WO2014156430A1 - 真贋判定システム，特徴点登録装置およびその動作制御方法，ならびに照合判定装置およびその動作制御方法 - Google Patents

真贋判定システム，特徴点登録装置およびその動作制御方法，ならびに照合判定装置およびその動作制御方法 Download PDF

Info

Publication number
WO2014156430A1
WO2014156430A1 PCT/JP2014/054491 JP2014054491W WO2014156430A1 WO 2014156430 A1 WO2014156430 A1 WO 2014156430A1 JP 2014054491 W JP2014054491 W JP 2014054491W WO 2014156430 A1 WO2014156430 A1 WO 2014156430A1
Authority
WO
WIPO (PCT)
Prior art keywords
image
correlation value
product
feature point
feature points
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Ceased
Application number
PCT/JP2014/054491
Other languages
English (en)
French (fr)
Japanese (ja)
Inventor
徹郎 江畑
與那覇　誠
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Fujifilm Corp
Original Assignee
Fujifilm Corp
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Fujifilm Corp filed Critical Fujifilm Corp
Publication of WO2014156430A1 publication Critical patent/WO2014156430A1/ja
Priority to US14/857,496 priority Critical patent/US10083371B2/en
Anticipated expiration legal-status Critical
Ceased legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q30/00—Commerce
    • G06Q30/018—Certifying business or products
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/22—Matching criteria, e.g. proximity measures
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/0002—Inspection of images, e.g. flaw detection
    • G06T7/0004—Industrial image inspection
    • G06T7/001—Industrial image inspection using an image reference approach
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/40—Extraction of image or video features
    • G06V10/42—Global feature extraction by analysis of the whole pattern, e.g. using frequency domain transformations or autocorrelation
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
    • G06V10/74—Image or video pattern matching; Proximity measures in feature spaces
    • G06V10/761—Proximity, similarity or dissimilarity measures
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00—Scenes; Scene-specific elements
    • G06V20/60—Type of objects
    • G06V20/66—Trinkets, e.g. shirt buttons or jewellery items
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00—Scenes; Scene-specific elements
    • G06V20/80—Recognising image objects characterised by unique random patterns
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/30—Subject of image; Context of image processing
    • G06T2207/30108—Industrial image inspection
    • G06T2207/30128—Food products

