COMPLETENESS CHECK OF A VALUE DOCUMENT

Abstract

The present invention relates to a method, a sensor, a sensor unit and a bank-note processing machine for checking the completeness and/or authenticity of value documents. A value document comprises at least one machine-readable feature substance in at the least two locations. According to the method, the value document is excited at least locally at measuring locations. Furthermore, a feature intensity with respect to the machine-readable feature substance is captured location-resolved at several different locations of the value document. The location-based feature intensities are classified location-based with the help of a threshold value. Furthermore, location-based limits of a location distribution to be expected of the machine-readable feature substance are determined. Finally, a location-based distribution of the classified feature intensities is assessed.

Claims

1.-20. (canceled)

21. A method for checking the completeness and/or authenticity of value documents, wherein at the least one value document comprises at least one machine-readable feature substance at two locations at the least, having the steps: at least locally exciting of the value document; location-resolved capturing of a feature intensity with respect to the machine-readable feature intensity at several different locations of the value document; location-based classification of the location-based feature intensities with the help of a threshold value; determining location-based limits of a location distribution to be expected of the machine-readable feature substance; and assessing a location-based distribution of the classified feature intensities.

22. The method according to claim 21, wherein the classification of the location-based feature intensities is effected with the help of location-dependent threshold values.

23. The method according to claim 21, wherein the step of the location-resolved capturing of remission values at several different locations of the value document.

24. The method according to claim 23, wherein the threshold value is configured as the location-dependent threshold value which is determined from a characteristic curve dependent on the remission value determined at the respective location.

25. The method according to claim 23, wherein the measuring locations of the remission values overlap with the measuring locations of the feature intensities and preferably are identical.

26. The method according to claim 21, wherein the step of computing a track completeness by a comparison of the number of the measuring locations having above-threshold feature intensity with the number of captured measuring locations within a convex envelope of the measuring locations having above-threshold feature intensity or, where applicable, within a convex envelope of the measuring locations having above-threshold remission value.

27. The method according to claim 21, wherein the step of checking a two-dimensional distribution of the classified feature intensities relative to a convex envelope of the measuring locations with above-threshold feature intensity or, provided that remission values were captured, relative to a two-dimensional distribution of the measurement values of the remission measurement, wherein preferably the checking of the two-dimensional distribution of the classified measuring locations comprises a computation of a column completeness.

28. The method according to claim 21, wherein the feature intensities are captured along at least one measuring track on the value document.

29. The method according to claim 21, wherein the location-resolved capturing of the feature intensities with respect to the machine-readable feature substance comprises the measurement of a spectral luminescence intensity of a luminescent substance and/or the spectral measurement of a Raman band of a Raman-active substance and/or a substance detectable by surface-enhanced Raman spectroscopy and/or the spectral measurement of an absorption band of a substance absorbing in the infrared spectral region and/or the measurement of the magnetic properties of a ferromagnetic substance.

30. The method according to claim 21, wherein a local authenticity of the value document is checked with the help of at least one feature value.

31. The method according to claim 21, wherein a number and a spatial distribution by measuring locations classified as below-threshold is compared with reference values.

32. The method according to claim 21, wherein the value document is moved during the measurement at a speed of 1-13 m/s.

33. The method according to claim 21, wherein the location-resolved capturing of the feature intensities of the machine-readable feature substance and/or, where applicable, the remission values on front side and back side of the value document, is effected in particular at same, opposing locations of the front side and back side.

34. The method according to claim 33, wherein the location-dependent threshold values are determined by a characteristic curve which is dependent on the feature intensity determined at the side opposing at the respective location of the value document.

35. The method according to claim 21, wherein location-resolved transmission values of the value document are captured.

36. The method according to claim 35, wherein the transmission measurement is effected through a time-shifted illumination within the framework of remission measurements on front side and back side and/or through a time-shifted illumination within the framework of the measurement of feature values on front side and back side.

