Public health information authenticity prediction method and device based on image recognition

Abstract

The invention relates to a public health information authenticity prediction method based on image recognition, and the method comprises the following steps: obtaining a to-be-detected picture uploaded by a user, and recording public health information in the to-be-detected picture; firstly, a metadata checking tool is called to judge whether a picture PS event exists in a to-be-detected picture or not, and if yes, a prediction result with suspected authenticity is directly output; if not, next judgment is carried out; then, calling an image evidence obtaining tool to judge whether a picture editing event exists in the to-be-detected picture or not, and if yes, directly outputting a prediction result with suspected authenticity; if not, next judgment is carried out; and executing an edge detection algorithm on the to-be-detected picture to obtain an edge image, performing edge blurring condition analysis on the edge image, if so, outputting a prediction result with suspected authenticity, and if not, outputting a prediction result with true image and ending detection. The method can perform authenticity analysis on the to-be-detected picture uploaded by the user and automatically output the authenticity prediction result, is convenient for the user to use, and is high in generalization performance.

Description

Technical Field

[0001] The present invention relates to the technical field related to the prediction of information authenticity, and particularly to a method and device for predicting the authenticity of public health information based on image recognition. Background Art

[0002] Public health information refers to data and information related to population health, including data in aspects such as disease surveillance, health behaviors, environmental factors, or some picture-based information announcements.

[0003] If some lawbreakers or unethical persons tamper with and publish public health information, it is very likely to cause public panic. Therefore, it is necessary to predict the authenticity of public health information.

[0004] Most of the existing authenticity predictions of picture-based public health information still rely on manual comparison with relevant content on official websites. This method has poor popularity and cannot be promoted for public use. Summary of the Invention

[0005] The purpose of the present invention is to at least solve one of the deficiencies of the prior art, and provide a method for predicting the authenticity of public health information based on image recognition.

[0006] To achieve the above purpose, the present invention adopts the following technical solutions:

[0007] Specifically, a method for predicting the authenticity of public health information based on image recognition is proposed, including the following:

[0008] Obtain a to-be-detected picture uploaded by a user, where the to-be-detected picture records public health information;

[0009] First, call a metadata verification tool to determine whether there is a picture PS event in the to-be-detected picture. If so, directly output a prediction result of doubtful authenticity; if not, proceed to the next judgment;

[0010] Then call an image forensics tool to determine whether there is a picture editing event in the to-be-detected picture. If so, directly output a prediction result of doubtful authenticity; if not, proceed to the next judgment;

[0011] Perform an edge detection algorithm on the to-be-detected picture to obtain an edge image, and analyze the edge blurring situation of the edge image. If the result is edge blurring, output a prediction result of doubtful authenticity; if the result is not edge blurring, output a prediction result of the image being real and end the detection.

[0012] Furthermore, specifically, calling a metadata verification tool to determine whether there is a picture PS event in the to-be-detected picture includes,

[0013] Call ExifTool, Photoshop or an online EXIF viewer to view the metadata, i.e., EXIF data, of the image to be detected;

[0014] If the EXIF data shows that the image has been created or modified by an image processing software, or the timestamp has been modified, or the EXIF data has been cleared or is missing

[0015] The existence of at least one of the above situations indicates that there is an image PS event in the image to be detected.

[0016] Furthermore, specifically, call an image forensics tool to determine whether there is an image editing event in the image to be detected, including,

[0017] Call JPEGsnoop, Fotoforensics, and GIMP tools in sequence to query for editing traces in the image to be detected. If any one of the tools determines that there are editing traces in the image to be detected, it indicates that there is an image editing event.

[0018] Furthermore, specifically, perform an edge detection algorithm on the image to be detected to obtain an edge image, including,

[0019] Grayscale the image to be detected to obtain a grayscale image;

[0020] Perform image preprocessing on the grayscale image to remove image noise and obtain a preprocessed image;

[0021] Perform edge detection on the preprocessed image based on a pre-selected edge detection operator to obtain an edge image.

[0022] Furthermore, specifically, the selected edge detection operator is the Laplacian operator.