Definitions

  • the present invention relates to an authenticity determination system, a feature point registration device and its operation control method, and a verification determination device and its operation control method.
  • Patent Documents 1 and 2 Use pattern matching to determine whether or not the judgment target tablet is genuine by searching for the same genuine tablet image as the judgment target tablet image from a number of pre-registered genuine tablet images Can do.
  • Patent Document 3 describes that feature points and feature amounts are extracted from an image file and a search image, and an image file having a feature amount that matches or is similar to the feature amount of the search image is described.
  • Japanese Patent Application Laid-Open No. 2004-228561 describes a method for calculating a correlation value between a reference image and a collation image and determining whether the paper document represented by the collation image is authentic.
  • JP 2005-258940 A Japanese Patent Laid-Open No. 9-178442 JP 2012-133484 A JP 2005-38389 A
  • This invention is intended to realize high-speed authentication.
  • the authenticity determination system includes an intrinsic feature point registration device and a verification determination device.
  • the genuine product feature point registration device includes a first correlation value calculating unit that calculates a correlation value between a partial image of a genuine product image and a template image, and the correlation value calculated by the first correlation value calculating unit is a first correlation value.
  • a genuine product feature point extracting means for extracting a plurality of feature points of the genuine product image that is equal to or greater than a threshold value, and a genuine product identification including the plurality of feature points of the genuine product image extracted by the genuine product feature point extracting means
  • Intrinsic product identification data storage means for storing data is provided.
  • the collation determining device includes a second correlation value calculating unit that calculates a correlation value between the partial image and the template image in the authenticity determination product image, and the correlation value calculated by the second correlation value calculating unit is a second threshold value.
  • the above-described authenticity determination feature point extracting means for extracting a plurality of feature points of the authenticity determination product image, and the geometric features of the plurality of feature points of the authenticity determination product image extracted by the authenticity determination feature point extraction means. Using the characteristics and the geometric characteristics of the plurality of feature points of the genuine product image stored in the genuine product identification data storage means of the feature point registration device, the authenticity judgment product image and the genuine product image Similarity calculation means for calculating the similarity is provided.
  • Authentic product feature point registration device is used to store (register) identification data (authentic product identification data) for the genuine product image.
  • Intrinsic product means a genuinely manufactured product (genuine or genuine product).
  • the genuine product identification data is acquired (created) by image processing using an genuine product image representing the genuine product, and includes a plurality of feature points (coordinate data) of the genuine product image.
  • a correlation value between a partial image in a partial region of the genuine product image and the template image is calculated, and a plurality of positions (coordinates) in the genuine product image where the calculated correlation value is equal to or greater than a first threshold value. Is determined as the feature point of the genuine product image.
  • the determined feature points are stored in the genuine product identification data storage means.
  • the collation determination device is used to extract a plurality of feature points (coordinates) for the authentication product image representing the authentication product.
  • the authenticity determination product means a determination target product for determining whether the product is the above-described genuine product or an illegally manufactured article (fake or non-authentic product).
  • a plurality of feature points extracted for the authenticity determined product image are also acquired (created) by image processing using the authenticity determined product image.
  • a correlation value between a partial image in a partial region of the authenticity determination product image and the template image is calculated, and a plurality of positions (in the authentication determination product image where the calculated correlation value is equal to or greater than a second threshold value ( Coordinate) is determined as a feature point of the authenticity determination product image.
  • the first threshold value and the second threshold value may be the same value or different values.
  • the similarity (a numerical value that quantitatively represents the degree of coincidence) between the authenticity determined product image and the genuine product image is calculated.
  • the geometric characteristics of the plurality of feature points include the interval between the plurality of feature points, the figure shape defined by connecting the plurality of feature points with straight lines, and the like.
  • the authenticity-determined product image is the same as or very similar to the authenticity product image, and it can be estimated that the authenticity-determined product is an authentic product. Conversely, if the calculated similarity is less than the predetermined value, it is estimated that the authenticity product image is not the same as the genuine product image (not similar), and the authenticity product is not an authentic product (a counterfeit product). can do.
  • the geometric characteristics of the plurality of feature points of the authenticity product image and the geometric characteristics of the multiple feature points of the authenticity product image are calculated without calculating the correlation value between the authenticity product image itself and the authentic product image itself.
  • the similarity between the authenticity product image and the authentic product image is calculated using the physical characteristics, and the authenticity of the authenticity determination product can be determined according to the calculated similarity, so that high-speed authentication can be performed. it can. Even if there are a large number of genuine product images, the authenticity determination can be completed quickly.
  • the same template image as that used for the genuine product image is also used for the authenticity determination product image.
  • the first correlation value calculation means included in the genuine product feature point registration device scans the template image on the genuine product image, and a plurality of the plurality of images corresponding to positions of the template image in the genuine product image.
  • the genuine product feature point registration device further generates means for generating two-dimensional correlation value data in which the plurality of correlation values are arranged in accordance with the position of the scanned template image.
  • the template image is scanned into a correlation value image represented by correlation value image data (luminance image data) using a correlation value in the correlation value two-dimensional array data as a luminance value, and the template image in the correlation value image is scanned.
  • a third correlation value calculating means for calculating a plurality of correlation values according to the position, wherein the genuine product feature point extracting means comprises the first correlation value calculating means.
  • the feature point of the genuine product image is extracted based on the correlation value calculated by the third correlation value calculating unit.
  • the template image is scanned with respect to the genuine product image to calculate a plurality of correlation values according to the position of the template image in the genuine product image, and the magnitude and distribution of the calculated correlation values (two-dimensional array) ) To generate a correlation value image.
  • the template image is scanned with respect to the generated correlation value image, and a plurality of correlation values corresponding to the position of the template image in the correlation value image are calculated.