37. The method according to claim 21, wherein at several measuring locations respectively a combined classification is performed with consideration of data tuple associated with the measuring locations, wherein the data tuple comprise at least one feature intensity as well as at the least one of the following components: a further feature intensity, a remission value, and/or a transmission value.

38. A sensor for capturing a feature intensity and/or a feature value, configured for carrying out a method according to claim 21.

39. A sensor unit having a sensor, wherein the sensor is configured for capturing at the least one feature intensity, a feature value, a remission value and/or a transmission value in particular according to claim 38 and/or wherein the sensor unit is configured to execute a method.

Description

[0081] Further features and advantages of the invention will result from the present description of embodiment examples of the invention as well as further alternative embodiments in connection with the following drawings, which show:

[0082] FIG. 1: A schematic representation of an embodiment of a method according to the invention;

[0083] FIG. 2a: A first diagram according to one embodiment for classifying at pixel level;

[0084] FIG. 2b: A further diagram according to one embodiment for classifying at pixel level;

[0085] FIG. 3: A schematic representation of a characteristic curve for threshold values for the classification at pixel level;

[0086] FIG. 4: A schematic representation of the time course of the illumination for remission or;

[0087] FIG. 5a: A schematic representation for classification at pixel level for both-sided feature measurement;

[0088] FIG. 5b: A schematic representation of a further characteristic curve for classification at pixel level with both-sided feature measurement;

[0089] FIG. 6: A curvature of feature intensity, remission value as well as a dynamically established threshold value for the classification at pixel level;

[0090] FIG. 7: A schematic representation of remission values of a bank note to be checked;

[0091] FIG. 8: A schematic representation of feature intensities of a bank note to be checked;

[0092] FIG. 9: A schematic representation of feature intensities of an incomplete bank note to be checked;

[0093] FIG. 10a: A schematic representation of location-based distribution of classified feature intensities;

[0094] FIG. 10b: A schematic representation of location-based distribution of classified feature intensities of an incomplete bank note;

[0095] FIG. 11: A representation of transmission values of a bank note;

[0096] FIG. 12: A further schematic representation of a pixel-wise classification; and

[0097] FIG. 13: A schematic representation of a combined classification of feature values.

[0098] In FIG. 1 a process flow for checking of a value document according to the invention is represented schematically.

[0099] In a first Step S1 a value document is supplied. The value document comprises at the least one machine-readable feature substance. The feature substance is arranged at two different locations at the least, preferably over a substantial region of the value document. Preferably the machine-readable feature substance extends partially in the total areal extent of the value document.

[0100] In a Step S2, the value document is excited at least locally preferably with electromagnetic radiation. The exciting can be effected by means of irradiating the entire value document. Preferably a regional, particularly preferably a pointwise, irradiating of the value document takes place. By means of a sensor unit, a feature value is captured location-resolved, in particular a feature intensity with respect to the machine-readable feature substance is captured (S3a) at several different locations of the value document. The capturing relates to, as a rule, the areal section of the value document which was excited by means of electromagnetic radiation, wherein preferably the excited section has an area equal to or greater than the captured region or point.

[0101] Preferably, substantially simultaneously to Step 3a, a remission value is captured in a location-resolved manner with respect to the feature values captured in Step 3a (S3b), wherein also several remission values can be captured which, for example, relate to different wavelengths.

[0102] In a Step S4, the feature values and the preferably captured remission value are evaluated in a location-resolved manner according to Steps S2, S3a and, where applicable, S3b. In the process, the feature values are compared with expected reference signals and respectively one feature intensity each is established for the feature values captured in a location-resolved manner. Preferably a normalization of the location-based feature intensities takes place.

[0103] Starting out from the evaluation from Step S4, a classification of the location-based feature intensities takes place in Step S5. The classification is effected based on a lower threshold value of the feature intensities (see FIG. 2a) or a combined employment of a lower and an upper threshold value of the feature intensities (see FIG. 2b) or an employment of different threshold values of the feature intensities, in particular in dependence on one or different remission values (FIG. 3).

