An artificial intelligence-based seal authenticity identification system

Abstract

The present invention relates to the field of information security technology, and particularly to a system for distinguishing the authenticity of seals based on artificial intelligence, including: a target data acquisition module for acquiring a target seal image and image illumination data, a standard data acquisition module for acquiring standard seal sample data and standard texture feature data, an image data processing module for performing grayscale analysis on the target seal image and adjusting the grayscale analysis process of the target seal image, and also for processing the target seal image, a first feature extraction module for extracting the texture features and seal features of the target seal image, a second feature extraction module for extracting the standard texture features and standard seal features, and an authenticity discrimination module for discriminating the authenticity of the target seal image. The present invention effectively improves the accuracy of distinguishing the authenticity of purple clay teapot seals.

Description

Technical Field

[0001] The present invention relates to the field of information security technology, and particularly to a seal authenticity discrimination system based on artificial intelligence. Background Art

[0002] With the development of social economy, the importance of seals in various legal documents has become increasingly prominent, and the act of forging seals also occurs from time to time. Traditional seal identification methods rely on expert experience, with low efficiency and prone to errors. Therefore, it is of great significance to develop an efficient and accurate seal authenticity discrimination system.

[0003] Chinese Patent Publication No. CN113705330A discloses a seal authenticity discrimination method and system, which specifically includes the following steps: obtaining an original seal pattern and a seal pattern to be compared; performing a rectification operation on the seal pattern to be compared; using a preset sampling area for local comparison and screening; taking special points of the seal pattern and using the connection lines of the special points for comparison operations. The present invention uses a preset sampling area for local comparison and screening and takes special points of the seal pattern, and uses the connection lines of the special points for comparison operations to perform multiple screening and comparison of the seal pattern to be compared from multiple angles, making the solution of the present application more accurate and mature compared with the existing seal recognition technology; thus, it can be seen that the method for discriminating the authenticity of the seal in the present invention only screens the seal graphic pattern and does not analyze the influencing factors and material textures of the seal image, resulting in inaccurate discrimination of the authenticity of the seal. Summary of the Invention

[0004] The purpose of the present invention is to provide a seal authenticity discrimination system based on artificial intelligence to solve at least one of the problems existing in the prior art.

[0005] To achieve the above purpose, the present invention adopts the following technical solutions:

[0006] A target data acquisition module for acquiring a target seal image and image illumination data;

[0007] A standard data acquisition module for acquiring standard seal sample data and standard texture feature data;

[0008] An image data processing module for performing grayscale analysis on the target seal image, adjusting the grayscale analysis process of the target seal image according to the grayscale values of each coordinate point of the target seal image, and optimizing the grayscale adjustment process of the target seal image according to the image illumination data. The image data processing module is also used to process the target seal image according to the grayscale analysis result of the target seal image;

[0009] A first feature extraction module for extracting the texture features and seal features of the target seal image according to the processing result of the target seal image;

[0010] A second feature extraction module for extracting standard texture features and standard seal features based on standard seal sample data and standard texture feature data;

[0011] An authenticity discrimination module for discriminating the authenticity of a target seal image based on the texture features of the target seal image, the seal features of the target seal image, the standard texture features, and the standard seal features.

[0012] Further, the image data processing module is provided with a grayscale analysis unit. The grayscale analysis unit establishes a rectangular coordinate system of the target seal image based on the pixel point coordinates in the target seal image and projects the pixel point coordinates into coordinate system coordinates. The grayscale analysis unit is further configured to record the grayscale value of each coordinate system coordinate point as g(x, y), where x is the abscissa of the coordinate system coordinate, y is the ordinate of the coordinate system coordinate, and g(x, y) is the grayscale value of the coordinate system coordinate (x, y);

[0013] The grayscale analysis unit performs grayscale analysis on the grayscale value g(x, y) of each coordinate system coordinate point: if g(x, y) < G × η, the grayscale analysis unit determines that the grayscale value of this coordinate system coordinate point is a high grayscale value; if g(x, y) ≥ G × η, the grayscale analysis unit determines that the grayscale value of this coordinate system coordinate point is a background grayscale value;

[0014] where G is the average grayscale value of the coordinate system coordinate points, and η is the background area ratio of the sample image;

[0015] The grayscale analysis unit performs grayscale processing on the target seal image according to the grayscale analysis result: setting the grayscale value of the coordinate system coordinate point with the grayscale analysis result being the background grayscale value to 255.

[0016] Further, the image data processing module is also provided with a contrast adjustment unit. The contrast adjustment unit calculates the grayscale value coefficient of variation σ according to the grayscale value g(x, y) of each coordinate system coordinate point in the target seal image:

[0017] σ = {[g(x1, y1) - E] 2 + [g(x1, y2) - E] 2 +... + [g(xN1, yN2) - E] 2} / (N1

[0018] × N2);

[0019] E = [g(x1, y1) + g(x1, y2) +... + g(xN1, yN2)] / (N1 × N2);

[0020] Among them, g(x1, y1) is the gray value of the coordinate point in the coordinate system with abscissa x1 and ordinate y1, g(x1, y2) is the gray value of the coordinate point in the coordinate system with abscissa x1 and ordinate y2, g(xN1, yN2) is the gray value of the coordinate point in the coordinate system with abscissa xN1 and ordinate yN2, N1 is the number of horizontal pixel points of the target seal image, and N2 is the number of vertical pixel points of the target seal image;

[0021] The contrast adjustment unit adjusts the gray analysis process of the target seal image according to the gray value coefficient of variation σ: if σ ≥ T, the contrast adjustment unit does not adjust the gray analysis process of the target seal image; if σ < T, the contrast adjustment unit adjusts the background area ratio η of the sample image to η', and sets η' = exp[(T - σ) / T];

[0022] Among them, T is a preset coefficient of variation.