  • the feature points of the genuine product image are emphasized in the correlation value image. Improve accuracy of similarity calculation between authenticity product image and authentic product image using geometric properties of multiple feature points of authenticity product image and geometric properties of multiple feature points of authentic product image can do.
  • the verification determination device scans the template image in the authenticity determination product image to calculate a plurality of correlation values according to the position of the template image in the authentication determination product image, and calculates the calculated plurality of correlation values.
  • a correlation value image using a correlation value as a luminance value is generated, and the template image is scanned in the generated correlation value image to calculate a plurality of correlation values according to the position of the template image in the correlation value image.
  • the correlation value image the feature points of the authenticity determination product image are emphasized.
  • the creation of the correlation value two-dimensional array data, generation of the correlation value image data, and calculation of the correlation value using the correlation value image and the template image may be repeated a plurality of times. It is possible to further emphasize the feature points of the genuine product image and the feature points of the authenticity determination product image.
  • the genuine product identification data stored in the genuine product identification data storage means of the feature point registration device includes data representing the genuine product image or a plurality of feature points of the genuine product image. Data representing a plurality of partial images may be included. By storing not only feature points but also data representing the genuine product image or data representing a plurality of partial images including each of the plurality of feature points of the genuine product image, the above-described authenticity judgment product image is stored. In addition to authenticity determination based on calculation of the similarity between the authenticity product image and the authenticity product image using the geometric characteristics of multiple feature points and the geometrical characteristics of multiple feature points of the authentic product image, Additional (more precise) authentication can be performed.
  • the collation determination device when the similarity calculated by the similarity calculation means is greater than or equal to a third threshold value, includes the partial image of the genuine product image and the authenticity determination product image.
  • a fifth correlation value calculating means for calculating a correlation value with the partial image.
  • data representing the genuine product image stored in the genuine product identification data storage means of the feature point registration device, or a plurality of feature points of the genuine product image Data representing a plurality of partial images including each of the above can be used.
  • the image data representing the authenticity determined product image may be given to the collation determining device when the correlation value is calculated by the fifth correlation value calculating means.
  • the correlation value between the genuine product image and the authenticity determined product image is calculated. That is, the similarity between the authenticity product image and the authenticity image using the geometric characteristics of the feature points of the authenticity product image and the geometrical characteristics of the feature points of the authenticity image described above. Only when it is determined that the genuine product image and the authenticity determination product image are relatively similar in the authenticity determination based on the calculation, the process proceeds to the calculation of the correlation value between the authentic product image and the authenticity determination product image. In particular, when there are a large number of genuine product images, since the correlation value between the genuine product image and the authenticity determination product image is not calculated brute force, high-speed authenticity determination is not greatly hindered. By calculating a correlation value between the genuine product image determined to be relatively similar to the authenticity determination product image and the authenticity determination product image, the accuracy of the authenticity determination can be improved.
  • the fifth correlation value calculating means included in the verification determination device scans a correlation value calculation region (scan window) in the genuine product image and calculates a correlation value in the authenticity determination product image. A region is scanned, and a correlation value between partial images at corresponding positions of the genuine product image and the authenticity determination product image in the correlation value calculation region is calculated. Correlation values are calculated for the entire genuine product image and the entire authenticity determination product image.
  • the fifth correlation value calculation means included in the verification determination device includes a plurality of feature points each including a plurality of feature points of the genuine product image included in the genuine product identification data for the genuine product image.
  • the correlation value between the partial images and the partial images of the authenticity determination product image at the position corresponding to the partial image is calculated.
  • the partial image of the genuine product image used for calculating the correlation value represents a plurality of partial images in which the genuine product identification data stored in the genuine product identification data storage unit includes each of the plurality of feature points of the genuine product image.
  • it may be a partial image represented by the plurality of partial image data itself, or the genuine product identification data stored in the genuine product identification data storage means represents the genuine product image. If it contains only data (image data representing the entire genuine product image), it may be created by extracting partial image data including each of a plurality of feature points from the data representing the genuine product image data. .
  • the partial image including the feature point of the genuine product image can be said to be a partial image suitable for calculating the correlation value between the genuine product image and the authenticity determination product.
  • the verification determination device stores a plurality of feature points of the authenticity determination product image extracted by the authenticity determination product feature point extraction unit and the authenticity product identification data storage unit of the authenticity product feature point registration device.
  • the authenticity-determined product image and the authenticity-determined product image are calculated based on a plurality of feature points of the genuine product image and the alignment parameters for eliminating the relative deviation between the authentic product image and the authenticity-determinated product image.
  • Alignment means for aligning the genuine product image is provided.
  • the correlation value between the authentic product image and the authenticity determination product can be calculated with high accuracy.
  • the present invention also provides a feature point registration device comprehensively defined as follows. That is, the feature point registration apparatus according to the present invention provides a first correlation value calculating means for calculating a correlation value between a partial image of a target image and a template image for each of a plurality of target images, and calculating the first correlation value.
  • a feature point extracting means for extracting a feature point of the target image having a correlation value calculated by the means equal to or greater than a first threshold, and a target image identification including the feature point of the target image extracted by the feature point extracting means
  • Identification data storage means for storing data for each of the plurality of target images is provided.
  • the first correlation value calculating means scans the template image on the target image and calculates a plurality of correlation values according to positions of the template image in the target image.
  • the registration device further includes means for creating correlation value two-dimensional array data in which the plurality of correlation values are arranged according to the position of the scanned template image, and the correlation value in the correlation value two-dimensional array data is used as a luminance value.