[0023] Furthermore, specifically, analyze the edge blurring situation of the edge image, including,

[0024] Calculate the local sharpness of each edge pixel point. The local sharpness is achieved by calculating the gradient change of its neighboring pixels. Determine whether the proportion of edge pixel points with local sharpness within a preset range to the total number of edge pixel points is greater than the proportion threshold. If so, it indicates normal; if not, it indicates edge blurring;

[0025] Calculate the local contrast of the edge image and determine whether there is a region with local contrast lower than the contrast threshold. If there is, it indicates edge blurring; if not, it indicates normal;

[0026] When there is any situation of edge blurring in the above situations, it is determined that the prediction result of the edge image is edge blurring.

[0027] Further, the method further includes that when outputting a prediction result with doubtful authenticity, a preset logo pattern is added to the picture to be detected for display, and the logo pattern is used to inform the user that the prediction result is of doubtful authenticity.

[0028] The present invention also provides a public health information authenticity prediction device based on image recognition, including the following:

[0029] A data acquisition module, configured to acquire a picture to be detected uploaded by a user, where the picture to be detected records public health information;

[0030] A first judgment module, configured to call a metadata verification tool to judge whether there is a picture PS event in the picture to be detected. If so, directly output a prediction result with doubtful authenticity; if not, proceed to the next judgment;

[0031] A second judgment module, configured to call an image forensics tool to judge whether there is a picture editing event in the picture to be detected. If so, directly output a prediction result with doubtful authenticity; if not, proceed to the next judgment;

[0032] A third judgment module, configured to perform an edge detection algorithm on the picture to be detected to obtain an edge image, and analyze the edge blur situation of the edge image. If the result is edge blur, output a prediction result with doubtful authenticity; if the result is not edge blur, output a prediction result that the image is real and end the detection.

[0033] The beneficial effects of the present invention are as follows:

[0034] The present invention provides a public health information authenticity prediction method based on image recognition. Considering that if the public health information in the form of pictures is false, it is generally a PS-edited or tampered picture. Based on this idea, first, call a metadata verification tool to judge whether there is a picture PS event in the picture to be detected, and then call an image forensics tool to judge whether there is a picture editing event in the picture to be detected. When it is impossible to determine whether it has been edited through conventional image processing tools, obtain the edge image of the picture to be detected through an edge detection image processing method, and then judge whether it is suspected of being edited according to the analysis of the edge blur situation of the edge image. If so, also output a prediction result with doubtful authenticity. The present invention can perform authenticity analysis on the picture to be detected uploaded by the user and automatically output an authenticity prediction result, which is convenient for users to use and has high popularization. Description of the Drawings

[0035] By describing the embodiments shown in the accompanying drawings in detail, the above and other features of the present disclosure will become more apparent. The same reference numerals in the drawings of the present disclosure denote the same or similar elements. Obviously, the drawings in the following description are only some embodiments of the present disclosure. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings. In the drawings:

[0036] Figure 1 The flowchart of the method for predicting the authenticity of public health information based on image recognition according to the present invention is shown. Specific embodiments

[0037] The concept, specific structure and technical effects of the present invention will be clearly and completely described below in conjunction with the embodiments and the drawings to fully understand the purpose, solution and effects of the present invention. It should be noted that, without conflict, the embodiments in the present application and the features in the embodiments can be combined with each other. The same reference numerals used throughout the drawings indicate the same or similar parts.

[0038] Embodiment 1, referring to Figure 1 , the present invention proposes a method for predicting the authenticity of public health information based on image recognition, including the following:

[0039] Step 110: Obtain the image to be detected uploaded by the user, and the image to be detected records public health information;

[0040] Step 120: First, call the metadata verification tool to determine whether there is an image PS event in the image to be detected. If so, directly output the prediction result of doubtful authenticity; if not, proceed to the next judgment;

[0041] Step 130: Then call the image forensics tool to determine whether there is an image editing event in the image to be detected. If so, directly output the prediction result of doubtful authenticity; if not, proceed to the next judgment;

[0042] Step 140: Perform an edge detection algorithm on the image to be detected to obtain an edge image, and analyze the edge blurring situation of the edge image. If the result is edge blurring, output the prediction result of doubtful authenticity. If the result is not edge blurring, output the prediction result of the image being real and end the detection.