[0104] The evaluation of a feature value and the classification of a feature intensity can be carried out temporally independent from the capturing of further feature values. Therefore, for a feature intensity preferably Step S4 can be effected immediately after Step S3a and/or Step S4 be effected for one or several feature intensities after the capturing of the several feature intensities according to S3a. Analogously, for a feature intensity preferably Step S5 can be effected immediately after Step S4 and/or Step S5 be effected for one or several feature intensities after the evaluating of the several feature intensities according to S4.

[0105] In Step S6, a location-based distribution of the feature intensities is determined starting out from the evaluation from Step S4 or alternatively starting out from the classification of the feature intensities from Step S5. Expected location-based limits of the distribution of the feature substance are derived from the location-based distribution. These location-based limits are established either from the distribution of the classified location-based feature intensities, for example by computation of the convex envelope of the above-threshold feature intensities, or are established by including further measurement values, in particular the remission values.

[0106] Thereupon, the location-based distribution of the classified feature intensities obtained in Step S5 is assessed in Step S7. The assessment is effected in particular with regard to the relative position of the pixels classified as above- or below-threshold to each other as well as with regard to the relative position of the pixels classified as below-threshold relative to the limits determined in S6 of the location distribution to be expected of the machine-readable feature substance.

[0107] On the basis of the assessment from Step S7, a completeness measure is finally established for the entire value document which can be drawn upon for the authenticity assessment or e.g. for sorting decisions in a bank-note processing machine.

[0108] In the diagrams cited now, the colors are employed, yellow with reference sign ge, green with reference sign g, black with reference sign s, red with reference sign r and blue with reference sign b. All specified color particulars are to be understood only by way of example and serve only for illustrative purposes. Of course, values or other designations can be employed instead of the color particulars.

[0109] In FIGS. 2a and 2b, an intensity field is represented respectively for a scanned pixel, wherein the threshold values for feature signals or remission signals employed for classifying at pixel level are entered according to an aspect of the present invention by way of example.

[0110] The classification of the pixels as authentic/false is effected by way of example with reference to FIG. 1 as follows. For assessing a value document as to authenticity and/or completeness, a classification is performed on pixel basis.

[0111] All measurement points or pixels which have remission values above a certain threshold R.sub.1 in the remission channel also have to deliver a sufficient feature intensity in the feature signal to be recognized as an authentic portion of the value document. In this way the feature intensity has to be higher than a lower threshold of the feature intensity M.sub.min. This classifying of all pixel while employing fixed threshold values can be clearly represented with the help of the 4-field board according to FIG. 2a.

[0112] FIG. 2b shows threshold values for feature intensity values or remission values for classifying at pixel level in a modified 4-quadrant diagram using a lower threshold M.sub.min, R.sub.1 and an upper threshold M.sub.max. In the process, all pixels which are sufficiently light (i.e. remission value R>remission threshold value R.sub.1) are assessed as green and deliver an sufficiently intense feature signal (feature intensity M>minimal feature intensity M.sub.min (lower threshold value of the feature intensity)). Pixels too dark (R<R.sub.1), as they can occur e.g. due to holes in the value document, are classified as black, whereas present regions of the value document (R>R.sub.1), i.e. a sufficiently high remission value is captured, and without sufficient feature signal is classified as suspect of forgery, in particular as a snippet forgery, as red. If regions are present having insufficient remission but sufficient feature intensity, these are classified yellow as a feature excess. This can occur e.g. with heavy soiling (with special spectral behavior of the illuminated areas) or in window regions with an invisible feature.

[0113] Furthermore, according to FIG. 2b an upper threshold for the expected feature intensity M.sub.max is employed. Here, all regions with an excess of feature signal can then be classified yellow. The combined evaluation of remission and feature intensity at pixel level allows in any case a simple consideration of otherwise problematic situations, such as a high running (i.e. y offset) or skew running of a value document in the processing machine as a result of a transport malfunction.

