WO2021179157A1 - 用于验证产品真伪的方法和装置 - Google Patents

用于验证产品真伪的方法和装置 Download PDF

Info

Publication number
WO2021179157A1
WO2021179157A1 PCT/CN2020/078556 CN2020078556W WO2021179157A1 WO 2021179157 A1 WO2021179157 A1 WO 2021179157A1 CN 2020078556 W CN2020078556 W CN 2020078556W WO 2021179157 A1 WO2021179157 A1 WO 2021179157A1
Authority
WO
WIPO (PCT)
Prior art keywords
product
image
micro
line segment
feature
Prior art date
Application number
PCT/CN2020/078556
Other languages
English (en)
French (fr)
Inventor
高煜
谢晖
杨莞琳
Original Assignee
罗伯特·博世有限公司
高煜
谢晖
杨莞琳
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 罗伯特·博世有限公司, 高煜, 谢晖, 杨莞琳 filed Critical 罗伯特·博世有限公司
Priority to DE112020005864.6T priority Critical patent/DE112020005864T5/de
Priority to CN202080098278.0A priority patent/CN115244542A/zh
Priority to PCT/CN2020/078556 priority patent/WO2021179157A1/zh
Publication of WO2021179157A1 publication Critical patent/WO2021179157A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06K—GRAPHICAL DATA READING; PRESENTATION OF DATA; RECORD CARRIERS; HANDLING RECORD CARRIERS
    • G06K19/00—Record carriers for use with machines and with at least a part designed to carry digital markings
    • G06K19/06—Record carriers for use with machines and with at least a part designed to carry digital markings characterised by the kind of the digital marking, e.g. shape, nature, code
    • G06K19/06009—Record carriers for use with machines and with at least a part designed to carry digital markings characterised by the kind of the digital marking, e.g. shape, nature, code with optically detectable marking
    • G06K19/06046—Constructional details
    • G06K19/06178—Constructional details the marking having a feature size being smaller than can be seen by the unaided human eye
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06K—GRAPHICAL DATA READING; PRESENTATION OF DATA; RECORD CARRIERS; HANDLING RECORD CARRIERS
    • G06K19/00—Record carriers for use with machines and with at least a part designed to carry digital markings
    • G06K19/06—Record carriers for use with machines and with at least a part designed to carry digital markings characterised by the kind of the digital marking, e.g. shape, nature, code
    • G06K19/08—Record carriers for use with machines and with at least a part designed to carry digital markings characterised by the kind of the digital marking, e.g. shape, nature, code using markings of different kinds or more than one marking of the same kind in the same record carrier, e.g. one marking being sensed by optical and the other by magnetic means
    • G06K19/083—Constructional details
    • G06K19/086—Constructional details with markings consisting of randomly placed or oriented elements, the randomness of the elements being useable for generating a unique identifying signature of the record carrier, e.g. randomly placed magnetic fibers or magnetic particles in the body of a credit card

