JP2007034748A - Article confirmation device and method thereof - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide an article confirmation device and method, capable of preventing reduction of accuracy in real and false confirmation of an article. <P>SOLUTION: In registration of an article 40 such as a card, a predetermined range 50 on the surface of a reference article 40 of the registration object is imaged, a plurality of images of small areas (cells) 15 in which an unreproducible fine characteristic on the reference article surface is easy to recognize are selected from the taken image of the reference article 40 as characteristic images, any one of the taken images of the reference article 40, the plurality of selected characteristic images, and characteristic information showing the unreproducible fine characteristic on the article surface contained in each of the selected characteristic images and position information showing each position of the selected characteristic images are registered. When an article 40 of a confirmation object is confirmed to be the reference object 40 or not, whether the article 40 of the confirmation object is the registered reference article 40 or not is determined based on the comparison result with each of the registered characteristics. <P>COPYRIGHT: (C)2007,JPO&INPIT

Description

  The present invention relates to an article confirmation apparatus and an article confirmation method for confirming the authenticity of an article.

Conventionally, various techniques have been proposed to confirm the authenticity of valuables such as credit cards and important documents. For example, as shown in FIG. 7, the article 60 is illuminated by the light source 54, and the article 60 itself originally has a surface of the article 60 from a predetermined observation area (the shooting range 70 in FIG. 7B). The non-reproducible minute features, for example, the pattern of the fine unevenness on the surface (FIG. 7B is characterized by the convex portion 60a on the article surface and the shadow extending in the opposite direction to the light source 54 with respect to the convex portion 60a. Etc.) is read by the camera 56 or the like and registered in advance, and minute features on the surface of the article 60 to be confirmed are read by the camera 56 or the like and compared with the features registered in advance. Document confirmation apparatuses and article confirmation apparatuses for confirming the authenticity of the object 60 to be confirmed are known (see, for example, Patent Documents 1 and 2).
JP 2004-153405 A JP 2005-010881 A

  However, in the prior art, since there is only one imaging range for extracting the feature of the article surface, if the shape of the article surface changes due to scratching or rubbing at that position, the captured image There is a possibility that the feature extracted from the feature will be different from the feature registered in advance, and even though it is the same article, an authentication error will occur and it will become impossible to identify correctly. In addition, the part where fine features that cannot be reproducible are easily recognized differs depending on the article, but since the imaging range was fixed in the past, the article in which the range where the feature is difficult to recognize was imaged. There was a risk that the accuracy of authenticity confirmation would be lowered.

  The present invention has been made to solve the above-described problems, and an object of the present invention is to provide an article confirmation apparatus and an article confirmation method that can prevent a decrease in accuracy in confirming the authenticity of an article.

  In order to achieve the above object, an article confirmation apparatus according to the present invention includes an imaging unit that images an article surface, and a reference article surface when the reference article surface is imaged by the imaging unit. Selection means for selecting a plurality of small-area images in which fine features that are not reproducible on the surface of the article are easily recognized as feature images, the captured image of the reference article surface, and a plurality of images selected by the selection means One of the feature information indicating the non-reproducible fine features of the article surface included in each of the feature images and the plurality of feature images selected by the selection unit, and a plurality of the feature images selected by the selection unit Registration means for registering position information indicating the position of each feature image, and a plurality of positions corresponding to the positions indicated by the registered position information on the surface of the article to be confirmed by the imaging means When each feature image of a small area is captured, each of the captured feature images is compared with each of the selected feature images of the registered reference article Based on the result or a comparison result between each of the features included in the plurality of captured feature images and each of the features included in the selected plurality of feature images of the registered reference article, Determining means for determining whether or not the article to be confirmed is the registered reference article.

  When registering a reference article, the surface of the reference article is imaged, and a plurality of small area images in which fine features that are not reproducible on the reference article surface are easily recognized are selected as feature images. Position information indicating the position of the selected plurality of feature images while registering any one of the image, the plurality of selected feature images, and the feature information indicating the features included in the selected plurality of feature images Register.

  When confirming the confirmation target article, a feature image of each of a plurality of small regions corresponding to the position indicated by the registered position information on the surface of the confirmation target article is captured, and a plurality of the confirmation target article A comparison result of comparing each of the feature images with each of the plurality of feature images of the reference article, or each of the features included in the plurality of feature images of the article to be confirmed and a plurality of features of the reference article Based on the comparison result with each of the features included in the image, it is determined whether or not the confirmation target article is a registered reference article.

  Even when the captured image of the reference article surface is registered, the position information of the feature image is also registered together with the captured image. Therefore, in the determination unit, the image of the position indicated by the position information is used as the reference article. If it is extracted from the captured image, it can be easily confirmed whether or not the article to be confirmed is a registered reference article.

