Article anti-counterfeiting method and system based on local feature visual information

Abstract

The invention relates to the field of anti-counterfeiting authentication, and discloses an article anti-counterfeiting method and system based on local feature visual information, and the method comprises the steps: endowing each product with an independent two-dimensional code, then selecting a specific local region on the product as an image feature region, extracting a general image of the image feature region and microspur local magnified images of two positions, and carrying out the recognition of the microspur local magnified images. And the features are stored in a cloud database to serve as anti-counterfeiting information of the article. A user scans a two-dimensional code on a product through mobile phone software to access a cloud database, retrieves information related to a product number, carries out comparative analysis through a macro picture uploaded by the user and a macro local magnified picture, analyzes a matching condition, and obtains a matching result. The feature deep analysis is performed on the uploaded macro picture and the macro local amplified picture in the cloud database through the convolutional neural network model, so that the authenticity of the product is identified, the anti-counterfeiting identification cost and the anti-counterfeiting identification popularization difficulty are reduced, and the accuracy and reliability of anti-counterfeiting inspection of the article are ensured.

Description

Technical Field

[0001] The present invention relates to the field of anti-counterfeiting authentication technology, and more specifically to an article anti-counterfeiting method and system based on local feature visual information. Background Art

[0002] With the rapid development of computer vision, Internet of Things, big data and other technologies, more technical means and possibilities are provided for the anti-counterfeiting system of items. These technologies can realize the rapid and accurate identification and tracking of item information, providing strong support for the anti-counterfeiting system. Therefore, the anti-counterfeiting method of items came into being, mainly through the use of advanced algorithms of artificial intelligence technology and big data analysis methods to achieve all-round and multi-angle anti-counterfeiting verification of products, so as to accurately identify the tiny details, special marks or hidden information on the surface of the product, and effectively distinguish between genuine and fake products;

[0003] However, the above process still has the following disadvantages:

[0004] First, the existing anti-counterfeiting methods lack the method of using mobile phone APP to scan the QR code of the item information, interacting the anti-counterfeiting identification operation with the user to perform visual anti-counterfeiting identification on the item, which cannot reduce the anti-counterfeiting cost and the difficulty of promotion;

[0005] Second, existing anti-counterfeiting methods lack multiple anti-counterfeiting verification steps to analyze and screen the items, and cannot guarantee the accuracy and reliability of the anti-counterfeiting inspection of items, which may cause misjudgment of the anti-counterfeiting identification of items.

[0006] Third, with the technological advancement of printing and other equipment, the existing anti-counterfeiting methods for items can make anti-counterfeiting features easy to copy, especially for high-value goods, and the anti-counterfeiting costs are high. Summary of the invention

[0007] In order to overcome the above-mentioned defects of the prior art, the present invention provides an article anti-counterfeiting method and system based on local feature visual information to solve the problems existing in the above-mentioned background technology.

[0008] The present invention provides the following technical solution: an anti-counterfeiting method for articles based on local feature visual information, comprising:

[0009] S1: During production, a QR code is printed on the product using a coding device, thereby giving each product an independent QR code;

[0010] S2: by selecting a specific local area on the product as the image feature area during production or generating a random image as the image feature area during production;

[0011] S3: photographing the selected image feature area by using a high-resolution imaging device, extracting a general image of the image feature area and two macro-enlarged local images at different positions, and uploading the extracted general image of the image feature area and two macro-enlarged local images at different positions to a cloud database for storage;

[0012] S4: Retrieve product information based on the QR code on the product. The user scans the QR code on the product with a mobile phone to access the cloud database and retrieve information related to the product number. The user can judge the authenticity of the product based on the retrieved product information.

[0013] S5: Perform advanced verification of the product based on the user's judgment results, compare and analyze the macro pictures uploaded by the user with the macro partial enlarged pictures, calculate the macro difference value, and thus analyze the matching situation;

[0014] S6: Based on the preliminary comparative analysis results, the uploaded macro images and the macro partial magnification images in the cloud database are further analyzed through the convolutional neural network model, and the similarity is calculated;

[0015] S7: used to identify the authenticity of the product, identify the authenticity of the product through the results of in-depth feature analysis, and transmit the anti-counterfeiting identification results to S8;

[0016] S8: The anti-counterfeiting identification result is transmitted to the mobile phone via the cloud server.