Definitions

  • the present invention relates to a method and device for verifying the authenticity of a product, in particular to a method and device for verifying the authenticity of a product identification based on the micro-point feature, line segment feature and/or image quality score of the product identification image.
  • the existing anti-counterfeiting technologies for products include digital anti-counterfeiting technology and texture anti-counterfeiting technology.
  • Digital anti-counterfeiting technology uses barcodes or two-dimensional codes to give products a unique identification (ID) for anti-counterfeiting verification and traceability functions. This digital anti-counterfeiting technology is easy to be copied and has poor security.
  • Texture anti-counterfeiting technology uses randomly generated natural textures as anti-counterfeiting features. This texture is physically non-reproducible and has non-reproducible features; however, the existing texture anti-counterfeiting technology lacks automatic identification capabilities for anti-counterfeiting features, and automatic identification capabilities require anti-counterfeiting Features are visually identifiable, or rely on the addition of fiber materials in the production process to form anti-counterfeiting features, which leads to increased costs of anti-counterfeit products and inconvenience to production.
  • the embodiments of the present invention provide a method and device for verifying the authenticity of a product, which can improve the accuracy of verifying the authenticity of the product.
  • the embodiment of the present invention provides a method for verifying the authenticity of a product.
  • the product identification of the product has randomly distributed microdots.
  • the method includes: extracting the product identification image of the product to be verified.
  • Another embodiment of the present invention provides a method for verifying the authenticity of a product, the product identification of the product has randomly distributed micro-dots, and the method includes: extracting an image of the product identification of the product to be verified The micro-dot features on the product identification; using an image quality evaluation algorithm to generate an image quality score for the image of the product identification; and verifying the verified product based on the image quality score and the extracted micro-dot features The authenticity of the product logo.
  • the device includes: a micro-dot feature extraction module for checking the authenticity of a product. Extract the micro-dot features on the product identification from the image of the product identification of the product; the line segment feature extraction module is used to use the line segment description method of image processing technology to extract the product identification from at least one area containing lines in the image The line segment feature; wherein the line segment feature is based on the line segment descriptor generated by the line segment description method; and a verification module for verifying the line segment based on the extracted micro-point feature and the line segment feature Verify the authenticity of the product's product identification.
  • Another embodiment of the present invention provides a device for verifying the authenticity of a product.
  • the product identification of the product has randomly distributed micro-dots.
  • the device includes: a micro-dot feature extraction module for checking the authenticity of a product. Extracting the micro-dot features on the product identification from the image of the product identification of the product; an image quality score generating module for generating an image quality score for the image of the product identification using an image quality evaluation algorithm; and a verification module for generating an image quality score based on the image of the product identification
  • the image quality score and the extracted micro-dot features are used to verify the authenticity of the product identification of the verified product.
  • Another embodiment of the present invention provides a device for verifying the authenticity of a product.
  • the product identification of the product has randomly distributed microdots.
  • the device includes: a memory for storing instructions;
  • the processor of the memory when the instruction is executed by the processor, the processor executes the method for verifying the authenticity of the product described in the foregoing embodiment.
  • Another embodiment of the present invention provides a computer-readable storage medium on which is stored executable instructions, which when executed by a computer cause the computer to execute the authenticity of the product described in the above embodiments. Pseudo method.
  • the image quality evaluation algorithm is used to score the image quality generated by the product identification image, which can improve the accuracy of verifying the authenticity of the product identification.
  • Fig. 1 shows a flowchart of a method for verifying the authenticity of a product according to a first embodiment of the present invention
  • Figure 2 (a) is a schematic diagram of embedding micro-dot features into a product QR code in an embodiment
  • Figure 2(b) is a schematic diagram of the positioning block in the two-dimensional code
  • Figures 3(a) and 3(b) show the image of the probability density function when the uniform distribution function is used as the random distribution function of the microdots, and the microdot distribution map obtained by sampling from the random distribution;
  • Fig. 4 shows a flowchart of a method for verifying the authenticity of a product according to a second embodiment of the present invention
  • Fig. 5 shows a flowchart of a method for verifying the authenticity of a product according to a third embodiment of the present invention
  • Fig. 6 shows a flowchart of a method for verifying the authenticity of a product according to a fourth embodiment of the present invention
  • Fig. 7 shows a flow chart of a method for verifying the authenticity of a product according to a fifth embodiment of the present invention.
  • FIG. 8 shows a block diagram of the structure of an apparatus for verifying the authenticity of a product according to a sixth embodiment of the present invention.
  • Fig. 9 shows a block diagram of the structure of an apparatus for verifying the authenticity of a product according to a seventh embodiment of the present invention.
  • Fig. 1 shows a flowchart of a method 100 for verifying the authenticity of a product based on micro-point features and line-segment features according to a first embodiment of the present invention.
  • the method 100 shown in FIG. 1 can be implemented by any computing device having computing capabilities.
  • the computing device can be, but is not limited to, a desktop computer, a notebook computer, a tablet computer, a server, or a smart phone.
  • the verification method 100 includes: extracting the micro-dot features on the product identification from the image of the product identification of the product being verified (step 101); Extract the line segment features of the product identification in at least one area; the line segment features are based on the line segment descriptor generated by the line segment description method (step 102); and the verified product is verified based on the extracted micro-point features and line segment features
  • the authenticity of the product identification (step 103). For example, by comparing the extracted micro-dot features and line-segment features with the micro-dot features and line-segment features in the authentic product identification image, the authenticity of the product identification of the product to be verified can be verified.
  • the verification step 103 includes: comparing the micro-dot features extracted from the product identification image with the micro-dot features in the authentic product identification image to obtain a micro-dot feature comparison result; and comparing the extracted line segment features Compare with the line segment features in the authentic product identification image to obtain the line segment feature comparison result; based on the line segment feature comparison result and the micro-point feature comparison result, compose the description vector about the product identification; and based on the description vector, use the classifier to determine the verified product The authenticity of the product identification.
  • the extracted line segment features include at least the printed features related to the printing of the area containing the lines in the product identification; the printed features are the paper used in the printing process of the genuine product identification.
  • Product identification printing can be the printing of digital files on physical paper or other carrying objects.
  • the combination of may make the details of the same digital image different after printing out.
  • This detail part reflects the printing characteristics. For example, due to the paper, ink, or printing equipment used, there will be subtle differences in the printed lines, such as fine jagged parts with different shapes or arrangements on the edges.
  • the two-dimensional code in the product identification contains multiple black blocks and white blocks.
  • the positioning block 2031 in the two-dimensional code 203 shown in Figure 2(a) and Figure 2(b) can include two black blocks and For a white block, all the boundaries between black and white blocks may be different under different printing conditions.
  • the printed color or gray scale will be different due to the influence of paper, ink or printing equipment. This printing difference is distributed throughout the printing area of the two-dimensional code. Using this difference, the printed features of the product identification can be extracted.
  • the printed features of the product identification of the counterfeit product produced by the replication technology are different from the product identification of the genuine product. There are subtle differences in the contours, and the appearance of the differences in the black or white areas is clearer. This makes it possible to use the printed features in the area surrounding the line segment to verify the authenticity of the product identification.
  • the line segment description including line segment features can describe the image gradient information of the area around the line segment.
  • This kind of image gradient information can reflect the difference between the genuine product identification and the counterfeit product identification.
  • usually the line segments in the authenticity logo image will be smoother, and the line segments in the fake product logo image will be rougher; or the white area in the genuine product label image is relatively clean, and the print in the white region in the fake product logo image More noise, etc. Therefore, through the use of line segment descriptors (also called line segment descriptors) and then through comparison, the visible difference of the printed area of the line segments is obtained. Therefore, the use of the line segments in the fixed area of the product identification can help distinguish true and false product identifications.
  • the step 102 of extracting line segment features includes: sampling the product identification image to construct a scale space of the image; performing line feature extraction on each layer in the scale space to extract line segments in the scale space ; Intercept the area surrounding at least one extracted line segment in the image, and divide the intercepted area into multiple strip sub-regions as line segment support regions (LSR); construct a line segment strip descriptor (LBD) for each strip sub-region, And get the stripe descriptor matrix.
  • LSR line segment support regions
  • LBD line segment strip descriptor
  • LBD Line Band Descriptor
  • the embodiment of the present invention is not limited to this, and other line segment description methods may also be used, such as the Scale Invariant Mean-Standard Deviation Line Segment Descriptor (SMLSD) algorithm.
  • SMLSD Scale Invariant Mean-Standard Deviation Line Segment Descriptor
  • the line segment feature comparison result is obtained through the following steps: calculate the mean value and variance of at least one LBD corresponding to the line as the LBD value; compare the calculated LBD value with the LBD reference value of the genuine identification image Compare, obtain the comparison result of line segment features.
  • Line segment description is a commonly used method in the field of image processing to generate a unique description of a specified line segment in an image. For example, the area around a line segment is divided into multiple sub-areas, the gradient size and direction of the pixel are calculated for each pixel in the sub-area, and then the calculated information is collected as a 256-dimensional line segment description vector.
  • the line segment feature is the line segment in the description image, which is related to the specific printing feature of the genuine anti-counterfeiting QR code produced by printing.
  • the line segment in the embodiment of the present invention may be a two-dimensional code of the product.
  • a two-dimensional code usually includes a coding part and a positioning part. Because the coding part does not have a sufficiently long stable line segment as the line segment description object, only the three large black squares of the positioning part are used as the area for the line segment description, as shown in Figure 2(a ) And the positioning block 2031 shown in Figure 2(b). As shown in Figure 2(b), for each positioning block, three sets of boundary contours can be found between the black and white blocks and the surrounding blank areas. Each set of boundary contours is composed of two horizontal lines.
  • the boundary line is composed of two vertical boundary lines, so 1-12 line segments can be extracted from it.
  • the number of line segments that can be extracted may change.
  • the line segment may be printed with multiple pixels in width, and the printing process will also affect the pixel value of the area adjacent to the line segment.
  • the captured image for the printed line segment will be sampled into a wide area.
  • the printed features in the fake logo image are different from the real product, for example, it may become rough or fuzzy.
  • the comparison between the genuine product and the fake product is for the line segments at the same position, and then respectively generate line segment description vectors for them. By counting the mean value of the line segment descriptors of the genuine product, it can be found that the line segment description vector of the fake or imitation product is far away from the average description vector of the genuine product.
  • the distance is calculated from the mean value (the mean value is also a vector) of the previous statistics for the positive sample. This distance is a vector, which can be added to the description vector that is ultimately used to distinguish true from false.
  • N is greater than or equal to 3
  • N is greater than or equal to 3
  • an N-level scale pyramid can be obtained.
  • a line segment detection algorithm such as the EDLine algorithm
  • four line segments can be extracted from each layer, and then composed as line segments according to the position and direction of each line segment.
  • N LineVec variables each LineVec variable contains N line segments in the N pyramid layers that are located in the same area and have the same direction.
  • a line support region (Line Support Region, LSR) needs to be calculated.
  • LineVec contains multiple line segments located in different pyramid layers, and these line segments must establish LSR in the image of the corresponding layer.
  • a line segment is the direction of the axis of abscissa is d L LSR new coordinate system, and d ⁇ d L ordinate axis orthogonal to the coordinate system of the new LSR, the midpoint of a line segment
  • the origin of the LSR coordinate system; each pixel on the strip is based on the newly established LSR to calculate the gradient direction; two Gaussian functions are used as the global weight coefficient f g and the local weight coefficient f L on the entire LSR, namely
  • i and k refer to the i-th row and k-th row of the Bj strip in the LSR
  • di represents the distance from the i-th row to the center row in the LSR
  • d'k represents the distance from the k-th row to the center row of the Bj strip
  • ⁇ g 0.5(m ⁇ w-1)
  • ⁇ l w.
  • the purpose of the global Gaussian window is to reduce the importance of the gradient away from the line segment, so as to ease the sensitivity to small changes in the vertical direction of the line segment.
  • the purpose of the local Gaussian window is to reduce the edge effect and avoid the sudden change of the descriptor when the pixel goes from one strip to the next.
  • ⁇ f g (k) f l (k) is the product of the global Gaussian weight and the local Gaussian weight of the corresponding position, j is the number of the strip, and n is the number contained in the strip Number of pixel rows.
  • BDM stripe description matrix
  • M is the mean vector of the corresponding v row corresponding to the m-th strip
  • S is the variance vector of the corresponding v row. Therefore, a strip will generate 8 description vectors, and finally m*8 vectors will be connected end to end to form a descriptor LBD.
  • a positioning block can be used as the verification area, and four internal line segments can be used as the verification line segment; for the four line segments, a segment support area (LSR) is generated; for each LSR generates the LBD value; repeats this process on the corresponding line segment of the corresponding positioning block of all positive samples to generate the LBD value, and calculates the average of all the corresponding LBD values as the LBD reference value.
  • LSR segment support area
  • the line segment of the corresponding position to be verified is intercepted; the LSR is generated for the line segment; the LBD value is calculated for each LSR; the LBD value is calculated with the previous example.
  • the difference between the calculated LBD reference values; the maximum value of the four difference values corresponding to the four line segments is taken as the output line segment feature comparison result.
  • the verification step 103 may further include: using a classifier trained by a machine learning algorithm to verify the authenticity of the product identification of the product to be verified; wherein the classifier adopts multiple authentic product identification images as positive samples , And use multiple fake identification images as negative samples for training.
  • the machine learning algorithm used to train the classifier is a machine learning algorithm that can classify feature vectors.
