M1 + b * M2> Y (i.e. MK> Y), is classified as clearly unfit with regard to the combined fitness measurement value MK (group unfit degree = 1), a bank note in which X <a * M1 + b * M2 <Y applies (ie X <MK <Y), lies in the uncertainty range U, in which it has a group unfit degree between 0 and 1 in relation to the combined fitness measurement value.

In Figure 7b it is shown how the group unfit degree G can be determined for the combined fitness measurement value MK, which was combined from the group of fitness measurement values M3 and M4. For this purpose, an unfit function F according to the invention with two threshold values X, Y and an uncertainty range U in between is set up for the combined fitness measurement value MK. With the help of the unfit function Figure 7b the result is the group unfit degree G. The unfit probability P of the respective banknote is then obtained by combining the group unfit degree G, which relates, for example, to the damage, with one or more other unfit degrees of individual fitness criteria and / or with one or more other groups - Unfit degrees, for example with a group unfit degree, which concerns the soiling of the bank note. The combination of all unfit degrees takes place, for example, by multiplying these unfit degrees according to formula (1) or by linear combination.

Since the number of fitness measured values is reduced by combining the fitness measured values into a single fitness measured value, the complexity (dimensionality) of the fitness test is reduced. This simplification of the fitness test is easier to understand and clearer for the user of the banknote machine. This makes it easier for a user to make manual adjustments to the rigor of the fitness test.

Claims (15)

1. A method for checking the fitness of value documents (A, B, C), comprising the following steps:

   - selecting at least two different fitness criteria (K1, K2) of the value documents which are characteristic for the state of the value documents,

   - checking the value documents by picking up measurement data of the value documents, wherein, from the measurement data picked up for each of the selected fitness criteria, a fitness measurement value (M1, M2) for the particular value document is determined,

   - determining respectively an unfit degree (G1, G2) for each of the selected fitness criteria (K1, K2) from the particular fitness measurement value of the particular value document (A, B, C) with the aid of an unfit function (F1, F2) which unambiguously assigns an unfit degree (G1, G2) to each fitness measurement value (M1, M2), wherein each unfit function is characterized by a first threshold value (X1, X2), a second threshold value (Y1, Y2), and an uncertainty range (U1, U2) being between the first and the second threshold value in which the particular unfit function either has a monotonously dropping or a monotonously rising course, and

   - combining the unfit degrees (G1, G2) of the different fitness criteria (K1, K2) into an unfit probability (P) which is specific for the particular value document (A, B, C), and

   - fitness classification of the particular value document on the basis of the unfit probability (P) which was determined for the particular value document.
2. The method according to claim 1, characterized in that the unfit function (F1, F2) assigns to the fitness measurement values (M1, M2) being in the uncertainty range respectively one unfit degree (G1, G2) which is greater than 0 and lower than 1 and between the first and second threshold value either assumes a monotonously dropping or monotonously rising, e.g. linear or nonlinear, course, wherein the unfit function assigns in particular an unfit degree of 0 to all fitness measurement values being beyond the first threshold value and assigns an unfit degree of 1 to all fitness measurement values being beyond the second threshold value.
3. The method according to any of the preceding claims, characterized in that the selected fitness criteria (K1, K2) relate to one or several of the following properties of the value documents: soiling, wear, damage, alien elements or limpness of the particular value document, wherein the selected fitness criteria preferably relate to at least two different ones of these properties.
4. The method according to any of the preceding claims, characterized in that at least two such fitness criteria are selected in which the frequency distribution of the fit value documents and the frequency distribution of the unfit value documents overlap each other as little as possible, wherein the two frequency distributions preferably have a maximum overlap of 30%.
5. The method according to any of the preceding claims, characterized in that upon combining the unfit degrees (G1, G2, ...) of the different fitness criteria (K1, K2) into the unfit probability (P) there is carried out a multiplication of the unfit degrees of the different fitness criteria, in particular that the unfit probability (P) is determined from the unfit degrees (G2, G2) according to the following formula: $$P=1-{\displaystyle \prod _j{\left(1-G_j\right)}^{k_j}=1-{\left(1-G1\right)}^{k_1}\cdot {\left(1-G2\right)}^{k_2}\cdot \mathrm{...}}$$

   
6. The method according to any of the preceding claims, characterized in that upon combining the unfit degrees (G1, G2, ...) of the different fitness criteria into the unfit probability (P) there is formed a linear combination of the unfit degrees of the different fitness criteria, in particular by adding up the unfit degrees (G1, G2, ...) of the different fitness criteria, where applicable with different weighting of the unfit degrees.
7. The method according to any of the preceding claims, characterized in that, for the fitness classification of the particular value document, the unfit probability (P) determined for the value document is compared with a fitness threshold (T) and the value document is classified as unfit, if the unfit probability (P) exceeds the fitness thresholds (T).
8. The method according to any of the preceding claims, characterized in that upon the fitness classification of the value documents of a value document group to be checked for fitness there is carried out an advance calculation, in which, for different values of the fitness thresholds (T), respectively the expected unfit portion of the particular value document group is determined, which indicates the portion of value documents which are classified as unfit in the fitness classification of the particular value document group, and that information is generated about how the unfit portion depends on the value of the fitness threshold (T), wherein this information in particular is communicated to the user of a value-document processing apparatus carrying out the method for checking the fitness, e.g. by outputting it at a user interface of the value-document processing apparatus.
9. The method according to any of the preceding claims, characterized in that prior to the fitness check the following steps are carried out:

   - providing a first group of fit value documents and a second group of unfit value documents, wherein the categorization of the value documents as fit or unfit was carried out in particular by a manual check by a person or by checking the value documents by means of a reference measuring system,