[0114] In a further-reaching embodiment, the remission signal is employed at pixel level to normalize the feature signal (only in the linear region) for the purpose of a correction of soiling or overprinting. In doing so, boundary effects are likewise taken into consideration if the value-document edge overlaps only partly with the measurement pixels and hence reduced feature and remission intensities are recognized.

[0115] Alternatively, the threshold required for the authenticity detection for the feature intensity can advantageously be dynamically adapted pixel-wise with the help of the measured remission signal. Here, a characteristic curve or a characteristic diagram for the authenticity detection is defined as is shown in FIG. 3.

[0116] FIG. 3 shows a characteristic curve for the threshold values for the classification at pixel level. The presence of a document for remission values R above a remission threshold R.sub.1 is recognized. This threshold can be fixed for all tracks uniformly, or preferably, parametrized individually for each track with the help of reference measurement values for white or black samples.

[0117] If a very dark region is registered on the document, a reduced threshold value is also applied for the feature intensity M (M.sub.1>M). If correspondingly lighter regions (R.sub.1<R<R.sub.2) are present, thus preferably the required feature intensity threshold is increased correspondingly between M.sub.1 and M.sub.4. At especially highly reflective places (R>R.sub.2) it can be assumed that here no normal paper-of-value substrate is present but rather a metallic reflector such as a hologram, security strip or the like. Because these are typically opaque for optical radiation, the threshold value for the feature signal is reduced accordingly down to M.sub.3, because the covered areas can in some cases deliver only a highly reduced signal contribution. If the spatial resolution of the feature sensor is not distinctly higher than the dimensions of the opaque structures, a masking will not be effected digitally but rather usually occur partially. This is taken into account by a gradual reduction of the feature threshold between M.sub.4 and M.sub.3 in the range R.sub.2<R<R.sub.3. For the purposes of a strong recognition of forgeries, a minimum of feature signal M.sub.2 can also be required with very high remission values R>R.sub.3. For a particularly strict assessment, M.sub.2=M.sub.3 can also be chosen. In these classifying variants, a hologram strip is marked in red. Alternatively, M.sub.2 can also be parametrized to very low values, which results in a classifying of reflective hologram strips as green.

[0118] At the boundary of the value document, red pixels can randomly occur due to an only partial overlap between value document and measurement pixel, which have to be treated or tolerated separately in the further assessment. Alternatively, the origination of these red boundary pixel can be prevented by a suitable parametrization of the threshold characteristic curve for R.sub.1 or M.sub.1. In so doing, M.sub.1 is set lower (relative to the maximum intensity) than R.sub.1, so that by the purely geometrical loss of intensity, which relates equally to remission as well as feature intensity, the situation cannot occur that indeed still R>R.sub.1, but already M<M.sub.1.

[0119] Upon the presence of several independently measurable feature substances, the feature measurement values can, of course, analogously as described in the case without remission measurement, be assessed individually as well as in a combined manner.

[0120] Pixel Completeness:

[0121] The first check for completeness is now performed on pixel basis: Within the recognized region of the value document, the number of measurement points or pixels classified as red may not exceed a certain threshold. In the strictest interpretation having the threshold 0 this means that not one single measuring location having insufficient feature intensity is allowed to be present, in order that the value document is recognized as complete. In other variants, individual red pixels can be tolerated.

[0122] Here, the ratio of the number of all green pixel relative to the number of all pixel within the extent of the value document can again be formed and checked against a minimum threshold. This corresponds to an area proportion or the area-based degree of completeness.

[0123] Track Completeness:

[0124] The track lengths determined from the remission measurements are employed respectively as a scale for assessing the track completeness. For computing the index for the track completeness, the number of the pixels is classified as green in this track is divided by the number of all pixel within this track length. Alternatively, one obtains a slightly stricter check criterion if for computing the index for the track completeness the number of the pixels classified as green in this track is divided by the number of pixels corresponding to the maximum length of the value document.