  • the positive samples can be two-dimensional code labels of multiple genuine products, and the negative samples can be copies of these positive samples obtained in various ways.
  • the negative samples have the same two-dimensional code labels, but the micro-dot features and printed features are different from those of the genuine two-dimensional code labels.
  • the description vector includes data related to at least one of the following items: the difference between the LBD value of the verified product identification and the LBD reference value of the authenticity identification image, and the extracted micro points
  • the step 101 of extracting micro-dot features includes: using image processing technology to extract at least one of the shape feature, position feature, gray-scale feature, and color feature of the micro-dot from the product identification image.
  • the product identification is a two-dimensional graphic code; the lines in the two-dimensional graphic code are the block boundary lines on the positioning block in the two-dimensional graphic code.
  • FIG. 2(a) and FIG. 2(b) show the two-dimensional code 203 and the positioning block 2031 therein.
  • the product identification may also be a barcode or other pattern containing lines.
  • the method for verifying the authenticity of the product may further include: using an image quality evaluation algorithm to generate an image quality score for the product identification image; Verify the authenticity of the product's product identification.
  • image processing technology There are many methods in image processing technology that can be used to evaluate the image quality of the target image, such as identifying whether the light and dark parts in the image are reasonable, whether it is too bright or too dark, and whether the picture is clear enough.
  • An image quality score is generated by evaluating the identification image obtained.
  • the image quality is related to the extraction of micro-point features, line-segment features and/or other image features in the image, and it will also affect the verification and recognition of micro-point features, line-segment features and/or other image features in the image. Therefore, the combination of image quality scoring in product identification verification will help to more accurately verify the authenticity of product identification.
  • Fig. 2(a) is a schematic diagram of embedding the micro-dot feature into the product two-dimensional code in an embodiment, and the black micro-dot feature 202 is not shown in the figure because its size is too small.
  • an algorithm is used to generate a specific high-dimensional random distribution map 201 of micro-points as the distribution characteristics of at least one of the location distribution, gray-scale distribution, color distribution and micro-morphology of all micro-point features.
  • Products in the same category or the same batch can follow a certain distribution characteristic, and each product has other different micro-point characteristics to show distinction. For example, different batches of products can use different random distribution maps, and different products of the same batch can use different microdots.
  • the algorithm is used to sample the random distribution map of the micro-dots, and a uniquely identifiable micro-dot feature 202 is generated for each product (or product identification or label), and then the generated micro-dot feature is combined according to a predetermined avoidance rule Embedded in the digital two-dimensional identification 203 of the product (such as the quick response matrix code, that is, two-dimensional code), and print the two-dimensional code with embedded micro-dot features on the surface of the product or the surface of the product packaging as the product identification, or print Manufactured on the surface of the product label to form a digital product identification (ID) with micro-dots.
  • a predetermined avoidance rule Embedded in the digital two-dimensional identification 203 of the product such as the quick response matrix code, that is, two-dimensional code
  • ID digital product identification
  • the avoidance rule can restrict at least one of the specific position distribution, grayscale distribution, and color distribution of the micro-dots.
  • the location distribution avoidance rule can ensure that only black or dark micro-dots are generated in the white block of the QR code
  • the gray-scale distribution or color distribution avoidance rule can ensure that the gray-scale or color of the micro-dots meets a certain gray-scale and saturation The degree limit will not interfere with the white blocks of the QR code.
  • the white micro-dot feature 202 can also be embedded in the black block of the two-dimensional code 203.
  • the avoidance rule restricts the micro-dots to be generated to the black block of the two-dimensional code, so that the two-dimensional code is added
  • the micro-point feature still meets the corresponding national standards and/or international standards.
  • the white micro-dots maintain the highest contrast in the black blocks of the two-dimensional code, and the white micro-dots are generated by short pauses in the printing inkjet during the printing process.
  • the composition of micro-point features includes the most basic two-dimensional coordinates (X, Y) as location features, and can also include other optional features, such as color, grayscale, shape, and so on.
  • the non-reproducibility and anti-counterfeiting performance of the micro-dots are first realized by the random distribution of the two-dimensional positions of the micro-dots.
  • the color, gray, or shape characteristics of the micro-dots can be used to further improve the anti-counterfeiting performance of the product.
  • Randomly distributed micro-dot features can also form randomly distributed micro-dot texture features.
  • the micro-point feature information on the product identification needs to be stored in the database for subsequent product authenticity verification.
  • the saved feature information of the micro-points includes, for example, randomly distributed location features and other features such as its color, gray scale, or shape.
  • Each two-dimensional code 203 in FIG. 2(a) includes three positioning blocks 2031, which are respectively located in the upper left corner, the upper right corner and the lower left corner.
  • the positioning block is also called the position detection pattern, which is used to mark the rectangular size of the two-dimensional code.
  • each positioning block 2031 includes a black rectangular solid block in the middle, a white square block in the middle, and a black square block on the outside. These three color blocks constitute three groups of containing blocks.
  • the block boundary line of each group of contours is composed of two horizontal lines and two vertical lines. These block boundary lines can be used as the object of the line segment description method in the embodiment. However, the present invention is not limited to this, and the object of the line segment description method can be any line segment in the product identification.
  • Figure 3(a) and Figure 3(b) show the image of the probability density function when the uniform distribution function is used as the random distribution function of the micro-point, and the micro-point sampled from the random distribution. Distribution.
  • the probability density function of the uniform distribution function is:
  • the Z coordinate is the probability density
  • the horizontal coordinate X and the vertical coordinate Y indicate the position (x, y) of the micro point.
  • the micro-dot distribution map in Fig. 3(b) was sampled from the random distribution map in Fig. 3(a) when the micro-dot coordinates (x, y) were generated.
  • the graph shows randomly distributed black micro-dots.
  • FIG. 4 shows a flowchart of a method 400 for verifying the authenticity of a product according to a second embodiment of the present invention.
  • the second embodiment is also an embodiment in which product identification is verified based on line segment features and micro-dot features.
  • the image or picture of the product identification of the product to be verified is first obtained (step 401). For example, after purchasing a product, the user can take a photo of the product identification part containing the barcode or QR code, and transmit the image of the product identification obtained by the photo to the verifier or verification device, so that the obtained image of the product identification can be verified. Pseudo verification.
  • the processing of an image containing a two-dimensional code includes two parts, namely, a line segment processing part (including steps 402-405) and a micro-point processing part (including steps 406-409).
  • the image needs to be preprocessed (steps 402 and 406), such as adjusting the brightness, cutting the effective part, and enhancing the contrast of the part of the image that contains the effective features (such as the two-dimensional code).
  • preprocessing methods in image processing technologies such as sharpening and image normalization.
  • the line segment description algorithm in the image processing technology is used to describe at least one line segment and the blank area adjacent to the line segment in the product identification image (step 403). Because the black and white blocks in the two-dimensional code encoding part change with the information of the two-dimensional code, there is no stable black and white blocks and boundary lines, so you can use the positioning block with stable color blocks in the two-dimensional code (as shown in Figure 2 ( a) and Figure 2(b)) on the line segment to generate a specific line segment description. Then, the generated line segment description is compared with the line segment description of the corresponding line segment in the genuine two-dimensional code image to obtain the line segment feature comparison result (step 404).
  • the line segment description of the corresponding line segment in the genuine two-dimensional code image is generated based on the genuine two-dimensional code image, which contains the printing features related to the printing of the corresponding area.
  • the line segment description of the two-dimensional code image of the verified product can be compared with the line segment description of a genuine two-dimensional code image to obtain the line segment feature comparison result, or it can be compared with multiple genuine two-dimensional codes.
  • the average value of the line segment descriptions at the same position in the image is compared to obtain the line segment feature comparison result.
  • the two-dimensional code copied in the fake logo loses subtle printed features in the corresponding line segment, so there will be a significant difference between the line segment description of the real two-dimensional code image, and this difference is essentially a black and white block
  • the decomposition part is due to the image difference surrounding the corresponding line segment caused by copying and different printing equipment paper.
  • the line segment feature comparison result is output as a description of the authenticity of the line segment.
  • the result of the comparison of the line segment feature is the difference between the LBD value of the identity of the product being verified and the LBD reference value of the authenticity image.
  • the LBD reference value of the authenticity image may be the LBD reference of multiple authenticity image. The average value of the value.
  • the description vector of the line segment in the two-dimensional code with the authenticity mark is already present, for example, the line segment description generated for the corresponding line segment in the two-dimensional code image copied in the counterfeit mark can be further obtained.
  • the vector difference Euclidean distance
  • the micro-point extraction algorithm is used to extract the micro-point features in the image (step 407), in which image processing technology can be used to read the random area where the product identification is verified.
  • the distributed micro-dot features include statistical data based on at least one of the location, size, color, or gray level of the micro-dots. For example, by counting the size of each micro-dot area (such as the number of pixels contained in each micro-dot), or counting the average RGB three-channel value of each area, the gray information of the micro-dot area in the image can be obtained.
  • step 409 extract the corresponding micro-point features of the authentic product of the pre-stored product identification in the database, and compare the read micro-point features with the micro-point features in the database (step 408), thereby outputting the micro-point feature comparison
  • the result (step 409) is used as one of the basis for judging the authenticity of the product identification.
  • the line segment feature comparison result output in step 405 and the micro point feature comparison result output in step 409 are used to form a description vector X (step 410).
  • a description vector X For example, take the output line segment feature comparison result as a feature dimension such as x1.
  • the comparison result of the micro-point feature can include several statistical data obtained after the matching and quantification of the micro-point feature in the comparison, such as the micro-point found in the target two-dimensional code matching the micro-point of the corresponding two-dimensional code in the database.
  • the percentage is taken as x2
  • the statistical parameters such as the mean and variance
  • the penalty for no matched micro-dots (mismatch) is taken as x5 etc.
  • the above information can be used to form a five-dimensional description vector about the identity of the product being verified.
  • image quality descriptions such as the sharpness of the image, the mean and variance of the light and dark distribution of the image can also be added to the description vector as other feature dimensions.
  • all collected positive and negative product identification images can be processed to obtain corresponding line segment features and LBD differences, as well as micro-point statistical features, and then corresponding description vectors can be obtained as samples
  • a part of the data set (such as 80%) can be used as a training set for training the classifier; the other part (such as 20%) can be used as a test set.
  • more popular machine learning algorithms can be used to train and test the classifiers.
  • the types of classifiers that can be selected include support vector machines (SVM), boost trees, decision trees, and shallow nerves. Network, k-nearest neighbor algorithm, random forest, etc.
  • the pre-trained classifier can discriminate and classify the verified product identification (step 411) based on the descriptive features in the description vector obtained in step 410 (including: line segment feature comparison results, and micro-point statistical features), thereby outputting information about the product being verified.
  • the authenticity determination result of the product identification is verified (step 412).
  • the combination of, for example, a product identification including a two-dimensional code, a micro-dot feature, and a line segment feature can further improve the verification accuracy of the authenticity of the product identification.
  • FIG. 5 shows a flowchart of a method 500 for verifying the authenticity of a product based on micro-point features and image quality scores according to a third embodiment of the present invention.
  • the verification method 500 includes: extracting randomly distributed micro-dot features on the product identification from the image of the product identification of the product being verified (step 501); using an image quality evaluation algorithm to generate an image quality score for the image of the product identification (Step 502); and verify the authenticity of the product identification of the verified product based on the image quality score and the extracted micro-point features (Step 503).
  • image processing technology can be used to evaluate the image quality of the logo image to generate an image quality score.
  • the image quality is related to the extraction of micro-point features and/or other image features in the image, and it will also affect the verification and recognition of the micro-point features and/or other image features in the image. Therefore, in the product identification verification Combined with the image quality score, it will help to verify the authenticity of the product label more accurately, especially to solve the problem of false product labels with good image quality and genuine product labels with poor image quality.
  • the authenticity of anti-counterfeit QR code images is related to the quality of image shooting, especially in mobile applications, mobile phones of various brands show different photo processing orientations; in different shooting scenarios, the same product label Due to the different processing orientations of different camera phones, the captured images show different image details. Therefore, if a simple and consistent threshold is used for the true and false identification and verification of all images, it may seriously reduce the accuracy of the true and false verification; and based on different image quality scores, for example, by using different true and false The discrimination threshold is used for authenticity verification, which can improve the accuracy of product label verification.
  • the verification step 503 includes: comparing the extracted micro-dot features with the micro-dot features in the authentic identification image to obtain a micro-dot feature comparison result; based on the image quality score and the micro-dot feature comparison result, Compose a description vector about the product identification; and based on the description vector, use a classifier to determine the authenticity of the product identification of the product being verified.
  • the verification step 503 further includes: using a classifier trained by a machine learning algorithm to verify the authenticity of the product identification of the verified product; wherein multiple authentic identification images are used as positive samples and multiple A fake identification image is used as a negative sample to train the classifier; the machine learning algorithm used to train the classifier is a machine learning algorithm that can classify feature vectors.
  • the positive samples can be two-dimensional code labels of multiple genuine products, and the negative samples can be copies of these positive samples obtained in various ways.
  • the negative samples have the same two-dimensional code label, but the micro-dot features and the correlation between the micro-dot features and the image quality are slightly different from those of the genuine two-dimensional code label.
