CN117725942A - Identification early warning method and system for label texture anti-counterfeiting - Google Patents
Identification early warning method and system for label texture anti-counterfeiting Download PDFInfo
- Publication number
- CN117725942A CN117725942A CN202410171122.0A CN202410171122A CN117725942A CN 117725942 A CN117725942 A CN 117725942A CN 202410171122 A CN202410171122 A CN 202410171122A CN 117725942 A CN117725942 A CN 117725942A
- Authority
- CN
- China
- Prior art keywords
- counterfeiting
- image
- label
- texture
- label image
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 31
- 239000000463 material Substances 0.000 claims abstract description 19
- 238000005520 cutting process Methods 0.000 claims abstract description 7
- 239000013598 vector Substances 0.000 claims description 55
- 238000004458 analytical method Methods 0.000 claims description 28
- 238000004364 calculation method Methods 0.000 claims description 23
- 238000000605 extraction Methods 0.000 claims description 20
- 238000007781 pre-processing Methods 0.000 claims description 11
- 238000013507 mapping Methods 0.000 claims description 9
- 230000002159 abnormal effect Effects 0.000 claims description 7
- 238000012216 screening Methods 0.000 claims description 7
- 230000011218 segmentation Effects 0.000 claims description 6
- 238000011156 evaluation Methods 0.000 claims description 5
- 238000003709 image segmentation Methods 0.000 claims description 5
- 238000012800 visualization Methods 0.000 claims description 5
- 238000006243 chemical reaction Methods 0.000 claims description 4
- 238000012545 processing Methods 0.000 abstract description 9
- 230000000694 effects Effects 0.000 abstract description 6
- 230000001915 proofreading effect Effects 0.000 description 4
- 238000012795 verification Methods 0.000 description 4
- 238000012937 correction Methods 0.000 description 3
- 238000012986 modification Methods 0.000 description 3
- 230000004048 modification Effects 0.000 description 3
- 230000008569 process Effects 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 2
- 230000006870 function Effects 0.000 description 2
- 230000000007 visual effect Effects 0.000 description 2
- 230000004913 activation Effects 0.000 description 1
- 238000013527 convolutional neural network Methods 0.000 description 1
- 238000013136 deep learning model Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000009826 distribution Methods 0.000 description 1
- 238000004049 embossing Methods 0.000 description 1
- 239000000284 extract Substances 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 230000008520 organization Effects 0.000 description 1
- 238000003860 storage Methods 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
- 230000001960 triggered effect Effects 0.000 description 1
Classifications
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/30—Computing systems specially adapted for manufacturing
Landscapes
- Image Analysis (AREA)
Abstract
The invention discloses a recognition and early warning method and a system for label texture anti-counterfeiting, belonging to the field of image processing, wherein the method comprises the following steps: connecting an online acquisition platform, acquiring a material texture picture of a target product, determining an initial acquisition image, dividing and cutting, determining an anti-counterfeiting label image and identifying a two-dimensional code; if the identification is successful, naming the two-dimensional code value, filing and storing in an anti-counterfeiting database; the user side scans the two-dimensional code, matches in the anti-counterfeiting database, and obtains a matched anti-counterfeiting label image; acquiring an actual product label image, and jointly matching the actual product label image to determine an anti-counterfeiting recognition result; and generating a warning popup window to perform product anti-counterfeiting warning. The anti-fake label is easy to copy or modify in the prior art, the anti-fake difficulty is low, the technical problem of lack of effective anti-fake identification early warning is solved, the anti-fake label is formed by utilizing the material texture of the product, the anti-fake difficulty is improved, and the technical effect of effective anti-fake identification early warning is achieved.
Description
Technical Field
The invention relates to the field of image processing, in particular to a recognition and early warning method and a recognition and early warning system for label texture anti-counterfeiting.
Background
At present, anti-counterfeit labels are widely used, and the existing anti-counterfeit labels are various in variety, such as bar code anti-counterfeit labels, radio frequency identification anti-counterfeit labels, holographic anti-counterfeit labels and the like, but the anti-counterfeit labels are easily copied or modified by illegal molecules, so that the anti-counterfeit difficulty is low, the effect of distinguishing genuine products is difficult to realize, and the effective anti-counterfeit identification early warning is lacked.
Disclosure of Invention
The identification early warning method and the identification early warning system for label texture anti-counterfeiting aim to solve the technical problems that in the prior art, an anti-counterfeiting label is easy to copy or modify, the counterfeiting difficulty is low, and effective anti-counterfeiting identification early warning is lacking.
In view of the above problems, the present application provides an identification and early warning method and system for label texture anti-counterfeiting.
In a first aspect of the present disclosure, an identification and pre-warning method for label texture anti-counterfeiting is provided, the method comprising: connecting an online acquisition platform, acquiring a material texture picture of a target product, and determining an initial acquisition image; based on the preset label gesture, the initial acquisition image is segmented and cut, an anti-counterfeiting label image is determined, and two-dimensional code recognition is performed; if the identification is successful, naming the two-dimensional code value of the anti-counterfeiting label image, and filing the anti-counterfeiting label image into an anti-counterfeiting database; the user side scans the two-dimensional code, takes the two-dimensional code value as an index, traverses the anti-counterfeiting database to match, acquires a matched anti-counterfeiting label image and visualizes a terminal display interface; acquiring an actual product label image, jointly matching the actual product label image, carrying out texture feature extraction and check decision based on a feature analysis model, and determining an anti-counterfeiting recognition result, wherein the feature analysis model is a twin network structure; based on the anti-counterfeiting recognition result, a warning popup window is generated to perform product anti-counterfeiting warning.
In another aspect of the present disclosure, there is provided an identification pre-warning system for label texture anti-counterfeiting, the system comprising: the material texture acquisition module is used for connecting with the online acquisition platform, acquiring a material texture picture of a target product and determining an initial acquisition image; the image segmentation and cutting module is used for carrying out segmentation and cutting on the initial acquisition image based on the preset label gesture, determining an anti-counterfeiting label image and carrying out two-dimensional code recognition; the image naming and archiving module is used for naming the two-dimensional code value of the anti-counterfeiting label image and archiving the two-dimensional code value into the anti-counterfeiting database if the identification is successful; the two-dimensional code value matching module is used for scanning the two-dimensional code by the user side, traversing the anti-counterfeiting database by taking the two-dimensional code value as an index to match, acquiring a matched anti-counterfeiting label image and visualizing a terminal display interface; the anti-counterfeiting recognition result module is used for collecting an actual product label image, jointly matching the anti-counterfeiting label image, carrying out texture feature extraction and check decision based on a feature analysis model, and determining an anti-counterfeiting recognition result, wherein the feature analysis model is of a twin network structure; and the product anti-counterfeiting alarm module is used for generating an alarm popup window based on an anti-counterfeiting recognition result to perform product anti-counterfeiting alarm.