  As described above, since a plurality of small area images in which fine features that are not reproducible on the reference article surface are easily recognized are selected as feature images from the captured image obtained by capturing the reference image, the features are easily recognized. Since a suitable feature image can be dynamically selected and the features included in the plurality of feature images or feature images are used for confirming the article to be confirmed, a single feature image or a single feature image is used. Compared with the case where the confirmation process is performed using only the features included in the item, it is possible to prevent a decrease in accuracy in confirming the authenticity of the article.

  In addition, the fine feature that cannot be reproduced on the surface of the article here means a fine feature that cannot be reproduced from the outside that can be read from the outside. In the case of an article covered with a transparent protective film (transparent layer), not only the non-reproducible fine features appearing in the transparent layer itself, but also the non-reproducible fine features appearing under the transparent layer Is also included.

  Further, the image of the small area selected by the selection unit may be an image of the entire small area, or may be an image of a specific part of the small area, for example, an image of the central part.

  The selection means divides the captured image of the reference article surface into a plurality of small areas, and the density between the micro areas when each of the small areas is divided into a plurality of micro areas. Can be obtained for each small area, and a plurality of small area images with small density variations can be selected as the feature images.

  For example, if there is a portion painted in a dark color such as a print-printed portion on the surface of the article, it is difficult to recognize the feature of minute irregularities that cannot be reproduced on the surface of the article. In other words, an image with little change in brightness due to printing or the like, or no change in brightness is more preferable as a feature image used for confirming the authenticity of the article because the feature of fine unevenness is easily recognized. Therefore, by selecting as a feature image an image of a small area that has a small variation in density between the micro areas when each of the plurality of small areas is divided into a plurality of micro areas, changes in light and darkness due to printing and printing etc. An image with a small amount (a fine feature that cannot be reproduced is easily recognized) can be selected.

  Here, the density is the density of the image, and the degree of density variation can also be expressed by the degree of lightness variation. Accordingly, the density variation referred to here also includes brightness variation.

  The method for obtaining the density variation is not particularly limited. For example, the density range (distribution range) obtained from the maximum value and the minimum value of the density of each minute region may be used as the density variation, A standard deviation may be used as a variation in density.

  The imaging unit images a predetermined range including each of a plurality of small regions corresponding to the position indicated by the registered position information on the surface of the article to be confirmed, and the imaging unit captures the predetermined range from the captured image of the predetermined range. A plurality of images corresponding to the positions indicated by the registered position information on the surface of the article to be confirmed are extracted as feature images by extracting images of the plurality of small regions corresponding to the positions indicated by the registered position information. It is possible to capture a feature image of each of the small regions.

  That is, the imaging unit images a predetermined range including each of a plurality of small areas corresponding to the position indicated by the registered position information of the reference article on the surface of the article to be confirmed, and the registered position information is obtained. By using the captured image, an image at the same position as the plurality of selected feature images of the reference article can be extracted as a feature image of the article to be confirmed. As a result, it is possible to capture all of the plurality of feature images with one imaging.

  The determination means, when the number of matches between each of the plurality of captured feature images of the surface of the article to be confirmed and each of the plurality of feature images of the registered reference article exceeds a majority, or The number of matches between each of the features included in the plurality of feature images captured on the surface of the article to be confirmed and each of the features included in the plurality of feature images of the registered reference article exceeds a majority. In this case, it is possible to determine that the confirmation target article is the registered reference article.

  Thereby, for example, even when the number of coincidence decreases due to noise included in the captured feature image, a change with time of the article, or the like, it is possible to make a correct determination without erroneous determination. Note that the term “match” here includes not only a case where they completely match, but also a case where the similarity is equal to or greater than a predetermined value.

  The article confirmation method of the present invention includes a first imaging step for imaging a reference article surface, and a small non-reproducible feature of the reference article surface that is easily recognized from the captured image captured in the first imaging step. A selection step of selecting a plurality of region images as feature images; a captured image captured in the first imaging step; a plurality of feature images selected in the selection step; and a plurality selected in the selection step Registration for registering any one of feature information indicating non-reproducible fine features of the article surface included in each of the feature images and position information indicating the position of each of the plurality of feature images selected in the selection step A second imaging step of capturing a feature image of each of a plurality of small regions corresponding to the position indicated by the registered position information on the surface of the article to be confirmed, and a plurality of images captured in the second imaging step Of feature images A comparison result comparing each of the plurality of selected feature images of the registered reference article, or each of the features included in the plurality of feature images captured in the second imaging step; Whether or not the article to be confirmed is the registered reference article based on the result of comparison with each of the features included in the selected plurality of feature images of the registered reference article. A determination step of determining.

  Since the article confirmation method of the present invention also operates in the same manner as the article confirmation apparatus of the present invention, it is possible to prevent a decrease in accuracy in confirming the authenticity of the article.

  As described above, according to the present invention, there is an excellent effect that it is possible to prevent a decrease in accuracy in confirming the authenticity of an article.

  Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.

[Random pattern]
First, prior to the description of the embodiment of the present invention, the non-reproducible fine unevenness characteristic (hereinafter referred to as random pattern) inherent in the article itself will be described.

  For example, non-woven fabrics are made by intricately intertwining fibers, and none of the patterns by the fibers are the same. That is, a random pattern due to fibers can be observed from the nonwoven fabric. Also, paper is made of intricately intertwined plant fibers, and a random pattern can be observed from paper as well as non-woven fabric. Even if the paper is the same type and the same lot, the pattern is different for each paper.

  Random patterns can also be observed from the surface of carbon-filled black rubber, the surface of ceramics for IC packages, and the surface of coatings (so-called lame coating) of UV-curable paint dispersed with fine metal particles, due to surface microcracks and material particles. Can be observed. Moreover, the random pattern made at the time of surface finishing, such as a hairline process and a sandblast process, can also be observed from the stainless steel material surface. In addition, since the leather has a natural wrinkle and random wrinkles formed on the surface, the wrinkles are observed as a random pattern.

  Thus, random patterns can be observed from various articles. This random pattern is not intentionally created, but is randomized in the formation of the article itself, in the manufacturing process, or after production, and it is unlikely that there are multiple articles with the same pattern. . Also, it seems difficult to deliberately create the same thing. That is, even if the article is manufactured and distributed through the same process, the random pattern is microscopically different for each article. In particular, the random pattern as described above is a microscopic fine pattern, and it is not easy to forge it. In addition, a random pattern of a soft material whose shape is likely to change, such as a rubber surface, a leather surface, or a non-woven fabric, is stable when no external force is applied.

  Note that the detailed description of the random pattern is described in Patent Document 1 (Japanese Patent Laid-Open No. 2004-153405) and Patent Document 2 (Japanese Patent Laid-Open No. 2005-010881) cited above as prior art documents. Further description of the random pattern is omitted here.

  The present invention utilizes a random pattern inherent to an article as described above as information for confirming (identifying / collating) each article. For reading such a fine random pattern, several methods such as a stylus method and an electron microscope observation method are conceivable. From the viewpoint of article protection, it is desirable that the sample is unprocessed and nondestructive. The method using light is excellent in this respect. Hereinafter, an article confirmation apparatus that reads light using a random pattern and confirms the authenticity of the article will be described.

[overall structure]
FIG. 1 is a schematic configuration diagram of an article confirmation device according to an embodiment of the present invention. As shown in FIG. 1, the article confirmation device 10 includes a CPU 12, a RAM 14, a ROM 16, an external I / F (interface) processing device 18, and an internal I / F processing device 20, which are connected to each other via a system bus 32. Has been.

  The CPU 12 executes a program stored in the ROM 16 and controls the operation of the entire article confirmation apparatus 10. The program stored in the ROM 16 captures the surface of the article to be registered, selects a plurality of small area images from the captured image as an image including a random pattern (hereinafter referred to as a feature image), and performs the selection. A registration processing program for extracting and registering feature quantities indicating features of a random pattern on the surface of the article from the acquired feature image, and capturing a plurality of feature images of the surface of the article to be confirmed, and randomizing the surface of the article from the captured feature image A confirmation processing program for determining authentication / non-authentication by extracting a feature amount indicating a pattern feature and comparing it with a pre-registered feature amount is included. The registration processing program includes a cell selection processing subroutine for selecting a plurality of small regions (cells) whose features are to be extracted from a captured image obtained by capturing the surface of the article to be registered.

  The random pattern on the surface of the article here refers to a fine feature that cannot be reproduced of the article that is visible from the outside. For example, in the case of an article that is covered with a transparent protective film to the extent that fine features of the lower layer appear to be imageable, the non-reproducible fine features appearing on the transparent protective film itself and transparent It includes both features of non-reproducible fine features appearing under the protective film. Specifically, a card such as a credit card that is currently widely used in financial institutions, for example, has a transparent layer such as a hard coat affixed to a paint on which a lamellar fine tint pattern appears, for example. Types are widely used. Therefore, in the case of such an article, both a fine random pattern such as irregularities appearing on the protective film itself and a fine ground pattern (random pattern) below the protective film can be included.

  In this embodiment, as an article to be registered and confirmed, a magnetic card (hereinafter referred to as a card) such as a cash card or a credit card in which a tape-like (striped) magnetic recording medium is attached to a plastic card is taken as an example. I will give you a description.

  The RAM 14 is a work memory, and includes an area for temporarily storing captured images and feature images, data indicating feature amounts extracted from the feature images, data indicating feature amounts acquired from the external device 34, and the like. .