[0017] Explanation: The macro magnification position is marked on the general map; and a general map and a macro local magnification map of the extracted image feature area are uploaded to the cloud database for storage as the anti-counterfeiting information of the item. Figure 1 Used for consumer anti-counterfeiting pictures, macro local magnification Figure 2 As a confidential anti-counterfeiting image of the manufacturer; users obtain the macro feature position based on the general image displayed by the mobile phone software, and take macro pictures of the object according to the software prompts;

[0018] Cloud-based macro local magnification obtained based on mobile phone APP Figure 1 Macro zoom shot taken on the APP Figure 1 , the user visually compares the difference between the two pictures on the APP to obtain verification, and at the same time, the macro picture uploaded by the user and the macro partial enlargement are Figure 1 Perform comparative analysis and calculate the macro difference value to analyze the matching situation for advanced verification.

[0019] Preferably, S1 generates a unique number for each product, converts the unique number into a QR code using a coding device, and prints the code on the product.

[0020] Preferably, in S2, a random image is generated during production as the image feature area.

[0021] Preferably, S2 captures an image of a selected area by using a high-resolution camera, and the specific local area includes the product serial number and a special pattern of the product.

[0022] Preferably, in S3, the selected image feature area and two different positions of the image feature area are photographed on the production line to obtain a general image of the image feature area, a macro local magnification image, and a macro local magnification image. Figure 1 And macro local magnification Figure 2 , and processes the captured images, including image calibration, contrast enhancement, and sharpening, extracts image data from the processed images, and uploads the extracted image data to a cloud database.

[0023] Preferably, when the user's mobile phone accesses the cloud database, S4 will immediately retrieve information related to the product number from the database to obtain the overall image and macro-enlarged image of the image feature area corresponding to the product. Figure 1 , and sent to the mobile phone, the user can visually compare the overall image of the product image feature area sent to the mobile phone and a macro local enlarged image to determine the authenticity of the product. Figure 1 When there is doubt or the product is a high-value commodity, the system will automatically prompt for advanced verification and guide the user through the advanced verification process. The user can follow the prompts and use the macro camera function of the mobile phone to take a picture of the designated location on the product and upload it to the cloud database.

[0024] Preferably, in step S5, when the cloud database receives the macro picture uploaded by the user, it immediately compares the macro local zoom in the cloud database with the macro local zoom in the cloud database. Figure 2 Compare and analyze to further determine the authenticity of the product by uploading macro images and zooming in on the macro parts stored in the cloud database. Figure 2 Perform comparative analysis and calculate the macro deviation value, and determine whether to conduct further analysis and comparison based on the macro deviation value.

[0025] The specific analysis method of the macro deviation value is as follows:

[0026] Step S511: Find the uploaded macro image and the macro local magnification stored in the cloud database by using a feature point detection algorithm Figure 2 The feature points in the image are analyzed and calculated, and the uploaded macro image and the macro local magnification stored in the cloud database are compared. Figure 2 The similarity deviation value is M represents the number of matching feature points in the uploaded macro image, and N represents the macro local magnification of the cloud database. Figure 2 The number of matching feature points in ;

[0027] Step S512: Analyze and calculate the uploaded macro image and the macro local magnification stored in the cloud database Figure 2 The color matching deviation value is C i Indicates the i-th color channel value corresponding to the uploaded macro image, D i A macro zoom of a cloud database Figure 2 The corresponding i-th color channel value, R max Indicates the maximum value of the color channel;

[0028] Step S513: Analyze and calculate the uploaded macro image and the macro local magnification stored in the cloud database Figure 2 The texture consistency deviation value is X represents the texture feature set in the uploaded macro image, and Y represents the macro local magnification in the cloud database. Figure 2 The texture feature set in , |X∩Y represents the number of features shared by the two sets, |X| and |Y| represent the size of their respective feature sets;

[0029] Step S514: comprehensively calculating the macro deviation value as V=α×S+β×CD+γ×TD, where α, β, and γ represent weight coefficients respectively;

[0030] The authenticity of the product can be determined by comparing the macro deviation value V with the preset macro deviation threshold θ. If the macro deviation value V> the preset macro deviation threshold θ, it indicates that the uploaded macro image is different from the macro local magnification stored in the cloud database. Figure 2 If the macro deviation value V ≤ the preset macro deviation threshold θ, it indicates that the uploaded macro image is different from the macro local magnification stored in the cloud database. Figure 2 If the match is high, further comparative analysis of the product authenticity is required.