One or more technical solutions provided in the present application have at least the following technical effects or advantages:
the on-line acquisition platform is connected to acquire the material texture picture of the target product, so that an initial acquisition image is determined, and the acquisition of the material texture image of the product is realized; based on the preset label gesture, the initial acquisition image is segmented and cut, an anti-counterfeiting label image is determined, two-dimensional code identification is carried out, and the anti-counterfeiting label and the two-dimensional code thereof are extracted; if the identification is successful, carrying out two-dimensional code value naming on the anti-counterfeiting label image, filing and storing the anti-counterfeiting label image into an anti-counterfeiting database, and establishing the anti-counterfeiting database by utilizing database storage and index technology; the user side scans the two-dimension code, takes the two-dimension code value as an index, traverses the anti-counterfeiting database for matching, acquires a matched anti-counterfeiting label image, performs terminal display interface visualization, and acquires the anti-counterfeiting label image from the anti-counterfeiting database by adopting a query index technology; acquiring an actual product label image, jointly matching the actual product label image, carrying out texture feature extraction and check decision based on a feature analysis model, determining an anti-counterfeiting recognition result, and carrying out texture image feature extraction and matching judgment by using a deep learning model; based on the anti-fake recognition result, a warning popup window is generated to carry out product anti-fake warning, so that the technical scheme of anti-fake recognition early warning is realized, the technical problems that in the prior art, the anti-fake label is easily copied or modified, the fake making difficulty is low, and the effective anti-fake early warning is lacking are solved, the technical effects of forming the anti-fake label by utilizing the material texture of the product per se, improving the fake making difficulty and realizing the effective anti-fake recognition early warning are achieved.
The foregoing description is only an overview of the technical solutions of the present application, and may be implemented according to the content of the specification in order to make the technical means of the present application more clearly understood, and in order to make the above-mentioned and other objects, features and advantages of the present application more clearly understood, the following detailed description of the present application will be given.
Drawings
Fig. 1 is a schematic flow chart of an identification early warning method for label texture anti-counterfeiting according to an embodiment of the present application;
fig. 2 is a schematic flow chart of generating a warning pop-up window in the identification early warning method for label texture anti-counterfeiting according to the embodiment of the present application;
fig. 3 is a schematic structural diagram of an identification and pre-warning system for label texture anti-counterfeiting according to an embodiment of the present application.
Reference numerals illustrate: the system comprises a material texture acquisition module 11, an image segmentation and clipping module 12, an image naming and archiving module 13, a two-dimensional code value matching module 14, an anti-counterfeiting recognition result module 15 and a product anti-counterfeiting warning module 16.
Detailed Description
The technical scheme provided by the application has the following overall thought:
the embodiment of the application provides a recognition and early warning method and a system for label texture anti-counterfeiting. Firstly, carrying out online acquisition and acquisition of texture images of product materials to obtain an initial acquisition image, and establishing a data base for identification of subsequent product tags. And secondly, dividing and identifying the initial acquisition image based on the preset label gesture, extracting an anti-counterfeiting label area, identifying a two-dimensional code in the label, and realizing the basis of label positioning and code value acquisition. And then, naming and archiving the extracted anti-counterfeiting label image, and establishing an anti-counterfeiting database with the query index function to support subsequent label matching. Then, the user side scans the two-dimensional code, and searches the correctly matched label image in the anti-counterfeiting database, thereby realizing extraction and presentation of the anti-counterfeiting information of the label. And then, carrying out depth feature contrast on the actual label image of the product and the matched anti-counterfeiting label image, and carrying out true and false identification based on a set feature analysis model to realize anti-counterfeiting identification of the label and obtain an anti-counterfeiting identification result. And then, carrying out anti-counterfeiting prompt or early warning feedback according to the anti-counterfeiting recognition result, and completing the whole closed loop process from label acquisition to anti-counterfeiting feedback.
Having described the basic principles of the present application, various non-limiting embodiments of the present application will now be described in detail with reference to the accompanying drawings.
Example 1
As shown in fig. 1, an embodiment of the present application provides an identification and early warning method for label texture anti-counterfeiting, where the method includes:
connecting an online acquisition platform, acquiring a material texture picture of a target product, and determining an initial acquisition image;
in the embodiment of the application, firstly, an online acquisition platform is connected, and the online acquisition platform comprises components such as image acquisition equipment, an image acquisition controller, an image acquisition workstation and the like. The image acquisition equipment is a device capable of acquiring image information, such as a high-definition digital camera, a scanner and the like; the image acquisition controller performs parameter configuration and acquisition instruction issuing on the image acquisition equipment; the image acquisition workstation is used for storing and processing image data. Then, placing the target product of which the anti-counterfeiting label needs to be collected in an effective working area of the image collecting equipment, adjusting the position and the posture of the target product, ensuring that the label area is completely exposed, and facilitating subsequent segmentation processing. And then controlling the image acquisition equipment to carry out image scanning acquisition on the target product label to obtain an initial acquisition image containing label image information. And then, outputting the initial acquisition image to an image acquisition workstation to finish the determination of the initial acquisition image. The initial acquisition image reserves the original material information and texture characteristics of the target product label, and lays a foundation for anti-counterfeiting identification processing.
Based on a preset label posture, cutting the initial acquisition image, determining an anti-counterfeiting label image and identifying a two-dimensional code;
in the embodiment of the application, firstly, a standard label posture template is preset according to the standard position and the attaching direction of a target product label, and is used as a preset label posture and is used as a positioning reference for subsequent segmentation. Then, loading an initial acquisition image, detecting a label area in the image, extracting a label outline and calculating geometric characteristic parameters including label area, a boundary box, corner coordinates and the like. And then, according to the extracted geometric characteristic parameters, combining a preset label posture, and correcting the label position in the initial acquired image by adopting a posture correction algorithm to enable the label posture to reach a standard state. And then, on the corrected and adjusted image, image segmentation is carried out according to the outline format of the label, and the anti-counterfeit label image containing the complete label content is cut out. And simultaneously, carrying out two-dimensional code recognition on the obtained anti-counterfeiting label image to obtain anti-counterfeiting code information contained in the label. Thus, the positioning segmentation of the initial acquisition image and the extraction of the anti-counterfeiting label image are completed, and the two-dimensional code information in the label is read, so that a foundation is laid for the subsequent anti-counterfeiting identification processing.