[0023] Furthermore, the image data processing module is also provided with a light optimization unit, and the light optimization unit is used to optimize the gray adjustment process of the target seal image according to the image illumination data;

[0024] The light optimization unit calculates the illumination index α according to the image illumination data;

[0025] The light optimization unit compares the illumination index α with each preset illumination index, and optimizes the gray adjustment process of the target seal image according to the comparison result: when α < A1, the light optimization unit determines that the illumination index of the target seal image is low, and optimizes the preset coefficient of variation T to T', and sets T' = T × exp{-(A1 - α) / A1}; when A1 ≤ α < A2, the light optimization unit determines that the illumination index of the target seal image is normal and does not perform optimization; when α ≥ A2, the light optimization unit determines that the illumination index of the target seal image is abnormal, and optimizes the preset coefficient of variation T to T", and sets T" = T × (A2 - α) / A2; the light optimization unit optimizes the contrast of the illumination conditions of different target seal images by optimizing the gray adjustment process of the target seal image, improving the accuracy of the adjustment process;

[0026] Among them, A1 is the first preset illumination index, A2 is the second preset illumination index, and A1 < A2.

[0027] Furthermore, the first feature extraction module is provided with a gradient analysis unit, and the gradient analysis unit analyzes the gradient of the target seal image according to the processing result of the target seal image;

[0028] The gradient analysis unit clusters the coordinate points in the coordinate system of the target seal image after grayscale processing with the grayscale value g(x,y) to obtain clusters of each coordinate point;

[0029] The gradient analysis unit takes the average grayscale value of the coordinate points in the coordinate system contained in each cluster of coordinate points as the marked grayscale value of each cluster of coordinate points, denoted as D(i), and sets D(i) = [G(i,1) + G(i,2) +... + G(i,J(i))] / J(i), where i = 1, 2... M, M is the number of clusters of coordinate points, G(i,1) is the grayscale value of the first coordinate point in the i-th coordinate cluster, G(i,2) is the grayscale value of the second coordinate point in the i-th coordinate cluster, G(i,J(i)) is the grayscale value of the J(i)-th coordinate point in the i-th coordinate cluster, and J(i) is the number of coordinate points in the i-th coordinate cluster;

[0030] The gradient analysis unit sorts each cluster of coordinate points in descending order according to the marked grayscale value D(i), and classifies each cluster of coordinate points according to the sorting result: the gradient analysis unit divides the first 20% of the sorting result into the seal pixel cluster, and divides the clusters of coordinate points outside the first 20% of the sorting result into the texture pixel cluster.

[0031] Further, the first feature extraction module is also provided with a texture feature analysis unit, and the texture feature analysis unit analyzes the texture features of the target seal image according to the gradient analysis result of the target seal image;

[0032] The texture feature analysis unit takes the continuous coordinate points in the coordinate system within the texture pixel cluster as a texture pattern;

[0033] The texture feature analysis unit counts the number P of texture patterns in the target seal image and calculates the texture density ρ, and sets ρ = P / S;

[0034] where S is the area of the target seal;

[0035] The texture feature analysis unit divides the target seal image into equidistant rectangles to obtain rectangular division regions, and counts the number p(k) of texture patterns in each rectangular division region;

[0036] The texture feature analysis unit calculates the texture roughness γ according to the number p(k) of texture patterns in each rectangular division region;

[0037] The texture feature analysis unit takes the texture roughness γ and the texture density ρ as the texture feature analysis result of the target seal image.

[0038] Furthermore, the first feature extraction module is also provided with a seal feature analysis unit, and the seal feature analysis unit analyzes the seal features of the target seal image according to the gradient analysis result of the target seal image;

[0039] The seal feature analysis unit generates a seal shape feature vector group H according to the coordinate points in the coordinate system within the seal pixel clustering cluster;

[0040] The seal feature analysis unit uses the seal shape feature vector group H as the seal feature analysis result of the target seal image.

[0041] Furthermore, the authenticity discrimination module is provided with a shape feature matching unit, and the shape feature matching unit is used to perform shape feature matching between the seal features of the target seal image and the standard seal features;

[0042] The shape feature matching unit calculates the offset coefficient β between the seal shape feature vector group H and the standard seal shape feature vector group h, and sets β = [|H(1) - h(1)| + |H(2) - h(2)| +... + |H(V) - h(V)|] / V;

[0043] Wherein, H(1) is the first sub-vector within the seal shape feature vector group H, h(1) is the first sub-vector within the standard seal shape feature vector group h, H(2) is the second sub-vector within the seal shape feature vector group H, h(2) is the second sub-vector within the standard seal shape feature vector group h, H(V) is the Vth sub-vector within the seal shape feature vector group H, h(V) is the Vth sub-vector within the standard seal shape feature vector group h, and V is the number of sub-vectors of the seal shape feature vector group H;

[0044] The shape feature matching unit compares the offset coefficient β with a preset offset value R, and analyzes the shape matching state according to the comparison result, where:

[0045] When β < R, the shape matching unit determines that the shape matching state is a low fitting match;

[0046] When β ≥ R, the shape matching unit determines that the shape matching state is a normal fitting match.