  • machine learning algorithms the accuracy of the classifier in distinguishing the authenticity of products can be continuously improved. For example, by using a machine learning algorithm to make the classifier learn to use different true and false discrimination thresholds under different image quality scores to perform true and false verification.
  • Various image quality evaluation algorithms in image processing technology can be used to generate an image quality score.
  • the following takes the spatial domain reference-free image quality assessment (BRISQUE) algorithm as an example, but the present invention is not limited to this.
  • the BRISQUE algorithm includes: normalizing the input original image to extract natural scene statistics (NSS) from the original image, calculating feature vectors, and using machine learning algorithms (such as support vector machines) to estimate image quality scores.
  • NSS natural scene statistics
  • the distribution of pixel intensities of natural images is different from the distribution of pixel intensities of distorted images.
  • the distribution difference is more obvious.
  • the pixel intensity of the natural image follows the Gaussian distribution (Bell curve), while the pixel intensity of the unnatural or distorted image does not follow the Gaussian distribution (the Bell curve). From this, the image can be scored for image quality.
  • the Mean Subtracted Contrast Normalization (MSCN) method is a method for normalizing images.
  • W and H are the width and height of the input image, and the image intensity I(i,j) at the pixel (i,j) is converted to brightness according to the following calculation formula To calculate the MSCN parameters:
  • ⁇ (i,j) and ⁇ (i,j) are the local average and local variance, respectively.
  • w is the Gaussian blur window
  • ⁇ is the result of Gaussian blurring of the original image I
  • ⁇ is the result of subtracting the Gaussian blur image ⁇ from the original image and then taking the square root of the Gaussian blur.
  • the feature vector is calculated using the calculated MSCN parameter and the four adjacent pixel product parameters.
  • the MSCN parameters For example, by fitting the MSCN parameters to the generalized Gaussian distribution (GGD) to calculate the first two eigenvector elements (shape parameter and variance) of the 36 ⁇ 1 eigenvector, fit each of the four adjacent pixel product parameters to Asymmetric Generalized Gaussian Distribution (AGGD) is used to calculate the four eigenvector elements of the eigenvector (shape parameter, mean, left variance, right variance).
  • GGD generalized Gaussian distribution
  • Asymmetric Generalized Gaussian Distribution Asymmetric Generalized Gaussian Distribution
  • the existing image quality data set already contains artificial scoring and evaluation of undistorted images and images with different degrees of distortion.
  • the image quality score is y
  • the descriptor generated by the BRISQUE algorithm is x.
  • Any linear regression algorithm such as SVR (Support Vector Regression) can be trained to obtain a regression model. After processing the input image to generate descriptors, this regression model is handed over to the trained model to obtain a score describing the degree of image distortion.
  • an image quality score of 0 represents no distortion
  • 100 represents severe distortion
  • the output image quality score is between 0-100.
  • the image quality score can also be normalized to a value between 0 and 1, as a data item in the description vector.
  • the image is first converted into a feature vector. Then, the feature vectors and output (quality score) of all images in the training data set are put into a support vector machine (SVM) for training.
  • SVM support vector machine
  • LIBSVM support vector machine toolkit
  • the feature vector is first scaled to -1 to 1, and the trained model is used to predict image quality. After training the model to predict, the final quality score of each distortion can be obtained.
  • the description vector may include data related to at least one of the following: the matching rate between the extracted micro-point feature and the pre-saved micro-point feature, and the micro-point on the matching rate
  • the image quality score is generated based on image quality data identifying at least one sub-region in the image, and the image quality data is related to at least one of the following image distortion types: Gaussian superimposed noise distortion, color component Distortion, illumination distortion, spatial correlation distortion, impulse distortion, image motion blur distortion, low-light image noise distortion, image file compression distortion and contrast distortion.
  • image distortion types Gaussian superimposed noise distortion, color component Distortion, illumination distortion, spatial correlation distortion, impulse distortion, image motion blur distortion, low-light image noise distortion, image file compression distortion and contrast distortion.
  • the input reference image can be used to evaluate the quality of the existing image, or the image quality evaluation without the reference image can be performed directly.
  • image quality evaluation an image evaluation algorithm is used to process the input image to generate an image quality score that represents the quality of the image taken.
  • image quality evaluations include whether the image is clear, the amount of noise in the image, and whether the light is moderate.
  • the image quality assessment technology can quantify these intuitive evaluations into multiple types of image distortion.
  • the image distortion types listed above are only a part of the common types, and the present invention is not limited to using the above image distortion types to generate an image quality score.
  • the step 501 of extracting micro-dot features includes: using image processing technology to extract at least one of the shape feature, position feature, gray-scale feature, and color feature of the micro-dot from the logo image.
  • Fig. 6 shows a flowchart of a method for verifying the authenticity of a product based on micro-point features and image quality scores according to a fourth embodiment of the present invention.
  • the image or picture of the product identification of the product to be verified is first obtained (step 601).
  • the steps 602-605 are the micro-point processing part, which are respectively the same as the steps 406-409 in the second method embodiment shown in FIG. 4, and will not be described in detail here.
  • Steps 606-608 are the image quality evaluation part.
  • the image needs to be pre-processed (step 606), such as adjusting the brightness of the part of the image that contains effective features (such as a two-dimensional code), intercepting the effective part, enhancing the contrast, sharpening the image, and Commonly used preprocessing methods in image processing technologies such as picture normalization.
  • the image quality description of the image is extracted from the preprocessed image (step 607), where various image description methods in image processing technology can be used to describe the target image quality, such as whether the light and shade distribution is reasonable or whether it is too bright Too dark, the image is not clear enough or contains distortion.
  • Step 607 may be the process of generating the image quality description using the BRISQUE algorithm as described above, and will not be described in detail here.
  • the image quality description is output.
  • the micro-point feature comparison result output in step 605 (different statistical data can be normalized) and the image quality description output in step 608 are used to form a description vector X (step 610).
  • a description vector X For example, take the output image quality score as a feature dimension such as x1.
  • the comparison result of the micro-point feature can include several statistical data obtained after the matching and quantification of the micro-point feature in the comparison, such as the micro-point found in the target two-dimensional code matching the micro-point of the corresponding two-dimensional code in the database.
  • the percentage is taken as x2
  • the statistical parameters (such as mean and variance) of the pixel distance between the matched micro-points in the image coordinate system and the micro-dots in the database are taken as x3 and x4
  • the penalty for no matched micro-dots (mismatch) is taken as x5 etc.
  • the above information can be used to compose a five-dimensional description vector X about the identification of the product being verified.
  • all collected images of product identifications of positive samples and negative samples can be processed to obtain corresponding image quality scores and micro-point statistical features, and then corresponding description vectors can be obtained as sample data sets.
  • One part (such as 80%) can be used as a training set to train the classifier; the other part (such as 20%) can be used as a test set.
  • more popular machine learning algorithms can be used to train and test the classifiers.
  • the types of classifiers that can be selected include support vector machines (SVM), boost trees, decision trees, and shallow nerves. Network, k-nearest neighbor algorithm, random forest, etc.
  • the pre-trained classifier can discriminate and classify the verified product identification (step 611) based on the descriptive features in the description vector obtained in step 610 (including: micro-point statistical features and image quality scores), thereby outputting information about the verified product
  • the authenticity determination result of the product identification step 612.
  • the combination of, for example, a product identification including a two-dimensional code, a micro-dot feature, and an image quality description can further improve the accuracy of verification of authenticity product identification, especially It can solve the problem of incorrect verification of fake product labels with good image quality and genuine product labels with poor image quality.
  • the image quality of fake product labels can still be accurately identified as fake products. No matter how bad the image quality of a genuine product label is, it can still be verified as a genuine label, thus allowing users to use various mobile phones or cameras to take product identification images in various external environments.
  • Fig. 7 shows a flowchart of a method for verifying the authenticity of a product according to a fifth embodiment of the present invention.
  • the method 700 is to verify the authenticity of the product based on the micro-point feature, the line segment feature, and the image quality score.
  • the image or picture of the product identification of the product to be verified is first obtained (step 701).
  • Steps 702-705 in FIG. 7 are the line segment processing parts, which are respectively the same as steps 402-405 in the second method embodiment shown in FIG. 4, and output the comparison results of the line segment features of the product identification image after processing.
  • Steps 706-709 in FIG. 7 are the micro-point processing part, which are respectively the same as steps 406-409 in the second method embodiment shown in FIG.
  • Steps 710-712 in FIG. 7 are the image quality evaluation parts, which are respectively the same as steps 606-608 in the fourth method embodiment shown in FIG. Narrated.
  • a description vector X is formed (step 713).
  • the output line segment feature comparison result can include several statistical data obtained after the matching and quantification of the micro-point feature in the comparison, such as the micro-point found in the target two-dimensional code matching the micro-point of the corresponding two-dimensional code in the database.
  • the percentage is taken as x2
  • the statistical parameters such as mean and variance
  • the penalty for no matched micro-dots (mismatch) is taken as x5 etc.
  • the above information can be used to compose a six-dimensional description vector X about the identification of the product to be verified.
  • all collected positive and negative product identification images can be processed to obtain corresponding micro-point statistical features, line segment features, and image quality scores, and then corresponding description vectors can be obtained as sample data Set, one part (such as 80%) can be used as a training set for training the classifier; the other part (such as 20%) can be used as a test set. Based on this sample data set, more popular machine learning algorithms can be used to train and test the classifiers.
  • the types of classifiers that can be selected include support vector machines (SVM), boost trees, decision trees, and shallow nerves. Network, k-nearest neighbor algorithm, random forest, etc.
  • the pre-trained classifier can discriminate and classify the verified product identification (step 714) based on the descriptive features (including: micro-point statistical features, line segment features, and image quality scores) in the description vector obtained in step 713 (step 714), and output Regarding the authenticity determination result of the product identification to be verified (step 715).
  • Fig. 8 shows a block diagram of the structure of an apparatus for verifying the authenticity of a product according to a sixth embodiment of the present invention.
  • the device 800 includes: a micro-point feature extraction module (801) for extracting randomly distributed micro-point features on the product identification from the product identification image of the product to be verified; a line segment feature extraction module (802) for the use of image processing technology
  • the line segment description method extracts the line segment feature in the product identification from at least one area containing the line in the image; the line segment feature is based on the line segment descriptor generated by the line segment description method; and the verification module (803) is used to extract the line segment feature based on the extracted micro Point features and line segment features are used to verify the authenticity of the product identification of the product being verified.
  • the device 800 may further include: an image quality score generating module (804), configured to use an image quality evaluation algorithm to generate an image quality score for the image of the product identification; the verification module (803) is based on the image quality score generating module (804) )
  • the generated image quality score, the micro-point feature extracted by the micro-point feature extraction module (801), and the line segment feature extracted by the line-segment feature extraction module (802) are used to verify the authenticity of the product identification of the product being verified.
  • Fig. 9 shows a block diagram of the structure of an apparatus for verifying the authenticity of a product according to a seventh embodiment of the present invention.
  • the device 900 includes: a micro-point feature extraction module (901) for extracting the micro-point feature on the product identification from the image of the product identification of the verified product; an image quality score generating module (902) for using an image quality evaluation algorithm Generate an image quality score for the image of the product identification; and a verification module (903) for verifying based on the image quality score generated by the image quality score generation module (902) and the micropoint features extracted by the micropoint feature extraction module (901) The authenticity of the product identification of the product being verified.
  • a micro-point feature extraction module (901) for extracting the micro-point feature on the product identification from the image of the product identification of the verified product
  • an image quality score generating module for using an image quality evaluation algorithm Generate an image quality score for the image of the product identification
  • a verification module (903) for verifying based on the image quality score generated by the image
  • the devices 800 and 900 shown in FIGS. 8 and 9 can be implemented by software, hardware, or a combination of software and hardware, and can be designed to include corresponding modules to implement the above-mentioned methods for verifying product authenticity of the present invention. Examples.
  • Embodiments of the present invention may also provide a device for verifying the authenticity of a product, which includes: a memory for storing instructions; and a processor coupled to the memory, which when executed by the processor causes the processor to execute The product verification method according to the above-mentioned embodiment of the present invention.
  • a database may also be stored in the memory, and the database may include at least the micro-dot features and line-segment features of the authentic product identification, for comparison with the micro-dot features and line-segment features of the verified product identification.
  • the memory of this embodiment may also store a sample library, which includes a plurality of authentic identification images as positive samples and a plurality of fake identification images as negative samples.
  • the processor is configured to use at least a part of the samples in the sample library to train a classifier for verifying product identification.
  • a set of feature vectors of positive/negative samples are randomly taken from the sample library in the memory and provided to the classifier, and the classifier will perform authenticity classification processing; then, according to the known truth of the sample Calculate the loss value of the classification result and determine whether the loss value is less than the preset threshold; when it is determined that the loss value is still greater than or equal to the threshold, the parameters of the classifier are updated according to the loss value, and then the classifier is based on The updated parameters continue to be processed according to those authenticity classifications.
  • the above process is continuously looped until the loss value is judged to be less than the threshold, and the training of the classifier is stopped.
  • the authenticity can be greatly improved by adopting, for example, product identification including two-dimensional code or barcode, micro-point feature, line segment feature, and image quality description.
  • product identification including two-dimensional code or barcode, micro-point feature, line segment feature, and image quality description.
  • the verification accuracy of product identification allows users who purchase products to use various mobile phones or cameras to take product identification images and perform accurate verification under various lighting conditions.