[0125] A further check criterion is the number of neighboring red pixels within the length of the value document and within a track. If this exceeds the defined threshold, the track is counted as incomplete. For the parametrization of this threshold, the maximum width of red regions occurring in authentic value documents, such as e.g. the maximum extent of hologram patches or similar, is expediently taken into consideration.

[0126] Analogously to the above-described procedure upon the determination of the track completeness without remission measurement, measuring tracks in boundary location can be assessed differently than mid tracks here too, although the corresponding position uncertainties are much lower here because of the remission measurement.

[0127] Two-Dimensional Completeness:

[0128] In the preferred case that the sensor has several measuring tracks, here too, as already described above, the two-dimensional distribution of the feature intensity or the two-dimensional distribution of the classified pixels is evaluated.

[0129] By means of the convex envelope around the pixels having above-threshold remission, holes or opaque spots can be localized within the value document. In the process, the occurrence of larger holes is checked in a targeted manner. For this purpose, red below-threshold neighboring pixels within the extent determined by the convex envelope of the value document are searched for and two-dimensionally contiguous regions are counted and identified/marked. If e.g. more than 2, 3, 5, . . . (resolution-dependent) contiguous red pixels are present, thus a potentially missing region is recognized. Thereupon the position and geometrical extent of the red regions are analyzed and matched to patterns occurring in the known manner such as e.g. a transparent window or a metallic hologram strip. In particular the form, maximum width and relative position to the edges or corners of the value document is checked as to plausibility and upon deviations is classified as incomplete.

[0130] Here too, an assessment can be performed for the efficient recognition of forgeries or incomplete value documents having vertical tampering structures with regard to the column completeness. Here, the number of the red pixels is established column by column and compared with a threshold value. If this threshold is now exceeded (by e.g. 2 or 3) in a column, the value document is rejected as incomplete.

[0131] For those forgery classes in which in the boundary-region sections of the authentic value document were replaced by e.g. a photocopy, an real qualitative advantage arises from the combined evaluation of the remission and the feature intensity: These forgeries can now be reliably recognized by the exact determination of the actual extent of the value document. In the process, in particular a targeted check can be performed for the presence of boundary columns classified as red (which were established by counting the red pixels in column direction). In the process, preferably the outermost two columns are assessed in order not to overrate or falsely assess the red boundary pixels randomly occurring from edge effects.

[0132] In one embodiment having highly different resolving power of the measurement in (x) track direction and y direction (track number), this is taken into consideration by the fact that red neighboring pixels are counted in line direction in a targeted manner and multiple pixel in this direction are assessed as particularly severely. In particular, the maximally occurring width of a hologram strip (or similar security features such as metal color) can be taken into consideration by the fact that value documents with a greater number of red pixels in the higher resolved measurement direction than a defined threshold value are directly classified as incomplete.

[0133] Both-Sided Measurement

[0134] In particularly preferred variants, the authenticity sensor comprises two partial sensors which allow a both-sided measurement of the feature intensity on each value document. In the process, preferably a remission channel is also available on at least one sideor particularly preferably on both sideswith which (track) length as well as exact position and alignment of the value document are determined.

[0135] In one embodiment, the two partial sensors are controlled centrally to synchronize the time courses of the excitation or measured value acquisition for both partial sensors. Alternatively, two individual independent sensors are used for front side or back side which are synchronized in a master/slave configuration by one of the two sensors (master). For example, this master sensor sets the operating mode and pre-specifies time delays to be adhered to for the measurement pulses and/or measured value acquisition after a trigger signal.

[0136] Furthermore, preferably different sensor architectures can be used for the master or slave sensor. Thus, for example, one of the sensors can be equipped with a more elaborate measuring technology than the other sensor and check the feature values with a higher precision or a higher spectral resolution.

