Paper sheet identification device, paper sheet processing device, paper sheet identification method, and paper sheet identification program

Abstract

This invention provides a paper sheet identification device, a paper sheet processing device, a paper sheet identification method, and a paper sheet identification program that can accurately distinguish between various photoluminescent inks. [Solution] A paper sheet identification device comprising: a light source capable of irradiating paper sheets printed with a photoluminescent ink to be identified with at least excitation light; a light receiving unit that receives photoluminescence emitted from the paper sheets irradiated with the excitation light and outputs a photoluminescence detection signal; and an identification unit that identifies the paper sheets based on a photoluminescence image of the photoluminescent ink to be identified created from the photoluminescence detection signal output from the light receiving unit and a trained model.

Description

Technical Field

[0001] The present disclosure relates to a paper sheet discrimination device, a paper sheet processing device, a paper sheet discrimination method, and a paper sheet discrimination program.

Background Art

[0002] Conventionally, photoluminescence compounds are known as security elements attached to paper sheets such as banknotes. Photoluminescence compounds are excited by ultraviolet light or the like and produce fluorescence emission or phosphorescence emission. As methods for detecting those characteristics, for example, those disclosed in the following documents are known.

[0003] Patent Document 1 describes a paper sheet authenticity determination device that determines the authenticity of paper sheets to which a phosphor is added. This paper sheet authenticity determination device discriminates the type of paper sheet based on characteristics other than fluorescence emission characteristics, irradiates the paper sheet with excitation light having different wavelengths in order, measures the intensity of the emission light of the phosphor added to the paper sheet for each wavelength in a predetermined range, and obtains fluorescence emission characteristic data as a result. Then, the authenticity of the paper sheet is determined using the fluorescence emission characteristic data of genuine paper sheets stored in advance for each type of paper sheet or a threshold value obtained therefrom and the acquired fluorescence emission characteristic data.

Prior Art Documents

Patent Documents

[0004]

Patent Document 1

[0006] The authenticity determination method described in Patent Document 1 does not apply to fluorescent inks that have multiple peaks upon irradiation with a specific excitation light. In other words, the authenticity determination method described in Patent Document 1 does not take into account the correlation of fluorescence detection intensities in each wavelength band.

[0007] This disclosure is made in view of the above-mentioned circumstances and aims to provide a paper sheet identification device, a paper sheet processing device, a paper sheet identification method, and a paper sheet identification program that can accurately distinguish various photoluminescent inks. [Means for solving the problem]

[0008] To solve the above-mentioned problems and achieve the objective, (1) a paper sheet identification device according to a first aspect of the present disclosure comprises: a light source capable of irradiating at least excitation light onto paper sheets printed with photoluminescent ink to be identified; a light receiving unit that receives photoluminescence emitted from the paper sheets irradiated with the excitation light and outputs a photoluminescence detection signal; and an identification unit that identifies the paper sheets based on a photoluminescence image of the photoluminescent ink to be identified created from the photoluminescence detection signal output from the light receiving unit and a trained model.

[0009] (2) In the paper sheet identification device described in (1) above, the trained model may receive the photoluminescence image or a value based on the photoluminescence image as input and output the similarity of the photoluminescence image to the training data.

[0010] (3) In the paper sheet identification device described in (1) or (2) above, the trained model may be unsupervisedly trained using one type of photoluminescent image produced by genuine photoluminescent ink.

[0011] (4) In the paper sheet identification device described in (1) or (2) above, the trained model may be supervised-learned using multiple types of photoluminescent images produced by multiple types of photoluminescent inks, including genuine photoluminescent ink.

[0012] (5) In the paper sheet identification device described in any of (1) to (4) above, the wavelength of the excitation light irradiated from the light source may be shorter than the wavelength of the photoluminescence emitted from the paper sheets and received by the light receiving unit.

[0013] (6) In the paper sheet identification device described in any of (1) to (5) above, the excitation light irradiated from the light source may be ultraviolet light.

[0014] (7) Furthermore, a paper sheet processing device according to a second aspect of the present disclosure is equipped with a paper sheet identification device as described in any of (1) to (6) above.

[0015] (8) A paper sheet identification method according to a third aspect of the present disclosure comprises the steps of: irradiating a paper sheet printed with a photoluminescent ink to be identified with at least excitation light from a light source; receiving photoluminescence emitted from the paper sheet irradiated with the excitation light with a light receiving unit and outputting a photoluminescence detection signal; and identifying the paper sheet based on a photoluminescence image of the photoluminescent ink to be identified created from the photoluminescence detection signal output from the light receiving unit and a trained model.

[0016] (9) Furthermore, a paper sheet identification program according to a fourth aspect of the present disclosure causes a paper sheet identification device to perform the following processes: irradiating a paper sheet printed with a photoluminescent ink to be identified with at least excitation light from a light source; receiving the photoluminescence emitted from the paper sheet irradiated with the excitation light with a light receiving unit and outputting a photoluminescence detection signal; and identifying the paper sheet based on a photoluminescence image of the photoluminescent ink to be identified created from the photoluminescence detection signal output from the light receiving unit and a trained model. [Effects of the Invention]

[0017] According to this disclosure, it is possible to provide a paper sheet identification device, a paper sheet processing device, a paper sheet identification method, and a paper sheet identification program that can accurately distinguish various photoluminescent inks. [Brief explanation of the drawing]