  The external I / F processing device 18 is an interface for exchanging data with the external device 34. The article confirmation apparatus 10 transmits data indicating the feature quantity of the random pattern on the article surface, or transmits the external I / F, in order to cause the external apparatus 34 to register the article to be registered via the external I / F processing apparatus 18. Data indicating the registered feature amount of the article is received from the external apparatus 34 via the F processing apparatus 18 for use in the confirmation process in the article confirmation apparatus 10.

  A light source 22, a camera 24, a magnetic head control device 26, and a card transport device 30 are connected to the internal I / F processing device 20.

  The light source 22 illuminates the card 40 inserted in the article confirmation apparatus 10 by irradiating light in a predetermined wavelength range. The lighting timing of the light source 22 (light irradiation timing) is controlled by the CPU 12 via the internal I / F processing device 20. As the light source 22, for example, an LED, a halogen lamp, a fluorescent lamp, a xenon discharge tube, or the like can be used.

  The camera 24 images a predetermined imaging range 50 on the surface of the card 40 illuminated by the light emitted from the light source 22 (see FIG. 4A). In the present embodiment, the camera 24 and the imaging range 50 of the camera 24 are fixed. The imaging timing of the camera 24 is controlled by the CPU 12 via the internal I / F processing device 20.

  A magnetic head 28 is connected to the magnetic head controller 26. The magnetic head 28 is controlled by the magnetic head control device 26, and when the card 40 is inserted from an insertion port (not shown) and conveyed to the reading position of the magnetic head 28, information is read from the magnetic recording medium affixed to the card 40. read out.

  The card transport device 30 includes a transport roller 30a and a motor 30b. When the card 40 is inserted from an insertion port (not shown), the transport roller 30a is rotated to transport the inserted card 40 to a predetermined position, and the processing using the inserted card 40 is completed. Then, the transport roller 30a is rotated in the reverse direction so that the processed card 40 is transported and discharged from the insertion slot.

  The external device 34 that exchanges data with the article confirmation device 10 receives the identification code of the card 40, the data indicating the feature amount, and the position information (described later) from the article confirmation device 10 when the identification code is received. Is stored in the database in association with the data indicating the feature amount and the position information, and when the request for the data indicating the feature amount and the position information is received from the article confirmation apparatus 10, the request is received. The data indicating the feature amount and the position information are read from the database and transmitted to the article confirmation apparatus 10.

[Operation of article confirmation device]
Next, the operation of the article confirmation device 10 will be described. In order to determine the authenticity of the card 40 by the article confirmation device 10, it is necessary to register in advance the characteristics of the random pattern that the real card 40 has. With reference to FIGS. 2-5, the registration process performed in the article | item confirmation apparatus 10 is demonstrated.

  When the registration target card 40 is inserted from an insertion port (not shown), the card transport device 30 transports the card 40 to the reading position of the magnetic head 28. At this time, the registration processing program shown in FIG. 2 is started.

  In step 100, the magnetic head controller 26 controls the magnetic head 28 to read out and obtain the identification code from the magnetic recording medium of the card 40. This identification code is a unique code for identifying the card 40 and is recorded in advance on the magnetic recording medium. After obtaining the identification code, the card transport device 30 transports the card 40 to a predetermined position.

  Subsequently, in step 102, a predetermined imaging range 50 on the surface of the card 40 conveyed to a predetermined position is imaged with the light source 22 turned on and the card surface 40 illuminated (FIG. 4A). )reference.). The light irradiated from the light source 22 illuminates unevenness of the random pattern on the surface of the card 40, and a captured image including the random pattern is captured. In this case, an imaging range 50 is set in advance as an imaging range 50 excluding an edge portion and an end portion that are easily deformed when the conveyance roller 30a of the card conveyance device 30 contacts or the card 40 is used. After imaging by the camera 24, the light source 22 is turned off.

  In Step 104, a cell selection processing subroutine for selecting a plurality of small regions (cells) whose features are to be extracted from a captured image obtained by capturing the surface of the article to be registered is executed.

  FIG. 3 is a flowchart showing the flow of the cell selection processing subroutine. In step 200, as shown in FIGS. 4A and 4B, the captured image is divided into virtual small regions (cells) 51 of N × M pixels. Here, the size and area of each cell 51 are set to be several times or more the size and area necessary for feature amount extraction. In FIG. 4B, symbols (a, b,..., X) are added at the end to distinguish each cell 51.

  In step 202, as shown in the cell 51b of FIG. 4B, each of the cells 51a to 51x is further divided into n × m minute cells 52.

  In step 204, the density (lightness) of each divided microcell 52 is calculated.

  In step 206, the standard value (average value) of the density of each micro cell 52 is obtained for each cell 51, and the standard deviation of the micro cell 52 is obtained for each cell 51.

  In step 208, the cells 51 are sorted based on the standard deviation obtained for each cell 51.