[0137] The two partial measurements of front side and back side are thereupon evaluated combined. In the process, the measuring data are associated with the respective measuring locations on the value document, the location-based data tuples of (Remission, Feature1, Feature2) or (Remission1, Remission2, Feature1, Feature2) are formed and evaluated.

[0138] Preferably, the position or clocking of the two measurements (front, back) are coordinated with each other such that the value document is measured at the same pixel positions on front side and back side. Particularly preferably, the measurement takes place respectively (almost) simultaneously, i.e. a measurement point is captured at a place of the value document from the front and from the back side at almost the same time.

[0139] Beside the simpler and more unambiguous assessment of the thus obtained measurement values, this offers the advantage that a usually unpreventable crosstalk between front-side measurement and back-side measurement does not lead to artifacts and spurious signals, but rather reinforces the feature signal to be measured.

[0140] In the process, the illumination of the first partial sensor can be utilized advantageously also for a transmission measurement using the detector part of the second partial sensor if the two illumination light pulses have a small time offset, so that the transmission signal can be recorded temporally separate from the remission signal 2. This time sequence of the light pulses or detections is represented schematically in FIG. 3. In this case, Transmission, Remission1, Remission2 as well as Feature1, Feature2 are available for each measuring location as a base of data for the completeness evaluation. This makes the complete completeness assessment possible even for existing opaque (metallic) or transparent (window) security features which can otherwise hinder the completeness check of certain parts of the value document.

[0141] The illumination for the remission measurement (alternatively: feature measurement) of the front side and the back side are effected slightly time-shifted, so that detector 2 can determine the transmitted portion of the illumination 1 independently and undisturbed from the illumination 2 as it is shown in FIG. 3.

[0142] In the simplest case, upon the evaluation the sum (or the mean value or the maximum) from Feature1 and Feature2 is formed at each measuring location and is thereupon classified and assessed according to the above-described procedures.

[0143] A more exact assessment is reached if individual thresholds are applied for Feature1 and Feature2. These can depend on remission as well as on the respectively other feature value. A corresponding characteristic diagram then takes the place of the just described characteristic curve for the pixel-wise red/green assessment. This can be adapted/parametrized exactly to the typical optical effects occurring in authentic value documents.

[0144] FIGS. 5a and 5b show a characteristic diagram for the threshold values for classification at pixel level upon both-sided feature measurement. In FIG. 5a, a classification is effected on account of static threshold values of feature values (M.sub.1,min, M.sub.2,min). Using the characteristic diagram in FIG. 5b, a classification is effected taking into consideration interference effects such as e.g. reflection at metallic surface structures applied to one side.

[0145] If, for example, a (reflective hence opaque) metallic strip is applied to a bank note on a side B1, it is to be expected that indeed on one side the FeatureValue1 is very low, the FeatureValue2 to be expected, however, is increased due to the occurring reflections compared with the immediate environment (or compared with the mean value over the entire bank note). This can be represented by the corresponding parameterization of the threshold characteristic diagram. Conversely, upon an overprinting on page B1 with black, spectrally broadband-absorbing (carbon black) color, the remission value and FeatureValue1 are low, whereas FeatureValue2 is at normal level.

[0146] The parameterization of the classifier is advantageously depending on the location, i.e. e.g. relative to the leading edge, relative to the corners, or concrete position within the convex envelope, etc. This allows a correct treatment of absorbable and reflective disturbance in dependence on (position- and denomination-dependent) the effects possibly occurring in these regions. In both cases the corresponding region can in any case be reliably assessed as authentic due to the both-sided feature measurement in spite of the insufficient feature intensity on one side.

[0147] This allows the gapless proof of the completeness independent of the bank note design, even in difficult situations with (one-sidedly occurring) covers/shadowing by opaque elements such as aluminum-coated hologram strips. With it, regions of the value document can also be reliably checked for completeness/authenticity which cannot be assessed by only one-sided measurement.