[0047] Furthermore, the authenticity discrimination module is also provided with a gradient feature matching unit, and the gradient feature matching unit is used to perform gradient feature matching between the seal features of the target seal image and the standard seal features: if (ρ - ρ’) × (γ - γ’) / (ρ’ × γ’) < L, the gradient matching unit determines that the gradient feature matches; if (ρ - ρ’) × (γ - γ’) / (ρ’ × γ’) ≥ L, the gradient matching unit determines that the gradient feature does not match.

[0048] Further, the authenticity discrimination module is also provided with an output unit, which is used to discriminate the authenticity of the target seal image according to the shape feature matching result and the gradient feature matching result: if the shape matching state is normal fitting matching and the gradient feature matches, the output unit determines that the authenticity discrimination result of the target image is true; if the shape matching state is low fitting matching and the gradient feature does not match, the output unit determines that the authenticity discrimination result of the target image is false; if the shape matching state is normal fitting matching and the gradient feature does not match, the output unit determines that the authenticity discrimination result of the target image is material process mismatch; if the shape matching state is low fitting matching and the gradient feature matches, the output unit determines that the authenticity discrimination result of the target image is false;

[0049] The output unit outputs the authenticity discrimination result to the user.

[0050] Compared with the prior art, the beneficial effects of the present invention are as follows: by obtaining the required data information in this embodiment through the target data acquisition module and the standard data acquisition module, the accuracy and integrity of data information acquisition are improved; by processing the acquired target image through the image data processing module, the accuracy of image data processing is improved; by extracting features from the target image through the first feature extraction module, the accuracy of feature extraction of the target seal image is improved; by extracting features from the standard seal sample image through the second feature extraction module, the accuracy of extraction of standard texture features and standard seal features is improved; by discriminating the authenticity of the target seal image through the authenticity discrimination module, the accuracy of authenticity discrimination of the purple clay teapot seal is improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following will briefly introduce the drawings required for description in the embodiments. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0052] Figure 1 It is a schematic structural diagram of the seal authenticity discrimination system based on artificial intelligence in this embodiment.

[0053] Figure 2 It is a schematic structural diagram of the image data processing module in this embodiment.

[0054] Figure 3 It is a schematic structural diagram of the first feature extraction module in this embodiment.

[0055] Figure 4 It is a schematic structural diagram of the authenticity discrimination module in this embodiment. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0056] To more clearly illustrate the present invention, the present invention will be further described below in conjunction with preferred embodiments and the accompanying drawings. Similar components in the drawings are denoted by the same reference numerals. Those skilled in the art should understand that the content specifically described below is illustrative rather than restrictive, and should not be used to limit the protection scope of the present invention.

[0057] It should be noted that although terms such as first, second, and third may be used in the embodiments of the present application for description, these descriptions should not be limited to these terms. These terms are only used to distinguish the descriptions. For example, without departing from the scope of the embodiments of the present application, the first may also be referred to as the second, and similarly, the second may also be referred to as the first.

[0058] Specifically, a seal authenticity discrimination system based on artificial intelligence described in this embodiment is applied to the authenticity discrimination of purple clay teapot seals; the purple clay teapot seal described in this embodiment is specifically the bottom-width seal of the purple clay teapot; in this embodiment, by extracting shape features and texture features from the standard seal sample image and the target seal image of the purple clay teapot seal image, and performing authenticity discrimination based on the extraction results, the accuracy of the authenticity discrimination of the purple clay teapot seal is improved.

[0059] Please refer to Figure 1 as shown, which is a schematic structural diagram of the seal authenticity discrimination system based on artificial intelligence in this embodiment, including:

[0060] A target data acquisition module for acquiring a target seal image and image illumination data;

[0061] A standard data acquisition module for acquiring standard seal sample data and standard texture feature data.

[0062] Specifically, the target seal image is the seal image data to be verified for authenticity; the standard seal sample data is the standard seal sample image, and the standard seal sample image is the mirror image of the original seal scanned image; the standard texture feature data is the standard texture image of the purple clay teapot; the image illumination data is the shooting illumination intensity of the target seal image.

[0063] It can be understood that the standard seal image is the binary-processed rectangular grayscale image data, that is, the grayscale value of the seal content is 0, and the grayscale value of the non-seal content is 255; the target seal image is a rectangular grayscale image with pixel grayscale values in the range of [0, 255]; this embodiment does not specifically limit the acquisition methods of the target seal image and the standard seal sample data, and those skilled in the art can freely set them as long as the acquisition requirements of the target seal image and the standard seal sample data are met. For example, the standard seal sample data can be obtained through user interaction input, and the target seal image can be obtained by taking a photo with a high-definition camera.