[0031] Preferably, S6 extracts the feature vector p of the uploaded macro image and the feature vector q of the macro local magnification image in the cloud database by using the trained convolutional neural network model, including color features, texture features and shape features, and calculates the similarity between the two feature vectors as p j and q j are the values ​​of the feature vectors p and q in the jth dimension respectively, and m represents the dimension of the vector.

[0032] Preferably, S7 further identifies the authenticity of the product based on the result of the in-depth feature analysis, by setting a similarity threshold K, and comparing the similarity threshold K with the similarity d(p, q) between the two feature vectors, so as to further identify the authenticity of the product. If d(p, q)<K, it is considered that the similarity of the two feature vectors is high, and the product is further determined to be authentic. If d(p, q)≥K, it is considered that the similarity of the two feature vectors is poor, and the product is further determined to be counterfeit.

[0033] Preferably, when the S8 receives the anti-counterfeiting identification result, the cloud server will encapsulate the anti-counterfeiting identification result into a data format suitable for transmission, and then send the result data to the user's mobile phone through the wireless network. The mobile phone APP receives the identification result and displays it to the user. The user can view the authenticity identification result of the product through the mobile phone application, and the user can perform corresponding operations based on the identification result, including confirming the purchase, contacting the seller, and reporting counterfeits.

[0034] To achieve the above object, the present invention provides the following technical solution: an article anti-counterfeiting system based on local feature visual information, implementing the above article anti-counterfeiting method based on local feature visual information, comprising:

[0035] Production line coding module: During production, a coding device is used to print a QR code on the product, thereby giving each product an independent QR code;

[0036] Local feature selection module: selects a specific local area on the product as the image feature area during production or generates a random image as the image feature area during production;

[0037] Image information extraction module: by using high-resolution imaging equipment to shoot the selected image feature area, extract a general image of the image feature area and two macro-enlarged local images at different positions, and upload the extracted general image of the image feature area and two macro-enlarged local images at different positions to a cloud database for storage;

[0038] Information retrieval module: retrieve product information based on the QR code on the product. Users scan the QR code on the product with their mobile phone to access the cloud database and retrieve information related to the product number. Users can judge the authenticity of the product based on the retrieved product information;

[0039] Preliminary comparison module: Advanced verification of products based on user judgment results, based on macro pictures uploaded by users and macro partial enlargement Figure 2 Conduct comparative analysis and calculate the macro difference value to analyze the matching situation;

[0040] Feature in-depth analysis module: Based on the preliminary comparative analysis results, the uploaded macro images and the macro partial magnification images in the cloud database are further analyzed through the convolutional neural network model, and the similarity is calculated;

[0041] Anti-counterfeiting identification module: used to identify the authenticity of products, identify the authenticity of products through the results of in-depth feature analysis, and transmit the anti-counterfeiting identification results to the result feedback module;

[0042] Result feedback module: transmit the anti-counterfeiting identification results to the mobile phone through the cloud server.

[0043] Technical effects and advantages of the present invention:

[0044] The present invention uses a coding device to print a two-dimensional code on a product, thereby assigning an independent two-dimensional code to each product, and then selects a specific local area on the product as an image feature area, and uses a high-resolution imaging device to shoot the selected image feature area, extracts a general image of the image feature area and macro-enlarged images of two positions, and retrieves product information according to the two-dimensional code on the product. The user scans the two-dimensional code on the product with a mobile phone to access the cloud database and retrieve information related to the product number. The user can judge the authenticity of the product through the retrieved product information, and the macro picture uploaded by the user and the macro-enlarged image can be used to determine the authenticity of the product. Figure 2 A comparative analysis is conducted to analyze the matching situation, and a convolutional neural network model is used to further analyze the features of the uploaded macro pictures and the macro partial enlarged pictures in the cloud database to identify the authenticity of the product. Finally, the anti-counterfeiting identification results are transmitted to the mobile phone through the cloud server. By using the mobile phone APP to scan the product information QR code, the anti-counterfeiting identification operation is interacted with the user, which is beneficial to reducing the cost of anti-counterfeiting identification and the difficulty of promoting anti-counterfeiting identification. By conducting anti-counterfeiting verification and screening of products multiple times, the accuracy and reliability of anti-counterfeiting inspection of items can be guaranteed. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 It is a diagram of the method steps of the present invention.