Landscapes

  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Image Analysis (AREA)

Abstract

本发明涉及一种用于验证产品真伪的方法和装置。该方法包括：从被验证产品的产品标识的图像中提取产品标识上随机分布的微点特征；使用图像处理技术的线段描述方法，从图像中包含线条的至少一个区域提取产品标识的线段特征；线段特征是基于利用线段描述方法所生成的线段描述符；以及基于所提取的微点特征和线段特征来验证被验证产品的产品标识的真伪。本发明的实施例通过利用标识图像的微点特征、线段特征和/或图像质量描述进行真伪产品标识的验证，可以显著改善验证准确度。

Description

用于验证产品真伪的方法和装置 技术领域
本发明涉及一种用于验证产品的真伪的方法和装置，特别是涉及一种基于产品标识图像的微点特征、线段特征和/或图像质量评分来验证产品标识真伪的方法和装置。
背景技术
假冒伪劣产品对于生产者和消费都造成巨大的损失，因此需要通过使用安全可靠的防伪技术来加以控制。现有的针对产品的防伪技术包括数字防伪技术和纹理防伪技术。
数字防伪技术利用条形码或二维码来给产品一个唯一的身份识别(ID)用于防伪验证和可追溯性功能，这种数字防伪技术易于被复制，且安全性差。
纹理防伪技术使用随机生成的自然纹理作为防伪特征，这种纹理是物理不可复制的，具有不可再现的特征；但是现有的纹理防伪技术缺乏对防伪特征的自动鉴别能力，而自动鉴别能力要求防伪特征具有可视识别性，或者依赖于在生产过程中添加纤维材料来形成防伪特征，从而导致防伪产品的成本上升以及不便于生产。
目前已经出现了将条形码或二维码与印制的微点特征相结合的新技术，用于进一步改善产品标识的防伪性能，同时又能简化防伪产品的生产工艺并降低生产成本。然而，在产品标识的验证过程中首先需要获取被验证产品的产品标识的图像，例如需要用户使用手机或数码相机拍摄产品标识的图像，而由于手机或相机的拍摄功能、拍摄环境(例如光线)以及拍摄水平(例如拍摄角度、拍摄距离、相机稳定 性)等因素存在差异，使得图像质量受到不同程度的影响，从而使验证结果出现偏差，例如将作为真品的产品标识的图像验证为伪品，或者将作为伪品的产品标识的图像验证为真品。
发明内容
针对现有技术的以上问题中至少之一，本发明的实施例提供用于验证产品的真伪的方法和设备，其能够改善验证产品真伪的准确度。
本发明的实施例提供一种用于验证产品的真伪的方法，所述产品的产品标识上具有随机分布的微点，所述方法包括：从被验证产品的产品标识的图像中提取所述产品标识上的微点特征；使用图像处理技术的线段描述方法，从所述图像中包含线条的至少一个区域提取所述产品标识的线段特征；其中，所述线段特征是基于利用所述线段描述方法所生成的线段描述符；以及基于所提取的所述微点特征和所述线段特征来验证所述被验证产品的产品标识的真伪。
本发明的另一实施例提供一种用于验证产品的真伪的方法，所述产品的产品标识上具有随机分布的微点，所述方法包括：从被验证产品的产品标识的图像中提取所述产品标识上的微点特征；使用图像质量评估算法针对所述产品标识的图像生成图像质量评分；以及基于所述图像质量评分和所提取的所述微点特征来验证所述被验证产品的产品标识的真伪。
本发明的又一实施例提供一种用于验证产品的真伪的装置，所述产品的产品标识上具有随机分布的微点，所述装置包括：微点特征提取模块，用于从被验证产品的产品标识的图像中提取所述产品标识上的微点特征；线段特征提取模块，用于使用图像处理技术的线段描述方法，从所述图像中包含线条的至少一个区域提取所述产品标识的线段特征；其中，所述线段特征是基于利用所述线段描述方法所生成的线段描述符；以及验证模块，用于基于所提取的所述微点特征和所述 线段特征来验证所述被验证产品的产品标识的真伪。
本发明的又一实施例提供一种用于验证产品的真伪的装置，所述产品的产品标识上具有随机分布的微点，所述装置包括：微点特征提取模块，用于从被验证产品的产品标识的图像中提取所述产品标识上的微点特征；图像质量评分生成模块，用于使用图像质量评估算法针对所述产品标识的图像生成图像质量评分；以及验证模块，用于基于所述图像质量评分和所提取的所述微点特征来验证所述被验证产品的产品标识的真伪。
本发明的又一实施例提供一种用于验证产品的真伪的设备，所述产品的产品标识上具有随机分布的微点，所述设备包括：用于存储指令的存储器；以及耦合到所述存储器的处理器，所述指令在由所述处理器执行时使得所述处理器执行上述实施例所述的用于验证产品真伪的方法。
本发明的又一实施例提供一种计算机可读存储介质，其上存储由可执行指令，所述可执行指令在由计算机执行时使得所述计算机执行上述实施例所述的用于验证产品真伪的方法。
根据本发明的实施例的方案，在验证产品标识的真伪时不仅基于产品标识上的微点特征，而且还基于使用图像处理技术的线段描述方法从产品标识图像中提取的线段特征和/或使用图像质量评估算法针对产品标识图像生成的图像质量评分，由此可以提高验证产品标识的真伪的准确度。
附图说明
本发明的其它特征、特点、益处和优点通过结合以下附图的详细描述将变得更加显而易见，其中：
图1示出根据本发明的第一实施例的用于验证产品真伪的方法流程图；
图2(a)是在实施例中将微点特征嵌入到产品二维码中的示意图；
图2(b)是二维码中的定位块的示意图；
图3(a)和图3(b)示出采用均匀分布函数作为微点的随机分布函数时的概率密度函数的图像、以及从随机分布中抽样得到的微点分布图；
图4示出根据本发明的第二实施例的用于验证产品真伪的方法流程图；
图5示出根据本发明的第三实施例的用于验证产品真伪的方法流程图；
图6示出根据本发明的第四实施例的用于验证产品真伪的方法流程图；
图7示出根据本发明的第五实施例的用于验证产品真伪的方法流程图；
图8示出根据本发明的第六实施例的用于验证产品真伪的装置的结构方框图；以及
图9示出根据本发明的第七实施例的用于验证产品真伪的装置的结构方框图。
具体实施方式
以下结合附图进一步描述本发明的实施例。
图1示出根据本发明的第一实施例的用于基于微点特征和线段特征验证产品真伪的方法100的流程图。图1所示的方法100可以由具有计算能力的任何计算设备来实现。该计算设备可以是但不局限于台式计算机、笔记本电脑、平板电脑、服务器或智能手机等。
如图1所示，验证方法100包括：从被验证产品的产品标识的图像中提取产品标识上的微点特征(步骤101)；使用图像处理技术的线段描述方法，从产品标识图像中包含线条的至少一个区域提取产品 标识的线段特征；其中的线段特征是基于利用所述线段描述方法所生成的线段描述符(步骤102)；以及基于所提取的微点特征和线段特征来验证被验证产品的产品标识的真伪(步骤103)。例如，通过将所提取的微点特征和线段特征与真品产品标识图像中的微点特征和线段特征进行比较，可以验证被验证产品的产品标识的真伪。
在本发明的实施例中，验证步骤103包括：将从产品标识图像中所提取的微点特征与在真品标识图像的微点特征进行比较，获得微点特征比较结果；将所提取的线段特征与在真品标识图像的线段特征进行比较，获得线段特征比较结果；基于线段特征比较结果和微点特征比较结果，组成关于产品标识的描述向量；以及基于描述向量，使用分类器判断被验证产品的产品标识的真伪。
在本发明的实施例中，所提取的线段特征至少包括与产品标识中包含线条的区域的印制相关的印制特征；印制特征是与真品的产品标识的印制过程中使用的纸张、油墨、印制设备中的至少一项相关联的特征。
产品标识印刷可以是将数字文件印刷到物理纸张或其他承载物品上，同样的数字文件打印时由于打印机设置不同，印刷机种类不同，油墨或者碳粉或者上色剂不同，纸张特性不同等这些复杂的组合，都可能会使得将同样的数字图像打印出来以后，其细节部分是不一样的。这种细节部分就体现了印制特征。例如，由于所采用的纸张、油墨或印制设备的不同，所印制出来的线条会存在细微的差别，如在边缘出现形状或排列不同的细微锯齿部分。例如，在产品标识中的二维码包含多个黑色块和白色块，如图2(a)和图2(b)所示的二维码203中的定位块2031可包括两个黑色块和一个白色块，其中所有的黑白色块分界处在不同的印制情况下都可能是有差别的。另外，受到纸张、油墨或印制设备的影响，印制的颜色或灰度也会有差别。这种印制差别分布在整个二维码的印刷区域。而利用这种差别就可以提取产品标 识的印制特征。通过复制技术制作的伪品的产品标识所具有的印制特征与真品的产品标识是不同的，例如在真品和伪品的二维码中的定位块2031的黑白色区块分界处边界线的轮廓存在细微的差异，且比黑色或白色区块内部存在的差异的显现更为清晰。由此使得可以利用围绕线段的区域中的印制特征来验证产品标识的真伪。
包含线段特征的线段描述可以描述线段周边区域的图像梯度信息。这种图像梯度信息可以反映真品标识与伪品标识之间的差异。在示例中，通常真品标识图像中的线段会比较平滑，仿制伪品标识图像中的线段会比较粗糙；或者真品标签图像中的白色区域比较干净，仿制伪品标识图像中的白色区域内的印刷噪点较多，等等。所以通过使用线段描述符(也称为线段描述子)进而通过比对，取得线段印刷区域的可见差异，因此利用产品标识中固定区域的线段可以帮助区分真假产品标识。
在本发明的实施例中，提取线段特征的步骤102包括：对产品标识图像进行采样，以构建图像的尺度空间；对尺度空间中的每一层进行线特征提取，以在尺度空间中提取线段；在图像中截取围绕至少一个所提取的线段的区域，并将截取区域划分成多个条带子区域作为线段支持区域(LSR)；针对每一个条带子区域构建线段条带描述符(LBD)，并获得条带描述符矩阵。在该实施例中利用的线段描述方法是比较常用的线段条带描述符(Line Band descriptor，LBD)算法。但是本发明的实施例并不局限于此，也可以利用其它的线段描述方法，如尺度不变均值-标准差线段描述符(Scale invariant mean-standard deviation line segment descriptor，SMLSD)算法等。
在本发明的实施例中，通过以下步骤获得线段特征比较结果：计算与线条相对应的至少一个LBD的均值和方差，作为LBD值；将所计算的LBD值与真品标识图像的LBD参考值进行比较，获得线段特征比较结果。
以下以LBD算法为例，说明在本发明的实施例中提取线段特征和获得线段特征比较结果的一种具体实施方式。
线段描述是图像处理领域中一种常用的对图像中的指定线段生成唯一描述的方法。例如将一个线段周围的区域划分为多个子区域，针对子区域中的每个像素计算这个像素的梯度大小和方向，然后把这些计算得到的信息集合起来作为一个例如256维的线段描述向量。
线段特征是在描述图像中的线段，其与印刷生产的真品防伪二维码所具有的特定印制特征相关联。在本发明的实施例中的线段可以是产品的二维码。二维码通常包括编码部分和定位部分，编码部分因为没有足够长的稳定线段作为线段描述对象，所以只使用定位部分的三个大的黑色方块部分作为做线段描述的区域，如图2(a)和图2(b)所示的定位块2031。如图2(b)所示，针对每一个定位块，可以在其中的黑白色块之间、以及与其周边的空白区之间共找到三组边界轮廓，每一组边界轮廓都由两条横边界线和两条竖边界线组成，因而从中可以提取1～12条线段，每个二维码共有三个大的定位块，因此可以使用最多3*12＝36条线段。随着二维码的规格不同，可提取的线段数可能会发生变化。
应该注意的是，在实际印制过程中不可能印刷理想的一维线段，可能印制出线段宽度为多个像素，印刷过程还会影响与线段相邻区域的像素值。而针对印制线段的拍摄图像，会被采样成为一个有宽度的区域。伪品标识图像中的印制特征与真品相比是不一样的，例如可能会变得粗糙或模糊。真品与伪品的比较是针对同一位置的线段，然后分别对其生成线段描述向量。通过统计真品的线段描述子的均值，可以发现伪品或仿制品的线段描述向量和真品平均的描述向量距离很远。因此通过从对被验证产品标识图像中提取对应位置的线段并生成描述向量，再与以前针对正样本统计的均值(均值也是一个向量)求距离。这个距离是一个矢量，可以加入最终用于真假判别的描述向量。
在作为线段描述方法的LBD算法中，首先需要对输入的二维码图像构建N层图像金字塔(一般N大于或等于3)作为尺度空间。通过一组尺度因子和高斯模糊对图像进行N个下采样，就可以得到N层尺度金字塔。然后，利用线段检测算法(如EDLine算法)在尺度空间的每一层中提取线段，例如在理想情况下可在每一层中提取四条线段，然后根据各线段的位置与方向，分别组成作为线段特征的N个LineVec变量，每个LineVec变量包含N个金字塔层中的N条位于对应相同区域并具有一致方向的线段。针对LineVec中的每一条线段，需要进行线段支持区域(Line Support Region，LSR)的计算。LineVec中包含了多个位于不同金字塔层的线段，这些线段都要在对应层的图像中建立LSR。
通过线段构成的LSR线段支持区域可以计算线段描述符。围绕所要描述的线段截取一个图像区域，并把这个区域划分成m个条带状子区域，每个条带状子区域的宽度是w(例如，m＝5，w＝3，宽度单位：像素)。根据线段的方向建立新的LSR坐标系，线段的方向为d L是新LSR坐标系的横坐标轴，与d L正交的d ⊥为新LSR坐标系的纵坐标轴，线段的中点为LSR坐标系的原点；在条带上的每个像素都基于新建立的LSR计算梯度方向；在整个LSR上分别使用两个高斯函数作为全局权重系数f g和局部权重系数f L，即
Figure PCTCN2020078556-appb-000001
Figure PCTCN2020078556-appb-000002
其中，i和k是指在LSR中Bj条带的第i行和第k行，di表示LSR中第i行到中心行的距离，d’k表示第k行到Bj条带中心行的距离，σ g＝0.5(m·w-1)，σ l＝w。全局高斯窗口的目的是为降低远离线段的梯度的重要性，以此缓和在线段垂直方向上微小变化的敏感度。局部高斯窗口的目的是为降低边缘效应，避免了像素从一个条带到下一 个条带时描述符的突然改变。