  In step 210, a plurality of cells (as many as a predetermined number) are selected from the smaller standard deviation. FIG. 4C is a diagram showing the distribution of each sampled microcell 52 as a histogram with the horizontal axis representing the minimum to maximum density values for each cell 51. As shown in this density histogram, a cell 51 having a small density variation (small standard deviation) is adopted as a cell for extracting a feature amount, and conversely, a cell 51 having a large variation is not adopted.

  Since it is difficult to extract (recognize) a random pattern of unevenness from a dark-colored part such as a print-printed part on the surface of the article, the surface roughness (irregularity) is not an image change due to surface print-printing. ) Is selected as the cell 51 from which the feature amount is extracted, so that the feature amount is extracted from the minute image change due to (). Thereby, the cell 51 in which the random pattern is easily recognized can be selected.

  The number of cells selected as cells for extracting feature quantities is not particularly limited as long as it is two or more, and may be determined according to the material and size of the article to be registered and confirmed, or the article confirmation It can also be determined according to the performance of the device 10 and the required accuracy.

  In addition, here, the standard deviation of the density of the micro cell is obtained as representing the density variation, but not limited to this, the density range (distribution range) obtained from the maximum value and the minimum value of the density of each micro cell. Alternatively, the dispersion of the concentration of each minute cell may be obtained. It is also possible to create a density histogram as described above, and determine density variations from the degree of bias.

  After the execution of the cell selection processing subroutine is completed, the process proceeds to step 106, and an image of the central region 53 (see FIG. 4B) of the cell 51 selected in the cell selection processing subroutine is extracted as a feature image. The size of the region 53 and the position in the cell 51 are set in advance.

  In step 108, a subroutine for extracting a feature amount from the image data of the feature image of the selected cell 51 is executed. FIG. 5 is a flowchart showing the flow of processing of the feature amount extraction subroutine. Note that a conventionally known technique can be employed for extracting the feature amount. However, in this embodiment, Patent Document 1 (Japanese Patent Laid-Open No. 2004-153405) and Patent Document previously cited as prior art documents are used. 2 (Japanese Patent Laid-Open No. 2005-010881) is applied.

  First, in step 300, the captured feature image is divided (quantized) into meshes of appropriate size (number of meshes d = vertical M × horizontal N), and each mesh is represented by a certain density value (density level q). (Sampling) and converting into image data of a mosaic image.

  In step 302, the image data after quantization sampling is set to a vector (x = (x 1, x 2,... Xd) t (t represents transposition), where j j is the density of the mesh and x j is the density. (Feature vector). This feature vector is used as a feature amount.

  Each element of this feature vector gives the density of the corresponding image area. The obtained pattern is represented as one point on the feature space spanned by the feature vector. As described above, since each article has a random pattern that differs microscopically, the feature vector also represents a unique feature. That is, the feature of the random pattern of each article can be represented by this feature vector. Instead of the feature vector, a variance-covariance matrix or correlation matrix may be obtained from all the vectors, and this may be used as the feature quantity.

  After the feature quantity extraction subroutine at step 108 is completed, the routine proceeds to step 110, where it is determined whether or not feature quantities have been extracted from the feature images of all cells selected in the cell selection processing subroutine. If a negative determination is made here, the process returns to step 106, a feature image is extracted for the next cell, and the feature amount extraction subroutine of step 108 is executed using the feature image. These processes are repeated until feature values are extracted for all cells selected in the cell selection process subroutine. When feature amounts are extracted for all cells selected in the cell selection processing subroutine and an affirmative determination is made in step 110, the process proceeds to step 112. In step 112, the data indicating the feature amount extracted from the feature image of each cell and the position information indicating the position of each feature image are associated with each other, and the data and the position information indicating each associated feature amount are described above. The information is registered in the external device 34 in association with the identification code read from the magnetic recording medium of the card 40.

  Instead of the position information indicating the position of each feature image, position information indicating the position of each cell corresponding to each feature image may be registered in the external device 34. As described above, since each feature image is extracted from a predetermined region (position) in each cell, the position of each feature image can be easily derived from the position of each cell. Confirmation processing can be performed.

  Next, an operation for confirming the authenticity of the confirmation target card 40 will be described. FIG. 6 is a flowchart showing the flow of the confirmation processing program executed in the article confirmation apparatus 10.

  When the card 40 to be confirmed is inserted from an insertion port (not shown), the card transport device 30 transports the card 40 to the reading position of the magnetic head 28. At this time, the confirmation processing program shown in FIG. 6 is started.

  In step 400, the magnetic head controller 26 controls the magnetic head 28 to read out and acquire the identification code from the magnetic recording medium of the card 40. After obtaining the identification code, the card transport device 30 transports the card 40 to a predetermined position.

  In step 402, data indicating the feature amount registered in correspondence with the acquired identification code and position information are acquired from the external device 34.

  In step 404, as in the registration process, the light source 22 is turned on, the surface of the card 40 is illuminated, and a feature image in the same imaging range 50 as the imaging range in which the registration target card 40 is imaged in the registration process is captured by the camera 24. Take an image with.