[0046] Figure 2 It is a system structure block diagram of the present invention. DETAILED DESCRIPTION

[0047] The technical solution of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the present invention. In addition, the forms of the various structures recorded in the following embodiments are merely illustrative. The present invention involves an anti-counterfeiting method and system for articles based on local feature visual information. It is not limited to the various structures recorded in the following embodiments. All other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0048] like Figure 1 This embodiment provides an anti-counterfeiting method for articles based on local feature visual information, including:

[0049] S1: During production, a QR code is printed on the product using a coding device, thereby giving each product an independent QR code.

[0050] In this embodiment, S1 generates a unique number for each product, converts the unique number into a QR code using a coding device, and prints it on the product.

[0051] It should be noted that the specific steps for printing the product QR code include:

[0052] Step 1: Select the inkjet printer suitable for the production line, install or configure the corresponding inkjet printer software on the inkjet printer to generate and transmit QR code data, and connect the inkjet printer to the cloud database;

[0053] Step 2: Establish a unique numbering rule for the product, including production date, batch number, and serial number, and generate a unique number for each product through a cloud database;

[0054] Step 3: Convert the unique number of the product into a QR code format and optimize the QR code according to the printing equipment and the surface material of the product;

[0055] Step 4: Calibrate the inkjet printer and set the printing parameters;

[0056] Step 5: Start the inkjet printer. When the product reaches the printer through the conveyor belt of the production line, the sensor is triggered to activate the inkjet printer. The inkjet printer then prints the QR code at the designated location of the product according to the preset parameters.

[0057] S2: During production, a specific local area on the product is selected as the image feature area or a random image is generated during production as the image feature area.

[0058] In this embodiment, in S2, a random image is generated during production as an image feature area; and then an image of a selected area is captured by using a high-resolution camera, wherein the specific local area includes a product serial number and a special pattern of the product.

[0059] S3: By using a high-resolution imaging device to shoot the selected image feature area, an overall image of the image feature area and macro-enlarged images of two locations are extracted, and the overall image of the extracted image feature area and the macro-enlarged images of two locations are uploaded to a cloud database for storage.

[0060] In this embodiment, the S3 takes photos of the selected image feature area and two different positions of the image feature area on the production line to obtain the overall image of the image feature area and the macro local magnification. Figure 1 And macro local magnification Figure 2 , and processes the captured images, including image calibration, contrast enhancement, and sharpening, extracts image data from the processed images, and uploads the extracted image data to a cloud database.

[0061] It should be specifically noted that by installing high-resolution imaging equipment on the production line to photograph the image feature area during the product production process, the obtained overall image is used to display the appearance of the entire image feature area selected by the product, and image data is extracted from it, and the obtained macro local magnification Figure 1 It is used to take a macro shot of the first position in the image feature area and extract detailed local image data from it. Figure 2 It is used to take a macro photo of a second position within the image feature area and extract another detailed local image data from it, and then upload the extracted image data to a cloud database in a certain format and structure, and each image is associated with a unique QR code of the product;

[0062] The macro magnification position is marked on the general map; and a general map of the extracted image feature area and a macro local magnification map are uploaded to the cloud database for storage as the anti-counterfeiting information of the item. Figure 1 Used for consumer anti-counterfeiting pictures, macro local magnification Figure 2 As a confidential anti-counterfeiting image of the manufacturer; users obtain the macro feature position based on the general image displayed by the mobile phone software, and take macro pictures of the object according to the software prompts;

[0063] Cloud-based macro local magnification obtained based on mobile phone APP Figure 1 Macro zoom shot taken on the APP Figure 1 , the user visually compares the difference between the two pictures on the APP to obtain verification, and at the same time, the macro picture uploaded by the user and the macro partial enlargement are Figure 1 Perform comparative analysis and calculate the macro difference value to analyze the matching situation for advanced verification.