   - checking the fit and the unfit value documents of the first and second group by picking up measurement data of these value documents with the aid of a measuring device,

   - determining at least one fitness measurement value (M1, M2) for each of the value documents from the measurement data of the particular value document,

   - determining a first frequency distribution of the particular fitness measurement value for the first group of the fit value documents and a second frequency distribution of the particular fitness measurement value for the second group of the unfit value documents,

   - employing the first and second frequency distribution of the particular fitness measurement value (M1, M2) to select the fitness criteria (K1, K2) to be employed in the fitness check of the value documents and/ or to determine the unfit function (U1, U2) of the particular fitness criterion (K1, K2).
10. The method according to any of the preceding claims, characterized in that for the value documents of at least one value document group to be checked for fitness the following steps are carried out after the fitness classification of the value documents of the value document group:

    - ascertaining the unfit portion of the value document group, which indicates the portion of value documents which are classified as unfit in the fitness classification of the value document group,

    - checking the unfit portion for at least one specification determined for the unfit portion,

    - changing the unfit function (U1, U2) of one or several of the employed fitness criteria (K1, K2) in dependence on the ascertained unfit portion of the value document group, wherein, if the ascertained unfit portion fulfills the determined specification, the unfit function is left unchanged, and if the unfit portion does not fulfil the determined specification, the unfit function is changed and the following steps a)-f) are carried out, within the framework of a simulation, using the changed unfit function:

    a) determining anew the unfit degrees (G1, G2) of the particular value document for the at least two different fitness criteria (K1, K2) from the particular fitness measurement value using the changed unfit function of the particular fitness criterion,

    b) combining anew the unfit degrees of the different fitness criteria into an unfit probability (P) of the particular value document and

    c) classifying anew the fitness of the particular value document with the help of the particular unfit probability (P),

    d) ascertaining anew the unfit portion of the one or several value document groups, which indicates the portion of value documents which are classified as unfit in the fitness classification of the particular value document group,

    e) checking anew the unfit portion for the determined specification,

    f) changing anew the unfit function of one or several of the employed fitness criteria in dependence on the ascertained unfit portion of the value document group, wherein, if the ascertained unfit portion fulfills the determined specification, the unfit function is left unchanged, and if the ascertained unfit portion does not fulfil the determined specification, the unfit function is changed and the steps a)-f) are repeated within the framework of the simulation.
11. The method according to claim 10, characterized in that as soon as the unfit portion fulfills the determined specification, the unfit function is left unchanged and the fitness classification (step c) last carried out is employed as the final fitness classification and/or that then the unfit function last employed (for the fitness classification in step c) is employed for the future fitness classification of further value document groups, in particular for further value document groups of the same value document type.
12. The method according to any of claims 10 to 11, characterized in that upon changing the unfit function (U1, U2) of the particular fitness criterion (K1, K2), the unfit function of this fitness criterion is changed in dependence on the unfit portion of the value document group such that the unfit portion upon the new fitness classification is changed in comparison to the unfit portion ascertained before, for example is increased or is decreased, wherein the steps a)-f) according to claim 10 in particular are repeated so often, until the newly ascertained unfit portion corresponds at least approximately to that unfit portion which was detected before in a manual fitness check or in an automatic fitness check with the aid of a value-document processing apparatus for this value document group.
13. The method according to any of the preceding claims, characterized in that at least one of the fitness measurement values (M1, M2) is an aggregated fitness measurement value (MK), in which there are aggregated at least two different fitness measurement values (M3, M4), e.g. by linear combination of these fitness measurement values (M1, M2), and that at least one of the unfit degrees (G1, G2) is a group unfit degree (G) which indicates the fitness of the particular value document with respect to at least two different fitness criteria, wherein the group unfit degree (G) is determined with the aid of an unfit function (F) which was formulated for the aggregated fitness measurement value (MK).
14. The method according to claim 13, characterized in that for the value documents there is respectively determined a first group unfit degree (G) for a first group of at least two fitness criteria which respectively relate to the soiling of the particular value document, and that there is determined a second group unfit degree for a second group of at least two fitness criteria which respectively relate to the damage of the particular value document, wherein the unfit probability (P) of the particular value document in particular is determined by combining the first group unfit degree (G) relating to the damage with the second group unfit degree relating to the soiling of the bank note, and where applicable with one or several further unfit degrees and/ or group unfit degrees.
15. An apparatus for checking the fitness of value documents (A, B, C) by the method according to one of the preceding claims, comprising:

    - a measuring device (41) for picking up measurement data of the value documents, and

    - an evaluation device (40) for the fitness classification of the value documents on the basis of the measurement data picked up,

    characterized in that the evaluation device is configured for

    - selecting at least two different fitness criteria (K1, K2) of the value documents, which are characteristic for the state of the value documents,

    - determining, from the measurement data picked up for each of the selected fitness criteria, a fitness measurement value (M1, M2) for the particular value document,

    - determining, for each of the selected fitness criteria, respectively one unfit degree (G1, G2) from the particular fitness measurement value of the particular value document with the aid of an unfit function (F1, F2), wherein, the unfit function (F1, F2) unambiguously assigns an unfit degree (G1, G2) to each fitness measurement value (M1, M2), and wherein each unfit function is characterized by a first threshold value (X1, X2), a second threshold value (Y1, Y2), and an uncertainty range (U1, U2) being between the first and the second threshold value in which the particular unfit function either has a monotonously dropping or a monotonously rising course, and

    - combining the unfit degrees (G1, G2) of the different fitness criteria (K1, K2) of the particular value document into an unfit probability (P) which is specific for the particular value document (A, B, C), and

    - carrying out a fitness classification of the particular value document (A, B, C) on the basis of the unfit probability (P) which was determined for the particular value document.

Images