If the identification is successful, carrying out two-dimensional code value naming on the anti-counterfeiting label image, and filing and storing the anti-counterfeiting label image into an anti-counterfeiting database;
in the embodiment of the application, when two-dimensional code identification is performed, if the two-dimensional code identification is successful, the code value in the extracted two-dimensional code is obtained. And then, uniquely naming the two-dimensional code value to form a main index mark. And then, in the set anti-counterfeiting database, taking the main index mark as a file name, establishing an anti-counterfeiting file, and writing the corresponding anti-counterfeiting label image into the file for archiving. Meanwhile, an index table is created in the anti-counterfeiting database, and the mapping relation between the file identification and the file path is registered. Thus, two-dimensional code value extraction, naming, archiving and database index establishment of the anti-counterfeiting label image are completed, and an archive foundation is established for product anti-counterfeiting identification.
The user side scans the two-dimensional code, takes the two-dimensional code value as an index, traverses the anti-counterfeiting database to match, acquires a matched anti-counterfeiting label image and visualizes a terminal display interface;
in the embodiment of the application, firstly, a consumer or a detector uses a mobile terminal provided with an anti-counterfeiting identification application program to scan a two-dimensional code of a product to be verified to obtain a two-dimensional code value. And then submitting the extracted two-dimension code value to an application background server, and carrying out matching inquiry in the anti-counterfeiting database by using the two-dimension code value as an index by the application background server. And then traversing the database index table, searching a file identifier matched with the submitted two-dimensional code value, and obtaining a corresponding anti-counterfeit label image file path. And accessing the file path, extracting a matched anti-counterfeit label image, and transmitting the matched anti-counterfeit label image to a mobile terminal of a user submitting a check request. After receiving the matching anti-counterfeit label image, the mobile terminal performs visual display in the interface of the terminal by means of the application program to complete the presentation of the matching anti-counterfeit label image, and performs prepositive work for subsequent anti-counterfeit comparison and identification.
Acquiring an actual product label image, combining the matched anti-counterfeiting label image, and carrying out texture feature extraction and check decision based on a feature analysis model to determine an anti-counterfeiting recognition result, wherein the feature analysis model is of a twin network structure;
in the embodiment of the application, firstly, a user terminal shoots an image of a target product to obtain an overall product image, and the acquired overall product image is segmented and corrected to extract an actual product label image containing complete label content. Then, a matching anti-counterfeiting label image corresponding to the two-dimensional code value is extracted and is combined with an actual product label image to form an image pair. And then, constructing and starting a feature analysis model, wherein the model adopts a twin network structure, namely the feature analysis model comprises two identical branches to respectively process the matched anti-counterfeit label image and the actual product label image, and the parameters of the two branches are identical to the structure. And then, the combined matching anti-counterfeiting label image and the actual product label image are simultaneously input into a feature analysis model, and feature extraction is carried out on the matching anti-counterfeiting label image and the actual product label image respectively, so that feature vectors corresponding to the two images are obtained. And then comparing the output characteristic vectors, outputting an anti-counterfeiting recognition result, and determining the authenticity of the actual product.
And generating a warning popup window based on the anti-counterfeiting recognition result to perform product anti-counterfeiting warning.
Further, as shown in fig. 2, the embodiment of the present application further includes:
identifying the anti-counterfeiting identification result and determining an anti-counterfeiting identification sequence;
integrating pseudo code single columns based on the anti-counterfeiting mark sequence to generate the warning popup;
and visually displaying and alarming the alarming popup window on a display interface of the user terminal.
In one possible embodiment, first, an anti-counterfeit identification result is extracted, where the anti-counterfeit identification result includes a plurality of data bits representing different anti-counterfeit feature item information, for example, data bit 1 is label boundary integrity, data bit 2 is label embossed print, and data bit 3 is label ground color section. Then, according to the preset state comparison table, the meaning of the anti-counterfeiting feature represented by each data bit is identified, for example, when a certain data bit is in a '1' state, the anti-counterfeiting feature is abnormal, and when the data bit is in a '0' state, the anti-counterfeiting feature is normal. And then, recognizing the characteristic state of each data bit to obtain an anti-counterfeiting mark sequence representing the anti-counterfeiting mark of the label, wherein if the anti-counterfeiting mark sequence is 011, the label boundary integrity is normal, the label embossing mark is abnormal, and the label base color is abnormal.
Then, a mapping dictionary of the anti-counterfeiting characteristic items and corresponding pseudo codes is preset, for example: label boundary integrity-01, label embossed print-02, label base color-03. Then, loading a preset warning popup window template, wherein the template comprises a plurality of pseudo code sites and text descriptions. Meanwhile, a determined anti-counterfeiting mark sequence is obtained, each anti-counterfeiting feature item in the sequence is traversed in sequence, and for each feature item, a corresponding pseudo code is determined according to a preset mapping dictionary and written into a pseudo code position in a warning popup template. And after the anti-counterfeiting mark sequence is traversed, forming a complete pseudo code sequence to obtain a warning popup window containing an anti-counterfeiting feature item interpretation text. And then transmitting the warning popup window to the user terminal, and popping up the warning popup window by the user terminal through a display interface of the user terminal, wherein the warning popup window is positioned in the center of the display interface, and the highlighting mark is carried out by using a striking color frame. Meanwhile, the identification audio is sent out, the frame of the flashing warning popup window drives the user to pay attention to, the combined warning effect of light and sound is realized, and the product anti-counterfeiting warning is carried out. The user can manually close the popped warning popup window or automatically close the warning popup window after the specified display time is over.
Further, the embodiment of the application further includes:
if the identification is unsuccessful, carrying out failure acquisition marking on the anti-counterfeiting label image, and storing the failure acquisition label image into a failure acquisition library;
and when the failed acquisition inventory is in the newly added marked image, suspending the system operation, and carrying out manual identification and identification of the abnormal point.