[0064] S4: Retrieve product information based on the QR code on the product. The user scans the QR code on the product with a mobile phone to access the cloud database and retrieve information related to the product number. The user can judge the authenticity of the product based on the retrieved product information.

[0065] In this embodiment, when the user's mobile phone accesses the cloud database, S4 will immediately retrieve information related to the product number from the database to obtain the overall image and macro-enlarged local image of the image feature area corresponding to the product. Figure 1 , and sent to the mobile phone, the user can visually compare the overall image of the product image feature area sent to the mobile phone and a macro local enlarged image to determine the authenticity of the product. Figure 1 When there is doubt or the product is a high-value commodity, the system will automatically prompt for advanced verification and guide the user through the advanced verification process. The user can follow the prompts and use the macro camera function of the mobile phone to take a picture of the designated location on the product and upload it to the cloud database.

[0066] It should be specifically noted that the user scans the product QR code with his mobile phone and downloads the anti-counterfeiting identification APP as prompted. After the APP is running, it will automatically read or require the consumer to enter the unique number of the product and send it to the cloud server. After the cloud server receives the product number, it will retrieve the information corresponding to the number in the database. After finding the matching information, the server will send the corresponding product general image and a macro picture to the consumer's mobile phone APP. The consumer compares the general image displayed by the APP with the product-specific feature image. When the user clicks on a specific area on the image or tries to proceed to the next step, the system detects the user's doubts or recognizes the high-value attributes of the product, and automatically pops up an advanced verification prompt.

[0067] S5: Advanced product verification based on user judgment results, based on macro images uploaded by users and macro partial magnification Figure 2 Perform comparative analysis and calculate the macro difference value to analyze the matching situation.

[0068] In this embodiment, when the cloud database receives the macro picture uploaded by the user, S5 immediately compares it with the macro local zoom in the cloud database. Figure 2 Compare and analyze to further determine the authenticity of the product by uploading macro images and zooming in on the macro parts stored in the cloud database. Figure 2 Perform comparative analysis and calculate the macro deviation value, and determine whether to conduct further analysis and comparison based on the macro deviation value.

[0069] The specific analysis method of the macro deviation value is as follows:

[0070] Step S511: Find the uploaded macro image and the macro local magnification stored in the cloud database by using a feature point detection algorithm Figure 2 The feature points in the image are analyzed and calculated, and the uploaded macro image and the macro local magnification stored in the cloud database are compared. Figure 2 The similarity deviation value is M represents the number of matching feature points in the uploaded macro image, and N represents the macro local magnification of the cloud database. Figure 2 The number of matching feature points in ;

[0071] Step S512: Analyze and calculate the uploaded macro image and the macro local magnification stored in the cloud database Figure 2 The color matching deviation value is C i Indicates the i-th color channel value corresponding to the uploaded macro image, D i A macro zoom of a cloud database Figure 2 The corresponding i-th color channel value, R max Indicates the maximum value of the color channel;

[0072] Step S513: Analyze and calculate the uploaded macro image and the macro local magnification stored in the cloud database Figure 2 The texture consistency deviation value is X represents the texture feature set in the uploaded macro image, and Y represents the macro local magnification in the cloud database. Figure 2 The texture feature set in , |X∩Y represents the number of features shared by the two sets, |X| and |Y| represent the size of their respective feature sets;

[0073] Step S514: comprehensively calculating the macro deviation value as V=α×S+β×CD+γ×TD, where α, β, and γ represent weight coefficients respectively;

[0074] The authenticity of the product can be determined by comparing the macro deviation value V with the preset macro deviation threshold θ. If the macro deviation value V> the preset macro deviation threshold θ, it indicates that the uploaded macro image is different from the macro local magnification stored in the cloud database. Figure 2 If the macro deviation value V ≤ the preset macro deviation threshold θ, it indicates that the uploaded macro image is different from the macro local magnification stored in the cloud database. Figure 2 If the match is high, further comparative analysis of the product authenticity is required.