[0064] An image data processing module is used to perform grayscale analysis on a target seal image, adjust the grayscale analysis process of the target seal image according to the grayscale values of coordinate points in each coordinate system of the target seal image, and optimize the grayscale adjustment process of the target seal image according to image illumination data. The image data processing module is also used to process the target seal image according to the grayscale analysis result of the target seal image. The image data processing module is connected to the target data acquisition module; please refer to Figure 2 As shown, the image data processing module includes:

[0065] A grayscale analysis unit is used to perform grayscale analysis on a target seal image and perform grayscale processing on the target seal image according to the contrast analysis result;

[0066] The grayscale analysis unit establishes a plane rectangular coordinate system of the target seal image based on the pixel point coordinates in the target seal image, projects the pixel point coordinates into coordinate system coordinates, and the grayscale analysis unit is also used to record the grayscale value of each coordinate system coordinate point as g(x, y), where x is the abscissa of the coordinate system coordinate, y is the ordinate of the coordinate system coordinate, and g(x, y) is the grayscale value of the coordinate system coordinate (x, y);

[0067] The grayscale analysis unit performs grayscale analysis on the grayscale value g(x, y) of each coordinate system coordinate point: if g(x, y) < G×η, the grayscale analysis unit determines that the grayscale value of this coordinate system coordinate point is a high grayscale value; if g(x, y) ≥ G×η, the grayscale analysis unit determines that the grayscale value of this coordinate system coordinate point is a background grayscale value; where G is the average grayscale value of the coordinate system coordinate point, and η is the background area ratio of the sample image; the grayscale analysis unit improves the efficiency of image processing by classifying coordinate-type coordinate points to distinguish background pixel points and pixel points that can be used as the basis for authenticity discrimination;

[0068] The grayscale analysis unit performs grayscale processing on the target seal image according to the grayscale analysis result: sets the grayscale value of the coordinate system coordinate point with the grayscale analysis result of the background grayscale value to 255; the grayscale analysis unit improves the accuracy of target image feature analysis by performing grayscale processing on the target seal image to remove the interference of ordinary background pixel points on subsequent feature extraction.

[0069] Specifically, in this embodiment, no specific limitation is imposed on the value of the background area ratio η of the sample image, and those skilled in the art can freely set it as long as the value requirement of the background area ratio η of the sample image is satisfied. In this embodiment, the background area ratio of the sample image is 0.73. At the same time, in this embodiment, no specific limitation is imposed on the process of "establishing a plane rectangular coordinate system of the target seal image based on the pixel point coordinates in the target seal image" by the grayscale analysis unit. For example, the upper left starting point of the image can be set as the origin of the coordinate system, and the positive direction of the horizontal axis is from left to right, and the positive direction of the vertical axis is from top to bottom to establish the plane rectangular coordinate system of the target seal image.

[0070] Please continue to refer to Figure 2 As shown, the image data processing module is further provided with a contrast adjustment unit, and the contrast adjustment unit is used to calculate the grayscale value coefficient of variation σ according to the grayscale value g(x, y) of each coordinate point in the coordinate system of the target seal image:

[0071] σ = {[g(x1, y1) - E] 2 + [g(x1, y2) - E] 2 +... + [g(xN1, yN2) - E] 2} / (N1 × N2);

[0072] E = [g(x1, y1) + g(x1, y2) +... + g(xN1, yN2)] / (N1 × N2);

[0073] Among them, g(x1, y1) is the grayscale value of the coordinate point in the coordinate system with the abscissa x1 and the ordinate y1, g(x1, y2) is the grayscale value of the coordinate point in the coordinate system with the abscissa x1 and the ordinate y2, g(xN1, yN2) is the grayscale value of the coordinate point in the coordinate system with the abscissa xN1 and the ordinate yN2, N1 is the number of horizontal pixel points of the target seal image, and N2 is the number of vertical pixel points of the target seal image. The contrast adjustment unit numerically calculates the contrast of the target seal image by calculating the grayscale value coefficient of variation σ, so as to provide a numerical basis for the subsequent adjustment process and make the adjustment process more accurate.

[0074] The contrast adjustment unit adjusts the grayscale analysis process of the target seal image according to the grayscale value coefficient of variation σ: if σ ≥ T, the contrast adjustment unit does not adjust the grayscale analysis process of the target seal image; if σ < T, the contrast adjustment unit adjusts the background area ratio η of the sample image to η', and sets η' = exp[(T - σ) / T]; where T is the preset coefficient of variation. The contrast adjustment unit improves the accuracy of grayscale analysis by adjusting the grayscale analysis process of the target seal image and amplifying the judgment process of grayscale analysis of the target seal image with low contrast.

[0075] Specifically, in this embodiment, the value of the preset coefficient of variation T is not specifically limited, and those skilled in the art can freely set it as long as the value requirement of the preset coefficient of variation T is met. For example, the preset coefficient of variation T can be set to 15%.

[0076] Please continue to refer to Figure 2 As shown, the image data processing module is further provided with a light optimization unit, and the light optimization unit is used to optimize the gray-scale adjustment process of the target seal image according to the image illumination data;

[0077] The light optimization unit calculates the illumination index α according to the image illumination data, and the calculation formula of the illumination index α is as follows:

[0078] α = ln{1 + |w - W| / W|};

[0079] Where w is the illumination intensity when the target seal image is taken, and W is the preset image illumination intensity; the light optimization unit optimizes the reflection and light saturation conditions of the obtained target seal image by calculating the illumination index α, improving the accuracy of adjusting the gray-scale analysis process of the target seal image;

[0080] The light optimization unit compares the illumination index α with each preset illumination index, and optimizes the gray-scale adjustment process of the target seal image according to the comparison result: when α < A1, the light optimization unit determines that the illumination index of the target seal image is low, and optimizes the preset coefficient of variation T to T', and sets T' = T × exp{-(A1 - α) / A1}; when A1 ≤ α < A2, the light optimization unit determines that the illumination index of the target seal image is normal and does not perform optimization; when α ≥ A2, the light optimization unit determines that the illumination index of the target seal image is abnormal, and optimizes the preset coefficient of variation T to T", and sets T" = T × (A2 - α) / A2; where A1 is the first preset illumination index, A2 is the second preset illumination index, and A1 < A2; the light optimization unit optimizes the gray-scale adjustment process of the target seal image, optimizes the contrast of the illumination conditions of different target seal images, and improves the accuracy of the adjustment process.

