Electronic license forgery detection method and application in license monitoring

Abstract

The invention discloses an electronic license forgery detection method, which comprises the following steps of: performing median filtering and sharpening operation processing on acquired electronic license data, and constructing a residual error set comprising a sharpening residual error, a sharpening-median filtering residual error and a median filtering residual error by utilizing the processed electronic license; discrete cosine conversion is carried out on the residual set, and characteristics of frequency domain changes before and after the original image is captured are obtained; extracting a plurality of residual features in a spatial domain by using an autoregression model and a Markov model so as to evaluate detail differences between the electronic certificates before and after median filtering; finally, integrating the features from the spatial domain and the frequency domain to construct a feature set; and inputting the extracted feature set into a support vector machine for training to obtain a median filtering detector to execute a detection task. According to the method, the defects of traditional electronic certificate counterfeiting detection in low-resolution and high-compression image processing are well overcome, the detection rate is high, and the method can be well applied to certificate monitoring.

Description

Technical Field

[0001] The present invention relates to the field of multimedia content security and certificate monitoring and anti-counterfeiting, and specifically to an electronic certificate forgery detection method based on dual-domain multiple residual median filtering forensics and its application in certificate monitoring. Background Art

[0002] With the rapid advancement of science and technology, users without professional training can easily use image convenience software to create forged images; electronic certificates, as a digital form of personal identity information, qualification certification and other legal documents, have gradually become an indispensable part of modern society. However, the popularity of electronic certificates also exposes them to the risk of forgery and makes them a target for criminals. If these forged electronic certificates are abused, they will cause undue influence in the judicial and financial fields, and also have a great negative impact on the normal activities of the market. Forgers often use post-processing operations to cover up their modifications, and techniques such as median filtering, Gaussian filtering and mean filtering will significantly affect the detection effect of forged images, especially median filtering.

[0003] Median filtering is a nonlinear operation widely used in image noise removal. It improves the quality of the filtered image through smoothing. At the same time, the noise characteristics of the forged area are also weakened compared with the original image, resulting in partial blurring of edge information. For this reason, median filtering is often used as an anti-forensic method to hide potential forensic clues by blurring or eliminating subtle features in digital images. Therefore, research on detection technology based on median filtering is of great significance to enhance the forgery detection capability of electronic certificates.

[0004] The patent document with the publication number CN114757877A discloses a deep forgery detection method based on frequency domain filtering residual, which belongs to the technical field of image classification, including a preprocessing module and a classification module, including the following steps: by performing Haar wavelet transform on the image, a low-frequency information map of the image is obtained; the grayscale map and the low-frequency information map of the original image are subjected to residual operation to obtain the residual map of the medium and high frequency information of the original image; in the classification module, the original image and the residual map of the medium and high frequency information are spliced ​​and input into the convolutional neural network for classification processing. The features of the medium and high frequency domains of the image are obtained by Haar wavelet transform and residual, so as to mine the forgery traces weakened or contaminated in the RGB domain due to image compression in the frequency domain, and improve the accuracy of the detection method when detecting compressed images. The RGB image and the high-frequency residual image are spliced ​​together as the input of the convolutional neural network, making full use of the rich semantic information in the RGB domain and the detailed texture information in the medium and high frequency domains, so that the detection method has good robustness to image compression.

[0005] The patent document with publication number CN105868778A discloses a median filtering detection method based on frequency domain fluctuation characteristics: the acquired original image is subjected to discrete Fourier transform processing to obtain a frequency domain image of the original image; the frequency domain image is processed to obtain description points that can reflect the frequency domain fluctuation characteristics of the original image, and the processed image is divided into N annular areas, and then the total number of description points in each annular area is counted respectively, which is the constructed N-dimensional detection feature; the extracted N-dimensional detection feature is input into a support vector machine for training to obtain a support vector machine median filter detector, and then the support vector machine median filter detector is used to detect whether the original image has undergone median filtering.

[0006] Existing methods mostly focus on the utilization of spatial domain or frequency domain features, ignoring the correlation between them, and perform poorly when processing small-size images or low-quality JPEG images.