[0018] [Figure 1] This is a schematic diagram illustrating an example of the configuration of a paper sheet identification device according to Embodiment 1, and is a view from an oblique direction. [Figure 2] This is a schematic perspective view illustrating an example of the configuration of the light-receiving unit of the paper sheet identification device according to Embodiment 1. [Figure 3] This is a schematic diagram showing the wavelength characteristics of the color filter of the light-receiving section of the paper sheet identification device according to Embodiment 1. [Figure 4] This is a flowchart illustrating an example of the operation of the paper sheet identification device according to Embodiment 1. [Figure 5] This figure shows color images of the ink samples used in each verification test. [Figure 6] This graph shows the evaluation results using a single-class SVM in the first validation test, with the score for each sample. [Figure 7] This graph shows the evaluation results using multi-class SVM in the second validation test, indicating the estimated ink score for each sample. [Figure 8]Shows the classification result based on the score shown in FIG. 7. [Figure 9] It is a perspective schematic view showing an example of the appearance of a paper sheet processing apparatus according to Embodiment 2. [Figure 10] It is a cross-sectional schematic view for explaining an example of the configuration of an imaging unit included in the paper sheet identification apparatus according to Embodiment 2. [Figure 11] It is a block diagram for explaining an example of the configuration of the paper sheet identification apparatus according to Embodiment 2. [Figure 12] It is a plan schematic view for explaining an example of the configuration of a light receiving unit included in the paper sheet identification apparatus according to Embodiment 2. [Figure 13] It is a timing chart showing an example of the timing of light emission and light reception by the paper sheet identification apparatus according to Embodiment 2.

Embodiments for Carrying Out the Invention

[0019] Hereinafter, referring to the drawings, embodiments of a paper sheet identification apparatus, a paper sheet processing apparatus, a paper sheet identification method, and a paper sheet identification program according to the present disclosure will be described in detail. As the paper sheets targeted by the present disclosure, various paper sheets such as banknotes, checks, gift certificates, bills of exchange, forms, securities, card-shaped media, etc. are applicable. However, hereinafter, the present disclosure will be described by taking an apparatus for banknotes as an example.

[0020] Also, in this specification, photoluminescence is a concept including fluorescence and phosphorescence. However, hereinafter, the present disclosure will be described by taking fluorescence (photoluminescence that can be detected during excitation light irradiation) as an example of photoluminescence. That is, hereinafter, cases where "photoluminescence", "photoluminescence detection signal", "photoluminescence ink", and "photoluminescence image" are respectively "fluorescence", "fluorescence detection signal", "fluorescence ink", and "fluorescence image" will be described.

[0021] The paper sheet identification program may be pre-introduced into a paper sheet identification apparatus or a paper sheet processing apparatus, or may be recorded on a computer-readable recording medium or provided to an operator via a network.

[0022] Thus, the paper sheet identification device and paper sheet processing device according to this disclosure may include a storage unit composed of a semiconductor memory (RAM or ROM), a hard disk, or other storage device.

[0023] Furthermore, in the following explanation, the same reference numerals are used in common across different drawings for identical parts or parts with similar functions, and repeated explanations are omitted as appropriate. In addition, mutually orthogonal XYZ coordinate systems are shown as appropriate in the drawings illustrating the structure.

[0024] (Embodiment 1) The configuration of the paper sheet identification device according to this embodiment will be explained using Figure 1.

[0025] As shown in Figure 1, the paper sheet identification device 1 according to this embodiment detects fluorescence emitted from a banknote BN to be identified, and includes a light source 11 capable of irradiating at least excitation light onto a banknote BN printed with the fluorescent ink to be identified, a light receiving unit 13 that receives fluorescence emitted from the irradiated banknote BN and outputs a fluorescence detection signal, and an identification unit 23 that identifies the banknote BN using a fluorescence image of the fluorescent ink to be identified created from the fluorescence detection signal output from the light receiving unit 13.

[0026] Here, the banknote BN to be identified may be transported in the X direction within the XY plane. The Y direction may correspond to the main scanning direction of the light receiving unit 13, and the X direction may correspond to the sub-scanning direction of the light receiving unit 13.

[0027] The light source 11 irradiates the banknote BN with excitation light to excite the fluorescent ink to be identified. The light source 11 may be located on the same side as the light receiving unit 13 relative to the banknote BN.

[0028] The wavelength of the excitation light emitted from the light source 11 is not particularly limited and can be set appropriately depending on the fluorescent ink, but it may be shorter than the wavelength of the fluorescence emitted from the banknote BN (and its fluorescent ink) and received by the light receiving unit 13. Thus, the fluorescent ink used in this embodiment may absorb the energy of light in a specific wavelength band and emit that energy as light in a longer wavelength band.

[0029] More specifically, the excitation light emitted from the light source 11 may be ultraviolet light.

[0030] The light source 11 may be longer than the length of the banknote BN in the Y direction, and may illuminate the entire banknote BN in the Y direction in a straight line extending in the Y direction. In this case, the light source 11 may comprise a transparent, linear rod-shaped light guide and light-emitting elements (usually multiple, for example, LEDs (Light Emitting Diodes)) facing at least one of the end faces of the light guide, and may illuminate the banknote BN with light via the light guide.

[0031] The light-receiving unit 13 is configured to receive (detect) fluorescence emitted from the fluorescent ink of the banknote BN that is to be identified while the excitation light is irradiated. In this case, the light-receiving unit 13 can function as a sensor that is sensitive to at least the wavelength range of the fluorescence emitted from the fluorescent ink of the identification target. The light-receiving unit 13 may also be a sensor that is sensitive to a wavelength range that covers at least from the visible region to the infrared region (or near-infrared region). The light-receiving unit 13 outputs an electrical signal (which may be a digital signal) corresponding to the amount of incident light (amount of light received). That is, the fluorescence detection signal is an electrical signal corresponding to the amount of incident light of fluorescence emitted from the banknote BN during the period of excitation light illumination.

[0032] The light-receiving unit 13 may include one or more light-receiving elements, which may receive light, convert it into an electrical signal corresponding to the amount of incident light, and output it.

[0033] The light-receiving unit 13 may be longer than the length of the banknote BN in the Y direction, and may receive light that has been transmitted, reflected, or emitted along the entire Y direction of the banknote BN.

[0034] The light-receiving unit 13 may output an electrical signal corresponding to the amount of incident light as image data. In this case, the light-receiving unit 13 may have multiple pixels arranged in a row in the Y direction (main scanning direction). That is, the light-receiving unit 13 may output an electrical signal corresponding to the amount of incident light in multiple channels corresponding to multiple pixels (positions in the Y direction (main scanning direction)). Note that a channel (row) is a number assigned sequentially to the light-receiving element (image sensor) in the Y direction. In this case, the light-receiving unit 13 may output line data as image data, which is data relating to the light received simultaneously in each channel. By repeatedly irradiating the banknote BN with light from the light source 11 and receiving the light with the light-receiving unit 13 while transporting the banknote BN in the X direction (sub-scanning direction), image data of the entire banknote BN may be output.