  • a second correlation value calculating unit that scans the template image on the correlation value image represented by the used correlation value image data and calculates a plurality of correlation values according to the position of the template image in the correlation value image;
  • the feature point extracting means extracts feature points of the target image based on the correlation value calculated by the first correlation value calculating means. Ete, it extracts the feature points on the target image based on the correlation value calculated by said second correlation value calculating means.
  • the emphasized feature points can be stored in the identification data storage means.
  • the present invention also provides a method for controlling the operation of the feature point registration apparatus described above.
  • the present invention also provides a collation determination device used in the above-described authenticity determination system and an operation control method thereof.
  • the authentic tablet identification data is shown. It is a flowchart which shows operation
  • the state of the normalized correlation calculation process is shown.
  • a template image is shown. The other template image is shown. The other template image is shown. The other template image is shown. The other template image is shown. The other template image is shown. The other template image is shown. The other template image is shown.
  • An authentic tablet image is shown.
  • a correlation value image is shown. A correlation value image is shown. A part of the correlation value image is enlarged and shown. It is a block diagram which shows the whole structure of a collation determination system.
  • a mode that a correlation value is made between a collation image and three similar intrinsic tablet images is shown. The mode of the normalized correlation calculation of a collation image and a similar intrinsic tablet image is shown.
  • Fig. 1 shows the overall structure of the tablet registration system.
  • the tablet registration system is a system for registering data for identifying each of the tablets 4T1, 4T2, 4T3... Manufactured by a pharmaceutical company (tablet drug manufacturing company). It is installed in a production line for producing 4T3. A large number of tablets 4T1, 4T2, 4T3,..., After registration of identification data by the tablet registration system are packaged and shipped.
  • the tablets 4T1, 4T2, 4T3... Are the same kind of tablets, and all are so-called genuine (regular) tablets (hereinafter referred to as true tablets).
  • the tablets 4T1, 4T2, 4T3,... Are the same type, but the fine irregularities on the surface are different for each of the tablets 4T1, 4T2, 4T3,.
  • the tablet registration system includes a registration device 1, an imaging device 2, and a storage device 3.
  • the registration device 1 is a computer system including a CPU, a memory, a communication device, and the like, and a program that causes the computer system to function as a registration device that constitutes a tablet registration system is installed. Functions as 1.
  • the imaging device 2 includes an imaging device (CCD, CMOS, etc.) that images the intrinsic tablets 4T1, 4T2, 4T3,... (These surfaces), and represents the intrinsic tablets 4T1, 4T2, 4T3,. Output image data.
  • Image data representing the authentic tablets 4T1, 4T2, 4T3,... Output from the image sensor 2 is input to the registration device 1.
  • the registration device 1 uses identification data unique to each of the intrinsic tablets 4T1, 4T2, 4T3,... Using the input image data of the authentic tablets 4T1, 4T2, 4T3,. (Hereinafter referred to as the true tablet identification data).
  • the created genuine tablet identification data is stored in the storage device 3.
  • FIG. 2 shows an example of the authentic tablet identification data 3A stored in the storage device 3.
  • Authentic tablet identification data 3A is created for each of a number of authentic tablets 4T1, 4T2, 4T3,.
  • Each of the row data of the authentic tablet identification data 3A shown in FIG. 2 is the intrinsic tablet identification data unique to each of the multiple authentic tablets 4T1, 4T2, 4T3,.
  • Authentic tablet identification data 3A created for each of the intrinsic tablets 4T1, 4T2, 4T3,... Is an intrinsic tablet image ID (registered image ID), data representing feature points, and an intrinsic tablet output by the imaging device 2.
  • Each of the intrinsic tablet image IDs is each of the intrinsic tablets 4T1, 4T2, 4T3,... (Each of the intrinsic tablet image data output by imaging the intrinsic tablets 4T1, 4T2, 4T3,... By the imaging device 2).
  • the registration number is sequentially assigned by the registration device 1.
  • a feature point is a set of x and y coordinates that specify the location (address) of a plurality of feature points in an intrinsic tablet image.
  • FIG. 3 and 4 show the operation of the registration device 1 of the tablet registration system, and show the flow of processing for creating the above-described genuine tablet identification data 3A and storing it in the storage device 3.
  • FIG. 3 and 4 show the operation of the registration device 1 of the tablet registration system, and show the flow of processing for creating the above-described genuine tablet identification data 3A and storing it in the storage device 3.
  • Image data representing the true tablet image output by the imaging device 2 is input to the registration device 1 (step 31).
  • the process proceeds to a normalized correlation calculation process (steps 33 to 35).
  • FIG. 5 shows the state of the normalized correlation calculation process, and shows the relationship between the intrinsic tablet image 10 and the correlation calculation area (window) S.
  • FIG. 6 shows an example of a template image (local filter) 11 used for normalized correlation calculation processing.
  • a normalized correlation value r between the partial image in the correlation calculation area S that is a part of the intrinsic tablet image 10 and the template image 11 is calculated.
  • genuine tablet image 10 and correlation calculation area S are both rectangular.
  • intrinsic tablet image 10 has a size of 128 pixels ⁇ 128 pixels
  • correlation calculation area S has 9 pixels ⁇ 9 pixels. Each has a size.
  • the template image 11 shown enlarged in FIG. 6 has the same size of 9 pixels ⁇ 9 pixels as the correlation calculation area S.
  • the correlation value r is calculated by performing normalized correlation calculation using the partial image in the correlation calculation area S extracted from the intrinsic tablet image 10 and the template image 11.
  • Various known algorithms such as NCC (Normalized Cross-Correlation), ZNCC (Zero-mean Normalized Cross-Correlation), etc. can be used for the normalized correlation calculation processing for calculating the correlation value r.
  • the correlation value may be calculated using SSD (Sum? Of? SquaredDifference) or SAD (Sum? Of? Absolute? Difference).
  • the correlation calculation area S is moved in the intrinsic tablet image 10 by a predetermined distance (for example, one pixel) in the horizontal direction and the vertical direction, and between the partial image in the correlation calculation area S and the template image 11 each time the movement is performed.
  • a correlation value r is calculated.
  • the template image 11 shown in FIG. 6 is based on a two-dimensional normal distribution and has the highest luminance at the center, and the luminance gradually decreases concentrically as the distance from the center increases.
  • a correlation value r that is robust to rotation can be obtained.
  • a large correlation value r is calculated for a partial image with high luminance
  • a small correlation value r is calculated for a partial image with low luminance.
  • the calculated normalized correlation value r has a value in the range of ⁇ 1 to +1.
  • the template image 11a shown in FIG. 7 is based on a two-dimensional normal distribution and has the lowest luminance at the center, and the luminance gradually increases concentrically as the distance from the center increases.
  • a correlation value r robust to rotation is calculated.
  • a large correlation value r is calculated for a partial image with low luminance
  • a small correlation value r is calculated for a partial image with high luminance.
  • the luminance of any one of the four corners of the rectangle is different from the luminance of the remaining three corners.
  • the correlation calculation area S moves to the start point at the upper left corner of the input true tablet image 10 (see step 33, FIG. 5).
  • Data representing a partial image in the correlation calculation area S is extracted (step 34).