[0043] In Embodiment 1 of the present invention, considering that if the public health information in the form of pictures is false, it is generally a PS-edited or tampered picture. Based on this idea, first, a metadata verification tool is called to determine whether there is a picture PS event in the picture to be detected. Then, an image forensics tool is called to determine whether there is a picture editing event in the picture to be detected. When it is impossible to determine whether it has been edited through conventional image processing tools, an edge image of the picture to be detected is obtained through an edge detection image processing method, and then whether it is suspected of being edited is judged based on the analysis of the edge blur situation of the edge image. If so, a prediction result of doubtful authenticity is also output. The present invention can perform authenticity analysis on the picture to be detected uploaded by the user and automatically output an authenticity prediction result, which is convenient for users to use and has high popularity.

[0044] As a preferred embodiment of the present invention, specifically, calling a metadata verification tool to determine whether there is a picture PS event in the picture to be detected includes:

[0045] Calling ExifTool, Photoshop or an online EXIF viewing tool to view the metadata, i.e., EXIF data, of the picture to be detected;

[0046] If the EXIF data shows that the image was created or modified by an image processing software, or the timestamp was modified, or the EXIF data was cleared or missing

[0047] The existence of at least one of the above situations indicates that there is a picture PS event in the picture to be detected.

[0048] In this preferred embodiment, considering that the metadata (EXIF data) of an image usually contains information such as the shooting device information, shooting time, and shooting parameters of the image. After PS editing, the metadata may be modified or deleted.

[0049] How to check: Tools (such as ExifTool, the built-in function of Photoshop, or an online EXIF viewing tool) can be used to view the metadata of the image.

[0050] Contents to be checked:

[0051] If the EXIF data shows that the image was created or modified by software such as Photoshop, or the timestamp does not match the shooting time, this may be an indication that the image has been edited.

[0052] If the EXIF data is cleared or missing, it may also mean that the image has been processed.

[0053] As a preferred embodiment of the present invention, specifically, calling an image forensics tool to determine whether there is a picture editing event in the picture to be detected includes:

[0054] Call JPEGsnoop, Fotoforensics, and GIMP tools respectively to query for editing traces in the picture to be detected in sequence. If any one of the tools determines that there are editing traces in the picture to be detected, it indicates that there is a picture editing event.

[0055] In this preferred embodiment, considering that modern image forensics tools can automatically analyze images to find signs of forgery. These tools can usually identify the traces left during the image editing process.

[0056] Common tools:

[0057] JPEGsnoop: This tool can detect compression traces in JPEG images and determine whether the image has been modified.

[0058] Fotoforensics: This is an online image analysis tool that can analyze forgery traces in images, such as splicing, compression, cloning, etc.

[0059] GIMP (open source): It can help analyze whether there are forgery or editing traces in the image through some filters and image analysis tools.

[0060] As a preferred embodiment of the present invention, specifically, performing an edge detection algorithm on the picture to be detected to obtain an edge image includes:

[0061] Grayscale the picture to be detected to obtain a grayscale image;

[0062] Perform image preprocessing on the grayscale image to remove image noise to obtain a preprocessed image;

[0063] Perform edge detection on the preprocessed image based on a preselected edge detection operator to obtain an edge image.

[0064] As a preferred embodiment of the present invention, specifically, the selected edge detection operator is the Laplacian operator.

[0065] As a preferred embodiment of the present invention, specifically, analyzing the edge blur situation of the edge image includes:

[0066] Calculate the local sharpness of each edge pixel point, which is achieved by calculating the gradient change of its neighboring pixels, and determine whether the proportion of edge pixel points with local sharpness within a preset range to the total number of edge pixel points is greater than the proportion threshold. If so, it indicates normal; if not, it indicates edge blur;

[0067] Calculate the local contrast of the edge image, and determine whether there is a region with local contrast lower than the contrast threshold. If there is, it indicates edge blur; if not, it indicates normal;

[0068] If there is any blurred edge in the above situation, it is determined that the prediction result of the edge image is blurred.