[0148] In a preferred variant, the complete present data set of (Transmission, Remission1, Remission2, Feature Intensity1, Feature Intensity2) is classified and assessed in a combined manner. Besides regions having opaque, absorbent or reflective concealments, in particular also holes or window regions, can be reliably identified in the process by the transmission signal, and their position and extent can be checked in comparison to the values permissible for authentic value documents. Further embodiment examples are described hereinafter.

[0149] Example 1: Here, a spectrally resolving single-track luminescence sensor having remission measurement is employed for the completeness check. The sensor is operated on a bank-note processing machine at a transport speed of 11 m/s and is employed for the authenticity check as well as completeness check of bank notes having a luminescence sensor-coordinated luminescence marker incorporated in the paper. The bank notes have a reflective hologram strip on the front side in the right region.

[0150] FIG. 6 shows a feature curve (O), a remission curve (x) and the dynamically computed feature threshold (dashed) of an authentic and complete bank note. Remission intensity as well as feature intensity are significantly modulated. The completeness can nevertheless be established correctly by applying a remission-dependent threshold upon the classification of the feature intensity.

[0151] Example 2: Here, a spectrally resolving 11-track luminescence sensor having remission measurement is employed for the completeness check. The sensor is operated on a bank-note processing machine at a transport speed of 11 m/s and is employed for the authenticity check as well as completeness check of bank notes having a luminescence marker incorporated in the paper. The bank notes have a reflective hologram strip on the front side in the right region as well as a transparent window in the left region.

[0152] FIG. 7 shows a representation of the measured remission values of the bank note. High remission occurs in particular in the region of the reflective hologram strip, while very low remission is present in the transparent window.

[0153] FIG. 8 shows a representation of the feature intensity of the bank note. White corresponds to high intensity, while black corresponds to low values. In the region of the window (on the left) as well as the hologram strip (on the right) only very low feature intensity is detectable.

[0154] Example 3: For comparison, correspondingly prepared snippet forgeries having approx. 10% of forgery portion were measured.

[0155] FIG. 9 shows a representation of the feature intensity of an incomplete bank note with a diagonally inserted strip of a copy without feature.

[0156] FIG. 10a shows a pixel-wise classification of the bank note (FIG. 7-8) with dynamic threshold. The low feature intensity in the region of the hologram strip could be taken into consideration by the dynamic threshold, while the missing feature intensity is marked red in the window region in the absence of remission signal. (0=black, 1=red, 2=yellow, 3=green)

[0157] FIG. 10b shows a pixel-wise classification of the incomplete bank note (FIG. 9) with dynamic threshold. The low feature intensity in the region of the hologram strip could be corrected by the dynamic threshold, while the missing feature intensity is marked red in the window region in the absence of remission signal. The missing feature region is correctly recognized and likewise marked red. (0=black, 1=red, 2=yellow, 3=green)

[0158] Example 4: The bank note of FIG. 7-8 was again surveyed with a sensor construction having both-sided measurement. Feature1 (front), Feature2 (back), Remission) (front), Remission2 (back) were measured as well as the transmission.

[0159] FIG. 11 shows transmission data of the bank note

[0160] For classifying the measurement pixels, the front side and back side were classified separately with dynamic feature threshold were thereupon separately combined according to the following association of the class allocations established respectively on front side (Classification1) and back side (Classification2), into an overall classification for each pixel, as shown in FIG. 12.

[0161] The window region was thereupon recognized with the help of the high transmission>85 and was correspondingly classified as window (4).

[0162] FIG. 12 shows a pixel-wise classification of the measuring data of the complete test bank note established on both sides, with dynamic threshold and transmission measurement. (0=black, 1=red, 2=yellow, 3=green, 4=light blue) Here, in spite of the regarding measurement technique difficult architecture of the bank note having metallically reflective and transparent window regions, all regions are reliably checked for local authenticity and the completeness is correctly assessed.

[0163] In FIG. 13, a combination of feature values classified on both sides is schematically represented, according to which likewise an assessment of the value document or the bank note on authenticity and/or completeness is effected.