[0035] Thus, the light source 11 and the light receiving unit 13 may acquire an image of the entire banknote BN by continuously repeating the imaging process, with a predetermined cycle of imaging being considered as one period.

[0036] In this specification, one cycle refers to a control pattern in which the timing of turning on and off the light-emitting elements in each wavelength band, and the timing of signal reading are set. One cycle of this control pattern is considered one period, and by continuously repeating this, a fluorescence detection signal may be obtained from the entire sheet of paper. One cycle may also represent a periodic control pattern related to turning on, off, and receiving light set to acquire a reflective image and / or transmitted image of the sheet of paper.

[0037] A reflected image is an image based on light reflected by paper sheets, irradiated from a light source positioned on the same side as the light-receiving unit. A transmitted image is an image based on light transmitted through paper sheets, irradiated from a light source positioned on the opposite side of the light-receiving unit. Therefore, reflected and transmitted images are distinct from fluorescence images, which are based on fluorescence emitted from paper sheets.

[0038] The image data acquired by the light-receiving unit 13 consists of multiple pixels arranged in a matrix in the Y direction (main scanning direction) and the X direction (sub-scanning direction). The address of each pixel is identified by the channel (column) of the light-receiving unit 13 corresponding to its position in the Y direction and the line (row) corresponding to its position in the X direction. The line (row) is a number sequentially assigned to the line data output sequentially by the light-receiving unit 13.

[0039] Furthermore, the light-receiving unit 13 may receive light of multiple wavelength bands arriving from the banknote BN and output electrical signals (fluorescence detection signals) for each of the multiple wavelength bands. In this case, each pixel may be equipped with multiple light-receiving elements that selectively receive light of different wavelength bands from each other.

[0040] Multiple wavelength bands that the light-receiving unit 13 can selectively receive include red (R), green (G), blue (B), infrared (IR), and others.

[0041] As shown in Figure 2, the light-receiving unit 13 may include a first light-receiving element 31B having a color filter 32B, a second light-receiving element 31G having a color filter 32G, and a third light-receiving element 31R having a color filter 32R.

[0042] The light-receiving unit 13 may include a plurality of pixels 30 arranged in a row in the main scanning direction D1 (the direction perpendicular to the transport direction of the banknote BN, the Y direction), and each pixel 30 may include one first light-receiving element (image sensor) 31B, one second light-receiving element (image sensor) 31G, and one third light-receiving element (image sensor) 31R, and the first light-receiving element 31B, the second light-receiving element 31G, and the third light-receiving element 31R may be arranged in this order in a row in the main scanning direction D1.

[0043] Here, a light-receiving element (image sensor) means an element that detects (converts into an electrical signal) the intensity of light in a predetermined wavelength band, and may be configured to include a photodetector such as a photodiode, and a color filter (color resist) provided on the light-receiving surface of the photodetector that suppresses the transmission of light in wavelength bands other than the predetermined wavelength band to be detected (for example, the blue and infrared wavelength bands) (for example, the green and red wavelength bands).

[0044] As shown in Figure 2, the first light-receiving element 31B may include a photodetector 33 and a color filter 32B, the second light-receiving element 31G may include a photodetector 33 and a color filter 32G, and the third light-receiving element 31R may include a photodetector 33 and a color filter 32R.

[0045] Furthermore, as shown in Figure 3, color filter 32B transmits blue light and infrared light, color filter 32G transmits green light and infrared light, and color filter 32R transmits red light and infrared light. Therefore, the first light-receiving element 31B, the second light-receiving element 31G, and the third light-receiving element 31R each receive infrared light along with their corresponding visible light. Color filter 32B absorbs green light and red light, color filter 32G absorbs blue light and red light, and color filter 32R absorbs blue light and green light.

[0046] Thus, the light-receiving unit 13 may receive both the visible-range fluorescent component and the infrared-range fluorescent component emitted from the banknote BN (fluorescent ink) irradiated with excitation light, without separating them with each light-receiving element, and output a fluorescence detection signal that includes a signal value corresponding to the total light intensity of both components.

[0047] Even in such cases, the identification unit 23 makes it possible to distinguish between various fluorescent inks, including fluorescent inks that emit fluorescence in the visible and infrared regions, respectively.

[0048] Specifically, the identification unit 23 identifies banknote BNs based on a fluorescence image of the target fluorescent ink (hereinafter, this fluorescence image is also simply referred to as the "target fluorescence image") created from the fluorescence detection signal output from the light receiving unit 13, and a trained model. According to the trained model, it is possible to distinguish fluorescent inks from slight differences in fluorescence detection signals. Therefore, even if the light receiving unit 13 cannot separate fluorescence in the visible and infrared regions, the trained model can distinguish various fluorescent inks, including those that emit fluorescence in both the visible and infrared regions, based on the target fluorescence image. In other words, according to this embodiment, various fluorescent inks can be distinguished with high accuracy.

[0049] The fluorescent image to be identified is a color image of the fluorescent ink to be identified that has emitted fluorescence when irradiated with specific excitation light from the light source 11, and the fluorescent ink to be identified may be captured in the entire area of ​​the image. The shape of the fluorescent image to be identified is not particularly limited and can be set according to the printed area of ​​the fluorescent ink to be identified, and may be rectangular, for example.

[0050] More specifically, the trained model may be input with the target fluorescence image or values ​​based on the target fluorescence image, and output the similarity of the target fluorescence image to the training data.

[0051] Thus, the trained model may be input with the fluorescence image to be identified itself, that is, the RGB values ​​of the fluorescence image to be identified, or it may be input with values ​​based on the fluorescence image to be identified. Here, "values ​​based on the fluorescence image to be identified" are values ​​calculated from the image data of the fluorescence image to be identified, and may be, for example, representative values ​​of the fluorescence image to be identified. Specific examples of representative values ​​include, for example, the mean or median for each RGB value.