[0075] S6: Based on the preliminary comparative analysis results, the uploaded macro images and the macro local magnification images in the cloud database are further analyzed through the convolutional neural network model, and the similarity is calculated.

[0076] In this embodiment, S6 uses the trained convolutional neural network model to extract the feature vector p of the uploaded macro image and the feature vector q of the macro local magnification image in the cloud database, including color features, texture features, and shape features, and calculates the similarity between the two feature vectors as p j and q j are the values ​​of the feature vectors p and q in the jth dimension respectively, and m represents the dimension of the vector.

[0077] S7: used to identify the authenticity of the product, identify the authenticity of the product through the results of in-depth feature analysis, and transmit the anti-counterfeiting identification results to S8;

[0078] In this embodiment, S7 further identifies the authenticity of the product based on the result of the in-depth feature analysis, and sets a similarity threshold K, and compares the similarity threshold K with the similarity d(p, q) between the two feature vectors to further identify the authenticity of the product. If d(p, q) < K, it is considered that the similarity between the two feature vectors is high, and the product is further determined to be authentic. If d(p, q) ≥ K, it is considered that the similarity between the two feature vectors is poor, and the product is further determined to be counterfeit.

[0079] S8: The anti-counterfeiting identification result is transmitted to the mobile phone via the cloud server.

[0080] In this embodiment, when the S8 receives the anti-counterfeiting identification result, the cloud server will encapsulate the anti-counterfeiting identification result into a data format suitable for transmission, and then send the result data to the user's mobile phone through the wireless network. The mobile phone APP receives the identification result and displays it to the user. The user can view the authenticity identification result of the product through the mobile phone application, and the user can perform corresponding operations based on the identification result, including confirming the purchase, contacting the seller, and reporting counterfeits.

[0081] like Figure 2The embodiment shown provides an implementation system corresponding to an anti-counterfeiting method for articles based on local feature visual information, including a production line coding module, a local feature selection module, an image information extraction module, an information retrieval module, a preliminary comparison module, a feature in-depth analysis module, an anti-counterfeiting identification module and a result feedback module. The production line coding module is connected to the information retrieval module, the local feature selection module is connected to the image information extraction module, the image information extraction module is connected to the information retrieval module, the information retrieval module is connected to the preliminary comparison module, the preliminary comparison module is connected to the feature in-depth analysis module, the feature in-depth analysis module is connected to the anti-counterfeiting identification module, and the anti-counterfeiting identification module is connected to the result feedback module.

[0082] The production line coding module uses a coding device to print a QR code on the product during production, thereby giving each product an independent QR code;

[0083] The local feature selection module selects a specific local area on the product as the image feature area during production or generates a random image as the image feature area during production;

[0084] The image information extraction module uses a high-resolution imaging device to shoot the selected image feature area, extracts a general image of the image feature area and two macro-enlarged local images at different positions, and uploads the extracted general image of the image feature area and two macro-enlarged local images at different positions to a cloud database for storage;

[0085] The information retrieval module retrieves product information based on the QR code on the product. The user scans the QR code on the product with a mobile phone to access the cloud database and retrieve information related to the product number. The user can judge the authenticity of the product based on the retrieved product information;

[0086] The preliminary comparison module performs advanced verification of the product based on the user's judgment results, and Figure 2 Conduct comparative analysis and calculate the macro difference value to analyze the matching situation;

[0087] The feature in-depth analysis module performs further feature in-depth analysis on the uploaded macro image and the macro partial magnification image in the cloud database based on the analysis results of the preliminary comparison through a convolutional neural network model, and calculates the similarity;

[0088] The anti-counterfeiting identification module is used to identify the authenticity of the product, identify the authenticity of the product through the result of in-depth feature analysis, and transmit the anti-counterfeiting identification result to the result feedback module;

[0089] The result feedback module transmits the anti-counterfeiting identification result to the mobile phone through the cloud server