  In the confirmation process, when the same imaging range as that in the registration process is imaged, accurate alignment is necessary. However, the alignment method is not particularly limited. The methods described in Patent Document 1 (Japanese Patent Application Laid-Open No. 2004-153405) and Patent Document 2 (Japanese Patent Application Laid-Open No. 2005-010881) mentioned above can be applied. Further description of this alignment method is omitted here.

  In step 406, 0 is set to k. k is a variable for counting the number of matches.

  In step 408, an image of the area indicated by the position information acquired from the external device 34 is extracted as a feature image from the captured image.

  In step 410, a subroutine for extracting feature amounts from the image data of the extracted feature images is executed in the same manner as described above.

  In step 412, the feature amount (confirmation-side feature) of the article to be confirmed extracted from the feature image at the same position as the feature image from which the registration-side feature amount is extracted (registration-side feature amount) acquired from the external device 34. Amount) is compared with each other to determine whether or not they match. Here, if the degree of similarity between the registration-side feature quantity and the confirmation-side feature quantity is equal to or greater than the threshold value, it is determined that they match. For calculating the degree of similarity, a conventionally known technique can be employed. However, in this embodiment, Patent Document 1 (Japanese Patent Laid-Open No. 2004-153405) and Patent Document previously cited as prior art documents are used. 2 (Japanese Patent Laid-Open No. 2005-010881) is applied to calculate the distance between the registration-side feature quantity (feature vector) and the confirmation-side feature quantity (feature vector) to obtain the similarity. The closer the calculated distance, the higher the similarity (they are similar). Naturally, the distance may be obtained by reducing the dimension of the feature space by KL (Karhunen-Loeve) expansion or the like.

  Here, as the distance to be calculated as the similarity, it is possible to use a distance used in statistical discriminant analysis or cluster analysis, for example, city distance, Euclidean distance, standardized Euclidean distance, Minkowski distance, Mahalanobis distance, etc. Yes, by Masakatsu Murakami: Behavioral metric series "Science of Shinji" Asakura Shoten, 1996). The former four distances are all obtained as the distance between the feature vector of the article to be confirmed and the registered feature vector of the genuine article. The Mahalanobis distance is calculated from the feature vector of the article to be confirmed, the feature vector (average vector) of the registered genuine article, and the inverse matrix of the feature matrix (variance covariance matrix or correlation matrix).

  Note that the threshold value to be compared with the similarity is preferably set with a predetermined allowable range in consideration of a quantization sampling error or the like. That is, the size of the threshold value may be selected as appropriate according to the request for strict similarity determination or sweetening. Further, the allowable range may be varied depending on the type of article to be registered (confirmed).

  If it is determined in step 412 that the registration-side feature value matches the confirmation-side feature value, k is incremented by 1 in step 414, and the process proceeds to step 416. If it is determined in step 412 that the registration-side feature quantity and the confirmation-side feature quantity do not match, the process of step 414 is skipped and the process proceeds to step 416.

  In step 416, it is determined whether or not the feature amounts of all feature images corresponding to the position indicated by the position information have been confirmed. If a negative determination is made here, the processing from steps 408 to 414 is repeated, a feature image is extracted according to the next position information, and a feature amount (confirmation-side feature amount) is extracted from the extracted feature image. Compare with the registration-side feature value corresponding to the position. These processes are repeated until an affirmative determination is made in step 416.

  When an affirmative determination is made at step 416, the routine proceeds to step 418, where it is determined whether or not k exceeds a majority. If it is determined that k exceeds a majority, the process proceeds to step 420, where it is determined that the card 40 to be checked is genuine (registered card) and is authenticated. On the other hand, if it is determined that k is less than a majority, the process proceeds to step 422, and the confirmation target card 40 is determined to be unauthenticated (fake).

  When the determination result is obtained in step 420 and step 422, the determination result may be transmitted to the external device 34. When the article confirmation apparatus 10 is provided with a display device, the display device The determination result may be displayed on the screen.

  As described above, in the registration process, the surface of the registration target card 40 is imaged, and a small region (cell) image in which fine features (random patterns) that cannot be reproduced are easily recognized from the captured image of the registration target. Select multiple feature images (in the above-mentioned image in the center of the cell) as feature images, extract feature values from the selected feature images, register them together with the location information of each feature image, The surface of the target card 40 is imaged, a feature image at the same position as each of the feature images registered by the registration process is extracted from the captured image of the confirmation target, and the extracted feature images are further extracted. Since each feature amount is extracted and compared with each feature amount of the registered card 40, authentication / non-authentication is determined, so that a feature image suitable for extracting the feature amount can be dynamically selected, Furthermore, Since the confirmation process is performed using several feature amounts (feature images), the authenticity of the article (the card 40 in the above) is compared with the case where the confirmation process is performed using only the feature amounts extracted from a single feature image. It is possible to prevent a decrease in accuracy in confirmation.