[0052] In principle, trained models are input with data in the same format as the training data. That is, for example, if the training data is the fluorescence images themselves, the trained model will also be input with the fluorescence images themselves, and if the training data is the average value of the fluorescence images, the trained model will also be input with the average value of the fluorescence images.

[0053] Furthermore, "similarity to training data" is almost synonymous with "distance to training data," and both are indicators used to compare the fluorescence image to be identified with the training data. Similarity indicates the degree to which the fluorescence image to be identified and the training data are similar, while distance indicates the degree to which the fluorescence image to be identified and the training data are dissimilar. In other words, here, "the trained model outputs the similarity of the fluorescence image to be identified with the training data" also includes the mode in which the trained model outputs the distance of the fluorescence image to be identified with respect to the training data.

[0054] The identification unit 23 may perform discrimination or classification of the fluorescent image to be identified by comparing the similarity output from the trained model with predetermined reference data (e.g., a threshold), or it may determine the authenticity or presence of fluorescent ink on a banknote BN based on the result of the discrimination or classification.

[0055] The trained model may be unsupervised, using only one type of fluorescence image produced by a genuine fluorescent ink. In this case, the data of the fluorescent ink to be identified can be compared with the training data to evaluate the degree of similarity, i.e., the degree to which the data of the fluorescent ink to be identified is anomalous compared to the training data. This allows for the determination of whether the fluorescent ink to be identified is genuine or not, based on the degree of similarity (or anomalousity). Specifically, such models can be used, for example, a one-class SVM (Support Vector Machine) or an autoencoder.

[0056] The single fluorescence image used in unsupervised learning may be the image itself, or it may be a value based on that image, i.e., a value calculated from the image data. In the latter case, for example, it may be a representative value of the fluorescence image (e.g., the mean or median of each RGB value). For values ​​based on a single fluorescence image, for example, a one-class SVM can be applied, and for the single fluorescence image itself, for example, an autoencoder can be applied.

[0057] In unsupervised learning, the single type of fluorescence image used may be just one image or multiple images of the same type.

[0058] The trained model may be supervised and trained using multiple types of fluorescence images from multiple types of fluorescent inks, including genuine fluorescent ink. In this case, the data of the fluorescent ink to be identified can be compared with the training data, and it is possible to evaluate which class of the training data the fluorescent ink to be identified is most similar to (closest to) can be identified. This allows the fluorescent ink to be identified according to its similarity (distance) to each class. Specifically, such models can utilize, for example, multi-class SVMs (Support Vector Machines) or neural networks such as CNNs (Convolutional Neural Networks).

[0059] The multiple types of fluorescence images used in supervised learning may be the images themselves, or they may be values ​​based on those images, i.e., values ​​calculated from the image data. In the latter case, for example, they may be representative values ​​of the fluorescence images (e.g., the mean or median of each RGB value). For the values ​​based on each of the multiple types of fluorescence images, for example, a multi-class SVM can be applied, and for the multiple types of fluorescence images themselves, for example, a neural network such as a CNN can be applied.

[0060] The multiple types of fluorescence images used in supervised learning may consist of one image for each type, or they may include multiple images of the same type for one or more types of fluorescence images.

[0061] Furthermore, multiple types of fluorescent inks, including genuine fluorescent inks, used in supervised learning may consist only of multiple types of genuine fluorescent inks, or they may include one or more types of genuine fluorescent inks and one or more non-genuine inks (e.g., counterfeit fluorescent inks).

[0062] Furthermore, a trained model may be further trained. That is, a trained model may be further trained by applying different training data than that used during initial training.

[0063] The true fluorescent ink used in unsupervised and supervised learning is compared to the target fluorescent ink and emits fluorescence in a predetermined wavelength range while irradiated with excitation light (e.g., ultraviolet light). The true fluorescent ink may contain one or more, for example, two or more photoluminescent compounds.

[0064] Furthermore, the fluorescence properties of true fluorescent inks used in unsupervised and supervised learning are not particularly limited. For example, true fluorescent inks may include fluorescent inks that emit fluorescence in multiple wavelength bands, such as the visible and infrared regions, or fluorescent inks that emit fluorescence in a single wavelength band. Similarly, the fluorescence of true fluorescent inks may have peak wavelengths in multiple wavelength bands, such as the visible and infrared regions, or it may have a peak wavelength in only a single wavelength band. Here, the infrared region may also be the near-infrared region. Hereinafter, fluorescent inks with peak wavelengths in the visible and infrared regions may be referred to as special fluorescent inks.

[0065] This special fluorescent ink emits almost no light when exposed to visible light and transmits visible light, making it invisible to the human eye under natural light or typical artificial lighting. Furthermore, when the special fluorescent ink is irradiated with excitation light, the fluorescent components that emit light in the visible range may be visible to the human eye, but the fluorescent components that emit light in the infrared range remain invisible. Therefore, this special fluorescent ink can function as a highly secure security element.

[0066] The printed area of ​​the special fluorescent ink may, for example, be printed with an ink containing a mixture of a photoluminescent compound that fluoresces in the visible range and a photoluminescent compound that fluoresces in the infrared range, or an ink containing a photoluminescent compound that fluoresces in the visible range and an ink containing a photoluminescent compound that fluoresces in the infrared range may be applied in layers.

[0067] The special fluorescent ink may have a fluorescence spectrum that has a peak in at least one of the blue wavelength band, the green wavelength band, and the red wavelength band, or it may have a peak in only one of the blue, green, or red wavelength bands in the visible range.

[0068] The special fluorescent ink may have a fluorescence spectrum that peaks in the infrared region or in the near-infrared region.