[0069] In this preferred embodiment, considering that the Laplacian operator can detect the details of an image by calculating the second derivative of the image to find edges, the Laplacian operator is selected for edge detection.

[0070] In addition, edge sharpness is an index to measure the clarity of an edge. Blurred edges usually appear as "transition zones", while sharp edges have obvious discontinuities.

[0071] Method: Calculate the local sharpness of each edge point, which can be achieved by calculating the gradient change of neighboring pixels. If the gradient change in the neighborhood is large, it indicates that the edge is relatively clear; if the gradient change in the neighborhood is small, it indicates that the edge is blurred.

[0072] The sharpness formula is:

[0073]

[0074] where represents the gradient of the image at a certain pixel point.

[0075] In addition, for an edge image, areas with higher local contrast represent clear edges, while areas with low contrast usually represent blurred edges.

[0076] Contrast calculation: Contrast can be measured by calculating the standard deviation of pixel values within an image region:

[0077] contrast = std(I(x, y))

[0078] where std(I(x, y)) is the standard deviation of the local region of the image, reflecting the degree of change of pixel values. A smaller standard deviation indicates that the edge of this region may be blurred.

[0079] Combined with the above analysis process, the judgment situation of blurred edges can be analyzed.

[0080] As a preferred embodiment of the present invention, the method further includes, when outputting a prediction result with doubtful authenticity, adding a preset logo pattern to the image to be detected for display, and the logo pattern is used to inform the user that the prediction result is of doubtful authenticity.

[0081] In this preferred embodiment, considering the intuitive viewing situation of the user, through the above method, good visual display can be achieved, which is convenient for the user's intuitive perception of the prediction result.

[0082] Embodiment 2. The present invention also proposes a public health information authenticity prediction device based on image recognition, including the following:

[0083] A data acquisition module, configured to acquire a to-be-detected picture uploaded by a user, where the to-be-detected picture records public health information;

[0084] A first judgment module, configured to call a metadata verification tool to judge whether there is a picture PS event in the to-be-detected picture. If so, directly output a prediction result of doubtful authenticity; if not, proceed to the next judgment;

[0085] A second judgment module, configured to call an image forensics tool to judge whether there is a picture editing event in the to-be-detected picture. If so, directly output a prediction result of doubtful authenticity; if not, proceed to the next judgment;

[0086] A third judgment module, configured to perform an edge detection algorithm on the to-be-detected picture to obtain an edge image, analyze the edge blur situation of the edge image. If the result is edge blur, output a prediction result of doubtful authenticity; if the result is not edge blur, output a prediction result that the image is real and end the detection.

[0087] In this Embodiment 2, consistent with the public health information authenticity prediction method based on image recognition proposed by the present invention, considering that if the public health information in the form of pictures is false, it is generally a PS-edited or tampered picture. Based on this idea, first, call a metadata verification tool to judge whether there is a picture PS event in the to-be-detected picture, then call an image forensics tool to judge whether there is a picture editing event in the to-be-detected picture. When it is impossible to determine whether it has been edited through conventional image processing tools, obtain the edge image of the to-be-detected picture through the edge detection image processing method, and then judge whether it is suspected of being edited according to the analysis of the edge blur situation of the edge image. If so, also output a prediction result of doubtful authenticity. The present invention can perform authenticity analysis on the to-be-detected pictures uploaded by users and automatically output authenticity prediction results, which is convenient for users to use and has high popularization.

[0088] Although the description of the present invention has been quite detailed and particularly described several of the described embodiments, it is not intended to be limited to any of these details or embodiments or any particular embodiment, but rather should be regarded as providing a broad possible interpretation of these claims in light of the prior art by reference to the appended claims, thereby effectively covering the intended scope of the present invention. In addition, the present invention is described above with embodiments foreseeable by the inventor for the purpose of providing a useful description, and those non-substantive modifications to the present invention that are not currently foreseeable may still represent equivalent modifications of the present invention.

[0089] As described above, these are only the preferred embodiments of the present invention. The present invention is not limited to the above-described embodiments. As long as the same means are used to achieve the technical effects of the present invention, they should fall within the protection scope of the present invention. Within the protection scope of the present invention, various modifications and variations can be made to its technical solutions and / or implementation manners.