  • a normalized correlation operation is performed between the extracted partial image and the above-described template image (here, the template image 11 shown in FIG. 6 is used) to calculate a correlation value r (step 35).
  • step 36 It is determined whether or not the correlation calculation area S has reached the end point (lower right corner of the true tablet image 10) (step 36). If the end point has not been reached, the correlation calculation area S moves horizontally or vertically (scanning), and a partial image in the correlation calculation area S after the movement is extracted (NO in step 36, step 34). ). A correlation value r between the new extracted partial image and the template image 11 is calculated (step 35).
  • a two-dimensional array table storing a large number of calculated correlation values r is created (YES in step 36, step 38).
  • An array (row direction and column direction) of a large number of correlation values r in the two-dimensional array table corresponds to the position in the intrinsic tablet image 10 of the correlation calculation area S described above.
  • luminance image Data
  • luminance image data
  • the correlation value r has a numerical value in the range of ⁇ 1 to +1.
  • the smallest correlation value r among a large number of correlation values r stored in the two-dimensional array table is made to correspond to the luminance value 0
  • the largest correlation value r is made to correspond to the luminance value 255.
  • Correlation value images are created with 256 levels of brightness.
  • the correlation value r stored in the above-described two-dimensional array table is expressed in advance by 8-bit (0 to 255) data, the two-dimensional array table can be used as it is as correlation value image data.
  • FIG. 12 shows the authentic tablet image
  • FIG. 13 shows the correlation value image 10a created by passing the above-described normalized correlation operation once for the genuine tablet image 10 shown in FIG. 12, and
  • FIG. A correlation value image 10b obtained by performing the normalized correlation calculation four times is shown.
  • An image 10b is created.
  • the location (coordinates) of a pixel having a luminance value equal to or greater than a predetermined threshold using the created correlation value image is determined and extracted as a feature point in intrinsic tablet image 10 (step 42).
  • a plurality of feature points are extracted.
  • FIG. 15 shows a partially enlarged image 10b1 of the above-described correlation value image 10b (see FIG. 14).
  • the feature point extraction process when a plurality of pixels having a luminance equal to or higher than a predetermined threshold are set (adjacent), one set of feature points (coordinates) may be associated with the set pixel. In this case, adjacent pixels having a luminance equal to or higher than a predetermined threshold are grouped.
  • FIG. 15 shows three grouped collective pixel areas, one of which is denoted by reference numeral G1. For example, the coordinates of the center of gravity g1 of the collective pixel region G1 are treated as feature points for the collective pixel region G1. Instead of the center of gravity, the circumscribed rectangle of the collective pixel area G1 or the coordinates of the center of the inscribed rectangle may be used as the feature point of the collective pixel area G1.
  • the plurality of extracted feature points are associated with an intrinsic tablet image ID and image data, thereby producing intrinsic tablet identification data 3A (FIG. 2) including the intrinsic tablet image ID, feature points, and image data.
  • the created genuine tablet identification data 3A is stored in the storage device 3 as described above (step 43). By executing the same processing for each of a large number of genuine tablets 4T1, 4T2, 4T3,..., Each of the genuine tablets 4T1, 4T2, 4T3,. Registered in
  • the correlation value r is calculated using each of a plurality of (for example, two) template images of the plurality of types of template images 11, 11a to e, and feature points are extracted in the same manner as the above-described processing. You may do it. It is possible to increase the feature points extracted for one true tablet image. Further, even when one type of template image (for example, template image 11) is used, it is extracted from one intrinsic tablet image by distinguishing and handling each sign ( ⁇ or +) of the calculated correlation value r. Feature points can be increased.
  • FIG. 16 is a block diagram showing the overall configuration of the collation determination system.
  • the verification determination system uses the genuine tablet identification data 3A stored in the storage device 3 of the tablet registration system described above to determine whether the brought-in tablet 4D is an authentic tablet or not an authentic tablet (whether it is a counterfeit tablet). System for collating and judging.
  • the collation determination system includes a collation determination device 6, an imaging device 5, and a storage device 3.
  • the collation determination device 6 is a computer system including a CPU, a memory, a communication device, and the like, and the computer system performs collation determination by executing a program that causes the computer system to function as the collation determination device 6. It functions as a collation determination device 6 constituting the system.
  • the storage device 3 is shown as the same storage device as the storage device 3 constituting the above-described tablet registration system.
  • the storage device 3 of the tablet registration system is included in the verification determination system by connecting the storage device 3 of the tablet registration system to the verification determination device 6 constituting the verification determination system via a network (such as the Internet). Can do.
  • the authentic tablet identification data 3A stored in the storage device 3 of the tablet registration system may be downloaded (copied) and stored in a storage device included in the verification determination system.
  • FIG. 17 to 19 show the operation of the collation determination device 6 of the collation determination system. Processing for determining whether the tablet 4D is an authentic product or not an authentic product (whether it is a counterfeit product). The flow is shown.
  • the first determination process is a feature point included in the intrinsic tablet identification data 3A for each of a number of intrinsic tablets 4T1, 4T2, 4T3,... Stored in the storage device 3 in advance in the tablet registration system (see FIG. 2). ), An image similar to the image of the tablet 4D to be authenticated is output by being imaged by the imaging device 5 of the collation determination system, and the intrinsic tablet 4T1, 4T2, 4T3,. .. is a process for determining whether or not the image is present in the intrinsic tablet image obtained by imaging.
  • the process determines that the tablet 4D is an authentic tablet.
  • the tablet 4D to be authenticated is imaged by the imaging device 5, and data representing an image representing the tablet 4D (hereinafter referred to as a collation image 20) is output from the imaging device 5.
  • Data representing the verification image 20 is input to the verification determination device 6 (step 51).
  • step 52 The same processing as that of the registration device 1 of the tablet registration system described above is performed on the collation image 20, and feature points (coordinates) (see FIG. 2) of the collation image 20 are extracted (step 52).
  • a plurality of feature points are also extracted from the collation image 20.
  • the normalized correlation operation of the same algorithm as that of the normalized correlation operation executed in the tablet registration system is also executed in the collation determination system.
  • the same template image as that used in the tablet registration system is also used in the collation determination system.
  • a plurality of feature points of the intrinsic tablet image i read from the storage device 3 and a plurality of feature points extracted for the collation image 20 are used, and the similarity between the intrinsic tablet image i and the collation image 20 is used. Is calculated (step 55).
  • a geometric hashing method can be used to calculate the similarity between two images using the positions (coordinates) of a plurality of feature points.
  • two target models here, geometric representations that are invariant to translation, expansion (reduction) and rotation
  • multiple feature points coordinates