[0007] Therefore, it is an urgent problem to provide an electronic certificate forgery detection method that can be applied to certificate monitoring activities, especially an electronic certificate forgery detection method based on dual-domain multiple residual median filtering forensics. Summary of the invention

[0008] In order to solve the problem of poor performance of median filtering detection on small-size images or low-quality JPEG images, the present invention proposes an electronic certificate forgery detection method based on dual-domain multiple residual median filtering forensics, which aims to combine spatial domain and frequency domain features and introduce multiple residual analysis to improve the accuracy and robustness of forged image detection, thereby providing more effective protection for the credibility of electronic certificates and providing assistance for electronic certificate monitoring and authenticity verification.

[0009] The object of the present invention is achieved in that:

[0010] The present invention provides a method for detecting electronic certificate forgery based on dual-domain multiple residual median filtering forensics, comprising:

[0011] Step 1): convert the acquired electronic certificate image data into a grayscale image and then perform median filtering and sharpening operations;

[0012] Step 2): constructing a residual set including sharpening residual, sharpening-median filtering residual and median filtering residual using the processed image;

[0013] Step 3): Perform discrete cosine transformation on the residual set constructed in step 1) to obtain its features that capture the frequency domain changes before and after the original image; use autoregressive model and Markov model to extract multiple residual features in the spatial domain to evaluate the detail differences between the unchanged image and the median filtered image;

[0014] Step 4): Integrate the features of the spatial domain and the frequency domain to construct a feature set and input it into the support vector machine for training to obtain a median filter detector to perform the detection task.

[0015] The construction of the residual set described in step 2 includes:

[0016] For the image after median filtering (I MF ) is subtracted from the original image (I) to obtain the median filter residual (I MFR ), the formula is as follows:

[0017] I MFR =I MF -I

[0018] For the image after sharpening (I S ) is subtracted from the original image (I) to obtain the sharpened residual (I SR ), the formula is as follows:

[0019] I SR =I S -I

[0020] For the sharpening residual (I SR ) and the median filter residual (I MFR ) is subtracted to obtain the sharpening-median filter residual (I SMR ), which is equivalent to the image after median filtering (I MF ) and the image after sharpening (I S ) to make a difference, the formula is as follows:

[0021] I SMR =I SR -I MFR =I S -I MF ;

[0022] The step 3) of acquiring frequency domain features includes:

[0023] First, the three types of residuals in the residual set obtained in step 1 are processed by discrete cosine transformation respectively, and the maximum value, minimum value, average value and variance of the obtained discrete cosine transformation coefficient matrix are extracted as features.

[0024] The autoregressive model extraction feature described in step 3) includes:

[0025] Before inputting the three types of residuals in the residual set into the autoregressive model, the data dimension of the image residual is expanded to prepare for the subsequent transposition and vectorization processing. Next, the augmented image is transposed, and the original image and the transposed image are vectorized and merged into a new vector, which will then be used as the input of the autoregressive model; finally, the first ten-order results of the output of the autoregressive model are taken as features.

[0026] The Markov model extraction feature described in step 3) includes:

[0027] Use the first-order difference matrix in eight directions to calculate the changes between adjacent pixels and define the first-order difference array D h The (x,y) formula is as follows:

[0028] D h (x,y)=I(x+Δx,y+Δy)-I(x,y)

[0029] Where (x, y) represents the pixel position in the image, and Δx and Δy represent the offsets in the horizontal and vertical directions respectively; these offsets are taken from the set {(-1,-1), (-1,0), (-1,1), (0,-1), (0,1), (1,-1), (1,0), (1,1)}, corresponding to eight different directions, the first-order difference matrix is ​​subdivided into left, middle and right matrices as the input of the Markov model, and the threshold is set to 1; the results in the vertical, horizontal and diagonal directions are divided into two groups to calculate the average value, as follows:

[0030] F1=(F (-1,-1) +F (1,1) +F (-1,1) +F (1,-1) ) / 4

[0031] F2=(F (-1,0) +F (1,0) +F (0,-1) +F (0,1) ) / 4

[0032] The results obtained by averaging the three residuals finally form a Markov feature with a dimension of 108;

[0033] The training classification described in step 4) includes:

[0034] The acquired image data are divided into 20 mutually exclusive subsets, 19 of which are used as training sets and the remaining subsets are used as test sets in each validation. The process is repeated 20 times to finally obtain the average performance index of the algorithm on the entire data set. At the same time, a grid search is performed on the key parameters, the search range of the penalty factor C and the kernel function parameter γ is defined, and the cross-validation method is used to ensure the fairness of the performance evaluation.

[0035] The certificate forgery detection method is applied in electronic certificate monitoring.

[0036] Positive and beneficial effects: Based on the statistical differences of image residuals in different domains, the present invention proposes the concepts of sharpening residuals and sharpening-median filter residuals, and together with the median filter residuals, constitutes a residual set. The multi-model strategy is adopted to further enhance the feature extraction effect in the frequency domain and the spatial domain. The multi-model strategy is used to extract features using discrete cosine transform, autoregressive model and Markov model in the frequency domain and the spatial domain respectively, and the frequency domain features are combined with the spatial domain features to more keenly capture the difference between the image processed by the median filter and the original image, thereby improving the accuracy of the median filter detection. Specifically, multiple residual DCT statistical features are extracted in the frequency domain to capture the detailed differences between the unchanged image and the median filtered image. In the spatial domain, features are extracted from multiple residuals in combination with the autoregressive model and the Markov model. Finally, these features are combined into a feature set for classification. When facing small-size images or low-quality JPEG images, a high accuracy rate can still be maintained. At the same time, when facing the confusion of other operations, it can also have a good ability to distinguish the median filter operation and has strong robustness. This method has good performance on small-size images or low-quality JPEG images. It can help in the monitoring and inspection of electronic certificates in market activities and has been well applied. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 This is a flow chart of the detection feature extraction of the present invention. DETAILED DESCRIPTION

[0038] In order to more clearly illustrate the purpose, technical solutions and advantages of the present invention, the technical solutions in the embodiments of the present invention are described in detail below. The embodiments are only partial implementations of the present invention, not all. Based on these embodiments of the present invention, all other embodiments that can be obtained by ordinary technicians in the field without creative work should be included in the protection scope of the present invention. The specific implementation mode of the present invention is described in detail below in conjunction with the embodiment diagram.

[0039] Example 1

[0040] like Figure 1 As shown, an embodiment of the present invention provides a method for detecting electronic certificate forgery based on dual-domain multiple residual median filtering forensics, comprising the following steps:

[0041] Step 1): convert the acquired electronic certificate image data into a grayscale image and then perform median filtering and sharpening operations;

[0042] Step 2): constructing a residual set including sharpening residual, sharpening-median filtering residual and median filtering residual using the processed image, wherein the processing includes:

[0043] The acquired image is first converted into a grayscale image, and then the grayscale image is subjected to median filtering and sharpening operations. MF ) is subtracted from the original image (I) to obtain the median filter residual (I MFR ), the formula is as follows:

[0044] I MFR =I MF -I

[0045] For the image after sharpening (I S ) is subtracted from the original image (I) to obtain the sharpened residual (I SR ), the formula is as follows:

[0046] I SR =I S -I

[0047] For the sharpening residual (I SR ) and the median filter residual (I MFR ) is subtracted to obtain the sharpening-median filter residual (I SMR ), which is equivalent to the image after median filtering (I MF ) and the image after sharpening (I S ) to make a difference, the formula is as follows:

[0048] I SMR =I SR -I MFR =I S -I MF ;

[0049] Step 3: Perform discrete cosine transformation on the residual set to obtain its characteristics of capturing the frequency domain changes before and after the original image. The discrete cosine transformation formula is as follows:

[0050]

[0051] in M and N are the sizes of the matrix rows and columns respectively; for the obtained discrete cosine transform coefficient matrix, the mean value, variance, maximum value and minimum value are extracted as frequency domain features;

[0052] Multiple residual features in the spatial domain are extracted using autoregressive and Markov models to assess the detail differences between the unaltered image and the median filtered image. The details are as follows:

[0053] Before the three residuals in the residual set are input into the autoregressive model, the data dimension of the image residual is first expanded to facilitate transposition and vectorization. Then, the expanded image is transposed, and the original image and the transposed image are vectorized and merged into new vectors, which will be used as the input of the autoregressive model. Finally, the first ten order results are extracted from the model output as autoregressive features.