[0069] In this specification, "blue" generally refers to light (color) with a wavelength of approximately 400 nm to 500 nm, and may also refer to light (color) having a peak wavelength in this wavelength range. "Green" generally refers to light (color) with a wavelength of approximately 500 nm to 600 nm, and may also refer to light (color) having a peak wavelength in this wavelength range. "Red" generally refers to light (color) with a wavelength of approximately 600 nm to 750 nm, and may also refer to light (color) having a peak wavelength in this wavelength range. Furthermore, "infrared light" generally refers to light with a wavelength of 750 nm or more, and may also refer to light having a peak wavelength in this wavelength range. "Near-infrared light" generally refers to light with a wavelength of approximately 750 nm to 1500 nm, and may also refer to light having a peak wavelength in this wavelength range.

[0070] Next, the operation of the paper sheet identification device 1 according to this embodiment will be explained using Figure 4.

[0071] As shown in Figure 4, first, the light source 11 irradiates the banknote BN printed with the fluorescent ink to be identified with at least excitation light (step S01).

[0072] Next, the light-receiving unit 13 receives the fluorescence emitted from the banknote BN that has been irradiated with excitation light and outputs a fluorescence detection signal (step S02).

[0073] Subsequently, the identification unit 23 identifies the banknote BN based on the fluorescent image of the target fluorescent ink created from the fluorescence detection signal output from the light receiving unit 13 and the trained model (step S03), and the operation of the paper sheet identification device 1 ends.

[0074] The identification unit 23 may also function by executing a corresponding program by the control unit, which will be described later.

[0075] Here, we will describe a verification test in which the trained model was used to distinguish between fluorescent inks.

[0076] Each verification test used samples S1 to S11, each printed on blank paper with 11 different inks. Figure 5 shows color images of samples S1 to S11 taken after being irradiated with ultraviolet light. Although Figure 5 is shown in grayscale, color images were used in the actual verification tests. In addition, 10 color images were prepared for each sample in each verification test. The inks in samples S1 to S2 are not fluorescent inks and do not fluoresce in the visible or infrared regions. The inks in samples S3 to S4 are fluorescent inks that have an emission peak in the green wavelength band in the visible region (however, the peak intensity differs between samples S3 and S4), but do not fluoresce in the infrared region. The inks in samples S5 to S6 are fluorescent inks that have an emission peak in the near-infrared region (however, the peak intensity differs between samples S5 and S6), but do not fluoresce in the visible region. Samples S7-S11 are special fluorescent inks that have emission peaks in the green wavelength band in the visible range and also in the near-infrared range (however, the peak intensities differ for samples S7-S11). In other words, the color images of samples S3-S11 are color fluorescent images.

[0077] First, as a first validation test, unsupervised learning of a one-class SVM was applied to the representative values ​​(specifically, the average values ​​for each RGB value) of the color image of sample S7, and the representative values ​​(specifically, the average values ​​for each RGB value) of the color images of samples S1 to S11 were input to that one-class SVM. The resulting scores for each sample are shown in Figure 6. Here, each score indicates the similarity (abnormality) of each input sample to sample S7, and in Figure 6, a higher score indicates a greater dissimilarity. As shown in Figure 6, the scores of samples other than sample S7 are clearly higher than the score of sample S7, so it can be seen that if a threshold is set between these scores, for example between -0.8 and -0.6, it is possible to accurately determine whether each input sample is the special fluorescent ink of sample S7 or not.

[0078] Next, as a second validation test, supervised learning of a multi-class SVM was applied to the representative values ​​(specifically, the average values ​​for each RGB value) of the color images of each sample S1 to S11, and the representative values ​​(specifically, the average values ​​for each RGB value) of the color images of samples S1 to S11 were input to the multi-class SVM. The results are shown in Figure 7 for the scores of each sample. In Figure 7, the samples input to the multi-class SVM are shown vertically, and the scores of each input sample for each sample S1 to S11 are shown horizontally, with brighter colors indicating higher scores. Here, each score indicates the similarity of each input sample to each sample S1 to S11, and in Figure 7, a higher score indicates greater similarity. Furthermore, Figure 8 shows the results of classifying each score from Figure 7 output from the multi-class SVM using a predetermined threshold. As a result, each input sample (true class) and the classified sample (predicted class) match for all 10 images of each sample (no misclassification), indicating that various fluorescent inks can be accurately distinguished.

[0079] (Embodiment 2) The paper sheet processing device according to this embodiment may have, for example, the configuration shown in Figure 9. The paper sheet processing device 300 shown in Figure 9 is a small paper sheet processing device that is installed and used on a table, and comprises a paper sheet identification device (not shown in Figure 9) that performs banknote identification processing, a hopper 301 on which a plurality of banknotes to be processed are placed in a stacked state, two reject units 302 that discharge rejected banknotes such as counterfeit bills and bills of uncertain authenticity that are fed from the hopper 301 into the housing 304, an operation unit 303 for inputting instructions from the operator, four stacking units 306a to 306d for classifying and stacking banknotes whose denomination, authenticity, and condition have been identified within the housing 304, and a display unit 305 for displaying information such as the banknote identification counting results and the stacking status of each stacking unit 306a to 306d. Based on the results of the paper sheet identification device's determination of whether a banknote is genuine or damaged, genuine banknotes are stored in storage units 306a to 306c, and damaged banknotes are stored in storage unit 306d. The method for distributing banknotes to storage units 306a to 306d can be set arbitrarily.

[0080] Next, the configuration of the imaging unit, which is the main part of the paper sheet identification device according to this embodiment, will be described using Figure 10. As shown in Figure 10, the imaging unit 211 comprises an upper unit 110 and a lower unit 120 that are arranged facing each other. A gap is formed between the upper unit 110 and the lower unit 120, which are spaced apart in the Z direction, through which banknotes BN are transported in the X direction within the XY plane. This gap constitutes part of the transport path of the paper sheet processing device according to this embodiment. The upper unit 110 and the lower unit 120 are located on the upper side (+Z direction) and lower side (-Z direction) of the transport path, respectively. The Y direction corresponds to the main scanning direction of the imaging unit 211, and the X direction corresponds to the sub-scanning direction of the imaging unit 211.