  In the above embodiment, an example has been described in which the same imaging range as in the registration process is captured in the confirmation process, and a plurality of feature images are extracted from the captured image according to the registered position information. For example, a plurality of cameras 24 are provided, and at the time of confirmation processing, only the camera 24 that captures the feature image at the position indicated by the registered position information among the plurality of cameras 24 is operated to operate the feature image. You may make it image. When a plurality of cameras 24 are provided, position information indicating the positions of the cameras 24 that captured the selected plurality of feature images may be registered in the external device 24 during the registration process.

  In the above embodiment, the example in which the imaging range 50 on the surface of the card 40 is imaged at a time by the single camera 24 has been described. However, the camera 24 is moved in each divided area when the imaging range 50 is divided. However, it is also possible to pick up images one by one and finally pick up the entire imaging range 50. In this case, in the confirmation process, the movement position of the camera 24 may be controlled so as to capture only the divided area including the feature image.

  Furthermore, in the above-described embodiment, the example in which only one surface of the card 40 to be registered and confirmed is captured by the camera 24 has been described. However, both surfaces are imaged, and feature images are selected and extracted from the captured images on both surfaces. May be. Thereby, compared with the case where a feature image is selected from one surface of the card 40, a feature image in which the feature is more easily recognized can be selected, and the accuracy of authenticity is improved. When the article to be registered and confirmed is not an article having a front and back such as a card but a polyhedron, each face is imaged and a preferred feature image is selected and extracted from a plurality of captured images. You may do it.

  Also, in the above embodiment, when comparing the registered feature quantity (feature vector) and the feature quantity to be confirmed (feature vector), the similarity is judged based on the distance and the two match. However, it may be determined from the angle formed by the registered feature vector and the feature vector to be confirmed.

  In the above-described embodiment, the example in which the feature amount is extracted and registered during the registration process has been described. However, the selected feature image is registered without extracting the feature amount during the registration process, and the feature image is registered during the confirmation process. The registered feature image may be read to extract the feature amount, and compared with the feature amount of the feature image to be confirmed.

  Furthermore, a feature image can be registered instead of the feature amount, and the similarity can be obtained by comparing the registered feature image with the feature image to be confirmed (image matching method). In addition to comparing feature images in real space for identification and collation, the feature images may be converted to the frequency domain by two-dimensional Fourier transform and compared in Fourier space. In this case, the feature image obtained from the card 40 registered in advance and the feature image obtained from the card 40 to be confirmed are synthesized in Fourier space, and the correlation intensity image is obtained by inverse Fourier transform, and the peak The similarity between two images can be evaluated from the value. For example, when the magnitude of the amplitude peak is equal to or greater than a predetermined threshold value, it can be determined that the registered feature quantity and the feature quantity to be confirmed are similar (matched).

  Further, instead of the feature amount, a captured image including each of the feature images may be registered. If the position information of a plurality of selected feature images (each cell) is registered corresponding to the captured image, a plurality of feature images are extracted from the captured image registered during the confirmation process based on the position information. Therefore, it is possible to compare the feature image to be registered or the feature amount extracted from the feature image with that to be confirmed.

  In the above embodiment, the example in which the registration target card 40 is registered by storing the data indicating the feature amount in the external device 34 has been described. However, the feature amount and the feature image are stored in the article confirmation device 10. A storage device may be provided and stored in this storage device.

  Furthermore, the article confirmation apparatus 10 is connected to the imaging apparatus having an imaging unit that illuminates the card 40 and captures a characteristic image, and the imaging apparatus is connected to the imaging apparatus via a communication line or a cable and is captured by the imaging apparatus. You may comprise with the processing apparatus which performs a registration process and a confirmation process using an image. Specifically, in the registration process, the imaging device images the surface of the registration target card 40, selects a plurality of feature images, and transmits the selected plurality of feature images and their position information to the processing device. In the confirmation process, a plurality of feature images of the card 40 to be confirmed are taken according to the position information received from the processing device 34, and the plurality of feature images are transmitted to the processing device. On the other hand, in the registration process, the processing device calculates a feature amount from the feature image of the registration target card 40 received from the imaging device in the same manner as in the above embodiment, and associates data indicating the feature amount with the position information in the database. And so on. In the confirmation process, a feature amount is calculated from a plurality of feature images of the confirmation target card 40 received from the imaging device in the same manner as in the above embodiment, and the calculation result and a feature amount of the card 40 registered in advance are calculated. And authenticity of the card 40 to be confirmed is determined. Further, the process up to extraction of the feature amount may be performed on the imaging apparatus side. Even with such a configuration, the same operational effects as the above-described embodiment can be obtained.