In a possible implementation mode, if the two-dimensional code identification is unsuccessful, performing failure acquisition marking on the current anti-counterfeit label image, adding an attribute label of identification failure to the anti-counterfeit label image which is not successfully identified, and collecting the anti-counterfeit label image marked with the identification failure into a failure acquisition library. When the new added anti-counterfeit label image marked as 'recognition failure' in the failure collection library is detected, an abnormal removal mechanism is triggered, namely, the current automatic collection and recognition flow is stopped, a prompt is popped up, a recognition technician is required to perform manual intervention, the new added label images in the failure collection library are checked and analyzed one by one, the occurrence of the conditions of label damage, label offset and the like which cause the recognition failure are determined, corresponding identifications are given, the system is adjusted and optimized in a targeted manner, the reappearance of similar errors is avoided, and the reliability and the stability of the system are maintained.
Further, the embodiment of the application further includes:
the anti-counterfeiting identification mode comprises a system anti-counterfeiting decision and a user side decision;
the terminal display interface visualization is carried out on the matching anti-counterfeiting label image, and the user side selects a preprocessing mode based on the anti-counterfeiting recognition mode;
based on the preprocessing mode, anti-counterfeiting recognition decision is carried out on the actual product label image.
In one possible implementation, first, two anti-counterfeit identification methods are provided, including a system anti-counterfeit decision and a user terminal decision, for selection by the user terminal. The system anti-counterfeiting decision mode is that the background of the system inputs an actual product label image uploaded by a user through an anti-counterfeiting recognition engine, automatically completes image preprocessing, feature extraction and other calculations, and outputs an anti-counterfeiting recognition result of label authenticity; the user terminal decision mode is to visually present the matching anti-counterfeiting label image and the actual product label image on the user terminal interface at the same time, and the user can judge and compare the two label images by himself, so as to give an autonomous anti-counterfeiting recognition result.
Then, the user side obtains the matching anti-counterfeiting label image obtained in the anti-counterfeiting database and displays the matching anti-counterfeiting label image in an interface of the user side. Meanwhile, a selection frame prompt of the anti-counterfeiting recognition mode is flicked out of the display interface, and a user is required to select a subsequent anti-counterfeiting recognition mode, including a system anti-counterfeiting decision or a user anti-counterfeiting decision. When the user selects the anti-counterfeiting recognition mode, a preprocessing mode aiming at the anti-counterfeiting recognition mode appears in the display interface. The preprocessing mode of the system anti-counterfeiting decision is to upload an actual product label image; the preprocessing mode of the user terminal decision is that the user refers to the conditions of monitoring and comparing label difference, insufficient text definition and the like, and the user selects to re-shoot or directly submit judgment.
And then, according to the anti-counterfeiting recognition mode selected by the user, calling a corresponding preprocessing mode, and triggering the anti-counterfeiting recognition processing flow of the background. If the system anti-counterfeiting decision is selected, the user uploads the actual product label image, the actual product label image is directly input into a background anti-counterfeiting recognition model to start processing, and after feature extraction, feature comparison and authenticity analysis are completed, an anti-counterfeiting recognition result is output. If the user anti-counterfeiting decision is selected, waiting for the user to compare and judge the displayed actual product label image by referring to the matched anti-counterfeiting label image, and confirming and submitting the anti-counterfeiting recognition result by the user.
Further, the embodiment of the application further includes:
the feature analysis model comprises a first convolution branch and a second convolution branch which are parallel, and a post-calibration decision unit;
inputting the actual product label image into the first convolution branch, inputting the matching anti-counterfeiting label into the second convolution branch, carrying out image convolution feature recognition, and determining a group of texture feature vectors and two groups of texture feature vectors;
and transferring the group of texture feature vectors and the two groups of texture feature vectors to the checking decision unit, performing similarity calculation and evaluation marking, and determining the anti-counterfeiting recognition result.
In a preferred embodiment, the feature analysis model adopts a convolutional neural network structure and comprises two parallel convolutional branches, namely a first convolutional branch and a second convolutional branch, and the feature extraction is respectively carried out on the matched anti-counterfeit label image and the actual product label image through the two independent convolutional branches, and meanwhile, a later-stage proofreading decision unit is arranged to realize feature matching judgment. The first convolution branch and the second convolution branch have the same hierarchical organization architecture and parameter configuration, are composed of convolution layers, activation layers and the like, and each branch route is responsible for feature extraction of a corresponding input image; the checking decision unit receives the extracted features from the two branches, calculates distance measurement in the feature space, gives out the matching relation of the two images on the feature level, and forms a judging result aiming at whether the matching anti-counterfeiting label image and the actual product label image belong to the same product label. The feature extraction task is completed through parallel branches, the decision unit completes the matching evaluation task, the division cooperation realizes the whole anti-counterfeiting identification process, and the parallel computing capacity and the regularity of the feature analysis model are improved.
And then, inputting the obtained matching anti-counterfeiting label image and the obtained actual product label image into a feature analysis model, inputting the actual product label image into a first convolution branch by the feature analysis model, and inputting the matching anti-counterfeiting label image into a second convolution branch. And then, the first convolution branch and the second convolution branch start parallel operation, the first convolution branch performs layer-by-layer convolution calculation on an actual product label image, gradually merges and extracts the actual product label image, converts the actual product label image into a group of vectors representing visual texture features of the label to obtain a group of texture feature vectors, and meanwhile, the second convolution branch performs independent feature extraction on a matched anti-counterfeit label image to form two groups of texture feature vectors. And then, inputting the obtained one group of texture feature vectors and two groups of texture feature vectors into a proofreading decision unit, searching one-to-one correspondence between the vectors by the proofreading decision unit, carrying out dimension-by-dimension difference calculation to obtain the relative difference between each vector element, reflecting the difference of the two input images on the feature factors, and calculating the matching similarity scores between the two groups of feature vectors according to a similarity calculation formula by combining with preset weight coefficients. And setting a similarity threshold interval, and evaluating the marks of similar matching, such as complete matching, partial matching, non-matching and the like, according to the position of the matching similarity score in the range of the given interval, so as to generate an anti-counterfeiting recognition result.