  • the LLHA method (LocallyLLikely Arrangement Hashing) may be used instead of the geometric hashing method. Also by the LLHA method, a numerical value representing the similarity between two images is calculated based on the geometric characteristics of a plurality of feature points (coordinates) of the two images. In geometric hashing and LLHA, the positional deviation relationship, enlargement (reduction) relationship, and rotation angle relationship between the true tablet image i and the collation image 20 are also detected in the similarity calculation process. That is, in geometric hashing and LLHA, deviation amounts between a plurality of feature points of the intrinsic tablet image i and a plurality of feature points of the matching image 20 corresponding thereto are obtained. The alignment parameters (movement parameter, enlargement / reduction parameter, rotation parameter) for eliminating the relative deviation are also calculated together with the similarity.
  • step 56 It is determined whether the calculated similarity is greater than or equal to a predetermined threshold (step 56). If the degree of similarity between genuine tablet image i and collation image 20 is less than a predetermined threshold (NO in step 56), it is determined that true tablet image i is not similar to collation image 20 (different), and counter i is incremented. (NO in step 56, step 57). The feature points (coordinates) of the next true tablet image i stored in the storage device 3 are read out, and the above-mentioned similarity is calculated again (steps 54 and 55).
  • the genuine tablet image ID (see FIG. 2) of the genuine tablet image i is temporarily stored in the memory of the collation determination device 6. (Step 58).
  • step 59 Whether or not the counter i matches the total number of data M of the authentic tablet identification data 3A stored in the storage device 3, that is, whether or not the calculation of the similarity between all the authentic tablet images i and the collation images 20 has been completed. Judgment is made (step 59). If there is an intrinsic tablet image i for which similarity calculation has not been performed, the counter i is incremented (NO in step 59, step 57), the feature point of the next intrinsic tablet image i and the feature of the collation image 20 Similarity calculation using points is performed (step 55).
  • step 58 the true tablet image ID of the true tablet image i whose similarity with the collation image 20 is equal to or greater than a predetermined threshold is sequentially stored in the memory.
  • the process proceeds to the second determination process.
  • a plurality of genuine tablet image IDs are stored in the memory at step 58, that is, when a plurality (J) of genuine tablet images similar to the collation image 20 (hereinafter referred to as similar genuine tablet images j) are found. Will be explained.
  • the collation image 20 is positioned according to the positional deviation relationship, the enlargement / reduction relationship, and the rotation angle relationship (positioning parameters (movement parameters, enlargement / reduction parameters, rotation parameters)) detected in the above-described similarity calculation (step 55). (Parallel) movement, enlargement or reduction, and rotation are performed, whereby the similar intrinsic tablet image j and the collation image 20 are aligned (shift correction) (step 62). Instead of the collation image 20, the similar intrinsic tablet image j may be aligned.
  • step 63 to 65 Proceed to normalized correlation calculation (steps 63 to 65).
  • the above-described template image (local filter) 11 (FIG. 6) is not used, but the similar intrinsic tablet image j and the collation image 20 are used.
  • the correlation value r between the collation image 20 and a plurality of similar genuine tablet images (three similar genuine tablet images 10A, 10B, and 10C are shown in FIG. 20) will be described below. Will be calculated.
  • FIG. 21 shows a state of normalized correlation calculation between the similar intrinsic tablet image 10A and the collation image 20.
  • Correlation calculation areas (windows) S1 and S2 are set in the similar genuine tablet image 10A and the collation image 20, respectively, and partial images 10P and 20P in the correlation calculation areas S1 and S2 are extracted, respectively. Normalized correlation calculation is performed between the extracted partial images 10P and 20P.
  • any of the above-described known algorithms such as NCC and ZNCC can be used.
  • the correlation calculation areas S1 and S2 are moved to the start point in each of the similar genuine tablet image 10A and the collation image 20 (step 63). Partial images 10P and 20P in the normalized correlation calculation areas S1 and S2 are extracted (step 64). A correlation value between the partial image 10P in the correlation calculation area S1 of the similar intrinsic tablet image 10A and the partial image 20P in the correlation calculation area S2 of the collation image 20 is calculated (step 65).
  • step 66 It is determined whether or not the correlation calculation areas S1 and S2 are located at the end points (step 66). If the end point has not been reached (NO in step 66), the correlation calculation areas S1 and S2 are moved horizontally or vertically (step 67), and the correlation value between the partial images is calculated again (steps 64 and 65).
  • the counter j is incremented to proceed to the calculation of the correlation value between the next similar genuine tablet image j and the collation image 20 (YES in step 66, NO in step 68, step 69, steps 61 to 65). .
  • a number of correlation values respectively calculated between the plurality of similar genuine tablet images j and collation images 20 Is calculated (step 71), and it is determined whether the average correlation value having the largest value is greater than or equal to a predetermined threshold value (step 72).
  • the similar genuine tablet image for example, the similar genuine tablet image 10B shown in FIG. 20
  • the authenticity determination target tablet 4D used to capture the verification image 20 is determined to be an authentic tablet (YES in step 72, step 73). For example, a determination result indicating that the tablet 4D is an authentic product is displayed on the display screen of a display device connected to the verification determination device 6.
  • the true tablet image most similar to the collation image 20 among the many true tablet images stored in the storage device 3 is not the same as the collation image 20
  • the genuine tablet image that is the same as the collation image 20 is not stored in the storage device 3, and therefore it is determined that the authenticity determination target tablet 4D used for imaging is not an authentic tablet (a forged tablet) (step). NO at 72, step 74). For example, a warning that the tablet 4D is a counterfeit is displayed on the display screen of the display device connected to the verification determination device 6.
  • the correlation value (average correlation value) between the similar genuine tablet image j and the verification image 20 is calculated using the entire similar genuine tablet image j and the entire verification image 20 (that is, overall pattern verification).
  • the partial image including the feature point of the similar genuine tablet image j is correlated with the similar authentic tablet image j and the collation image 20 by using the feature point of the similar authentic tablet image j (see FIG. 2).
  • the correlation value is calculated only between, for example, the partial image centered on the feature point of the similar intrinsic tablet image j and the partial image of the collation image 20 at the position corresponding thereto.
  • the partial image centered on the feature point of the similar authentic tablet image j is the correlation value of the similar authentic tablet image j and the collation image 20. It can be said that it is a partial image suitable for calculation.
  • the accuracy of the correlation value calculation is almost sacrificed. Therefore, the time for calculating the correlation value between the similar genuine tablet image j and the collation image 20 can be shortened.
  • the authentic tablet identification data 3A of the storage device 3 not only image data representing the whole authentic tablet but only data representing a partial image centered on the feature point may be stored.
  • Registration device (correlation value calculation means, feature point extraction means, two-dimensional array data creation means) 2,5 Imaging device 3 Storage device 3A Genuine tablet identification data 4T1, 4T2, 4T3 Genuine tablet (genuine product) 4D tablet (authentication product) 6 Collation determination device (correlation value calculation means, feature point extraction means, similarity calculation means, alignment means) 10, 10A, 10B, 10C True tablet image 10a, 10b Correlation value image 11, 11a, 11b, 11c, 11d, 11d, 11e Template images