[0081] As shown in Figure 10, the upper unit 110 includes two reflective light sources 111, a condensing lens 112, a light receiving unit 113, and a UV-cut film 115. The reflective light sources 111 sequentially irradiate the main surface (hereinafter referred to as surface A) of the banknote BN on the light receiving unit 113 side with light having different wavelength bands, specifically infrared light, white light including red, green, and blue light, and ultraviolet light as excitation light for fluorescence. The condensing lens 112 collects the light emitted from the reflective light sources 111 and reflected from surface A of the banknote BN, the light emitted from the transmitting light source 124 provided in the lower unit 120 and transmitted through the banknote BN, and the fluorescence emitted from surface A of the banknote BN. The light receiving unit 113 receives the light collected by the condensing lens 112 and converts it into an electrical signal. After amplifying the electrical signal, it performs A / D conversion to digital data and outputs it. Here, the light received by the light-receiving unit is also called incident light, and the light emitted by the light source is also called emitted light. The UV-cut film 115 prevents ultraviolet light emitted from the reflective light source 111 and reflected from side A of banknote BN from being received by the light-receiving unit 113 via the condensing lens 112.

[0082] The lower unit 120 includes two reflective light sources 121 and one transmissive light source 124, a condensing lens 122, a light receiving unit 123, and a UV-cut film 125. The reflective light sources 121 irradiate the main surface (hereinafter referred to as the B-side) of the banknote BN on the light receiving unit 123 side with illumination light having different wavelength bands, specifically infrared light, white light including red, green, and blue light, and ultraviolet light as excitation light for fluorescence. The condensing lens 122 focuses the light emitted from the reflective light sources 121 and reflected from the B-side of the banknote BN, as well as the fluorescence emitted from the B-side of the banknote BN. The light receiving unit 123 receives the light focused by the condensing lens 122 and converts it into an electrical signal. After amplifying the electrical signal, it performs A / D conversion to digital data and outputs it. The UV-cut film 125 prevents ultraviolet light emitted from the reflective light source 121 and reflected from the B-side of the banknote BN from being received by the light-receiving unit 123 via the condensing lens 122.

[0083] The light source 124 for transmission is positioned on the optical axis of the focusing lens 112 of the upper unit 110. A portion of the light emitted from the light source 124 passes through the banknote BN and is focused by the focusing lens 112 of the upper unit 110 and detected by the light receiving unit 113. The light source 124 may sequentially irradiate the B side of the banknote BN with light having different wavelength bands, or it may irradiate them simultaneously.

[0084] In this specification, light with different wavelength bands (irradiated light, incident light, etc.) refers, for example, to light with different colors in the case of visible light, and to light with wavelength bands that overlap only partially or that do not overlap in the case of infrared and ultraviolet light.

[0085] Each light source 111, 121, and 124 includes a line-shaped light guide (not shown) extending in a direction perpendicular to the plane of the paper in Figure 10 (main scanning direction D1), and a plurality of LED elements (not shown) provided at both ends (or one end) of the light guide.

[0086] Each light source 111, 121 may include an LED element that emits infrared light with a peak wavelength of 750 nm or more, an LED element that emits red light (R) with a peak wavelength of 600 nm or more and less than 750 nm, an LED element that emits green light (G) with a peak wavelength of 500 nm or more and less than 600 nm, an LED element that emits blue light (B) with a peak wavelength of 400 nm or more and less than 500 nm, and an LED element that emits ultraviolet light (UV) with a peak wavelength of less than 400 nm. One light source 111 is placed on the upstream and downstream sides in the transport direction, flanking the condensing lens 112, and one light source 121 is placed on the upstream and downstream sides in the transport direction, flanking the condensing lens 122.

[0087] The light source 124 may include multiple LED elements that emit light having different peak wavelengths. The peak wavelength refers to the wavelength at which the light emission intensity is maximum.

[0088] As shown in Figure 2, each light-receiving unit 113, 123 is equipped with a plurality of pixels 30 arranged in a line in the main scanning direction D1 (the direction perpendicular to the transport direction of the banknote BN, the Y direction). Each pixel 30 is equipped with one first light-receiving element (image sensor) 31B having a color filter 32B, one second light-receiving element (image sensor) 31G having a color filter 32G, and one third light-receiving element (image sensor) 31R having a color filter 32R. The first light-receiving element 31B, the second light-receiving element 31G, and the third light-receiving element 31R are arranged in this order in a line in the main scanning direction D1.

[0089] The upper unit 110 and the lower unit 120 each repeatedly capture images of the banknote BN being transported in the transport direction and output signals corresponding to the amount of light received, thereby enabling the imaging unit 211 to acquire an image of the entire banknote BN. Specifically, the imaging unit 211 acquires a transmitted image and a reflected image of side A of the banknote BN based on the output signal of the upper unit 110, and acquires a reflected image of side B of the banknote BN based on the output signal of the lower unit 120.

[0090] Furthermore, the imaging unit 211 acquires a fluorescence detection signal across the entire banknote BN on both side A and side B of the banknote BN. In other words, the imaging unit 211 can acquire fluorescence images of both side A and side B of the banknote BN.

[0091] Next, the configuration of the paper sheet identification device according to this embodiment will be described using Figure 11. As shown in Figure 11, the paper sheet identification device 200 according to this embodiment includes a detection unit 210, a control unit 220, and a storage unit 230.

[0092] The control unit 220 is a controller that controls each part of the paper sheet identification device 200, and is composed of a program for realizing various processes stored in the memory unit 230, a CPU (Central Processing Unit) that executes the program, and various hardware (e.g., FPGA (Field Programmable Gate Array)) controlled by the CPU. The control unit 220 controls each part of the paper sheet identification device 200 based on signals output from each part of the paper sheet identification device 200 and control signals from the control unit 220, according to the program stored in the memory unit 230. In addition, the control unit 220 has the functions of a light source control unit 221, a sensor control unit 224, an image generation unit 225, and an identification unit 223, according to the program stored in the memory unit 230.