Further, the embodiment of the application further includes:
obtaining a similarity calculation formula:
；
wherein,is the characteristic difference of the kth texture characteristic in a group of texture characteristic vectors and two groups of texture characteristic vectors,weight configuration for kth texture feature, +.>N is the total texture feature amount, which is the similar conversion coefficient;
performing point-to-point mapping on the group of texture feature vectors and the two groups of texture feature vectors, and calculating feature difference;
calculating the similarity coefficient of the actual product label image and the matching anti-counterfeiting label image by combining the characteristic difference based on the similarity calculation formula;
and determining an anti-counterfeiting recognition result based on the similarity coefficient.
In a preferred embodiment, firstly, texture feature parameters of a label image are preset, the texture feature parameters comprise n texture features, a texture feature template of the label is formed, and each texture feature has the function of detecting and identifying respectively aiming at different texture information. Importance weights between each texture feature are then defined, noted asThe contribution of the kth texture feature to the determination of the difference between the labels is embodied. Furthermore, the similarity transformation coefficient is initialized>For adjusting the scale of variation of the similarity index. Next, the characteristic delta parameter is set>The numerical deviation of the feature vector of the kth texture feature in the two label images is represented. Then, on the basis, a label image similarity calculation formula is constructed as follows:
；
wherein,is the characteristic difference of the kth texture characteristic in a group of texture characteristic vectors and two groups of texture characteristic vectors,weighting of kth texture featuresPut (I) at>And n is the total texture characteristic quantity, and s is the similarity coefficient of the actual product label image and the matching anti-counterfeiting label image.
Then, the proofreading decision unit receives a group of texture feature vectors and two groups of texture feature vectors, wherein each group of vectors contains n texture features, each feature sequence number i in the group of texture feature vectors is traversed, a feature sequence number j corresponding to the texture feature vectors is searched in the two groups of texture feature vectors, a one-to-one mapping pairing relation is established, namely the ith feature and the jth feature are mutually used as reference indexes, and point-to-point mapping of the feature vectors is realized. Then, after mapping pairing, the value a of the ith feature in a group of texture feature vectors and the value b of the jth feature in two groups of texture feature vectors are obtained, and feature difference is calculated to represent the difference of the two label images on the texture features. And the correction decision unit calls a preset similarity calculation formula, introduces the obtained characteristic difference into the similarity calculation formula, loads the weight value of each texture characteristic and the similarity conversion coefficient, and calculates the similarity coefficient of the actual product label image and the matching anti-counterfeiting label image. And then, presetting a similarity judging threshold value between the label images, and judging that the similarity is large if the similarity judging threshold value is set to 0.8, namely that the similarity is larger than or equal to 0.8 and the similarity is smaller than 0.8. Then, comparing the similarity coefficient of the actual product label image and the matching anti-counterfeiting label image with a similarity judging threshold value, judging whether the similarity coefficient is higher than 0.8, and if the similarity coefficient is higher than or equal to 0.8, outputting an anti-counterfeiting recognition result that the actual product label image is similar to the matching anti-counterfeiting label image, belonging to the same single product, and passing anti-counterfeiting verification; if the similarity coefficient is smaller than 0.8, the anti-counterfeiting recognition result is that the actual product label image and the matched anti-counterfeiting label image have obvious difference, and the anti-counterfeiting recognition result may not belong to the same single product, and the anti-counterfeiting verification fails.
Further, the embodiment of the application further includes:
setting a first anti-counterfeiting identification node based on contour features and a second anti-counterfeiting identification node based on texture trend;
based on the first anti-counterfeiting identification node, contour feature vector screening and similarity calculation are carried out, and a first similarity coefficient is determined;
identifying the first similarity coefficient, and if the first similarity coefficient meets a threshold value standard, performing texture feature vector screening and similarity calculation based on the second anti-counterfeiting identification node, and determining a second similarity coefficient;
the first similarity coefficient and the second similarity coefficient are added to the similarity coefficients.
In a preferred embodiment, considering that the anti-counterfeiting identification of the label image can be identified step by step from the global feature to the local feature, two levels of anti-counterfeiting judgment nodes, namely a first anti-counterfeiting identification node and a second anti-counterfeiting identification node, are arranged. The first anti-counterfeiting identification node performs preliminary screening identification on contour information such as shape boundaries, size and dimension, right angle concentration and the like of two label images based on the overall contour features of the label images; the second anti-counterfeiting identification node accurately judges consistency of the two labels in terms of color texture, line texture and the like from a dense view angle based on microscopic texture distribution and trend detail characteristics in the label image. In this way, the first anti-fake identification node is used as a global base line, unqualified contours are filtered, and the second anti-fake identification node is used as local precision, so that accuracy of true and false discrimination is improved. The first anti-fake identification node and the second anti-fake identification node are matched together to construct a pyramid-shaped framework of texture anti-fake identification.
And then loading the actual product label image and the matched anti-counterfeiting label image, calling a first anti-counterfeiting identification node, respectively extracting contour feature vectors such as contour lines, shape curvatures and the like from the actual product label image and the matched anti-counterfeiting label image, and calculating similarity coefficients of the two label images on the contour feature level to be used as a first similarity coefficient. Then, a threshold standard of the first similarity coefficient is set, for example, the threshold standard is set to be 0.9, whether the numerical value of the first similarity coefficient meets the threshold requirement is judged, if the first similarity coefficient is smaller than 0.9, the first anti-fake identification node cannot pass, the anti-fake identification result of the actual product label image and the matching anti-fake label image is directly output to be not passed, the second anti-fake identification node is not required to be checked, and the anti-fake identification efficiency is improved. If the first similarity coefficient is greater than or equal to 0.9, the appearance outlines of the two label images are considered to have similarity as a whole, and the verification of the first anti-fake identification node is passed. And at the moment, continuously calling a second anti-counterfeiting identification node to perform finer texture feature analysis on the actual product label image and the matched anti-counterfeiting label image, extracting texture feature vectors, calculating feature similarity, and obtaining a second similarity coefficient. Then, adding the first similarity coefficient and the second similarity coefficient into the similarity coefficient, wherein the first similarity coefficient reflects the similarity scores of the two label images at the overall outline characteristic level, the second similarity coefficient reflects the similarity scores of the two label images at the local detail texture characteristic level, the similarity coefficient comprehensively representing the matching similarity of the labels is obtained by summarizing the first similarity coefficient and the second similarity coefficient,
in summary, the identification and early warning method for label texture anti-counterfeiting provided by the embodiment of the application has the following technical effects:
connecting an online acquisition platform, acquiring a material texture picture of a target product, determining an initial acquisition image, acquiring a material texture image of the product, and establishing a data base for subsequent processing. Based on the preset label gesture, the initial acquisition image is segmented and cut, an anti-counterfeiting label image is determined, two-dimensional code recognition is carried out, a label area is positioned and extracted, and the two-dimensional code value in the label is recognized. If the identification is successful, the two-dimensional code value naming is carried out on the anti-counterfeit label image, the anti-counterfeit label image is built and stored in an anti-counterfeit database, and an inquired anti-counterfeit label image database is built. The user side scans the two-dimension code, takes the two-dimension code value as an index, traverses the anti-counterfeiting database to match, acquires the matched anti-counterfeiting label image, performs terminal display interface visualization, and retrieves the matched label image from the database for verification by the user. And acquiring an actual product label image, jointly matching the actual product label image, carrying out texture feature extraction and checking decision based on a feature analysis model, determining an anti-counterfeiting recognition result, realizing the true and false judgment of the actual product label image and realizing true and false recognition. Based on the anti-counterfeiting recognition result, a warning popup window is generated to perform product anti-counterfeiting warning, and effective anti-counterfeiting recognition early warning is realized.