[0081] Specifically, in this embodiment, the values of the first preset illumination index A1 and the second preset illumination index A2 are not specifically limited, and those skilled in the art can freely set them as long as the value requirements of the first preset illumination index A1 and the second preset illumination index A2 are met. For example, the first preset illumination index A1 can be set to 0.2, and the second preset illumination index A2 can be set to 0.4.

[0082] Please continue to refer to Figure 1As shown, the artificial intelligence-based seal authenticity identification system further includes a first feature extraction module, which is used to extract the texture features and seal features of the target seal image according to the processing result of the target seal image. The first feature extraction module is connected to the image data processing module; Please refer to Figure 3 As shown, the first feature extraction module includes a gradient analysis unit, which is used to analyze the gradient of the target seal image according to the processing result of the target seal image;

[0083] The gradient analysis unit clusters the coordinate points of the coordinate system of the target seal image after gray processing with the gray value g(x,y) to obtain each coordinate point clustering cluster; The gradient analysis unit performs clustering analysis on the processed target image, so as to use the subsequent sorting result as the gradient representation method of the target seal image, improving the accuracy of feature extraction of the target seal image;

[0084] The gradient analysis unit takes the average gray value of the coordinate points of the coordinate system contained in each coordinate point clustering cluster as the marked gray value of each coordinate point clustering cluster, denoted as D(i), and sets D(i) = [G(i,1) + G(i,2) +... + G(i,J(i))] / J(i), where i = 1, 2... M, M is the number of coordinate point clustering clusters, G(i,1) is the gray value of the first coordinate point in the i-th coordinate clustering cluster, G(i,2) is the gray value of the second coordinate point in the i-th coordinate clustering cluster, G(i,J(i)) is the gray value of the J(i)-th coordinate point in the i-th coordinate clustering cluster, and J(i) is the number of coordinate points of the coordinate system in the i-th coordinate clustering cluster;

[0085] The gradient analysis unit sorts each coordinate point clustering cluster in descending order according to the marked gray value D(i), and classifies each coordinate point clustering cluster according to the sorting result: The gradient analysis unit divides the first 20% of the sorting result into seal pixel clustering clusters, and divides the coordinate point clustering clusters outside the first 20% of the sorting result into texture pixel clustering clusters; The gradient analysis unit classifies the clustering result to distinguish the texture pattern and the seal image, facilitating subsequent texture feature extraction and seal feature extraction.

[0086] Specifically, "the first 20% of the sorting result" in this embodiment specifically refers to the coordinate point clustering clusters within the sorting range of [0, 20%].

[0087] Please continue to refer to Figure 3 As shown, the first feature extraction module further includes a texture feature analysis unit, which is used to analyze the texture features of the target seal image according to the gradient analysis result of the target seal image. The texture feature analysis unit is connected to the gradient analysis unit;

[0088] The texture feature analysis unit regards the continuous coordinate system coordinates within the texture pixel clustering cluster as a texture pattern;

[0089] The texture feature analysis unit counts the number P of texture patterns in the target seal image, calculates the texture density ρ, and sets ρ = P / S, where S is the area of the target seal; the texture feature analysis unit calculates the texture density and takes it as part of the texture feature, so that the authenticity discrimination of the seal incorporates the discrimination content of the texture of the purple clay pot, improving the accuracy of authenticity discrimination;

[0090] The texture feature analysis unit divides the target seal image into equidistant rectangles to obtain each rectangular division area, and counts the number p(k) of texture patterns in each rectangular division area;

[0091] The texture feature analysis unit calculates the texture roughness γ according to the number p(k) of texture patterns in each rectangular division area. The texture roughness γ is defined as an index representing the uniform distribution state of the texture in the target seal image, and the texture roughness γ is set to {[p(1) - EP] 2 + [p(2) - EP] 2 +... + [p(K) - EP] 2} / EP, where EP = [Σp(k)] / K; among them, p(1) is the number of texture patterns in the first division area, p(2) is the number of texture patterns in the second division area, p(K) is the number of texture patterns in the Kth division area, p(k) is the number of texture patterns in the kth division area, k is a mathematical subscript, and K is the number of rectangular division areas;

[0092] The texture feature analysis unit takes the texture roughness γ and the texture density ρ as the texture feature analysis results of the target seal image.

[0093] Specifically, the process of "dividing the target seal image into equidistant rectangles" in this embodiment is specifically to divide the target seal image into equidistant rectangles with similar rectangles of the target seal image. The ratio of the similar rectangle to the original target seal image in this embodiment is 1:4, and those skilled in the art can also set it freely. For example, if more accurate texture features are required, the ratio of the similar rectangle to the original target seal image can be set to 1:6, etc.

[0094] Please continue to refer to Figure 3 As shown, the first feature extraction module further includes a shape feature analysis unit, which is used to analyze the seal features of the target seal image according to the gradient analysis result of the target seal image. The shape feature analysis unit is connected to the gradient analysis unit;

[0095] The seal feature analysis unit generates a seal shape feature vector group H according to the coordinate points of the coordinate system within the seal pixel clustering cluster;

[0096] The seal feature analysis unit uses the seal shape feature vector group H as the seal feature analysis result of the target seal image; by generating a vector group that can represent the seal shape and using it as the seal feature, the seal feature analysis unit improves the accuracy of seal feature analysis.