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Multimedia (AREA)
  • Business, Economics & Management (AREA)
  • Evolutionary Computation (AREA)
  • Artificial Intelligence (AREA)
  • Marketing (AREA)
  • General Business, Economics & Management (AREA)
  • Strategic Management (AREA)
  • Finance (AREA)
  • Economics (AREA)
  • Development Economics (AREA)
  • Accounting & Taxation (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Quality & Reliability (AREA)
  • Computing Systems (AREA)
  • Databases & Information Systems (AREA)
  • Health & Medical Sciences (AREA)
  • General Health & Medical Sciences (AREA)
  • Medical Informatics (AREA)
  • Software Systems (AREA)
  • Data Mining & Analysis (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • General Engineering & Computer Science (AREA)
  • Evolutionary Biology (AREA)
  • Image Analysis (AREA)
  • Medical Treatment And Welfare Office Work (AREA)
  • Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
PCT/JP2014/054491 2013-03-26 2014-02-25 真贋判定システム，特徴点登録装置およびその動作制御方法，ならびに照合判定装置およびその動作制御方法 Ceased WO2014156430A1 (ja)

Priority Applications (1)

Application Number Priority Date Filing Date Title
US14/857,496 US10083371B2 (en) 2013-03-26 2015-09-17 Authenticity determination system, feature point registration apparatus and method of controlling operation of same, and matching determination apparatus and method of controlling operation of same

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
JP2013063274A JP6063315B2 (ja) 2013-03-26 2013-03-26 真贋判定システム，特徴点登録装置およびその動作制御方法，ならびに照合判定装置およびその動作制御方法
JP2013-063274 2013-03-26

Related Child Applications (1)

Application Number Title Priority Date Filing Date
US14/857,496 Continuation US10083371B2 (en) 2013-03-26 2015-09-17 Authenticity determination system, feature point registration apparatus and method of controlling operation of same, and matching determination apparatus and method of controlling operation of same

Publications (1)

Publication Number Publication Date
WO2014156430A1 true WO2014156430A1 (ja) 2014-10-02

Family

ID=51623427

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/JP2014/054491 Ceased WO2014156430A1 (ja) 2013-03-26 2014-02-25 真贋判定システム，特徴点登録装置およびその動作制御方法，ならびに照合判定装置およびその動作制御方法

Country Status (3)

Country Link
US (1) US10083371B2 (enExample)
JP (1) JP6063315B2 (enExample)
WO (1) WO2014156430A1 (enExample)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108009560A (zh) * 2016-11-02 2018-05-08 广州图普网络科技有限公司 商品图像相似类别判定方法及装置
EP3242269A4 (en) * 2014-12-31 2019-01-02 Sang Rae Park Image analysis method and apparatus, and computer readable device

Families Citing this family (11)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2016186678A (ja) * 2015-03-27 2016-10-27 セイコーエプソン株式会社 インタラクティブプロジェクターおよびインタラクティブプロジェクターの制御方法
CN106486458B (zh) * 2015-08-31 2019-03-15 台达电子企业管理（上海）有限公司 多功率芯片的功率封装模块及功率芯片单元的制造方法
JP6587210B2 (ja) * 2015-11-18 2019-10-09 富士ゼロックス株式会社 印刷装置およびプログラム
CN105574873B (zh) * 2015-12-16 2018-05-22 西安空间无线电技术研究所 一种数量可控的图像特征点检测方法
US9754397B1 (en) 2017-04-07 2017-09-05 Mirage Worlds, Inc. Systems and methods for contextual augmented reality sharing and performance
US10504288B2 (en) 2018-04-17 2019-12-10 Patrick Piemonte & Ryan Staake Systems and methods for shared creation of augmented reality
CN114667431B (zh) * 2019-11-15 2025-10-28 索尼集团公司 信息处理装置和信息处理方法
CN111027573B (zh) * 2019-12-09 2023-05-12 广西荔玉高速公路有限公司 基于盲取证技术的图像真伪鉴别方法
CN112767337B (zh) * 2021-01-12 2023-08-04 杭州海康威视数字技术股份有限公司 一种烟盒检测方法和装置
CN112507315B (zh) * 2021-02-05 2021-06-18 红石阳光（北京）科技股份有限公司 基于智慧大脑的人员通行检测系统
US12462434B2 (en) * 2021-11-03 2025-11-04 Adobe Inc. Autonomous aesthetics-aware detection of counterfeit products

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JPH04324583A (ja) * 1991-04-24 1992-11-13 Fujitsu Ltd ムービング・ウインドウ型指紋画像照合方法及び照合装置
JPH0944676A (ja) * 1995-08-01 1997-02-14 Toyota Motor Corp 顔面検出装置
JP2006059282A (ja) * 2004-08-24 2006-03-02 Fujitsu Ltd 生体特徴情報に基づく認証方法
WO2012121166A1 (ja) * 2011-03-04 2012-09-13 日本電気株式会社 流通管理システム、流通管理方法及びそれらに用いられる装置、ラベル、プログラム