Example two
Based on the same inventive concept as the identification and pre-warning method for label texture anti-counterfeiting in the foregoing embodiment, as shown in fig. 3, an embodiment of the present application provides an identification and pre-warning system for label texture anti-counterfeiting, the system includes:
the material texture acquisition module 11 is used for connecting an online acquisition platform, acquiring a material texture picture of a target product and determining an initial acquisition image;
the image segmentation and clipping module 12 is used for carrying out segmentation and clipping on the initial acquisition image based on a preset label gesture, determining an anti-counterfeiting label image and carrying out two-dimensional code recognition;
the image naming archiving module 13 is used for naming the two-dimensional code value of the anti-counterfeiting label image and archiving the two-dimensional code value into an anti-counterfeiting database if the identification is successful;
the two-dimensional code value matching module 14 is used for scanning the two-dimensional code by the user side, traversing the anti-counterfeiting database by taking the two-dimensional code value as an index to match, acquiring a matched anti-counterfeiting label image and visualizing a terminal display interface;
the anti-counterfeiting recognition result module 15 is used for collecting an actual product label image, combining the matched anti-counterfeiting label image, carrying out texture feature extraction and correction decision based on a feature analysis model, and determining an anti-counterfeiting recognition result, wherein the feature analysis model is of a twin network structure;
and the product anti-counterfeiting alarm module 16 is used for generating an alarm popup window based on the anti-counterfeiting identification result to perform product anti-counterfeiting alarm.
Further, the embodiment of the application further comprises a manual identification module, and the module comprises the following execution steps:
if the identification is unsuccessful, carrying out failure acquisition marking on the anti-counterfeiting label image, and storing the failure acquisition label image into a failure acquisition library;
and when the failed acquisition inventory is in the newly added marked image, suspending the system operation, and carrying out manual identification and identification of the abnormal point.
Further, the embodiment of the application further comprises an anti-counterfeiting identification decision module, and the module comprises the following execution steps:
the anti-counterfeiting identification mode comprises a system anti-counterfeiting decision and a user side decision;
the terminal display interface visualization is carried out on the matching anti-counterfeiting label image, and the user side selects a preprocessing mode based on the anti-counterfeiting recognition mode;
based on the preprocessing mode, anti-counterfeiting recognition decision is carried out on the actual product label image.
Further, the anti-counterfeit identification result module 15 includes the following steps:
the feature analysis model comprises a first convolution branch and a second convolution branch which are parallel, and a post-calibration decision unit;
inputting the actual product label image into the first convolution branch, inputting the matching anti-counterfeiting label into the second convolution branch, carrying out image convolution feature recognition, and determining a group of texture feature vectors and two groups of texture feature vectors;
and transferring the group of texture feature vectors and the two groups of texture feature vectors to the checking decision unit, performing similarity calculation and evaluation marking, and determining the anti-counterfeiting recognition result.
Further, the anti-counterfeit identification result module 15 further comprises the following steps:
obtaining a similarity calculation formula:
；
wherein,is the characteristic difference of the kth texture characteristic in a group of texture characteristic vectors and two groups of texture characteristic vectors,weight configuration for kth texture feature, +.>N is the total texture feature amount, which is the similar conversion coefficient;
performing point-to-point mapping on the group of texture feature vectors and the two groups of texture feature vectors, and calculating feature difference;
calculating the similarity coefficient of the actual product label image and the matching anti-counterfeiting label image by combining the characteristic difference based on the similarity calculation formula;
and determining an anti-counterfeiting recognition result based on the similarity coefficient.
Further, the anti-counterfeit identification result module 15 further comprises the following steps:
setting a first anti-counterfeiting identification node based on contour features and a second anti-counterfeiting identification node based on texture trend;
based on the first anti-counterfeiting identification node, contour feature vector screening and similarity calculation are carried out, and a first similarity coefficient is determined;
identifying the first similarity coefficient, and if the first similarity coefficient meets a threshold value standard, performing texture feature vector screening and similarity calculation based on the second anti-counterfeiting identification node, and determining a second similarity coefficient;
the first similarity coefficient and the second similarity coefficient are added to the similarity coefficients.
Further, the product anti-counterfeiting alarm module 16 comprises the following steps:
identifying the anti-counterfeiting identification result and determining an anti-counterfeiting identification sequence;
integrating pseudo code single columns based on the anti-counterfeiting mark sequence to generate the warning popup;
and visually displaying and alarming the alarming popup window on a display interface of the user terminal.
Any of the steps of the methods described above may be stored as computer instructions or programs in a non-limiting computer memory and may be called by a non-limiting computer processor to identify any of the methods to implement embodiments of the present application, without unnecessary limitations.
Further, the first or second element may not only represent a sequential relationship, but may also represent a particular concept, and/or may be selected individually or in whole among a plurality of elements. It will be apparent to those skilled in the art that various modifications and variations can be made in the present application without departing from the scope of the application. Thus, if such modifications and variations of the present application fall within the scope of the present application and the equivalents thereof, the present application is intended to cover such modifications and variations.