[0097] Specifically, in this embodiment, the process of "generating a seal shape feature vector group according to the coordinate points in the coordinate system within the seal pixel clustering cluster" is specifically to perform vector calculation on the coordinate points in the coordinate system within the continuous seal pixel clustering cluster. For example, if the coordinate point a(c,d) in the coordinate system within the seal pixel clustering cluster is continuous with the coordinate point b(n,m) in the coordinate system within the seal pixel clustering cluster, the vector calculation process is: z = (n - c, m - d), where z is the result of vector calculation, and the seal shape feature vector group is the set of the results of vector calculation for all continuous coordinate points in the coordinate system within the seal pixel clustering cluster.

[0098] Please continue to refer to Figure 1 As shown, the seal authenticity discrimination system based on artificial intelligence further includes a second feature extraction module, which is used to extract standard seal features and standard texture features according to the standard seal sample image and standard texture feature data, and obtain standard seal feature h and standard texture features; the standard texture features include standard roughness γ' and standard texture density ρ'.

[0099] The second feature extraction module stores the extraction results of the standard seal features and standard texture features.

[0100] Specifically, the process of "extracting standard seal features and standard texture features according to the standard seal sample image and standard texture feature data" in this embodiment is the same as the implementation process and data structure of "extracting the texture features and seal features of the target seal image" in this embodiment, and will not be elaborated here.

[0101] Please continue to refer to Figure 1 As shown, the seal authenticity discrimination system based on artificial intelligence further includes an authenticity discrimination module, which is used to discriminate the authenticity of the target seal image according to the texture features of the target seal image, the seal features of the target seal image, the standard texture features, and the standard seal features, and output the authenticity discrimination result to the user. The authenticity discrimination module is connected to the first feature extraction module and the second feature extraction module; please refer to Figure 4 As shown, the authenticity discrimination module includes a shape feature matching unit, which is used to perform shape feature matching on the texture features of the target seal image and the standard texture features.

[0102] The shape feature matching unit calculates the offset coefficient β between the seal shape feature vector group H and the standard seal shape feature vector group h, and sets β = [|H(1) - h(1)| + |H(2) - h(2)| +... + |H(V) - h(V)|] / V;

[0103] Wherein, H(1) is the first sub-vector in the seal shape feature vector group H, h(1) is the first sub-vector in the standard seal shape feature vector group h, H(2) is the second sub-vector in the seal shape feature vector group H, h(2) is the second sub-vector in the standard seal shape feature vector group h, H(V) is the Vth sub-vector in the seal shape feature vector group H, h(V) is the Vth sub-vector in the standard seal shape feature vector group h, and V is the number of sub-vectors in the seal shape feature vector group H;

[0104] The shape feature matching unit compares the offset coefficient β with the preset offset value R, and analyzes the shape matching status according to the comparison result, where:

[0105] When β < R, the shape matching unit determines that the shape matching status is a low fitting match;

[0106] When β ≥ R, the shape matching unit determines that the shape matching status is a normal fitting match.

[0107] Specifically, in this embodiment, the value of the preset offset value R is not specifically limited, and those skilled in the art can freely set it as long as it meets the value requirements of the preset offset value R. For example, the preset offset value R can be set to 2.

[0108] Please continue to refer to Figure 4 As shown, the authenticity discrimination module is also provided with a gradient feature matching unit, which is used to perform gradient feature matching between the seal features of the target seal image and the standard seal features: if (ρ - ρ’) × (γ - γ’) / (ρ’ × γ’) < L, the gradient matching unit determines that the gradient feature matches; if (ρ - ρ’) × (γ - γ’) / (ρ’ × γ’) ≥ L, the gradient matching unit determines that the gradient feature does not match; where L is the preset comprehensive matching index; the gradient feature matching unit analyzes the gradient features to judge the texture gradient matching result, improving the accuracy of authenticity discrimination.

[0109] Specifically, in this embodiment, the value of the preset comprehensive matching index L is not specifically limited, and those skilled in the art can freely set it as long as it meets the value requirements of the preset comprehensive matching index L. For example, the preset comprehensive matching index L can be set to 0.4.

[0110] Please continue to refer to Figure 4As shown, the authenticity discrimination module further includes an output unit. The output unit is used to perform authenticity discrimination on the target seal image according to the shape feature matching result and the gradient feature matching result, and output the authenticity discrimination result to the user. The output unit is connected to the gradient feature matching unit and the shape feature matching unit;

[0111] The output unit performs authenticity discrimination on the target seal image according to the shape feature matching result and the gradient feature matching result: If the shape matching state is normal fitting match and the gradient feature matches, the output unit determines that the authenticity discrimination result of the target image is true; If the shape matching state is low fitting match and the gradient feature does not match, the output unit determines that the authenticity discrimination result of the target image is false; If the shape matching state is normal fitting match and the gradient feature does not match, the output unit determines that the authenticity discrimination result of the target image is material process mismatch; If the shape matching state is low fitting match and the gradient feature matches, the output unit determines that the authenticity discrimination result of the target image is false; By jointly analyzing the shape matching result and the gradient matching result, the output unit judges the authenticity result of the target seal image, improving the accuracy of authenticity discrimination;

[0112] The output unit outputs the authenticity discrimination result to the user.