[0093] The detection unit 210 includes a magnetic detection unit 212 and a thickness detection unit 213, in addition to the imaging unit 211 described above, along the banknote transport path. The imaging unit 211 captures images of the banknotes as described above and outputs an image signal (image data). The magnetic detection unit 212 is equipped with a magnetic sensor (not shown) for measuring magnetism, and detects magnetism such as magnetic ink and security threads printed on the banknotes using the magnetic sensor. The magnetic sensor is a magnetic line sensor in which multiple magnetic detection elements are arranged in a line. The thickness detection unit 213 is equipped with a thickness detection sensor (not shown) for measuring the thickness of the banknotes, and detects tape, double feeding, etc. using the thickness detection sensor. The thickness detection sensor detects the amount of displacement when banknotes pass through rollers facing each other across the transport path using sensors provided on each roller.

[0094] The memory unit 230 is composed of a non-volatile storage device such as a semiconductor memory or a hard disk, and stores various programs and data for controlling the paper sheet identification device 200. The memory unit 230 may also store a learned model used for identification processing by the identification unit 223. In addition, the memory unit 230 stores imaging parameters such as the wavelength band of the illumination light emitted from each light source 111, 121, and 124 during one cycle of imaging by the imaging unit 211, the timing of turning each light source 111, 121, and 124 on and off, the value of the forward current flowing through the LED elements of each light source 111, 121, and 124, and the timing of reading signals from the upper unit 110 and the lower unit 120, respectively.

[0095] One cycle of imaging refers to an imaging pattern in which the wavelength range of the light emitted from each light source 111, 121, and 124, as well as the timing of turning each light source 111, 121, and 124 on and off, and reading the signal, are set. One cycle of imaging constitutes one period, and by continuously repeating this process, an image of the entire banknote is acquired.

[0096] The light source control unit 221 performs dynamic lighting control of each light source 111, 121, and 124 in order to capture individual banknote images using each light source 111, 121, and 124. Specifically, the light source control unit 221 controls the lighting and extinguishing of each light source 111, 121, and 124 based on the timing set in the imaging parameters. This control is performed using a mechanical clock that changes according to the banknote transport speed and a system clock that is always output at a constant frequency regardless of the banknote transport speed.

[0097] The sensor control unit 224 controls the timing of reading signals from the upper unit 110 and the lower unit 120 based on the timing set in the imaging parameters, and reads signals from the upper unit 110 and the lower unit 120 in synchronization with the timing of the on and off of each light source 111, 121, and 124. This control is performed using the mechanical clock and the system clock. The sensor control unit 224 then sequentially stores the read signals, i.e., line data, in the ring buffer (line memory) of the storage unit 230.

[0098] Here, line data refers to data based on signals obtained from a single image capture by each of the upper unit 110 and the lower unit 120, and corresponds to one row of data in the horizontal direction (the direction perpendicular to the banknote transport direction, the Y direction) of the acquired image.

[0099] The image generation unit 225 has the function of generating images based on various signals related to banknotes acquired from the detection unit 210. Specifically, the image generation unit 225 first decomposes the data (image signals) stored in the ring buffer into data for each light irradiation and reception condition. Then, according to the characteristics of each decomposed data, the image generation unit 225 performs correction processing such as dark output cut, gain adjustment, and bright output level correction to generate various image data of banknotes and store them in the storage unit 230.

[0100] The identification unit 223 uses the fluorescence detection signal acquired by the imaging unit 211 to identify the banknote BN.

[0101] More specifically, the identification unit 223 extracts the region containing the fluorescent ink to be identified from the color fluorescent image of the entire banknote BN created by the image generation unit 225, and inputs the extracted image or a representative value calculated from the extracted image (for example, the average value for each RGB value) into the trained model. This extracted region may be set according to the denomination of the banknote. The trained model outputs a similarity score indicating how similar the input data (extracted image data or its representative value) is to the training data. The identification unit 223 then identifies the banknote BN by comparing the output similarity score with predetermined reference data (for example, a threshold). For example, it may determine whether the fluorescent ink to be identified is genuine fluorescent ink or not, or it may determine (classify) whether the fluorescent ink to be identified is one of several types of fluorescent ink, including genuine fluorescent ink.

[0102] Here, other examples of the light source and light receiving section for reflection in the upper unit 110 and the lower unit 120, and their control methods will be described.

[0103] As shown in Figure 12, each pixel 30 of each light-receiving section 113, 123 may include a first light-receiving element (image sensor) 31B having a color filter 32B, a second light-receiving element (image sensor) 31G having a color filter 32G, a third light-receiving element (image sensor) 31R having a color filter 32R, and a fourth light-receiving element (image sensor) 31C having a clear filter 32C.

[0104] As shown in Figure 3, color filter 32B transmits blue light and infrared light, color filter 32G transmits green light and infrared light, and color filter 32R transmits red light and infrared light. Therefore, the first light-receiving element 31B, the second light-receiving element 31G, and the third light-receiving element 31R each receive infrared light along with their corresponding visible light. Color filter 32B absorbs green light and red light, color filter 32G absorbs blue light and red light, and color filter 32R absorbs blue light and green light.

[0105] In contrast, the clear filter 32C transmits light from at least the short-wavelength limit of the visible range to a predetermined wavelength in the infrared range (e.g., 1100 nm).

[0106] Furthermore, the fourth light-receiving element 31C does not necessarily need to be equipped with a filter such as a clear filter 32C or a color filter.

[0107] Furthermore, each of the reflective light sources 111 and 121 includes an LED element that emits red light (R) with a peak wavelength of 600 nm or more and less than 750 nm, an LED element that emits green light (G) with a peak wavelength of 500 nm or more and less than 600 nm, an LED element that emits blue light (B) with a peak wavelength of 400 nm or more and less than 500 nm, and an LED element that emits ultraviolet light (UV) with a peak wavelength of less than 400 nm.

[0108] Then, according to the timing chart shown in Figure 13, the upper unit 110 and the lower unit 120 each repeatedly take images of the banknote BN being transported in the transport direction for one cycle, and the imaging unit 211 acquires an image of the entire banknote BN.