  Moreover, in the said embodiment, although the feature image was imaged using the camera 24, it is not limited to this, You may image a feature image using a scanner.

  Moreover, although the example which performs the process which selects the cell which is easy to recognize the characteristic for every card 40 was demonstrated, not only this but the process which selects a cell for every classification of the card | curd 40 previously The execution results are stored in a predetermined storage device, and during the registration process, a plurality of feature images are selected by reading out and using the processing results stored in advance for each type of the card 40. May be.

  Further, in the above-described embodiment, as an article for confirming authenticity, a card such as a credit card or a cash card has been described as an example. However, the present invention is not limited to this, and the article itself is not made intentionally. Any article that has a fine random pattern that is randomly created in the manufacturing process or after manufacturing, and that is not easily changed in shape without external force applied, is related to the present invention. The authenticity can be confirmed using the article confirmation apparatus and the article confirmation method.

1 is a schematic configuration diagram of an article confirmation device according to an embodiment of the present invention. It is a flowchart which shows the flow of a process of a registration process program. It is a flowchart which shows the flow of a process of a cell selection process subroutine. It is explanatory drawing explaining a cell selection process. It is a flowchart which shows the flow of a process of a feature-value extraction subroutine. It is a flowchart which shows the flow of a confirmation process program. It is explanatory drawing explaining the conventional document confirmation apparatus and the goods confirmation apparatus, (A) is a side view of goods, (B) is a top view of goods.

Explanation of symbols

10 article confirmation device 12 CPU
14 RAM
16 ROM
18 External I / F processing device 20 Internal I / F processing device 22 Light source 24 Camera 34 External device 40 Card 50 Imaging range 51 Cell 52 Small cell

Claims (5)

Imaging means for imaging the surface of the article;
When a reference article surface is picked up by the imaging means, a plurality of small area images from which the fine non-reproducible features of the reference article surface are easily recognized are selected as feature images from the picked-up picked-up image. A selection means;
The captured image of the reference article surface, the plurality of feature images selected by the selection unit, and the article surface included in each of the plurality of feature images selected by the selection unit cannot be reproduced. Registration means for registering any one of feature information indicating fine features and position information indicating each position of a plurality of feature images selected by the selection means;
When the feature image of each of the plurality of small regions corresponding to the position indicated by the registered position information on the surface of the article to be confirmed is captured by the imaging unit, each of the captured feature images And a comparison result of comparing each of the selected plurality of feature images of the registered reference article, or each of the features included in the plurality of captured feature images and the registered reference Determining means for determining whether or not the article to be confirmed is the registered reference article based on a comparison result with each of the features included in the selected plurality of feature images of the article;
An article confirmation apparatus including:   The selection means divides the captured image of the reference article surface into a plurality of small regions, and the density between the micro regions when each of the small regions is divided into a plurality of micro regions. The article confirmation apparatus according to claim 1, wherein a variation is obtained for each small region, and a plurality of small region images with small variations in density are selected as the feature images.   The imaging means captures a predetermined range including each of a plurality of small areas corresponding to the position indicated by the registered position information on the surface of the article to be confirmed, and the registration is performed from the captured image of the captured predetermined range. By extracting each image of the plurality of small regions corresponding to the position indicated by the position information as a feature image, a plurality of positions corresponding to the positions indicated by the registered position information on the surface of the article to be confirmed The article confirmation apparatus according to claim 1, wherein a feature image of each small region is captured.   The determination means, when the number of matches between each of the plurality of captured feature images of the confirmation target article surface and each of the plurality of feature images of the registered reference article exceeds a majority, Alternatively, the number of matches between each of the features included in the plurality of captured feature images of the surface of the article to be confirmed and each of the features included in the plurality of feature images of the registered reference article is a majority. The article confirmation apparatus according to any one of claims 1 to 3, wherein when the number exceeds, the article to be confirmed is determined to be the registered reference article. A first imaging step of imaging a reference article surface;
A selection step of selecting a plurality of small region images from which the fine features that cannot be reproduced on the reference article surface are easily recognized from the captured images captured in the first imaging step, as feature images;
The captured image captured in the first imaging step, the plurality of feature images selected in the selection step, and the fine non-reproducible surface of the article included in each of the plurality of feature images selected in the selection step A registration step of registering any one of feature information indicating a unique feature and location information indicating the location of each of the plurality of feature images selected in the selection step;
A second imaging step of imaging a feature image of each of a plurality of small regions corresponding to the position indicated by the registered position information on the surface of the article to be confirmed;
A comparison result of comparing each of the plurality of feature images captured in the second imaging step with each of the selected plurality of feature images of the registered reference article, or in the second imaging step Based on a comparison result between each of the features included in the plurality of captured feature images and each of the features included in the selected plurality of feature images of the registered reference article, the confirmation target A determination step of determining whether an article is the registered reference article;
An article confirmation method including:

Images