Claims (8)
1. The identification early warning method for label texture anti-counterfeiting is characterized by comprising the following steps:
connecting an online acquisition platform, acquiring a material texture picture of a target product, and determining an initial acquisition image;
based on a preset label posture, cutting the initial acquisition image, determining an anti-counterfeiting label image and identifying a two-dimensional code;
if the identification is successful, carrying out two-dimensional code value naming on the anti-counterfeiting label image, and filing and storing the anti-counterfeiting label image into an anti-counterfeiting database;
the user side scans the two-dimensional code, takes the two-dimensional code value as an index, traverses the anti-counterfeiting database to match, acquires a matched anti-counterfeiting label image and visualizes a terminal display interface;
acquiring an actual product label image, combining the matched anti-counterfeiting label image, and carrying out texture feature extraction and check decision based on a feature analysis model to determine an anti-counterfeiting recognition result, wherein the feature analysis model is of a twin network structure;
and generating a warning popup window based on the anti-counterfeiting recognition result to perform product anti-counterfeiting warning.
2. The method of claim 1, wherein after determining the security tag image and performing two-dimensional code recognition, comprising:
if the identification is unsuccessful, carrying out failure acquisition marking on the anti-counterfeiting label image, and storing the failure acquisition label image into a failure acquisition library;
and when the failed acquisition inventory is in the newly added marked image, suspending the system operation, and carrying out manual identification and identification of the abnormal point.
3. The method of claim 1, wherein after obtaining the matching security tag image and visualizing the terminal display interface, comprising:
the anti-counterfeiting identification mode comprises a system anti-counterfeiting decision and a user side decision;
the terminal display interface visualization is carried out on the matching anti-counterfeiting label image, and the user side selects a preprocessing mode based on the anti-counterfeiting recognition mode;
based on the preprocessing mode, anti-counterfeiting recognition decision is carried out on the actual product label image.
4. The method of claim 1, wherein the performing texture feature extraction and collation decisions based on the feature analysis model comprises:
the feature analysis model comprises a first convolution branch and a second convolution branch which are parallel, and a post-calibration decision unit;
inputting the actual product label image into the first convolution branch, inputting the matching anti-counterfeiting label into the second convolution branch, carrying out image convolution feature recognition, and determining a group of texture feature vectors and two groups of texture feature vectors;
and transferring the group of texture feature vectors and the two groups of texture feature vectors to the checking decision unit, performing similarity calculation and evaluation marking, and determining the anti-counterfeiting recognition result.
5. The method of claim 4, wherein performing similarity calculation and evaluation marking comprises:
obtaining a similarity calculation formula:
；
wherein,is the first texture feature vector of a group of texture feature vectors and the second texture feature vector of a group of texture feature vectorsCharacteristic differences of the k texture characteristics, +.>Weight configuration for kth texture feature, +.>N is the total texture feature amount, which is the similar conversion coefficient;
performing point-to-point mapping on the group of texture feature vectors and the two groups of texture feature vectors, and calculating feature difference;
calculating the similarity coefficient of the actual product label image and the matching anti-counterfeiting label image by combining the characteristic difference based on the similarity calculation formula;
and determining an anti-counterfeiting recognition result based on the similarity coefficient.
6. The method of claim 5, wherein said calculating a similarity coefficient of said actual product label image to said matching security label image comprises:
setting a first anti-counterfeiting identification node based on contour features and a second anti-counterfeiting identification node based on texture trend;
based on the first anti-counterfeiting identification node, contour feature vector screening and similarity calculation are carried out, and a first similarity coefficient is determined;
identifying the first similarity coefficient, and if the first similarity coefficient meets a threshold value standard, performing texture feature vector screening and similarity calculation based on the second anti-counterfeiting identification node, and determining a second similarity coefficient;
the first similarity coefficient and the second similarity coefficient are added to the similarity coefficients.
7. The method of claim 1, wherein a warning pop is generated based on the anti-counterfeit identification result, the method comprising:
identifying the anti-counterfeiting identification result and determining an anti-counterfeiting identification sequence;
integrating pseudo code single columns based on the anti-counterfeiting mark sequence to generate the warning popup;
and visually displaying and alarming the alarming popup window on a display interface of the user terminal.
8. A recognition and early warning system for label texture anti-counterfeiting, which is characterized by being used for implementing the recognition and early warning method for label texture anti-counterfeiting according to any one of claims 1-7, wherein the system comprises:
the material texture acquisition module is used for connecting an online acquisition platform, acquiring a material texture picture of a target product and determining an initial acquisition image;
the image segmentation and cutting module is used for carrying out segmentation and cutting on the initial acquisition image based on a preset label gesture, determining an anti-counterfeiting label image and carrying out two-dimensional code recognition;
the image naming and archiving module is used for naming the two-dimensional code value of the anti-counterfeiting label image and archiving the two-dimensional code value into an anti-counterfeiting database if the identification is successful;
the two-dimensional code value matching module is used for a user side to scan the two-dimensional code, traverse the anti-counterfeiting database by taking the two-dimensional code value as an index to match, acquire a matching anti-counterfeiting label image and visualize a terminal display interface;
the anti-counterfeiting recognition result module is used for collecting an actual product label image, combining the matched anti-counterfeiting label image, carrying out texture feature extraction and check decision based on a feature analysis model, and determining an anti-counterfeiting recognition result, wherein the feature analysis model is of a twin network structure;