[0109] In other words, the four phases are repeated as one cycle, and in phases 1 to 3, each of the reflective light sources 111 and 121 sequentially lights up LED elements that emit red light (R), LED elements that emit green light (G), and LED elements that emit blue light (B), while the first to third light-receiving elements sequentially receive the red reflected light (RR), green reflected light (RG), and blue reflected light (RB). In addition, in each of phases 1 to 3, the fourth light-receiving element also sequentially receives the red reflected light (RR), green reflected light (RG), and blue reflected light (RB). In Phase 4, each of the reflective light sources 111 and 121 lights up an LED element that emits ultraviolet (UV) light, while the first to third photodetectors receive red fluorescence (RR-UV), green fluorescence (RG-UV), and blue fluorescence (RB-UV), respectively, and the fourth photodetector receives fluorescence from the visible to infrared range (RC-UV).

[0110] This makes it possible to double the resolution of the reflected image data in the main scanning direction compared to when the fourth photodetector 31C is not provided. Specifically, for example, the resolution of the reflected image data can be set to main scanning direction × sub-scanning direction: 200 dpi × 100 dpi for each of R, G, and B. In this case, the resolution of the fluorescence image data will be main scanning direction × sub-scanning direction: 100 dpi × 100 dpi for each of R, G, B, and C (clear).

[0111] (Variation 1) In the above embodiment, the case of detecting fluorescence as photoluminescence was described, but phosphorescence (photoluminescence that can be detected after the excitation light is turned off) may also be used. In that case, the phosphorescence emitted from the banknote to be identified is received by the light receiving unit after the ultraviolet light used as excitation light is turned off, and a phosphorescence detection signal is output. This phosphorescence detection signal can be used to perform identification processing in the same way as the fluorescence detection signal. For example, banknotes can be identified based on a phosphorescence image of the phosphorescent ink to be identified, created from the phosphorescence detection signal output from the light receiving unit, and a trained model. This makes it possible to determine the authenticity of various phosphorescent inks that emit phosphorescence in a predetermined wavelength band after being irradiated with ultraviolet light as excitation light. For example, special phosphorescent inks that have peak wavelengths in the visible and infrared regions can be identified. Similar to special fluorescent inks, the phosphorescence component of these special phosphorescent inks that emits light in the infrared region is not visible to the human eye, so they can function as a highly secure security element.

[0112] Although embodiments have been described above with reference to the drawings, this disclosure is not limited to the embodiments described above. Furthermore, the configurations of each embodiment may be combined or modified as appropriate without departing from the spirit of this disclosure. [Industrial applicability]

[0113] As described above, this disclosure is a useful technology for accurately distinguishing various photoluminescent inks. [Explanation of Symbols]

[0114] 1,200: Paper sheet identification device 11, 111, 121, 124: Light source 13, 113, 123: Light receiving section 23, 223: Identification section 30 pixels 31B, 31G, 31R, 31C: Photodetector 32B, 32G, 32R, 32C: Color filters 33: Photodetector 110: Upper unit 112, 122: Focusing lens 115, 125: UV-cut film 120: Lower unit 210: Detection unit 211: Imaging Department 212: Magnetic detection unit 213: Thickness detection unit 220: Control Unit 221: Light source control unit 224: Sensor Control Unit 225: Image generation unit 230: Storage section 300: Banknote Processing Device 301: Hoppa 302: Rejection Department 303:Operation unit 304: Enclosure 305: Display section 306a~306d: Accumulation section BN:Banknote

Claims

1. A light source capable of irradiating at least excitation light onto paper sheets printed with the photoluminescent ink to be identified, A light receiving unit that receives photoluminescence emitted from the paper sheets irradiated with the excitation light and outputs a photoluminescence detection signal, An identification unit that identifies the paper sheets based on a photoluminescence image of the photoluminescence ink to be identified, created from the photoluminescence detection signal output from the light receiving unit, and a trained model. A paper sheet identification device characterized by being equipped with the following features.

2. The trained model receives the photoluminescent image or values ​​based on the photoluminescent image as input and outputs the similarity of the photoluminescent image to the training data. The paper sheet identification device according to claim 1, characterized in that it is a paper sheet identification device.

3. The aforementioned trained model was unsupervised and trained using one type of photoluminescent image produced by genuine photoluminescent ink. The paper sheet identification device according to claim 1 or 2, characterized in that it is a paper sheet identification device.

4. The aforementioned trained model was supervised and trained using multiple types of photoluminescent images produced by multiple types of photoluminescent inks, including genuine photoluminescent inks. The paper sheet identification device according to claim 1 or 2, characterized in that it is a paper sheet identification device.

5. The wavelength of the excitation light emitted from the light source is shorter than the wavelength of the photoluminescence emitted from the paper sheets and received by the light-receiving unit. A paper sheet identification device according to any one of features 1 to 4.

6. The excitation light emitted from the aforementioned light source is ultraviolet light. A paper sheet identification device according to any one of the features 1 to 5.

7. A paper sheet processing device characterized by comprising a paper sheet identification device according to any one of claims 1 to 6.

8. The steps include: irradiating the paper sheets printed with the photoluminescent ink to be identified with at least excitation light from a light source; The steps include: receiving the photoluminescence emitted from the paper sheets irradiated with the excitation light with a light-receiving unit and outputting a photoluminescence detection signal; The steps include identifying the paper sheets based on a photoluminescence image of the photoluminescence ink to be identified, created from the photoluminescence detection signal output from the light receiving unit, and a trained model. A method for identifying paper sheets, characterized by comprising the following features.

9. The process involves irradiating the paper sheets printed with the photoluminescent ink to be identified with at least excitation light from a light source, The process involves receiving the photoluminescence emitted from the paper sheets irradiated with the excitation light using a light-receiving unit and outputting a photoluminescence detection signal. A process for identifying the paper sheets is performed based on a photoluminescence image of the photoluminescence ink to be identified, created from the photoluminescence detection signal output from the light receiving unit, and a trained model. A paper sheet identification program characterized by causing a paper sheet identification device to execute the following.

Images