对于每一个条带(BD)需要计算如下参数：
Figure PCTCN2020078556-appb-000003
Figure PCTCN2020078556-appb-000004
其中k表示条带Bj的第k行，λ＝f g(k)f l(k)为对应位置的全局高斯权重与局部高斯权重的乘积，j为条带编号，n为条带中包含的像素行数。由以上计算的四个参数可以构建条带描述矩阵(BDM)
Figure PCTCN2020078556-appb-000005
根据构建条带描述矩阵BDM可以得到以下最终的线段描述符：
Figure PCTCN2020078556-appb-000006
其中M为对应第m个条带的对应v行的均值向量，S为对应v行的方差向量，因此一个条带会产生8个描述向量，最终会有m*8个向量首尾连接形成描述子LBD。
在一个示例过程中，对于真品二维码样本(即正样本)，可使用一个定位块作为验证区域，使用内部的四条线段作为验证线段；对于四条线段生成线段支持区域(LSR)；对每个LSR生成LBD值；在所有正样本的对应定位块的对应线段上重复这个过程生成LBD值，求取所有对应LBD值的均值作为LBD参考值。
在另一个示例过程中，对于输入的被验证产品二维码图像，截取其中待验证的对应位置的线段；针对线段生成LSR；对于每个LSR 求取LBD值；计算LBD值与先前的示例所计算的LBD参考值之间的差值；取对应于四个线段的四个差值中的最大值作为输出的线段特征比较结果。
在本发明的实施例中，验证步骤103还可包括：使用经机器学习算法训练的分类器来验证被验证产品的产品标识的真伪；其中的分类器通过采用多个真品标识图像作为正样本、以及采用多个伪品标识图像作为负样本来训练。用于训练分类器的机器学习算法是可对特征向量进行分类的机器学习算法。正样本可以是多个真品的二维码标签，负样本可以是使用各种方式获得这些正样本的复制品。负样本具有相同的二维码标签，但其中的微点特征、印制特征都与真品的二维码标签相比存在细节差异。通过利用机器学习算法，可以使分类器区分产品真伪的准确度不断得以提高。
在本发明的实施例中，描述向量包括与以下各项中的至少一项相关的数据：被验证产品标识的LBD值与真品标识图像的LBD参考值之间的差值、所提取的微点特征与预先保存的微点特征之间的匹配率、匹配上的微点在图像坐标系中距离预先保存的微点的像素距离的统计参数、被验证产品的产品标识图像中与预先保存的微点特征不匹配的微点数目、以及被验证产品标识图像的质量评分。
在本发明的实施例中，提取微点特征的步骤101包括：使用图像处理技术从产品标识图像中提取微点的形状特征、位置特征、灰度特征、颜色特征中的至少一项。
在本发明的实施例中，产品标识是二维图形码；二维图形码中的线条是二维图形码中的定位块上的区块边界线。例如，图2(a)和图2(b)中示出了二维码203和其中的定位块2031。在本发明的实施例中，产品标识也可以是条形码或其它包含线条的图案。
在本发明的实施例中，验证产品真伪的方法还可包括：使用图像质量评估算法针对产品标识图像生成图像质量评分；以及基于图像质 量评分、所提取的微点特征和线段特征来验证被验证产品的产品标识的真伪。
在图像处理技术上有很多方法可以用于评估目标图像的图像质量，比如标识图像中的明暗分部是不是合理，是不是过亮过暗，图片是不是足够清晰。通过评估获取标识图像生成图像质量评分。而图像质量与图像中的微点特征、线段特征和/或其它图像特征的提取是有关联性的，也会影响对图像中的微点特征、线段特征和/或其它图像特征的验证识别，因此在产品标识验证中结合图像质量评分，将有助于更加精准地验证产品标识的真伪。
图2(a)是在实施例中将微点特征嵌入到产品二维码中的示意图，其中的黑色微点特征202因其尺寸太小而在图中未显现。在微点特征的生成过程中，首先通过算法生成微点的特定高维随机分布图201，作为所有微点特征的位置分布、灰度分布、颜色分布和微观形态中至少之一的分布特性，在相同类或相同批次的产品可以共同遵循某一分布特性，其中各个产品又具有其它不同的微点特征以示区分。例如，不同批次的产品可采用不同的随机分布图，相同批次的不同产品则采用不同的微点。然后，利用该算法对微点的随机分布图进行采样，针对每个产品(或产品标识或标签)生成具有唯一标识性的微点特征202，再根据预定的回避规则将所生成的微点特征嵌入到产品的数字二维标识203(如快速响应矩阵码，即二维码)中，并将嵌入有微点特征的二维码印制在产品表面或产品包装的表面作为产品标识，或印制在产品标签的表面上，形成具有微点的数字式产品标识(ID)。回避规则可以限制微点的特定位置分布、灰度分布和颜色分布中的至少一项。例如，位置分布回避规则可确保仅在二维码的白色块中生成黑色或深色微点，灰度分布或颜色分布回避规则可分布确保微点的灰度或颜色满足某种灰度和饱和度限制，不会干扰二维码的白色块。这些回避规则共同作用而确保二维码本身的读取不会受到嵌入微点特征 的影响，并且使得二维码在添加了微点特征之后仍然满足相应的国家标准和/或国际标准。
在一些实施例中，也可以将白色微点特征202嵌入到二维码203的黑色块中，回避规则将要生成的微点仅限制在二维码的黑色块中，使得二维码在添加了微点特征之后仍然满足相应的国家标准和/或国际标准。白色微点在二维码的黑色块中保持最高的对比度，并且在印制过程中通过印刷喷墨中的短停顿来产生白色微点。
微点特征的构成包括最基本的二维坐标(X,Y)作为位置特征，还可以包括其他可选特征，如颜色、灰度、形状等等。通常，微点特征的不可再现性和防伪性能首先是通过微点的二维位置的随机分布来实现。而微点的颜色、灰度或形状特征可用于进一步提高产品的防伪性能。随机分布微点特征也可以形成随机分布的微点纹理特征。
在完成产品标识的制作后或是在制作过程中，需要将产品标识上的微点特征信息保存数据库中，以用于后续的产品真伪验证。保存的微点特征信息例如包括随机分布的位置特征、以及如其颜色、灰度或形状等其它特征。
在图2(a)中的每个二维码203包含三个定位块2031，分别位于左上角、右上角和左下角。定位块也被称为位置探测图形，用于标记二维码的矩形大小。如图2(b)所示，每个定位块2031包括位于中部的黑色矩形实心块、位于中间的白色方框块和位于外部的黑色方框块，这三个色块构建三组包含区块边界线的轮廓，每一组轮廓的区块边界线都由两条横线和两条竖线组成。这些区块边界线可以作为实施例中的线段描述方法的对象。但本发明并不局限于此，线段描述方法的对象可以是产品标识中的任何线段。
作为微点特征的一个示例，图3(a)和图3(b)示出采用均匀分布函数作为微点的随机分布函数时的概率密度函数的图像、以及从随机分布中抽样得到的微点分布图。其中均匀分布函数的概率密度函数为：
PDF(x,y)＝const
在图3(a)的概率密度函数的图像中，Z向坐标为概率密度，横向坐标X和纵向坐标Y指示微点的位置(x,y)。图3(b)的微点分布图是生成微点坐标(x,y)时从图3(a)的随机分布图中抽样得到的，图中显示随机分布的黑色微点。
图4示出根据本发明的第二实施例的用于验证产品真伪的方法400的流程图。第二实施例也是基于线段特征和微点特征来验证产品标识的实施例。在方法400中，首先获取被验证产品的产品标识的图像或图片(步骤401)。例如，用户在购买产品后，可以对包含条形码或二维码的产品标识部分进行拍照，将拍照获得的产品标识的图像传送给验证方或验证设备，以便针对所获得的产品标识的图像进行真伪验证。在本实施例中，对于包含二维码的图像的处理包括两个部分，即，线段处理部分(包括步骤402-405)、以及微点处理部分(包括步骤406-409)。在这两个处理部分中，首先需要对图像进行预处理(步骤402和406)，例如对图像中包含有效特征(例如二维码)的部分区域进行明暗调节、有效部分截取、对比度增强，图像锐化、图片归一化等图像处理技术中的常用预处理方法。
在图像线段处理部分中，在预处理步骤402之后，使用图像处理技术中的线段描述算法来描述产品标识图像中至少一条线段及线段相邻的空白区域(步骤403)。由于二维码编码部分中的黑白色块是随着二维码信息变化，不具备稳定的黑白色块和边界线，因此可以使用二维码中具有稳定色块的定位块(如图2(a)和图2(b)所示)上的线段来生成特定的线段描述。然后，将生成的线段描述与真品二维码图像中的对应线段的线段描述相比较，得到线段特征比较结果(步骤404)。
真品二维码图像中的对应线段的线段描述是根据真品二维码图像来生成的，其中包含与对应区域的印制相关的印制特征。在伪品标 识中，复制真品标识的二维码时会导致包含线段的区域部分中的印制特征丢失。在步骤404的比较中，可以利用被验证产品二维码图像的线段描述与一个真品二维码图像的线段描述进行比较，以得到线段特征比较结果，也可以是通过与多个真品二维码图像中相同位置的线段描述的平均值进行比较来得到线段特征比较结果。伪品标识中复制的二维码在对应的线段部分由于会损失细微的印制特征，因此与真品二维码图像的线段描述之间会有明显的差异，这种差异本质上是黑白色块分解部分因为复制和不同的印刷设备纸张等带来的围绕对应线段周围的图像差异。
在步骤405，输出线段特征比较结果作为线段真伪程度的描述。在一个示例中，线段特征比较结果是被验证产品标识的LBD值与真品标识图像的LBD参考值之间的差值，其中的真品标识图像的LBD参考值可以是多个真品标识图像的LBD参考值的平均值。在另一个示例中，由于已具有真品标识的二维码中线段的描述向量，例如可以进一步求出针对伪品标识中复制的二维码图像中对应线段生成的线段描述与真品标识图像中的线段描述之间的向量差(欧氏距离)。
在微点处理部分中，在预处理步骤406之后，利用微点提取算法提取图像中的微点特征(步骤407)，其中可利用图像处理技术来读取被验证的产品标识所处区域内随机分布的微点特征，包含基于微点的位置、大小、颜色或灰度等中至少之一的统计数据。例如，通过统计每个微点区域的大小(如每个微点包含的像素数目)，或者统计每个区域的平均RGB三通道值，得到图像中微点区域的灰度信息。然后，在数据库中提取预先保存的产品标识真品所具有的相应微点特征，并将所读取的微点特征与数据库中的微点特征进行比较(步骤408)，由此输出微点特征比较结果(步骤409)，作为判断产品标识的真伪的依据之一。
然后，利用步骤405输出的线段特征比较结果、以及步骤409输 出的微点特征比较结果(其中不同的统计数据可做归一化处理)，组成描述向量X(步骤410)。例如，将输出的线段特征比较结果作为一个特征维度如x1。微点特征比较结果可包括在对比中微点特征的匹配量化处理后得到的若干个统计数据，如在目标二维码中找到的微点与数据库中的对应二维码的微点匹配上的百分比作为x2，匹配上的微点在图像坐标系中距离数据库中的微点的像素距离的统计参数(如均值和方差)作为x3和x4，没有匹配上的微点的惩罚(误匹配)作为x5等。利用以上这些信息可以组成关于被验证产品标识的一个五维的描述向量。另外，也可以在描述向量中增加图像的锐度、图像的明暗分布的均值和方差等图像质量描述作为其它的特征维度。
在本发明的实施例中，可以针对所有收集的正样本和负样本的产品标识的图像进行处理而得到相应的线段特征和LBD差值、以及微点统计特征，进而得到相应的描述向量作为样本数据集合，其中一部分(如80％)可作为训练集，用于训练分类器；另一部分(如20％)可作为测试集。基于该样本数据集合，可以采用比较流行的机器学习算法对分类器进行训练和测试，可选择的分类器种类例如有支持向量机(SVM)、提升树(Boost Tree)、决策树、浅层神经网络、k近邻算法、随机森林等。预先训练好的分类器可以基于步骤410得到的描述向量中的描述特征(包括：线段特征比较结果，以及微点统计特征)对被验证产品标识进行判别分类(步骤411)，由此输出关于被验证产品标识的真伪判定结果(步骤412)。
在训练图像样本准备过程中，可以使用市面上的多种不同手机对若干真品产品标识在不同的光照环境下拍照，将其所得图像作为正样本；使用市面上的多种不同手机对制作的非真标签在不同的光照环境下拍照，将其所得图像作为负样本；然后，随机地把正负样本按照比例划分为训练样本集和测试样本集。通过利用不断积累的产品标识样本对分类器进行不断的训练，从而可以使分类器在实际产品标识验证 中给出更加精准的真伪判别。
根据本发明的上述实施例所提供的产品标识验证方法，采用例如包含二维码的产品标识、微点特征、以及利用线段特征的组合，可以进一步改善真伪产品标识的验证准确度。
图5示出根据本发明的第三实施例的用于基于微点特征和图像质量评分验证产品真伪的方法500的流程图。如图5所示，验证方法500包括：从被验证产品的产品标识的图像中提取产品标识上随机分布的微点特征(步骤501)；使用图像质量评估算法针对产品标识的图像生成图像质量评分(步骤502)；以及基于图像质量评分和所提取的微点特征来验证被验证产品的产品标识的真伪(步骤503)。
可以利用图像处理技术中的很多方法来评估标识图像的图像质量，以生成图像质量评分。而图像质量与图像中的微点特征和/或其它图像特征的提取是有关联性的，也会影响对图像中的微点特征和/或其它图像特征的验证识别，因此在产品标识验证中结合图像质量评分，将有助于更加精准地验证产品标识的真伪，尤其是可以解决在针对拍摄图像质量好的假产品标签和拍摄图像质量差的真产品标签被错误验证之问题。
例如，对防伪二维码图像的真假判别与图像拍摄质量是相关的，特别是在移动端的应用中，各种品牌手机呈现出不同的照片处理取向；不同的拍摄场景下，同样的产品标签由于不同拍摄手机的不同处理取向而使拍摄图像呈现出不同的图像细节。因此，如果对于所有图像的真假识别验证中都采用简单一致的真假判别阈值，可能会严重的降低真假验证的准确率；而基于不同的图像质量评分，例如可以通过采用不同的真假判别阈值来进行真假验证，由此可以提高产品标签的验证精准度。
在本发明的实施例中，验证步骤503包括：将所提取的微点特征与在真品标识图像的微点特征进行比较，获得微点特征比较结果；基 于图像质量评分和微点特征比较结果，组成关于产品标识的描述向量；以及基于描述向量，使用分类器判断被验证产品的产品标识的真伪。