[0090] Finally: The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

[0091] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. An anti-counterfeiting method for articles based on local feature visual information, characterized in that: include: S1: During production, a QR code is printed on the product using a coding device, thereby giving each product an independent QR code; S2: by selecting a specific local area on the product as the image feature area during production or generating a random image as the image feature area during production; S3: By using a high-resolution imaging device to shoot the selected image feature area, a general image of the image feature area and two macro-enlarged local images at different positions are extracted, and the general image of the image feature area and the two macro-enlarged local images at different positions are uploaded to the cloud database for storage S4: Retrieve product information based on the QR code on the product. The user scans the QR code on the product with a mobile phone to access the cloud database and retrieve information related to the product number. The user can judge the authenticity of the product based on the retrieved product information. S5: Perform advanced verification of the product based on the user's judgment results, compare and analyze the macro pictures uploaded by the user with the macro partial enlarged pictures, calculate the macro difference value, and thus analyze the matching situation; S6: Based on the preliminary comparative analysis results, the uploaded macro images and the macro partial magnification images in the cloud database are further analyzed through the convolutional neural network model, and the similarity is calculated; S7: used to identify the authenticity of the product, identify the authenticity of the product through the results of in-depth feature analysis, and transmit the anti-counterfeiting identification results to S8; S8: The anti-counterfeiting identification result is transmitted to the mobile phone via the cloud server.

2. The method for anti-counterfeiting of articles based on local feature visual information according to claim 1, characterized in that: The S1 generates a unique number for each product, converts the unique number into a QR code using a coding device, and prints it on the product.

3. The method for anti-counterfeiting of articles based on local feature visual information according to claim 2, characterized in that: In S2, a random image is generated as an image feature area during production.

4. The method for anti-counterfeiting of articles based on local feature visual information according to claim 1, characterized in that: The S2 captures an image of a selected area by using a high-resolution camera, wherein the specific local area includes a product serial number and a special pattern of the product.

5. The method for anti-counterfeiting of articles based on local feature visual information according to claim 1, characterized in that: The S3 shoots the selected image feature area and two specific positions of the image feature area on the production line to obtain a general image of the image feature area, a macro local enlarged image 1 and a macro local enlarged image 2, respectively. The macro enlarged image is the anti-counterfeiting feature of the item, and processes the captured images, including image calibration, contrast enhancement and sharpening, extracts image data from the processed images, and uploads the extracted image data to a cloud database.

6. The method for anti-counterfeiting of articles based on local feature visual information according to claim 1, characterized in that: When the user accesses the cloud database through the mobile phone software, S4 will immediately retrieve information related to the product number from the database, obtain the general image of the image feature area corresponding to the product and the macro local enlarged image 1, and send them to the mobile phone. The user can visually compare and verify the macro image taken by the software with the macro image in the database to determine the authenticity of the product. If the user has doubts about the macro local enlarged image 1 or the product is a high-value commodity, the software can automatically prompt for advanced verification and guide the user to perform the advanced verification process. The user can follow the prompt information and use the macro camera function of the mobile phone to shoot the designated position on the product to form a macro local enlarged image 2 and upload it to the cloud database.

7. The method for anti-counterfeiting of articles based on local feature visual information according to claim 1, characterized in that: In S5, when the cloud database receives the macro picture uploaded by the user, it immediately compares and analyzes it with the second macro partial enlarged image in the cloud database, so as to further determine the authenticity of the product, by comparing and analyzing the uploaded macro picture with the second macro partial enlarged image stored in the cloud database, and calculating the macro deviation value, and judging whether to perform further analysis and comparison according to the macro deviation value; The specific analysis method of the macro deviation value is as follows: Step S511: Find the feature points in the uploaded macro image and the second macro partial enlarged image stored in the cloud database by using the feature point detection algorithm, and then analyze and calculate the similarity deviation value between the uploaded macro image and the second macro partial enlarged image stored in the cloud database. M represents the number of matching feature points in the uploaded macro image, and N represents the number of matching feature points in the macro partial enlargement image 2 in the cloud database; Step S512: Analyze and calculate the color matching deviation between the uploaded macro image and the macro partial enlarged image 2 stored in the cloud database. C i Indicates the i-th color channel value corresponding to the uploaded macro image, D i represents the i-th color channel value corresponding to the macro local magnification image 2 in the cloud database, R max Indicates the maximum value of the color channel; Step S513: Analyze and calculate the texture consistency deviation value between the uploaded macro image and the macro partial enlarged image 2 stored in the cloud database. X represents the texture feature set in the uploaded macro image, Y represents the texture feature set in the macro partial enlargement image 2 in the cloud database, |X∩Y represents the number of features shared by the two sets, |X| and |Y| represent the size of their respective feature sets respectively; Step S514: comprehensively calculating the macro deviation value as V=α×S+β×CD+γ×TD, where α, β, and γ represent weight coefficients respectively; The authenticity of the product is determined by comparing the macro deviation value V with the preset macro deviation threshold θ. If the macro deviation value V>the preset macro deviation threshold θ, it indicates that the uploaded macro image does not match the macro partial enlarged image 2 stored in the cloud database, and the product is determined to be counterfeit, and the determination result at this time is fed back to the mobile phone. If the macro deviation value V≤the preset macro deviation threshold θ, it indicates that the uploaded macro image has a high degree of match with the macro partial enlarged image 2 stored in the cloud database, and further comparative analysis of the authenticity of the product is required.