[0113] So far, the technical solution of the present invention has been described in combination with the preferred embodiments shown in the drawings. However, it is easy for those skilled in the art to understand that the protection scope of the present invention is obviously not limited to these specific embodiments. Without departing from the principle of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will all fall within the protection scope of the present invention.

Claims

1. A seal authenticity identification system based on artificial intelligence, characterized in that: include: A target data acquisition module is used to acquire a target seal image and image illumination data; A standard data acquisition module is used to acquire standard seal sample data and standard texture feature data; An image data processing module is used to perform grayscale analysis on the target seal image, and adjust the grayscale analysis process of the target seal image according to the grayscale values ​​of the coordinate points of each coordinate system of the target seal image, and optimize the grayscale adjustment process of the target seal image according to the image illumination data. The image data processing module is also used to process the target seal image according to the grayscale analysis result of the target seal image; A first feature extraction module is used to extract texture features and seal features of the target seal image according to the processing result of the target seal image; A second feature extraction module is used to extract standard texture features and standard seal features according to the standard seal sample data and the standard texture feature data; An authenticity identification module is used to identify the authenticity of the target seal image according to the texture features of the target seal image, the seal features of the target seal image, the standard texture features and the standard seal features; The first feature extraction module is provided with a gradient analysis unit, and the gradient analysis unit analyzes the gradient of the target seal image according to the processing result of the target seal image; The gradient analysis unit clusters the coordinate points of the target seal image after grayscale processing according to the grayscale value g(x,y) to obtain clusters of each coordinate point; The gradient analysis unit uses the average grayscale value of the coordinate system coordinate points contained in each coordinate point cluster as the marked grayscale value of each coordinate point cluster, recorded as D(i), and sets D(i)=[G(i,1)+G(i,2)+...+G(i,J(i))] / J(i), wherein i=1,2...M, M is the number of coordinate point clusters, G(i,1) is the grayscale value of the first coordinate system coordinate point in the i-th coordinate cluster, G(i,2) is the grayscale value of the second coordinate system coordinate point in the i-th coordinate cluster, G(i,J(i)) is the grayscale value of the J(i)-th coordinate system coordinate point in the i-th coordinate cluster, and J(i) is the number of coordinate system coordinate points in the i-th coordinate cluster; The gradient analysis unit sorts each coordinate point cluster in descending order according to the marked gray value D(i), and classifies each coordinate point cluster according to the sorting result: the gradient analysis unit classifies the first 20% of the sorting result into stamp pixel clusters, and classifies the coordinate point clusters outside the first 20% of the sorting result into texture pixel clusters; The first feature extraction module is further provided with a texture feature analysis unit, and the texture feature analysis unit analyzes the texture feature of the target seal image according to the gradient analysis result of the target seal image; The texture feature analysis unit regards the continuous coordinate system coordinates in the texture pixel cluster as a texture graph; The texture feature analysis unit counts the number P of texture patterns in the target seal image, and calculates the texture density ρ, setting ρ=P / S; Where S is the target stamp area; The texture feature analysis unit divides the target seal image into equidistant rectangles to obtain rectangular divided areas, and counts the number of texture graphics p(k) in each rectangular divided area; The texture feature analysis unit calculates the texture roughness γ according to the number of texture patterns p(k) in each rectangular divided area; Texture roughness γ is defined as an index representing the uniform distribution state of the texture in the target stamp image. Texture roughness γ is set to {[p(1)-EP] 2 +[p(2)-EP] 2 +...+[p(K)-EP] 2 } / EP, EP=[Σp(k)] / K; Wherein, p(1) is the number of texture graphics in the first divided area, p(2) is the number of texture graphics in the second divided area, p(K) is the number of texture graphics in the Kth divided area, p(k) is the number of texture graphics in the kth divided area, k is a mathematical index, and K is the number of rectangular divided areas; The texture feature analysis unit uses the texture roughness γ and the texture density ρ as the texture feature analysis results of the target seal image.

2. The seal authenticity identification system based on artificial intelligence according to claim 1 is characterized in that: The image data processing module is provided with a grayscale analysis unit, which establishes a plane rectangular coordinate system of the target seal image based on the pixel coordinates in the target seal image, and projects the pixel coordinates into coordinate system coordinates. The grayscale analysis unit is also used to record the grayscale value of each coordinate system coordinate point as g(x, y), wherein x is the abscissa of the coordinate system coordinate, y is the ordinate of the coordinate system coordinate, and g(x, y) is the grayscale value of the coordinate system coordinate (x, y); The grayscale analysis unit performs grayscale analysis on the grayscale value g(x,y) of each coordinate point in the coordinate system: if g(x,y)<G×η, the grayscale analysis unit determines that the grayscale value of the coordinate point in the coordinate system is a high grayscale value; if g(x,y)≥G×η, the grayscale analysis unit determines that the grayscale value of the coordinate point in the coordinate system is a background grayscale value; Where G is the average gray value of the coordinate system coordinate point, and η is the background area ratio of the sample image; The grayscale analysis unit performs grayscale processing on the target seal image according to the grayscale analysis result: the grayscale value of the coordinate point of the coordinate system whose grayscale analysis result is the background grayscale value is set to 255.