在本发明的实施例中，验证步骤503还包括：使用经机器学习算法训练的分类器来验证被验证产品的产品标识的真伪；其中通过采用多个真品标识图像作为正样本、以及采用多个伪品标识图像作为负样本来训练分类器；用于训练分类器的机器学习算法是可对特征向量进行分类的机器学习算法。正样本可以是多个真品的二维码标签，负样本可以是使用各种方式获得这些正样本的复制品。负样本具有相同的二维码标签，但其中的微点特征、和微点特征与图像质量的关联性都与真品的二维码标签相比存在细微差异。通过利用机器学习算法，可以使分类器区分产品真伪的准确度不断得以提高。例如，通过利用机器学习算法使分类器学习在不同的图像质量评分下采用不同的真假判别阈值来进行真假验证。
可以采用图像处理技术中的各种图像质量评估算法来生成图像质量评分。以下以空间域无参考图像质量评估(BRISQUE)算法为例，但本发明并不限于此。
BRISQUE算法包括：对输入的原始图像进行规范化以原始图像中提取自然场景统计(NSS)、计算特征向量及利用机器学习算法(如支持向量机)估计图像质量评分。
自然图像的像素强度的分布不同于失真图像的像素强度的分布。当对于输入的原始图像进行归一化像素强度处理并计算这些归一化强度上的分布时，这种分布差异更加明显。特别是在归一化之后，自然图像的像素强度遵循高斯分布(贝尔曲线)，而不自然或失真图像的像素强度不遵循高斯分布(贝尔曲线)。由此可以得到图像进行图像质量评分。
去均值对比归一化(Mean Subtracted Contrast Normalization， MSCN)方法是一种用于对图像进行规范化图的方法。
设W和H分别为输入图像的宽和高，按以下计算式将像素(i,j)处的图像强度I(i,j)变换为亮度
Figure PCTCN2020078556-appb-000007
以计算出MSCN参数：
Figure PCTCN2020078556-appb-000008
μ(i,j)和σ(i,j)分别为局部平均和局部方差。
μ＝W*I
Figure PCTCN2020078556-appb-000009
w为高斯模糊窗口，μ是原始图像I做高斯模糊后的结果，σ为原始图像减去高斯模糊图像μ后取平方再取高斯模糊后求平方根的结果。
由于自然图像与失真图像的差异不仅限于像素强度分布，还包括相邻像素之间的关系，因此还还需要求取在水平(H)、垂直(V)、左对角线(D1)、右对角线(D2)四个方向上与相邻像素的相邻关系，即如下的相邻像素乘积参数：
Figure PCTCN2020078556-appb-000010
Figure PCTCN2020078556-appb-000011
Figure PCTCN2020078556-appb-000012
Figure PCTCN2020078556-appb-000013
然后，使用上述计算的MSCN参数和四个相邻像素乘积参数来计算特征向量。例如，通过将MSCN参数拟合到广义高斯分布(GGD)来计算36×1特征向量的前两个特征向量元素(形状参数和方差)，将四个相邻像素乘积参数的每一个拟合到非对称广义高斯分布(AGGD)来计算特征向量的四个特征向量元素(形状参数，均值，左方差，右方差)。最终得到了18个特征向量元素作为用于描述整幅输入图像的图像描述子。
现有的图像质量数据集中已经包含了对无失真图像和不同程度 失真图像的人为打分评价，使用公开的图像质量评估数据集中对图像质量的评分为y，BRISQUE算法生成的描述子为x，配合任意的线性回归算法比如SVR(支持向量回归)即可训练得到一个回归模型，这个回归模型在处理输入图像生成描述子后交由已经训练好的模型，就可以得到描述图像失真程度的分数。在一个示例中，图像质量评分为0代表没有失真，100表示严重失真，输出的图像质量评分在0～100之间。图像质量评分也可以归一化到0至1之间，作为描述向量中的一项数据。
在一个典型的机器学习应用示例中，首先将图像转换为特征向量。然后，将训练数据集中的所有图像的特征向量和输出(质量得分)放入支持向量机(SVM)中进行训练。使用支持向量机工具包(LIBSVM)加载所有特征向量元素及图像质量得分进行训练，训练过程中，特征向量首先被缩放到-1到1，训练得出的模型用于预测图像质量。经过训练模型预测，可以得出每种失真的最终质量得分。
在本发明的实施例中，描述向量可包括与以下各项中的至少一项相关的数据：所提取的微点特征与预先保存的微点特征之间的匹配率、匹配上的微点在图像坐标系中距离预先保存的微点的像素距离的统计参数、被验证产品的产品标识的图像中与预先保存的微点特征不匹配的微点数目、以及图像质量评分。
在本发明的实施例中，图像质量评分是基于标识图像中的至少一个子区域的图像质量数据生成的，图像质量数据与以下图像失真类型中的至少一种相关：高斯叠加噪声失真、色彩成分失真、光照失真、空间相关性失真、脉冲失真、图像运动模糊失真、低光照图像噪声失真、图像文件压缩失真和对比度失真。
在图像质量评估中，可以利用输入参考图像来评估现有图像的质量，也可以直接进行无参考图像的图像质量评估。在图像质量评估中利用图像评估算法来处理输入图像，以生成表示图像拍摄质量的图像 质量评分。通常直观的评价包括图像是否清晰、图像中的噪点数量、光线明暗是否适中。而图像质量评估技术则可以把这些直观评价量化为多种类型的图像失真。以上所列举的图像失真类型只是常见的一部分类型，本发明并不局限于利用上述图像失真类型来生成图像质量评分。
在本发明的实施例中，提取微点特征的步骤501包括：使用图像处理技术从标识图像中提取微点的形状特征、位置特征、灰度特征、颜色特征中的至少一项。
图6示出根据本发明的第四实施例的用于基于微点特征和图像质量评分验证产品真伪的方法流程图。在方法600中，首先获取被验证产品的产品标识的图像或图片(步骤601)。其中的步骤602-605为微点处理部分，分别与图4所示的第二方法实施例中的步骤406-409相同，在此不再详述。步骤606-608为图像质量评估部分。在图像质量评估部分中，首先需要对图像进行预处理(步骤606)，例如对图像中包含有效特征(例如二维码)的部分区域进行明暗调节、有效部分截取、对比度增强，图像锐化、图片归一化等图像处理技术中的常用预处理方法。然后，在预处理后的图像中提取图像的图像质量描述(步骤607)，其中可以使用图像处理技术中的各种图像描述方法来描述目标图像质量，比如明暗分布是不是合理，是不是过亮过暗，图像是不是足够清晰或包含失真。步骤607可以是如上所述利用BRISQUE算法产生图像质量描述的过程，在此不再详述。在步骤608，输出图像质量描述。
然后，利用步骤605输出的微点特征比较结果(其中不同的统计数据可做归一化处理)、以及步骤608输出的图像质量描述，组成描述向量X(步骤610)。例如，将输出的图像质量评分作为一个特征维度如x1。微点特征比较结果可包括在对比中微点特征的匹配量化处理后得到的若干个统计数据，如在目标二维码中找到的微点与数据 库中的对应二维码的微点匹配上的百分比作为x2，匹配上的微点在图像坐标系中距离数据库中的微点的像素距离的统计参数(如均值和方差)作为x3和x4，没有匹配上的微点的惩罚(误匹配)作为x5等。利用以上这些信息可以组成关于被验证产品标识的一个五维的描述向量X。
在本发明的实施例中，可以针对所有收集的正样本和负样本的产品标识的图像进行处理而得到相应的图像质量评分、以及微点统计特征，进而得到相应的描述向量作为样本数据集合，其中一部分(如80％)可作为训练集，用于训练分类器；另一部分(如20％)可作为测试集。基于该样本数据集合，可以采用比较流行的机器学习算法对分类器进行训练和测试，可选择的分类器种类例如有支持向量机(SVM)、提升树(Boost Tree)、决策树、浅层神经网络、k近邻算法、随机森林等。预先训练好的分类器可以基于步骤610得到的描述向量中的描述特征(包括：微点统计特征，以及图像质量评分)对被验证产品标识进行判别分类(步骤611)，由此输出关于被验证产品标识的真伪判定结果(步骤612)。
根据本发明的上述实施例所提供的产品标识验证方法，采用例如包含二维码的产品标识、微点特征、以及利用图像质量描述的组合，可以进一步改善真伪产品标识的验证准确度，尤其是可以解决在针对拍摄图像质量好的假产品标签和拍摄图像质量差的真产品标签被错误验证之问题，也就是说，假产品标签的图像质量再好也仍然可以被准确识别为伪品，真产品标签的图像质量再差也仍然可以被验证为真品标签，因而允许用户可以使用各种手机或相机在各种外部环境中拍摄产品标识图像。
图7示出根据本发明的第五实施例的用于验证产品真伪的方法流程图。方法700是基于微点特征、线段特征和图像质量评分来验证产品真伪。在方法700中，首先获取被验证产品的产品标识的图像或 图片(步骤701)。图7中的步骤702-705为线段处理部分，分别与图4所示的第二方法实施例中的步骤402-405相同，在处理后输出产品标识图像的线段特征比较结果。图7中的步骤706-709为微点处理部分，分别与图4所示的第二方法实施例中的步骤406-409相同，在处理后输出产品标识的微点特征比较结果。图7中的步骤710-712为图像质量评估部分，分别与图6所示的第四方法实施例中的步骤606-608相同，在评估后输出产品标识的图像质量描述，在此不再详述。
然后，利用步骤705输出的线段特征比较结果、步骤709输出的微点特征比较结果(其中不同的统计数据可做归一化处理)、以及步骤712输出的图像质量描述，组成描述向量X(步骤713)。例如，将输出的线段特征比较结果作为一个特征维度如x1。微点特征比较结果可包括在对比中微点特征的匹配量化处理后得到的若干个统计数据，如在目标二维码中找到的微点与数据库中的对应二维码的微点匹配上的百分比作为x2，匹配上的微点在图像坐标系中距离数据库中的微点的像素距离的统计参数(如均值和方差)作为x3和x4，没有匹配上的微点的惩罚(误匹配)作为x5等。将输出的图像质量评分作为一个特征维度如x6。利用以上这些信息可以组成关于被验证产品标识的一个六维的描述向量X。
在本发明的实施例中，可以针对所有收集的正样本和负样本的产品标识的图像进行处理而得到相应的微点统计特征、线段特征以及图像质量评分，进而得到相应的描述向量作为样本数据集合，其中一部分(如80％)可作为训练集，用于训练分类器；另一部分(如20％)可作为测试集。基于该样本数据集合，可以采用比较流行的机器学习算法对分类器进行训练和测试，可选择的分类器种类例如有支持向量机(SVM)、提升树(Boost Tree)、决策树、浅层神经网络、k近邻算法、随机森林等。预先训练好的分类器可以基于步骤713得到的描述向量中的描述特征(包括：微点统计特征，线段特征，以及图像质 量评分)对被验证产品标识进行判别分类(步骤714)，由此输出关于被验证产品标识的真伪判定结果(步骤715)。
图8示出根据本发明的第六实施例的用于验证产品真伪的装置的结构方框图。装置800包括：微点特征提取模块(801)，用于从被验证产品的产品标识图像中提取产品标识上随机分布的微点特征；线段特征提取模块(802)，用于使用图像处理技术的线段描述方法，从图像中包含线条的至少一个区域提取产品标识中的线段特征；线段特征是基于利用线段描述方法所生成的线段描述符；以及验证模块(803)，用于基于所提取的微点特征和线段特征来验证被验证产品的产品标识的真伪。
根据该实施例，装置800还可包括：图像质量评分生成模块(804)，用于使用图像质量评估算法针对产品标识的图像生成图像质量评分；验证模块(803)基于图像质量评分生成模块(804)生成的图像质量评分、微点特征提取模块(801)所提取的微点特征和线段特征提取模块(802)提取的线段特征来验证被验证产品的产品标识的真伪。
图9示出根据本发明的第七实施例的用于验证产品真伪的装置的结构方框图。装置900包括：微点特征提取模块(901)，用于从被验证产品的产品标识的图像中提取产品标识上的微点特征；图像质量评分生成模块(902)，用于使用图像质量评估算法针对产品标识的图像生成图像质量评分；以及验证模块(903)，用于基于图像质量评分生成模块(902)生成的图像质量评分和微点特征提取模块(901)所提取的微点特征来验证被验证产品的产品标识的真伪。
图8和图9所示的装置800和900可以利用软件、硬件或软硬件结合的方式来实现，并且可以被设计成包括相应的模块以实施本发明的上述用于验证产品真伪的各方法实施例。
本发明的实施例还可提供一种用于验证产品的真伪的设备，其包 括：用于存储指令的存储器；以及耦合到存储器的处理器，这些指令在由处理器执行时使得处理器执行根据本发明的上述实施例的产品验证方法。存储器中还可存储有数据库，该数据库可至少包括真品产品标识的微点特征和线段特征，用于与被验证产品标识的微点特征和线段特征进行比对。
在该实施例的存储器中还可存储样本库，该样本库包括作为正样本的多个真品标识图像、以及作为负样本的多个伪品标识图像。处理器被配置用于利用样本库中的至少一部分样本来训练用于验证产品标识的分类器。
在分类器的训练过程中，首先从存储器中的样本库中随机取出一组正/负样本的特征向量并提供给分类器，由分类器进行真伪分类处理；然后，根据样本已知的真值计算分类结果的损失值，并判断该损失值是否小于预设的阈值；当判断该损失值仍大于或等于阈值时，则根据该损失值来更新分类器的参数，然后再由分类器根据更新的参数继续就那些真伪分类处理。不断循环上述过程，直到损失值被判断为小于阈值时，停止分类器的训练。
根据本发明的上述实施例所提供的产品验证方法和设备，采用例如包含二维码或条形码的产品标识、微点特征、线段特征以及利用图像质量描述的各种组合方案，可以大大改善真伪产品标识的验证准确度，允许购买产品的用户在各种光照条件下使用各种手机或相机拍摄产品标识图像并进行准确验证。
以上公开的本发明的实施例均为示例性的，而非限制性的。本领域技术人员应当理解，以上公开的各个实施例可以在不偏离发明实质的情况下做出各种变型、修改和改变，这些变型、修改和改变都应当落入在本专利的保护范围之内。本专利的保护范围应由所附的权利要求书来限定。