[0054] For the three residuals in the residual set, the first-order difference matrix in eight directions is used to calculate the changes between adjacent pixels. Define the first-order difference array D h The (x,y) formula is as follows:

[0055] D h (x,y)=I(x+Δx,y+Δy)-I(x,y)

[0056] Where (x, y) represents the pixel position in the image, and Δx and Δy represent the offsets in the horizontal and vertical directions respectively. These offsets are taken from the set {(-1,-1), (-1,0), (-1,1), (0,-1), (0,1), (1,-1), (1,0), (1,1)}, corresponding to eight different directions. The first-order difference matrix is ​​subdivided into left, middle and right matrices as the input of the Markov model, and the threshold is set to 1; the results in the vertical, horizontal and diagonal directions are divided into two groups to calculate the average value, as follows:

[0057] F1=(F (-1,-1) +F (1,1) +F (-1,1) +F (1,-1) ) / 4

[0058] F2=(F (-1,0) +F (1,0) +F (0,-1) +F (0,1) ) / 4

[0059] The results obtained by averaging the three residuals finally form a Markov feature with a dimension of 108.

[0060] Step 4): Integrate the features of the spatial domain and the frequency domain to construct a feature set and input it into the support vector machine for training to obtain a median filter detector to perform the detection task.

[0061] F=[F spatial ,F frequency ]

[0062] The acquired image data was divided into 20 mutually exclusive subsets, 19 of which were used as training sets and the remaining subsets were used as test sets in each validation. The process was repeated 20 times, and the average performance index of the algorithm on the entire dataset was finally obtained. At the same time, a grid search was performed on the key parameters, the search range of the penalty factor C and the kernel function parameter γ was defined, and the cross-validation method was used to ensure the fairness of the performance evaluation. The classification performance was evaluated by the detection accuracy (ACC), and the true positive rate (TPR) of the detector under a low false positive rate (FPR) was illustrated by the ROC curve.

[0063] The certificate forgery detection method is applied in electronic certificate monitoring.

[0064] Based on the statistical differences of image residuals in different domains, the present invention proposes the concepts of sharpening residuals and sharpening-median filter residuals, and together with the median filter residuals, constitutes a residual set. The multi-model strategy is adopted to further enhance the feature extraction effect of the frequency domain and the spatial domain. The multi-model strategy is used to extract features using discrete cosine transform, autoregressive model and Markov model in the frequency domain and the spatial domain respectively, and the frequency domain features are combined with the spatial domain features to more keenly capture the difference between the image processed by the median filter and the original image, thereby improving the accuracy of the median filter detection. Specifically, multiple residual DCT statistical features are extracted in the frequency domain to capture the detailed differences between the unchanged image and the median filter image. In the spatial domain, features are extracted from multiple residuals in combination with the autoregressive model and the Markov model. Finally, these features are combined into a feature set for classification. When facing small-sized images or low-quality JPEG images, a high accuracy rate can still be maintained. At the same time, when facing the confusion of other operations, the median filter operation can also have a good resolution ability and strong robustness. This method has good performance on small-sized images or low-quality JPEG images. It can help monitor and verify electronic certificates in market activities and be well used.

[0065] The above embodiments are only used to clarify the technical solutions of the present invention, not to limit them. A person skilled in the art may adjust the technical solutions of the embodiments, or replace some of the technical features therein. Many variations may be made without changing the core of the corresponding technical solutions, all of which shall be included in the protection scope of the present invention.