Family Cites Families (21)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP3417178B2 (ja) 1995-12-22 2003-06-16 松下電工株式会社 回路パターンの検査方法及びその検査装置
US6539106B1 (en) * 1999-01-08 2003-03-25 Applied Materials, Inc. Feature-based defect detection
JP4218171B2 (ja) * 2000-02-29 2009-02-04 株式会社日立製作所 走査電子顕微鏡，マッチング方法、及びプログラムを記録したコンピュータ読み取り可能な記録媒体
EP1282021A4 (en) * 2000-05-12 2006-02-15 Ishida Seisakusho PRODUCTION MANAGEMENT SYSTEM AND SYSTEM FOR VERIFYING OPERATING STATES OF PRODUCTION DEVICES
US7162073B1 (en) * 2001-11-30 2007-01-09 Cognex Technology And Investment Corporation Methods and apparatuses for detecting classifying and measuring spot defects in an image of an object
JP4103826B2 (ja) 2003-06-24 2008-06-18 富士ゼロックス株式会社 真偽判定方法、装置及びプログラム
US7263243B2 (en) * 2003-12-29 2007-08-28 Carestream Health, Inc. Method of image registration using mutual information
JP4242796B2 (ja) 2004-03-12 2009-03-25 パナソニック株式会社 画像認識方法及び画像認識装置
GB0408328D0 (en) * 2004-04-14 2004-05-19 Imp College Innovations Ltd Method of processing image data
EP1768063B8 (en) * 2004-07-13 2010-06-02 Glory Ltd. Image checking device, image checking method, and image checking program
EP1993072A1 (en) * 2007-05-18 2008-11-19 Siemens Aktiengesellschaft Method for comparison of 3D computer model and as-built situation of an industrial plant
JP5367244B2 (ja) * 2007-09-13 2013-12-11 株式会社東芝 目標検出装置および目標検出方法
JP2009216342A (ja) * 2008-03-12 2009-09-24 Mitsubishi Materials Corp 溶融ロータリーキルンの操業方法及び溶融ロータリーキルンの制御装置
JP4926116B2 (ja) * 2008-04-16 2012-05-09 株式会社日立ハイテクノロジーズ 画像検査装置
GB0913930D0 (en) * 2009-08-07 2009-09-16 Ucl Business Plc Apparatus and method for registering two medical images
JP5919296B2 (ja) * 2010-12-02 2016-05-18 コーニンクレッカ フィリップス エヌ ヴェＫｏｎｉｎｋｌｉｊｋｅ Ｐｈｉｌｉｐｓ Ｎ．Ｖ． イメージレジストレーション装置
JP5417312B2 (ja) 2010-12-20 2014-02-12 日本電信電話株式会社 画像検索装置、および画像検索プログラム
US9235766B2 (en) * 2011-10-20 2016-01-12 International Business Machines Corporation Optimizing the detection of objects in images
JP5937878B2 (ja) * 2012-04-24 2016-06-22 株式会社日立ハイテクノロジーズ パターンマッチング方法及び装置
JP5941782B2 (ja) * 2012-07-27 2016-06-29 株式会社日立ハイテクノロジーズ マッチング処理装置、マッチング処理方法、及びそれを用いた検査装置
US9122950B2 (en) * 2013-03-01 2015-09-01 Impac Medical Systems, Inc. Method and apparatus for learning-enhanced atlas-based auto-segmentation

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JPH04324583A (ja) * 1991-04-24 1992-11-13 Fujitsu Ltd ムービング・ウインドウ型指紋画像照合方法及び照合装置
JPH0944676A (ja) * 1995-08-01 1997-02-14 Toyota Motor Corp 顔面検出装置
JP2006059282A (ja) * 2004-08-24 2006-03-02 Fujitsu Ltd 生体特徴情報に基づく認証方法
WO2012121166A1 (ja) * 2011-03-04 2012-09-13 日本電気株式会社 流通管理システム、流通管理方法及びそれらに用いられる装置、ラベル、プログラム

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
EP3242269A4 (en) * 2014-12-31 2019-01-02 Sang Rae Park Image analysis method and apparatus, and computer readable device
CN108009560A (zh) * 2016-11-02 2018-05-08 广州图普网络科技有限公司 商品图像相似类别判定方法及装置
CN108009560B (zh) * 2016-11-02 2021-05-11 广州图普网络科技有限公司 商品图像相似类别判定方法及装置

Also Published As

Publication number Publication date
US20160004934A1 (en) 2016-01-07
US10083371B2 (en) 2018-09-25
JP2014191361A (ja) 2014-10-06
JP6063315B2 (ja) 2017-01-18

Similar Documents

Publication Publication Date Title
JP6063315B2 (ja) 真贋判定システム，特徴点登録装置およびその動作制御方法，ならびに照合判定装置およびその動作制御方法
CN109977886B (zh) 货架空置率计算方法及装置、电子设备、存储介质
JP5997120B2 (ja) 薬剤照合装置、薬剤照合システム、及び薬剤照合方法
CN110765992B (zh) 印章鉴别方法、介质、设备及装置
CN106778586A (zh) 离线手写签名鉴别方法及系统
CN102799850A (zh) 一种条形码识别方法和装置
US12067796B2 (en) Method for detecting fraud in documents
JP7232222B2 (ja) 情報処理システム
JP2015041164A (ja) 画像処理装置、画像処理方法およびプログラム
CN109409158A (zh) 一种基于二维码边缘粗糙度的防伪方法
JP5919212B2 (ja) 目視照合支援装置およびその制御方法
CN112017352B (zh) 证件鉴伪方法、装置、设备及可读存储介质
CN112036232A (zh) 一种图像表格结构识别方法、系统、终端以及存储介质
JP5669572B2 (ja) 指紋認証装置
CN118247226B (zh) 基于人工智能的危险品识别方法、装置、设备及存储介质
US10915772B2 (en) Apparatus and method for registering face pose for face recognition
JP2013061844A (ja) 真贋検証装置、真贋検証対象物、真贋検証方法、プログラム及び記憶媒体
Roy et al. Offline signature verification using grid based and centroid based approach
JP5217917B2 (ja) 物体検知追跡装置，物体検知追跡方法および物体検知追跡プログラム
Amerini et al. Copy-move forgery detection from printed images
JP2021157537A (ja) 二次元マーカの認識装置、方法、プログラム及びシステム
JP6890849B2 (ja) 情報処理システム
JP5810561B2 (ja) 個体識別装置、個体識別方法、及びプログラム
JP6204634B1 (ja) 形状弁別装置、形状弁別方法及び形状弁別プログラム
JP2003281593A (ja) 硬貨識別装置と硬貨登録装置と硬貨判定装置、並びに硬貨判定方法

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 14775803

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 14775803

Country of ref document: EP

Kind code of ref document: A1