Claims

1. A method for predicting the authenticity of public health information based on image recognition, characterized in that, The following are included: Obtain a to-be-detected picture uploaded by a user, where the to-be-detected picture records public health information; First, call a metadata verification tool to determine whether there is a picture PS event in the to-be-detected picture. If so, directly output a prediction result with doubtful authenticity; if not, proceed to the next judgment; Then call an image forensics tool to determine whether there is a picture editing event in the to-be-detected picture. If so, directly output a prediction result with doubtful authenticity; if not, proceed to the next judgment; Perform an edge detection algorithm on the to-be-detected picture to obtain an edge image, and analyze the edge blurring situation of the edge image. If the result is edge blurring, output a prediction result with doubtful authenticity; if the result is not edge blurring, output a prediction result that the image is real and end the detection.

2. The method for predicting the authenticity of public health information based on image recognition according to claim 1, wherein Specifically, calling a metadata verification tool to determine whether there is a picture PS event in the to-be-detected picture includes: Call ExifTool, Photoshop or an online EXIF viewer to view the metadata, i.e., EXIF data, of the to-be-detected picture; If the EXIF data shows that the image was created or modified by an image processing software, or the timestamp was modified, or the EXIF data was cleared or missing The existence of at least one of the above situations indicates that there is a picture PS event in the to-be-detected picture.

3. The method for predicting the authenticity of public health information based on image recognition according to claim 1, wherein Specifically, calling an image forensics tool to determine whether there is a picture editing event in the to-be-detected picture includes: Call JPEGsnoop, Fotoforensics and GIMP tools in sequence to query for editing traces in the to-be-detected picture. If any one of the tools determines that there are editing traces in the to-be-detected picture, it means there is a picture editing event.

4. The method for predicting the authenticity of public health information based on image recognition according to claim 1, wherein Specifically, performing an edge detection algorithm on the to-be-detected picture to obtain an edge image includes: Grayscale the to-be-detected picture to obtain a grayscale image; Perform image preprocessing on the grayscale image to remove image noise to obtain a preprocessed image; Perform edge detection on the preprocessed image based on a pre-selected edge detection operator to obtain an edge image.

5. The method for predicting the authenticity of public health information based on image recognition according to claim 4, wherein Specifically, the selected edge detection operator is the Laplacian operator.

6. The method for predicting the authenticity of public health information based on image recognition according to claim 1, wherein, Specifically, analyze the edge blurring situation of the edge image. It includes: Calculate the local sharpness of each edge pixel point. The local sharpness is realized by calculating the gradient change of its neighboring pixels, and judge whether the proportion of edge pixel points with local sharpness within a preset range in the total edge pixel points is greater than a proportion threshold. If so, it means normal; if not, it means edge blurring; Calculate the local contrast of the edge image, and judge whether there is a region with local contrast lower than a contrast threshold. If so, it means edge blurring; if not, it means normal; When there is any one situation of edge blurring in the above situations, determine that the prediction result of the edge image is edge blurring.

7. The method for predicting the authenticity of public health information based on image recognition according to claim 1, characterized in that, The method further includes that when outputting a prediction result with doubtful authenticity, add a preset logo pattern to the to-be-detected picture for display, and the logo pattern is used to inform the user that the prediction result is with doubtful authenticity.

8. A public health information authenticity prediction device based on image recognition, characterized in that, The following are included: A data acquisition module, which is used to acquire a to-be-detected picture uploaded by a user, where the to-be-detected picture records public health information; The first judgment module is used to call the metadata verification tool to judge whether there is a picture PS event in the picture to be detected. If so, directly output the prediction result of doubtful authenticity; if not, proceed to the next judgment; The second judgment module is used to call the image forensics tool to judge whether there is an image editing event in the picture to be detected. If so, directly output the prediction result of doubtful authenticity; if not, proceed to the next judgment; The third judgment module is used to perform an edge detection algorithm on the picture to be detected to obtain an edge image, and analyze the edge blurring situation of the edge image. If the result is edge blurring, output the prediction result of doubtful authenticity. If the result is not edge blurring, output the prediction result of the image being real and end the detection.

Images