Claims

1. A method for detecting forged electronic certificates, characterized in that: The following steps are involved: Step 1): convert the acquired electronic certificate image data into a grayscale image and then perform median filtering and sharpening operations; Step 2): constructing a residual set including sharpening residual, sharpening-median filtering residual and median filtering residual using the processed image; Step 3): Perform discrete cosine transformation on the residual set constructed in step 1) to obtain its features that capture the frequency domain changes before and after the original image; use autoregressive model and Markov model to extract multiple residual features in the spatial domain to evaluate the detail differences between the unchanged image and the median filtered image; Step 4): Integrate the features of the spatial domain and the frequency domain to construct a feature set and input it into the support vector machine for training to obtain a median filter detector to perform the detection task.

2. The method for detecting forged electronic certificates according to claim 1, characterized in that: Step 2) of constructing the residual set includes: For the image after median filtering (I MF ) is subtracted from the original image (I) to obtain the median filter residual (I MFR ), the formula is as follows: I MFR =I MF -I For the image after sharpening (I S ) is subtracted from the original image (I) to obtain the sharpened residual (I SR ), the formula is as follows: I SR =I S -I For the sharpening residual (I SR ) and the median filter residual (I MFR ) is subtracted to obtain the sharpening-median filter residual (I SMR ), which is equivalent to the image after median filtering (I MF ) and the image after sharpening (I S ) to make a difference, the formula is as follows: I SMR =I SR -I MFR =I S -I MF 。 3. The electronic certificate forgery detection method according to claim 1, characterized in that: The step 3) of acquiring frequency domain features includes: The three types of residuals in the residual set are processed by discrete cosine transformation respectively, and the maximum value, minimum value, average value and variance of the obtained discrete cosine transformation coefficient matrix are extracted as features.

4. The electronic certificate forgery detection method according to claim 1, characterized in that: The autoregressive model feature extraction described in step 3 includes: Before inputting the three residuals in the residual set into the autoregressive model, the residual matrix is ​​preprocessed as follows: First, the data dimension of the image residual is expanded to prepare for the subsequent transposition and vectorization processing; next, the augmented image is transposed, and the original image and the transposed image are vectorized and merged into a new vector; finally, these vectors will be used as the input of the autoregressive model; for the output of the autoregressive model, the first ten order results are taken as features.

5. The electronic certificate forgery detection method according to claim 1, characterized in that: The Markov model extraction feature described in step 3) includes: First, use the first-order difference matrix in eight directions to calculate the changes between adjacent pixels; define the first-order difference array D h The (x,y) formula is as follows: D h (x,y)=I(x+Δx,y+Δy)-I(x,y) Where (x, y) represents the pixel position in the image, and Δx and Δy represent the offsets in the horizontal and vertical directions respectively; these offsets are taken from the set {(-1,-1), (-1,0), 9-1,1), (0,-1), 90,1), 91,-1), (1,0), (1,1)}, corresponding to eight different directions; the first-order difference matrix is ​​subdivided into three matrices: left, middle, and right, as the input of the Markov model, and the threshold is set to 1; the results in the vertical, horizontal, and diagonal directions are divided into two groups to calculate the average value, as follows: F1=(F (-1,-1) +F (1,1) +F (-1,1) +F (1,-1) ) / 4 F2=(F (-1,0) +F (1,0) +F (0,-1) +F (0,1) ) / 4 The results obtained by averaging the three residuals finally form a Markov feature with a dimension of 108.

6. The electronic certificate forgery detection method according to claim 1, characterized in that: The training classification described in step 4) includes: The acquired image data are divided into 20 mutually exclusive subsets, 19 of which are used as training sets and the remaining subsets are used as test sets in each validation. The process is repeated 20 times to finally obtain the average performance index of the algorithm on the entire data set. At the same time, a grid search is performed on the key parameters, the search range of the penalty factor C and the kernel function parameter γ is defined, and the cross-validation method is used to ensure the fairness of the performance evaluation.

7. The certificate forgery detection method as described in claims 1-6 is applied to electronic certificate monitoring.

Images