8. The method for anti-counterfeiting of articles based on local feature visual information according to claim 1, characterized in that: The S6 extracts the feature vector p of the uploaded macro image and the feature vector q of the macro local magnification image in the cloud database by using the trained convolutional neural network model, including color features, texture features and shape features, and calculates the similarity between the two feature vectors as p j and q j are the values ​​of the feature vectors p and q in the jth dimension respectively, and m represents the dimension of the vector.

9. The method for anti-counterfeiting of articles based on local feature visual information according to claim 1, characterized in that: The S7 further identifies the authenticity of the product based on the result of the in-depth feature analysis, by setting a similarity threshold K, and comparing the similarity threshold K with the similarity d(p, q) between the two feature vectors, so as to further identify the authenticity of the product. If d(p, q)<K, it is considered that the similarity of the two feature vectors is high, and the product is further determined to be authentic. If d(p, q)≥K, it is considered that the similarity of the two feature vectors is poor, and the product is further determined to be counterfeit.

10. The method for anti-counterfeiting of articles based on local feature visual information according to claim 1, characterized in that: When the S8 receives the anti-counterfeiting identification result, the cloud server will encapsulate the anti-counterfeiting identification result into a data format suitable for transmission, and then send the result data to the user's mobile phone through the wireless network. The mobile phone APP receives the identification result and displays it to the user. The user can view the authenticity identification result of the product through the mobile phone application, and the user can perform corresponding operations based on the identification result, including confirming the purchase, contacting the seller, and reporting counterfeits.

11. An article anti-counterfeiting system based on local feature visual information, implementing an article anti-counterfeiting method based on local feature visual information as claimed in any one of claims 1 to 10, characterized in that: include: Production line coding module: During production, a coding device is used to print a QR code on the product, thereby giving each product an independent QR code; Local feature selection module: selects a specific local area on the product as the image feature area during production or generates a random image as the image feature area during production; Image information extraction module: by using high-resolution imaging equipment to shoot the selected image feature area, extract a general image of the image feature area and two macro-enlarged local images at different positions, and upload the extracted general image of the image feature area and two macro-enlarged local images at different positions to a cloud database for storage; Information retrieval module: retrieve product information based on the QR code on the product. Users scan the QR code on the product through the mobile phone APP to access the cloud database and retrieve information related to the product number. Users can judge the authenticity of the product based on the retrieved product information; Preliminary comparison module: Perform advanced verification of the product based on the judgment result of the mobile phone, compare and analyze the macro picture uploaded by the user with the macro partial enlargement picture 2, calculate the macro difference value, and thus analyze the matching situation; Feature in-depth analysis module: Based on the preliminary comparative analysis results, the uploaded macro images and the macro partial magnification images in the cloud database are further analyzed through the convolutional neural network model, and the similarity is calculated; Anti-counterfeiting identification module: used to identify the authenticity of products, identify the authenticity of products through the results of in-depth feature analysis, and transmit the anti-counterfeiting identification results to the result feedback module; Result feedback module: transmit the anti-counterfeiting identification results to the mobile phone through the cloud server.

Images