3. The seal authenticity identification system based on artificial intelligence according to claim 2 is characterized in that: The image data processing module is further provided with a contrast adjustment unit, which calculates the gray value variation coefficient σ according to the gray value g(x, y) of each coordinate point of the target seal image: σ6{[g(x1,y1)-E] 2 +[g(x1,y2)-E] 2 +...+[g(xN1,yN2)-E] 2 } / (N1 ×N2)； E=[g(x1,y1)+g(x1,y2)+...+g(xN1,yN2)] / (N1×N2); Among them, g(x1,y1) is the grayscale value of the coordinate point of the coordinate system with the horizontal coordinate x1 and the vertical coordinate y1, g(x1,y2) is the grayscale value of the coordinate point of the coordinate system with the horizontal coordinate x1 and the vertical coordinate y2, g(xN1,yN2) is the grayscale value of the coordinate point of the coordinate system with the horizontal coordinate xN1 and the vertical coordinate yN2, N1 is the number of horizontal pixels of the target seal image, and N2 is the number of vertical pixels of the target seal image; The contrast adjustment unit adjusts the grayscale analysis process of the target seal image according to the grayscale value variation coefficient σ: if σ≥T, the contrast adjustment unit does not adjust the grayscale analysis process of the target seal image; if σ<T, the contrast adjustment unit adjusts the background area ratio η of the sample image to η', and sets η'=exp[(T-σ) / T]; Where T is the preset coefficient of variation.

4. The seal authenticity identification system based on artificial intelligence according to claim 3 is characterized in that: The image data processing module is also provided with a light optimization unit, and the light optimization unit is used to optimize the grayscale adjustment process of the target seal image according to the image illumination data; The light optimization unit calculates the light index α according to the image light data; The light optimization unit compares the illumination index α with each preset illumination index, and optimizes the grayscale adjustment process of the target seal image according to the comparison result: when α<A1, the light optimization unit determines that the illumination index of the target seal image is low, and optimizes the preset variation coefficient T to T', and sets T'=T×exp{-(A1-α) / A1}; when A1≤α<A2, the light optimization unit determines that the illumination index of the target seal image is normal and does not perform optimization; when α≥A2, the light optimization unit determines that the illumination index of the target seal image is abnormal, and optimizes the preset variation coefficient T to T", and sets T"=T×(A2-α) / A2; the light optimization unit optimizes the contrast of illumination conditions of different target seal images by optimizing the grayscale adjustment process of the target seal image, thereby improving the accuracy of the adjustment process; Wherein, A1 is a first preset illumination index, A2 is a second preset illumination index, and A1<A2.

5. The seal authenticity identification system based on artificial intelligence according to claim 4 is characterized in that: The first feature extraction module is further provided with a seal feature analysis unit, and the seal feature analysis unit analyzes the seal feature of the target seal image according to the gradient analysis result of the target seal image; The seal feature analysis unit generates a seal shape feature vector group H according to the coordinate system coordinate points in the seal pixel clustering cluster; The seal feature analysis unit uses the seal shape feature vector group H as the seal feature analysis result of the target seal image.

6. The seal authenticity identification system based on artificial intelligence according to claim 5 is characterized in that: The authenticity identification module is provided with a shape feature matching unit, and the shape feature matching unit is used to perform shape feature matching between the seal feature of the target seal image and the standard seal feature; The shape feature matching unit calculates the offset coefficient β between the seal shape feature vector group H and the standard seal shape feature vector group h, and sets β=[|H(1)-h(1)|+|H(2)-h(2)|+...+|H(V)-h(V)|] / V; Wherein, H(1) is the first component vector in the seal shape feature vector group H, h(1) is the first component vector in the standard seal shape feature vector group h, H(2) is the second component vector in the seal shape feature vector group H, h(2) is the second component vector in the standard seal shape feature vector group h, H(V) is the Vth component vector in the seal shape feature vector group H, h(V) is the Vth component vector in the standard seal shape feature vector group h, and V is the number of component vectors in the seal shape feature vector group H; The shape feature matching unit compares the offset coefficient β with the preset offset value R, and analyzes the shape matching state according to the comparison result, wherein: When β≥R, the shape feature matching unit determines that the shape matching state is a low-fitting match; When β<R, the shape feature matching unit determines that the shape matching state is a normal fitting match.

7. The seal authenticity identification system based on artificial intelligence according to claim 6 is characterized in that: The authenticity identification module is further provided with a gradient feature matching unit, which is used to perform gradient feature matching between the seal feature of the target seal image and the standard seal feature: if (ρ-ρ')×(γ-γ') / (ρ'×γ')<L, the gradient feature matching unit determines that the gradient features match; if (ρ-ρ')×(γ-γ') / (ρ'×γ')≥L, the gradient feature matching unit determines that the gradient features do not match; Among them, γ' is the standard roughness, ρ' is the standard texture density, and L is the preset comprehensive matching index.

8. The seal authenticity identification system based on artificial intelligence according to claim 7 is characterized in that: The authenticity identification module is further provided with an output unit, and the output unit is used to identify the authenticity of the target seal image according to the shape feature matching result and the gradient feature matching result: if the shape matching state is a normal fitting match and a gradient feature match, the output unit determines that the authenticity identification result of the target image is true; if the shape matching state is a low fitting match and the gradient feature does not match, the output unit determines that the authenticity identification result of the target image is false; if the shape matching state is a normal fitting match and the gradient feature does not match, the output unit determines that the authenticity identification result of the target image is a material process mismatch; if the shape matching state is a low fitting match and the gradient feature matches, the output unit determines that the authenticity identification result of the target image is false; The output unit outputs the authenticity identification result to the user.

Images