Claims (21)

  1. 一种用于验证产品的真伪的方法，所述产品的产品标识上具有随机分布的微点，所述方法包括：
    从被验证产品的产品标识的图像中提取所述产品标识上的微点特征；
    使用图像处理技术的线段描述方法，从所述图像中包含线条的至少一个区域提取所述产品标识的线段特征；其中，所述线段特征是基于利用所述线段描述方法所生成的线段描述符；以及
    基于所提取的所述微点特征和所述线段特征来验证所述被验证产品的产品标识的真伪。
  2. 根据权利要求1的方法，其中，所述验证步骤包括：
    将所提取的微点特征与在真品标识图像的微点特征进行比较，获得微点特征比较结果；
    将所提取的线段特征与在真品标识图像的线段特征进行比较，获得线段特征比较结果；
    基于所述线段特征比较结果和所述微点特征比较结果，组成关于所述产品标识的描述向量；以及
    基于所述描述向量，使用分类器判断所述被验证产品的产品标识的真伪。
  3. 根据权利要求1的方法，其中，所提取的线段特征至少包括与所述产品标识中包含线条的区域的印制相关的印制特征；其中，所述印制特征是与正样本的产品标识的印制过程中使用的纸张、油墨、印制设备中的至少一项相关联的特征。
  4. 根据权利要求1的方法，其中，所述提取线段特征的步骤包括：
    对所述图像进行采样，以构建所述图像的尺度空间；
    对所述尺度空间中的每一层进行线特征提取，以在所述尺度空间中提取线段；
    在所述图像中截取围绕至少一个所提取的线段的区域，并将所述截取区域划分成多个条带子区域作为线段支持区域(LSR)；
    针对每一个条带子区域构建线段条带描述符(LBD)，并获得条带描述符矩阵。
  5. 根据权利要求4的方法，其中，通过以下步骤获得所述线段特征比较结果：
    计算与所述线条相对应的至少一个LBD的均值和方差，作为LBD值；
    将所述LBD值与真品标识图像的LBD参考值进行比较，获得线段特征比较结果。
  6. 根据权利要求2的方法，其中，所述验证步骤还包括：
    使用经机器学习算法训练的分类器来验证被验证产品的产品标识的真伪；
    其中，所述分类器通过采用多个真品标识图像作为正样本、以及采用多个伪品标识图像作为负样本来训练的；
    其中，用于训练所述分类器的机器学习算法是可对特征向量进行分类的机器学习算法。
  7. 根据权利要求2的方法，其中，所述描述向量包括与以下各项中的至少一项相关的数据：被验证产品标识的LBD值与真品标识 图像的LBD参考值之间的差值、所提取的微点特征与预先保存的微点特征之间的匹配率、匹配上的微点在图像坐标系中距离预先保存的微点的像素距离的统计参数、所述被验证产品的产品标识的图像中与预先保存的微点特征不匹配的微点数目、以及所述被验证产品标识图像的质量评分。
  8. 根据权利要求1的方法，其中，所述提取微点特征的步骤包括：
    使用图像处理技术从所述图像中提取所述微点的形状特征、位置特征、灰度特征、颜色特征中的至少一项。
  9. 根据权利要求1的方法，其中，所述产品标识是二维图形码；所述线条是所述二维图形码中的定位块上的区块边界线。
  10. 根据权利要求1至9中任一项的方法，还包括：
    使用图像质量评估算法针对所述产品标识的图像生成图像质量评分；以及
    基于所述图像质量评分、所提取的所述微点特征和所述线段特征来验证所述被验证产品的产品标识的真伪。
  11. 一种用于验证产品的真伪的方法，所述产品的产品标识上具有随机分布的微点，所述方法包括：
    从被验证产品的产品标识的图像中提取所述产品标识上的微点特征；
    使用图像质量评估算法针对所述产品标识的图像生成图像质量评分；以及
    基于所述图像质量评分和所提取的所述微点特征来验证所述被 验证产品的产品标识的真伪。
  12. 根据权利要求11的方法，其中，所述验证步骤包括：
    将所提取的微点特征与在真品标识图像的微点特征进行比较，获得微点特征比较结果；
    基于所述图像质量评分和所述微点特征比较结果，组成关于所述产品标识的描述向量；以及
    基于所述描述向量，使用分类器判断所述被验证产品的产品标识的真伪。
  13. 根据权利要求12的方法，其中，所述验证步骤还包括：
    使用经机器学习算法训练的分类器来验证被验证产品的产品标识的真伪；
    其中，所述分类器通过采用多个真品标识图像作为正样本、以及采用多个伪品标识图像作为负样本来训练的；
    其中，用于训练所述分类器的机器学习算法是可对特征向量进行分类的机器学习算法。
  14. 根据权利要求12的方法，其中，所述描述向量包括与以下各项中的至少一项相关的数据：所提取的微点特征与预先保存的微点特征之间的匹配率、匹配上的微点在图像坐标系中距离预先保存的微点的像素距离的统计参数、所述被验证产品的产品标识的图像中与预先保存的微点特征不匹配的微点数目、以及所述图像质量评分。
  15. 根据权利要求11的方法，其中，所述提取微点特征的步骤包括：
    使用图像处理技术从所述图像中提取所述微点的形状特征、位置 特征、灰度特征、颜色特征中的至少一项。
  16. 根据权利要求11的方法，其中，所述图像质量评分是基于所述图像中的至少一个子区域的图像质量数据生成的，所述图像质量数据与以下图像失真类型中的至少一种相关：
    高斯叠加噪声失真、色彩成分失真、光照失真、空间相关性失真、脉冲失真、图像运动模糊失真、低光照图像噪声失真、图像文件压缩失真和对比度失真。
  17. 一种用于验证产品的真伪的装置，所述产品的产品标识上具有随机分布的微点，所述装置包括：
    微点特征提取模块，用于从被验证产品的产品标识的图像中提取所述产品标识上的微点特征；
    线段特征提取模块，用于使用图像处理技术的线段描述方法，从所述图像中包含线条的至少一个区域提取所述产品标识的线段特征；其中，所述线段特征是基于利用所述线段描述方法所生成的线段描述符；以及
    验证模块，用于基于所提取的所述微点特征和所述线段特征来验证所述被验证产品的产品标识的真伪。
  18. 根据权利要求17的装置，还包括：
    图像质量评分生成模块，用于使用图像质量评估算法针对所述产品标识的图像生成图像质量评分；
    其中，所述验证模块基于所述图像质量评分、所提取的所述微点特征和所述线段特征来验证所述被验证产品的产品标识的真伪。
  19. 一种用于验证产品的真伪的装置，所述产品的产品标识上具 有随机分布的微点，所述装置包括：
    微点特征提取模块，用于从被验证产品的产品标识的图像中提取所述产品标识上的微点特征；
    图像质量评分生成模块，用于使用图像质量评估算法针对所述产品标识的图像生成图像质量评分；以及
    验证模块，用于基于所述图像质量评分和所提取的所述微点特征来验证所述被验证产品的产品标识的真伪。
  20. 一种用于验证产品的真伪的设备，所述产品的产品标识上具有随机分布的微点，所述设备包括：
    用于存储指令的存储器；以及
    耦合到所述存储器的处理器，所述指令在由所述处理器执行时使得所述处理器执行根据权利要求1至16中任一项所述的方法。
  21. 一种计算机可读存储介质，其上存储由可执行指令，所述可执行指令在由计算机执行时使得所述计算机执行根据权利要求1至16中任一项所述的方法。
PCT/CN2020/078556 2020-03-10 2020-03-10 用于验证产品真伪的方法和装置 WO2021179157A1 (zh)

Priority Applications (3)

Application Number Priority Date Filing Date Title
DE112020005864.6T DE112020005864T5 (de) 2020-03-10 2020-03-10 Verfahren und Einrichtung zur Verifizierung der Authentizität eines Produkts
CN202080098278.0A CN115244542A (zh) 2020-03-10 2020-03-10 用于验证产品真伪的方法和装置
PCT/CN2020/078556 WO2021179157A1 (zh) 2020-03-10 2020-03-10 用于验证产品真伪的方法和装置

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
PCT/CN2020/078556 WO2021179157A1 (zh) 2020-03-10 2020-03-10 用于验证产品真伪的方法和装置

Publications (1)

Publication Number Publication Date
WO2021179157A1 true WO2021179157A1 (zh) 2021-09-16

Family

ID=77670418

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2020/078556 WO2021179157A1 (zh) 2020-03-10 2020-03-10 用于验证产品真伪的方法和装置

Country Status (3)

Country Link
CN (1) CN115244542A (zh)
DE (1) DE112020005864T5 (zh)
WO (1) WO2021179157A1 (zh)

Cited By (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20220092609A1 (en) * 2020-09-22 2022-03-24 Lawrence Livermore National Security, Llc Automated evaluation of anti-counterfeiting measures
CN116092061A (zh) * 2023-04-12 2023-05-09 深圳企业云科技股份有限公司 标签验证方法和装置
CN116227524A (zh) * 2023-05-05 2023-06-06 杭州蚂蚁酷爱科技有限公司 防伪码的生成与验证方法、基于标签的防伪系统
CN117037159A (zh) * 2023-10-09 2023-11-10 网思科技股份有限公司 基于卷积神经网络的油画真伪鉴别方法、装置及存储介质
CN117057826A (zh) * 2023-10-12 2023-11-14 深圳市一航网络信息技术有限公司 一种基于区块链技术的产品溯源验证方法
CN118190954A (zh) * 2024-05-15 2024-06-14 青岛澳科顺诚包装有限公司 一种镭射膜防伪图案监控质检系统
CN118644259A (zh) * 2024-08-12 2024-09-13 南通飞隼信息科技有限公司 一种基于仿真品识别的数据挖掘方法及系统

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116934719B (zh) * 2023-07-26 2024-02-13 苏州德斯米尔智能科技有限公司 一种用于带式输送设备的自动检测系统

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101350072A (zh) * 2007-07-16 2009-01-21 樊晓青 一种自我认证信息的印刷防伪产品及防伪方法
CN104537544A (zh) * 2015-01-04 2015-04-22 蒲亦非 带覆盖层的基于背景纹理特征提取的商品二维码防伪方法及系统
CN109478317A (zh) * 2016-07-26 2019-03-15 高通股份有限公司 用于合成图像的系统和方法
CN110070159A (zh) * 2019-04-19 2019-07-30 宁波高新区意融科技有限公司 一种基于实物和实物图像对比的防伪标签及其鉴别方法
JP2020013525A (ja) * 2018-07-04 2020-01-23 凸版印刷株式会社 認証装置、認証システム、および、認証方法

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101350072A (zh) * 2007-07-16 2009-01-21 樊晓青 一种自我认证信息的印刷防伪产品及防伪方法
CN104537544A (zh) * 2015-01-04 2015-04-22 蒲亦非 带覆盖层的基于背景纹理特征提取的商品二维码防伪方法及系统
CN109478317A (zh) * 2016-07-26 2019-03-15 高通股份有限公司 用于合成图像的系统和方法
JP2020013525A (ja) * 2018-07-04 2020-01-23 凸版印刷株式会社 認証装置、認証システム、および、認証方法
CN110070159A (zh) * 2019-04-19 2019-07-30 宁波高新区意融科技有限公司 一种基于实物和实物图像对比的防伪标签及其鉴别方法

Cited By (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20220092609A1 (en) * 2020-09-22 2022-03-24 Lawrence Livermore National Security, Llc Automated evaluation of anti-counterfeiting measures
CN116092061A (zh) * 2023-04-12 2023-05-09 深圳企业云科技股份有限公司 标签验证方法和装置
CN116092061B (zh) * 2023-04-12 2023-09-22 深圳企业云科技股份有限公司 标签验证方法和装置
CN116227524A (zh) * 2023-05-05 2023-06-06 杭州蚂蚁酷爱科技有限公司 防伪码的生成与验证方法、基于标签的防伪系统
CN116227524B (zh) * 2023-05-05 2023-08-08 杭州蚂蚁酷爱科技有限公司 防伪码的生成与验证方法、基于标签的防伪系统
CN117037159A (zh) * 2023-10-09 2023-11-10 网思科技股份有限公司 基于卷积神经网络的油画真伪鉴别方法、装置及存储介质
CN117037159B (zh) * 2023-10-09 2024-03-19 网思科技股份有限公司 基于卷积神经网络的油画真伪鉴别方法、装置及存储介质
CN117057826A (zh) * 2023-10-12 2023-11-14 深圳市一航网络信息技术有限公司 一种基于区块链技术的产品溯源验证方法
CN118190954A (zh) * 2024-05-15 2024-06-14 青岛澳科顺诚包装有限公司 一种镭射膜防伪图案监控质检系统
CN118644259A (zh) * 2024-08-12 2024-09-13 南通飞隼信息科技有限公司 一种基于仿真品识别的数据挖掘方法及系统

Also Published As

Publication number Publication date
DE112020005864T5 (de) 2022-10-13
CN115244542A (zh) 2022-10-25

Similar Documents

Publication Publication Date Title
WO2021179157A1 (zh) 用于验证产品真伪的方法和装置
CN112686812B (zh) 银行卡倾斜矫正检测方法、装置、可读存储介质和终端
CN110838119B (zh) 人脸图像质量评估方法、计算机装置及计算机可读存储介质
JP2020525947A (ja) 操作された画像の検出
Gill et al. A review paper on digital image forgery detection techniques
CN109117885B (zh) 一种基于深度学习的邮票识别方法
KR20200118842A (ko) 신원 인증 방법 및 장치, 전자 기기 및 저장 매체
CN110427972B (zh) 证件视频特征提取方法、装置、计算机设备和存储介质
CN109740572A (zh) 一种基于局部彩色纹理特征的人脸活体检测方法
JP5574033B2 (ja) 画像認識システム及びその認識方法並びにプログラム
CN109409158B (zh) 一种基于二维码边缘粗糙度的防伪方法
Yan et al. An IoT-based anti-counterfeiting system using visual features on QR code
EP4171051B1 (en) Collation device, program, and collation method
Centeno et al. Recurrent comparator with attention models to detect counterfeit documents
CN104102907A (zh) 一种消除光照不均匀的lbp人脸识别方法
CN113468954B (zh) 基于多通道下局部区域特征的人脸伪造检测方法
Agarwal et al. The advent of deep learning-based image forgery detection techniques
WO2021102770A1 (zh) 用于验证产品的真伪的方法和设备
CN110766001B (zh) 基于cnn和rnn的银行卡卡号定位与端到端识别方法
CN116665254A (zh) 基于手形语义先验和ViT的非接触式掌纹识别方法
Khuspe et al. Robust image forgery localization and recognition in copy-move using bag of features and SVM
Kavelar et al. Reading the legends of roman republican coins
EP3982289A1 (en) Method for validation of authenticity of an image present in an object, object with increased security level and method for preparation thereof, computer equipment, computer program and appropriate reading means
CN110415424B (zh) 一种防伪鉴定方法、装置、计算机设备和存储介质
Xu et al. Coin recognition method based on SIFT algorithm

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: 20924902

Country of ref document: EP

Kind code of ref document: A1

122 Ep: pct application non-entry in european phase

Ref document number: 20924902

Country of ref document: EP

Kind code of ref document: A1