and the product anti-counterfeiting alarm module is used for generating an alarm popup window based on the anti-counterfeiting recognition result to perform product anti-counterfeiting alarm.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410171122.0A CN117725942A (en) | 2024-02-06 | 2024-02-06 | Identification early warning method and system for label texture anti-counterfeiting |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410171122.0A CN117725942A (en) | 2024-02-06 | 2024-02-06 | Identification early warning method and system for label texture anti-counterfeiting |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN117725942A true CN117725942A (en) | 2024-03-19 |
Family
ID=90203766
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202410171122.0A Pending CN117725942A (en) | 2024-02-06 | 2024-02-06 | Identification early warning method and system for label texture anti-counterfeiting |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117725942A (en) |
Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2018176185A1 (en) * | 2017-03-27 | 2018-10-04 | 中国科学院深圳先进技术研究院 | Texture synthesis method, and device for same |
| CN109472337A (en) * | 2018-11-05 | 2019-03-15 | 上海中商网络股份有限公司 | A kind of label anti-counterfeit method and apparatus based on random character |
| CN110070159A (en) * | 2019-04-19 | 2019-07-30 | 宁波高新区意融科技有限公司 | A kind of antifalsification label and its discrimination method based on the comparison of in kind and material picture |
| WO2019242329A1 (en) * | 2018-06-20 | 2019-12-26 | 北京七鑫易维信息技术有限公司 | Convolutional neural network training method and device |
| CN114219753A (en) * | 2021-10-27 | 2022-03-22 | 国网福建省电力有限公司检修分公司 | Power equipment surface defect detection method based on deep learning and terminal |
| WO2022151460A1 (en) * | 2021-01-18 | 2022-07-21 | 罗伯特·博世有限公司 | Laser label identification device and method |
| CN115690027A (en) * | 2022-10-25 | 2023-02-03 | 李浩楠 | General characteristic retrieval anti-counterfeiting method suitable for object self texture |
| CN116664961A (en) * | 2023-07-31 | 2023-08-29 | 东莞市将为防伪科技有限公司 | Intelligent identification method and system for anti-counterfeit label based on signal code |
| CN116935449A (en) * | 2023-07-27 | 2023-10-24 | 深圳大学 | Fingerprint image matching model training method, fingerprint matching method and related medium |
| CN117172804A (en) * | 2023-09-06 | 2023-12-05 | 西安国威时代科技有限公司 | System for identifying wine true and false |
-
2024
- 2024-02-06 CN CN202410171122.0A patent/CN117725942A/en active Pending
Patent Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2018176185A1 (en) * | 2017-03-27 | 2018-10-04 | 中国科学院深圳先进技术研究院 | Texture synthesis method, and device for same |
| WO2019242329A1 (en) * | 2018-06-20 | 2019-12-26 | 北京七鑫易维信息技术有限公司 | Convolutional neural network training method and device |
| CN109472337A (en) * | 2018-11-05 | 2019-03-15 | 上海中商网络股份有限公司 | A kind of label anti-counterfeit method and apparatus based on random character |
| CN110070159A (en) * | 2019-04-19 | 2019-07-30 | 宁波高新区意融科技有限公司 | A kind of antifalsification label and its discrimination method based on the comparison of in kind and material picture |
| WO2022151460A1 (en) * | 2021-01-18 | 2022-07-21 | 罗伯特·博世有限公司 | Laser label identification device and method |
| CN114219753A (en) * | 2021-10-27 | 2022-03-22 | 国网福建省电力有限公司检修分公司 | Power equipment surface defect detection method based on deep learning and terminal |
| CN115690027A (en) * | 2022-10-25 | 2023-02-03 | 李浩楠 | General characteristic retrieval anti-counterfeiting method suitable for object self texture |
| CN116935449A (en) * | 2023-07-27 | 2023-10-24 | 深圳大学 | Fingerprint image matching model training method, fingerprint matching method and related medium |
| CN116664961A (en) * | 2023-07-31 | 2023-08-29 | 东莞市将为防伪科技有限公司 | Intelligent identification method and system for anti-counterfeit label based on signal code |
| CN117172804A (en) * | 2023-09-06 | 2023-12-05 | 西安国威时代科技有限公司 | System for identifying wine true and false |
Non-Patent Citations (4)
| Title |
|---|
| GU Y Q等: "Gap-enhancedRaman tags for physically unclonable anticounterfeiting labels", 《NATURE COMMUNICATIONS》, vol. 11, no. 1, 30 November 2020 (2020-11-30), pages 1 - 13 * |
| 向绪友;周超;贺艺;刘学文;李亮;: "基于纹理图像的农产品无损标识与真伪识别技术研究", 湖南农业科学, no. 08, 27 August 2020 (2020-08-27) * |
| 张宏坤等: "基于轻量级网络的防伪标签检测算法", 《上海大学学报》, vol. 28, no. 3, 30 June 2022 (2022-06-30), pages 534 - 543 * |
| 李雨佳;李京兵;: "基于变换域的纹理防伪标签自动鉴别算法", 计算机仿真, no. 12, 15 December 2014 (2014-12-15) * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| JP6452186B2 (en) | Insurance compensation fraud prevention method, system, apparatus and readable recording medium based on coincidence of multiple photos | |
| CN109857889B (en) | Image retrieval method, device and equipment and readable storage medium | |
| KR101515256B1 (en) | Document verification using dynamic document identification framework | |
| CN105528604B (en) | A kind of bill automatic identification and processing system based on OCR | |
| KR101446376B1 (en) | Identification and verification of an unknown document according to an eigen image process | |
| Mian et al. | Three-dimensional model-based object recognition and segmentation in cluttered scenes | |
| CN109815364A (en) | A kind of massive video feature extraction, storage and search method and system | |
| CN1322471C (en) | Comparing patterns | |
| US11954928B2 (en) | Systems and methods for detecting text in images | |
| CN111435421B (en) | Traffic-target-oriented vehicle re-identification method and device | |
| CN113313149B (en) | Dish identification method based on attention mechanism and metric learning | |
| CN111144372A (en) | Vehicle detection method, device, computer equipment and storage medium | |
| CN111738979B (en) | Certificate image quality automatic checking method and system | |
| EP0923044B1 (en) | Method and means of matching documents based on spatial region layout | |
| CN117115569B (en) | Automatic object image identification and classification method and system based on machine learning | |
| CN112163600B (en) | Commodity identification method based on machine vision | |
| CN112836041A (en) | Personnel relationship analysis method, device, equipment and storage medium | |
| CN117725942A (en) | Identification early warning method and system for label texture anti-counterfeiting | |
| CN113743382B (en) | Shelf display detection method, device and system | |
| EP3958171A1 (en) | Automatic method to determine the authenticity of a product | |
| WO2006046729A1 (en) | Pattern recognition for sheet metal part models | |
| EP3475926B1 (en) | Detection of surface irregularities in coins | |
| CN112070146A (en) | Method for automatically identifying and classifying laboratory samples | |
| CN111429156A (en) | Artificial intelligence recognition system for mobile phone and application thereof | |
| EP3916656A1 (en) | Method and apparatus for tracking, damage detection and classi-fication of a shipping object using 3d scanning |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |