False image determination method, device, computer readable medium, and electronic device

Abstract

The application belongs to the technical field of computers, and particularly relates to a false image judgment method and device, a computer readable medium and an electronic device. The method comprises the following steps: performing feature extraction on a to-be-judged image to obtain image features of the to-be-judged image; performing feature matching on the image features of the to-be-judged image and inventory features of a historical image feature library to obtain target inventory features; obtaining additional information associated with the target inventory features, wherein the additional information comprises objects associated with the target inventory features; and when the total number of the objects associated with the target inventory features is greater than a first preset number, judging that the to-be-judged image is a false image. The embodiment of the application can realize intelligent detection of the to-be-judged image, identify false images in the to-be-judged image, thereby saving the auditing manpower for the false images, and improving the detection efficiency of the false images.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of computers, and particularly relates to a false image determination method and device, a computer readable medium, and an electronic device. BACKGROUND

[0002] Merchant access refers to the process of signing a contract between a merchant and a platform to become a merchant of the platform. However, during the inspection of the platform on the merchant, it is found that some image access credentials for merchant access are tampered with or counterfeited. For example, by modifying and replacing key information through image splicing, the merchant access is completed by counterfeiting the image, thereby causing harm to the security of the platform and adversely affecting the credibility of the platform and other merchants on the platform.

[0003] During the audit stage of merchant access, whether the image access credentials are counterfeit credentials can be audited by manual auditing. However, this detection method of relying on manual auditing of whether the image access credentials are counterfeit has low detection efficiency.

[0004] It should be noted that the information disclosed in the above background section is only used to strengthen the understanding of the background of the present application, and therefore can include information that does not constitute prior art known to those of ordinary skill in the art. SUMMARY

[0005] The purpose of the present application is to provide a false image determination method and device, a computer readable medium, and an electronic device, which at least partially overcome the technical problem of how to improve the detection efficiency of counterfeit images in the related art.

[0006] Other characteristics and advantages of the present application will become apparent from the following detailed description, or will be learned by practice of the present application.

[0007] According to an aspect of an embodiment of the present application, a false image determination method is provided, which comprises:

[0008] performing feature extraction on a to-be-determined image to obtain an image feature of the to-be-determined image;

[0009] performing feature matching on the image feature of the to-be-determined image and a stock feature of a historical image feature library to obtain a target stock feature, wherein the stock feature is an image feature of a historical image stored in the historical image feature library;

[0010] obtaining additional information associated with the target stock feature, the additional information comprising an object associated with the target stock feature;

[0011] determine that the image to be determined is a false image when a total number of objects associated with the target inventory feature is greater than a first preset number.

[0012] According to an aspect of some embodiments of the present application, a false image determination device is provided, which comprises:

[0013] a feature extraction module configured to perform feature extraction on an image to be determined to obtain an image feature of the image to be determined;

[0014] a feature matching module configured to perform feature matching between the image feature of the image to be determined and an inventory feature of a historical image feature library to obtain a target inventory feature, wherein the inventory feature is an image feature of a historical image stored in the historical image feature library;

[0015] an additional information acquisition module configured to acquire additional information associated with the target inventory feature, the additional information including objects associated with the target inventory feature;

[0016] a first false image determination module configured to determine that the image to be determined is a false image when a total number of objects associated with the target inventory feature is greater than a first preset number.

[0017] In some embodiments of the present application, based on the above technical solutions, the feature matching module comprises:

[0018] a feature matching unit configured to perform feature matching between the image feature of the image to be determined and each inventory feature in the historical image feature library to obtain a corresponding feature similarity;

[0019] a target inventory feature determination unit configured to determine the inventory feature corresponding to the feature similarity as the target inventory feature when the feature similarity is greater than a preset threshold.

[0020] In some embodiments of the present application, based on the above technical solutions, the feature matching unit comprises:

[0021] a feature metric distance calculation subunit configured to calculate a feature metric distance between the image feature of the image to be determined and each inventory feature of the historical image feature library;

[0022] a feature similarity calculation subunit configured to calculate a corresponding feature similarity according to the feature metric distance, wherein the feature similarity is inversely proportional to the feature metric distance.

[0023] In some embodiments of the present application, based on the above technical solutions, the feature extraction module comprises:

[0024] The to-be-judged image preprocessing unit is configured to preprocess the to-be-judged image, wherein the preprocessing includes processing the image to obtain a to-be-judged image with a fixed image size and a fixed pixel value range.

[0025] The image feature acquisition unit is configured to input the preprocessed to-be-judged image into a pre-trained model to acquire image features of the to-be-judged image through the pre-trained model.

[0026] In some embodiments of the present application, based on the above technical solutions, the false image judgment device further comprises:

[0027] The dataset image preprocessing unit is configured to perform the preprocessing on images in a scene recognition image dataset, wherein the scene recognition image dataset comprises images and scene category labels corresponding to the images.

[0028] The model training unit is configured to train a neural network model for scene classification by using the scene recognition image dataset, wherein the neural network model comprises a feature layer and a scene classification layer, the feature layer is used to extract features of images to obtain image features of a preset dimension, and the scene classification layer is used to determine a scene category corresponding to an image according to image features of the image.

[0029] The scene classification layer removal unit is configured to remove the scene classification layer from the trained neural network model to obtain the pre-trained model.

[0030] In some embodiments of the present application, based on the above technical solutions, the false image judgment device further comprises:

[0031] The historical image preprocessing unit is configured to perform the preprocessing on historical images, wherein the historical images are historical received images, and the historical images are associated with additional information corresponding to the historical images.

[0032] The stock feature acquisition unit is configured to input the preprocessed historical images into the pre-trained model to acquire stock features corresponding to the historical images through the pre-trained model.

[0033] The historical image feature library establishment unit is configured to associate the stock features corresponding to the historical images with the additional information corresponding to the historical images, and establish the historical image feature library according to the associated stock features and the corresponding additional information.

[0034] In some embodiments of the present application, based on the above technical solutions, the false image judgment device further comprises:

[0035] The real image determination unit is configured to determine the image to be determined as a real image when a total number of objects associated with the target inventory feature is less than or equal to the first preset number.

[0036] In some embodiments of the present application, based on the above technical solutions, the false image determination device further comprises:

[0037] The historical image feature library updating unit is configured to add the image feature of the image to be determined and additional information corresponding to the image to be determined into the historical image feature library.

[0038] In some embodiments of the present application, based on the above technical solutions, the additional information acquisition module comprises:

[0039] The index information acquisition unit is configured to acquire index information associated with the target inventory feature.

[0040] The additional information reading unit is configured to read additional information associated with the index information according to the index information, so as to determine the object associated with the index information.

[0041] The total number of objects calculation unit is configured to merge the same objects when there are the same objects among the objects associated with the index information, and acquire a total number of the objects associated with the index information after the same objects are merged.

[0042] In some embodiments of the present application, based on the above technical solutions, the additional information further comprises an object state corresponding to the object, and the object state comprises a blacklist state and a normal state. The false image determination device further comprises:

[0043] The second false image determination unit is configured to determine the image to be determined as a false image when a proportion of the object state corresponding to the object associated with the target inventory feature is higher than a preset proportion.

[0044] In some embodiments of the present application, based on the above technical solutions, the additional information further comprises a subject associated with the object. The false image determination device further comprises:

[0045] The third false image determination unit is configured to determine the image to be determined as a false image when a total number of the subjects associated with the object associated with the target inventory feature is greater than a second preset number.

[0046] In some embodiments of the present application, based on the above technical solutions, the image to be determined is a store image of an object to be audited.

[0047] According to one aspect of the embodiments of this application, a computer-readable medium is provided having a computer program stored thereon, which, when executed by a processor, implements the fake image determination method as described in the above technical solutions.

[0048] According to one aspect of the embodiments of this application, an electronic device is provided, the electronic device comprising: a processor; and a memory for storing executable instructions of the processor; wherein the processor is configured to execute the fake image determination method as described above by executing the executable instructions.

[0049] According to one aspect of the embodiments of this application, a computer program product or computer program is provided, which includes computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to perform the fake image determination method as described above.

[0050] In the technical solution provided in this application embodiment, the image features of the image to be judged are matched with the existing features of the historical image feature library to obtain the target existing features. When the total number of objects associated with the target existing features is greater than a first preset number, the image to be judged is determined to be a fake image, thereby realizing intelligent detection of the image to be judged, identifying fake images in the image to be judged, thereby saving manpower for reviewing fake images and improving the detection efficiency of fake images.

[0051] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and do not limit this application. Attached Figure Description

[0052] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application. It is obvious that the drawings described below are merely some embodiments of this application, and those skilled in the art can obtain other drawings based on these drawings without any inventive effort.

[0053] Figure 1 An exemplary system architecture block diagram illustrating the application of the technical solution of this application is shown schematically.

[0054] Figure 2 A flowchart illustrating the steps of a fake image determination method according to certain embodiments of this application is shown schematically.

[0055] Figure 3A step flowchart is illustratively shown for feature extraction of the to-be-judged image in some embodiments of the present application to obtain image features of the to-be-judged image.

[0056] Figure 4 A step flowchart is illustratively shown for some embodiments of the present application to input the pre-processed to-be-judged image into the pre-trained model to obtain the image features of the to-be-judged image before the pre-trained model.

[0057] Figure 5 A step flowchart is illustratively shown for some embodiments of the present application to perform feature matching between the image features of the to-be-judged image and the inventory features of the historical image feature library to obtain the target inventory feature.

[0058] Figure 6 A step flowchart is illustratively shown for some embodiments of the present application to perform feature matching between the image features of the to-be-judged image and each inventory feature in the historical image feature library to obtain the corresponding feature similarity.

[0059] Figure 7 A step flowchart is illustratively shown for some embodiments of the present application to perform feature matching between the image features of the to-be-judged image and the inventory features of the historical image feature library to obtain the target inventory feature before the pre-trained model.

[0060] Figure 8 A step flowchart is illustratively shown for some embodiments of the present application to perform feature matching between the image features of the to-be-judged image and the inventory features of the historical image feature library to obtain the target inventory feature before the pre-trained model.

[0061] Figure 9 A step flowchart is illustratively shown for some embodiments of the present application to obtain additional information associated with the target inventory feature.

[0062] Figure 10 A step flowchart is illustratively shown for some embodiments of the present application to obtain additional information associated with the target inventory feature.

[0063] Figure 11 A step flowchart is illustratively shown for some embodiments of the present application to perform feature matching between the image features of the to-be-judged image and the inventory features of the historical image feature library to obtain the target inventory feature before the pre-trained model.

[0064] Figure 12 A step flowchart is illustratively shown for some embodiments of the present application to perform feature matching between the image features of the to-be-judged image and the inventory features of the historical image feature library to obtain the target inventory feature before the pre-trained model.

[0065] Figure 13 A step flowchart is illustratively shown for some embodiments of the present application to perform feature matching between the image features of the to-be-judged image and the inventory features of the historical image feature library to obtain the target inventory feature before the pre-trained model. DETAILED DESCRIPTION

[0066] Example implementations will now be described more fully with reference to the accompanying drawings. Example implementations can be implemented in any

[0067] Moreover, the described features, structures, or characteristics can be combined in any suitable manner in one or more embodiments. In the following description, numerous specific details are provided to give a thorough understanding of embodiments of the application. One skilled in the relevant art will recognize, however, that the

[0068] The block diagrams in the drawings show only the functionality and arrangement of physical blocks may not necessarily correspond to physical entities in an implementation. For example, not all functions shown with regard to a functional block can be implemented, one or more functions can be added, or a function can be combined with another function; or a functionality can be implemented by dedicated hardware as well as by a suitable configuration of software functioning. Thus, the term "functional block" is used only to indicate the functionality of that block, not the manner in which the functionality is implemented.

[0069] The flow diagrams depicted herein are examples only and are not necessarily meant to include all of the steps, operations, or functionality described or to follow the order described. For example, some operations / steps can be broken down further, while some operations / steps can be combined or partially combined, and the order of execution can be changed according to actual implementation.

[0070] Artificial intelligence (AI) is the theory, method, technology and application system that use digital computers or digital computer controlled machines to simulate, extend and expand human intelligence, perceive environment, acquire knowledge and use knowledge to obtain the best results. In other words, artificial intelligence is a comprehensive technology of computer science, which attempts to understand the essence of intelligence and produce a new intelligent machine that can react in a similar way to human intelligence. Artificial intelligence is to study the design principles and implementation methods of various intelligent machines, so that machines have the functions of perception, reasoning and decision-making.

[0071] Artificial intelligence technology is a comprehensive discipline, involving a wide range of fields, both hardware and software technologies. Artificial intelligence basic technologies generally include technologies such as sensors, special artificial intelligence chips, cloud computing, distributed storage, big data processing technology, operation / interaction system, mechatronics, etc. Artificial intelligence software technology mainly includes computer vision technology, speech processing technology, natural language processing technology, and machine learning / deep learning, etc.

[0072] Computer Vision (CV) Computer vision is a science that studies how to make machines "see". More specifically, it refers to the use of cameras and computers to replace the human eye to identify and measure targets, and further process graphics to make computer processing more suitable for human observation or transmission to instrument detection. As a scientific discipline, computer vision researches related theories and technologies, and tries to establish artificial intelligence systems that can obtain information from images or multidimensional data. Computer vision technology usually includes image processing, image recognition, image semantic understanding, image retrieval, OCR, video processing, video semantic understanding, video content / behavior recognition, three-dimensional object reconstruction, 3D technology, virtual reality, augmented reality, simultaneous localization and mapping, and other technologies. It also includes common face recognition, fingerprint recognition, and other biometric identification technologies.

[0073] Machine Learning (ML) Machine learning is a multi-disciplinary subject that involves probability theory, statistics, approximation theory, convex analysis, algorithm complexity theory, and other disciplines. It is a specialized study of how computers simulate or implement human learning behavior to acquire new knowledge or skills, and reorganize existing knowledge structure to continuously improve their performance. Machine learning is the core of artificial intelligence and the fundamental approach to making computers intelligent. Its applications are widespread in various fields of artificial intelligence. Machine learning and deep learning usually include artificial neural networks, belief networks, reinforcement learning, transfer learning, inductive learning, and rule-based learning.

[0074] With the research and progress of artificial intelligence technology, artificial intelligence technology has been researched and applied in many fields, such as common smart home, smart wearable devices, virtual assistants, smart speakers, smart marketing, unmanned vehicles, autonomous vehicles, drones, robots, smart medical care, smart customer service, etc. With the development of technology, artificial intelligence technology will be applied in more fields and play an increasingly important role.

[0075] Blockchain is a new application mode of distributed data storage, peer-to-peer transmission, consensus mechanism, encryption algorithm, and other computer technologies. Blockchain, in essence, is a decentralized database, a chain of data blocks generated using cryptography. Each data block contains information about a batch of network transactions, used to verify the validity of the information (anti-fake) and generate the next block. Blockchain can include blockchain underlying platforms, platform product service layers, and application service layers.

[0076] The blockchain underlying platform can include user management, basic service, smart contract and operation processing modules. Among them, the user management module is responsible for the identity information management of all blockchain participants, including maintaining the generation of public and private keys (account management), key management, and the maintenance of the correspondence between the real identity and the blockchain address (permission management), and in the case of authorization, supervising and auditing the transaction of certain real identities, providing risk control rule configuration (risk audit); the basic service module is deployed on all blockchain node devices to verify the validity of business requests, and record to the storage after the consensus of the valid request, for a new business request, the basic service first interface adaptation analysis and authentication processing (interface adaptation), then encrypt the business information through the consensus algorithm (consensus management), after encryption, the complete and consistent transmission to the shared ledger (network communication), and record storage; the smart contract module is responsible for the registration and issuance of contracts, contract triggering and contract execution, developers can define contract logic through a certain programming language, publish to the blockchain (contract registration), according to the logic of the contract terms, call the key or other event triggers to execute, complete the contract logic, and also provide contract upgrade and cancellation functions; the operation module is mainly responsible for the deployment, configuration modification, contract setting, cloud adaptation in the product release process, and the real-time state visualization output in the product running, for example: alarm.

[0077] The platform product service layer provides basic capabilities and implementation framework of typical applications, and developers can stack business characteristics based on the basic capabilities to complete the blockchain implementation of business logic. The application service layer provides application services based on the blockchain scheme for business participants to use.

[0078] The scheme provided by the embodiments of the present application relates to computer vision technology and machine learning technology of artificial intelligence, which is specifically explained through the embodiments below.

[0079] Figure 1 An exemplary system architecture block diagram to which the technical scheme of the present application is applied is schematically shown.

[0080] As Figure 1 shown, the system architecture 100 can include a terminal device 110, a network 120 and a server 130. The terminal device 110 can include various electronic devices such as smartphones, tablets, notebooks, desktop computers, etc. The server 130 can be a standalone physical server, a server cluster or a distributed system composed of multiple physical servers, or a cloud server providing cloud computing services. The network 120 can be various connection types of communication media that can provide a communication link between the terminal device 110 and the server 130, for example, can be a wired communication link or a wireless communication link.

[0081] According to implementation requirements, the system architecture in the embodiments of the present application can have any number of terminal devices, networks and servers. For example, the server 130 can be a server group composed of multiple server devices. In addition, the technical solutions provided by the embodiments of the present application can be applied to the terminal device 110, can be applied to the server 130, or can be jointly implemented by the terminal device 110 and the server 130, and the present application does not make special limitations on this.

[0082] For example, the server 130 can be loaded with the false image determination method of the present application, and the user interacts with the server 130 through the client on the terminal device 110. Merchants need to use the functions of commercial payment, commodity transaction, etc. of the platform as an object, so they need to handle object access on the platform. In the object access process or in the object access appeal process, in order to ensure the authenticity of the image submitted by the object, the server 130 of the platform signed with the object audits the image submitted by the object through the terminal device 110. When the server 130 needs to audit the image submitted by the object, the server 130 can implement the false image determination method of the present application to realize intelligent detection of the to-be-determined image, identify the false image in the to-be-determined image, thereby saving the auditing manpower for the false image, and improving the detection efficiency of the false image.

[0083] In addition, if artificial auditing is used to audit the authenticity of the to-be-determined image, because there are many auditors, the false photos seen by each auditor cannot form information sharing, which will increase the difficulty of identifying false images. However, the historical image feature library of the present application stores the inventory features corresponding to the images received by the platform in history for object access, so when the to-be-determined image is detected and determined, the information source is more comprehensive, a more accurate determination result can be obtained, the missed identification rate of the false image can be reduced, and the detection accuracy of the false image can be improved.

[0084] The object can be a merchant. Thus, for the to-be-determined image in which the merchant name or the merchant number in the merchant store photo is replaced by PS (Photoshop) tampering, image splicing or the like, but other features of the merchant store photo remain the original features, the present application can accurately identify the similarity between the to-be-determined image and the historical image, and obtain the total number of merchants associated with the historical image similar to the to-be-determined image. When the total number of merchants is greater than a preset number, it indicates that the aggregation scale of the merchants using the similar image to handle merchant access is large, so that the to-be-determined image can be identified as a false image, and the merchant access handling is rejected to ensure the safety of the platform and prevent black production groups from disturbing the platform order by tampering with image type access credentials and applying for a large number of merchants in violation of regulations. Thus, the auditing manpower for the false image can be saved, and the detection efficiency of the false image can be improved.

[0085] The server can be a standalone physical server, a server cluster or a distributed system composed of multiple physical servers, or a cloud server providing cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, CDNs, and basic cloud computing services such as big data and artificial intelligence platforms. The terminal can be a smartphone, a tablet computer, a notebook computer, a desktop computer, a smart speaker, a smart watch, and the like, but is not limited thereto. The terminal and the server can be connected directly or indirectly through wired or wireless communication, which is not limited in the present application.

[0086] The false image determination method provided by the present application will be described in detail in combination with the specific embodiments.

[0087] Figure 2 The step flowchart of the false image determination method of some embodiments of the present application is schematically shown. The execution subject of the false image determination method can be a terminal device or a server, and the like, which is not limited in the present application. As shown in Figure 2 The false image determination method can mainly include the following steps S210-S240.

[0088] S210. Feature extraction is performed on the to-be-determined image to obtain image features of the to-be-determined image.

[0089] In specific embodiments, a CNN (Convolutional Neural Networks) model combined with CV (Computer Vision) can be used to extract features of the to-be-determined image to obtain image features of the to-be-determined image.

[0090] In some embodiments, the image features can be an array or a vector, and the like, which is not limited in the present application. In specific embodiments, the image features can be a 512-dimensional feature vector.

[0091] Figure 3 The step flowchart of the feature extraction of the to-be-determined image to obtain the image features of the to-be-determined image in some embodiments of the present application is schematically shown. As shown in Figure 3 On the basis of the above embodiments, the feature extraction of the to-be-determined image in step S210 to obtain the image features of the to-be-determined image can further include the following steps S310-S320.

[0092] S310. Preprocessing is performed on the to-be-determined image, wherein the preprocessing includes processing the image to obtain a to-be-determined image with fixed image size and fixed pixel value range.

[0093] S320. inputting the preprocessed to-be-judgment image into the pre-trained model to obtain the image feature of the to-be-judgment image through the pre-trained model.

[0094] In specific embodiments, the preprocessing can be to perform scaling processing on the image to obtain an image with an image size of 256 pixels * 256 pixels, and to perform normalization processing on each pixel value in the image to make the value range of the pixel value of the image be 0-255.

[0095] After preprocessing the to-be-judgment image, the to-be-judgment image can be input into the pre-trained model to obtain the image feature of the to-be-judgment image through the pre-trained model.

[0096] Figure 4 A part of the step flow chart before the embodiment of the present application inputs the preprocessed to-be-judgment image into the pre-trained model to obtain the image feature of the to-be-judgment image through the pre-trained model is schematically shown. As shown in Figure 4 On the basis of the above embodiment, before the step S320 of inputting the preprocessed to-be-judgment image into the pre-trained model to obtain the image feature of the to-be-judgment image through the pre-trained model, the following steps S410-S430 can be further included.

[0097] S410. preprocessing the images in the scene recognition image dataset, wherein the scene recognition image dataset includes images and scene category labels corresponding to the images;

[0098] S420. training the neural network model for scene classification using the scene recognition image dataset, wherein the neural network model includes a feature layer and a scene classification layer, the feature layer is used to extract features of the images to obtain image features with a preset dimension, and the scene classification layer is used to determine the scene categories corresponding to the images according to the image features of the images;

[0099] S430. removing the scene classification layer from the trained neural network model to obtain the pre-trained model.

[0100] Specifically, training the neural network model for scene classification using the scene recognition image dataset can be to take the images in the scene recognition image dataset as the input of the neural network model, extract features of the images through the feature layer to obtain image features with a preset dimension, and take the scene category labels of the images in the scene recognition image dataset as the target output of the images in the neural network model, and train the internal structure and parameters of the feature layer and the internal structure and parameters of the scene classification layer. Thus, by the images and the scene category labels corresponding to the images, the internal structure and parameters of the feature layer and the internal structure and parameters of the scene classification layer are trained, and the training of the neural network model is realized.

[0101] The preprocessing process of the images in the scene recognition image dataset is the same as the preprocessing process of the to-be-judged image in step S310. That is, in specific embodiments, the to-be-judged image is preprocessed to obtain a to-be-judged image with an image size of 256 pixels*256 pixels and a value range of pixel values of 0-255, and the images in the scene recognition image dataset are preprocessed so that the images in the scene recognition image dataset have an image size of 256 pixels*256 pixels and a value range of pixel values of 0-255. In this way, the image specifications of the training images used to train the neural network model are the same as or similar to the image specifications of the to-be-judged image, so that the pre-trained model can more accurately extract features of the to-be-judged image.

[0102] Specifically, the neural network model for scene classification can be a residual neural network model, which includes a skip connection structure that connects the input of the model to the output of the model by skipping network layers, so that the output of the model is expressed as a linear superposition of the input and a nonlinear transformation of the output, which can alleviate the gradient vanishing problem of the neural network model, so that the pre-trained model obtained can more accurately extract features of the to-be-judged image. In specific examples, the residual neural network can include neural network models such as ResNet (Residual Network) 50, ResNet101, ResNet152, VggNet (Visual Geometry Group Net), etc.

[0103] The feature layer is used to extract features of the image to obtain image features of a preset dimension. Specifically, the feature layer can include convolution layers, pooling layers, etc. The scene classification layer is used to determine the scene category corresponding to the image according to the image features of the image output by the feature layer. In some embodiments, the preset dimension can be 256 dimensions, 512 dimensions, etc.

[0104] In specific embodiments, the scene recognition image dataset can be a places365 dataset, which includes a total of 1.8 million training images from 365 scene categories and scene category labels corresponding to each training image. For example, the scene category labels can include: restaurant, coffee and tea shop, fruit shop, pharmacy, scenic spot, etc.

[0105] In some embodiments, a validation set can also be used to input into the pre-trained model to optimize the hyperparameters of the pre-trained model. In the validation set, each scene category can include 50 images. In some embodiments, a test set can also be used to input into the pre-trained model to evaluate the pre-trained model for further optimization of the pre-trained model. In the test set, each scene category can include 900 images.

[0106] S220. performing feature matching between the image feature of the image to be judged and the inventory feature of the historical image feature library to obtain a target inventory feature, wherein the inventory feature is an image feature of a historical image stored in the historical image feature library.

[0107] Specifically, the image feature of the image to be judged can be matched with each inventory feature in the historical image feature library one by one to obtain the target inventory feature.

[0108] Figure 5 An illustrative flowchart of the step of performing feature matching between the image feature of the image to be judged and the inventory feature of the historical image feature library to obtain a target inventory feature according to some embodiments of the present application is shown in FIG. 5. Figure 5 As shown in FIG. 5, on the basis of the above embodiments, the step of performing feature matching between the image feature of the image to be judged and the inventory feature of the historical image feature library to obtain a target inventory feature in step S220 can further include steps S510-S520.

[0109] S510. performing feature matching between the image feature of the image to be judged and each inventory feature in the historical image feature library to obtain a corresponding feature similarity;

[0110] S520. determining the inventory feature corresponding to the feature similarity as the target inventory feature when the feature similarity is greater than a preset threshold.

[0111] Figure 6 An illustrative flowchart of the step of performing feature matching between the image feature of the image to be judged and each inventory feature in the historical image feature library to obtain a corresponding feature similarity according to some embodiments of the present application is shown in FIG. 6. Figure 6 As shown in FIG. 6, on the basis of the above embodiments, the step of performing feature matching between the image feature of the image to be judged and each inventory feature in the historical image feature library to obtain a corresponding feature similarity in step S510 can further include steps S610-S620.

[0112] S610. calculating a feature metric distance between the image feature of the image to be judged and each inventory feature in the historical image feature library;

[0113] S620. Calculate a corresponding feature similarity according to the feature metric distance, wherein the feature similarity is inversely proportional to the feature metric distance.

[0114] The greater the feature metric distance between the image feature of the image to be judged and the inventory feature, the smaller the feature similarity between the image feature of the image to be judged and the inventory feature, which indicates that the image to be judged is less similar to the historical image corresponding to the inventory feature. The smaller the feature metric distance between the image feature of the image to be judged and the inventory feature, the greater the feature similarity between the image feature of the image to be judged and the inventory feature, which indicates that the image to be judged is more similar to the historical image corresponding to the inventory feature.

[0115] Specifically, the feature metric distance can be Euclidean distance or cosine distance.

[0116] For example, calculating the feature metric distance between the image feature of the image to be judged and each inventory feature of the historical image feature library can be calculating the cosine distance between the image feature of the image to be judged and each inventory feature of the historical image feature library. For example, the image feature of the image to be judged is a 512-dimensional vector v1, and a certain inventory feature of the historical image feature library is a 512-dimensional vector v2. The cosine distance d1 between the image feature of the image to be judged and each inventory feature of the historical image feature library can be obtained by the following calculation formula:

[0117]

[0118] Therefore, the cosine distance between the image feature of the image to be judged and each inventory feature of the historical image feature library can be calculated as the feature metric distance between the image feature of the image to be judged and each inventory feature of the historical image feature library, and the corresponding feature similarity can be calculated according to the feature metric distance.

[0119] Figure 7 A part of the step flow chart before the image feature of the image to be judged is matched with the inventory feature of the historical image feature library to obtain the target inventory feature in some embodiments of the present application is schematically shown. As shown in Figure 7 On the basis of the above embodiments, before the image feature of the image to be judged is matched with the inventory feature of the historical image feature library to obtain the target inventory feature in step S220, the following steps S710-S730 can be further included.

[0120] S710. Preprocess the historical image, wherein the historical image is a historical received image, and the historical image is associated with additional information corresponding to the historical image;

[0121] S720. Input the preprocessed historical image into the pre-trained model to obtain the inventory feature corresponding to the historical image through the pre-trained model;

[0122] S730. Correlate the inventory features corresponding to the historical image with the additional information corresponding to the historical image, and establish a historical image feature library according to the correlated inventory features and the corresponding additional information.

[0123] Specifically, the historical image is an image received by the platform in the history for object admission. When the platform receives the historical image, additional information corresponding to the historical image is also obtained. For example, the additional information can include an object, an object number, an object name, an object state corresponding to the object, a subject associated with the object, a subject number of the subject associated with the object, a subject name of the subject associated with the object, etc., which are collectively used to complete the object admission process. In this application, the object can be a merchant, a store, etc. That is, the additional information can include a merchant, a merchant number, a merchant name, a merchant state corresponding to the merchant, a subject associated with the merchant, a subject number of the subject associated with the merchant, a subject name of the subject associated with the merchant, etc., which are collectively used to complete the merchant admission process.

[0124] Among them, the subject can be an industrial and commercial subject. During the admission and activity of the merchant on the platform, the platform can also generate a merchant state corresponding to the merchant and add it to the additional information corresponding to the historical image.

[0125] The processing process of the preprocessing of the historical image is the same as the processing process of the preprocessing of the images in the scene recognition image data set and the processing process of the preprocessing of the to-be-judged image in step S310. That is, in specific embodiments, the to-be-judged image is preprocessed to obtain a to-be-judged image with an image size of 256 pixels*256 pixels and a value range of pixel values of 0-255; the images in the scene recognition image data set are preprocessed so that the images in the scene recognition image data set have an image size of 256 pixels*256 pixels and a value range of pixel values of 0-255; and the historical image is preprocessed to obtain a historical image with an image size of 256 pixels*256 pixels and a value range of pixel values of 0-255. Thus, the image specifications of the historical images corresponding to the inventory features in the historical image feature library are the same as or similar to the image specifications of the training images used to train the neural network model and the image specifications of the to-be-judged image, so that the matching result of the feature matching of the image features of the to-be-judged image and the inventory features of the historical image feature library can accurately reflect the similarity between the to-be-judged image and the historical image. Thus, it is beneficial to find the most similar historical image to the to-be-judged image from a large number of historical images of the platform and to obtain the additional information corresponding to the historical image.

[0126] In this embodiment, the preprocessed historical image is input into the pre-trained model to obtain the inventory feature of the historical image through the pre-trained model, that is, the feature extraction method of the historical image is consistent with the feature extraction method of the image to be determined, so that the matching result of the feature matching between the image feature of the image to be determined and the inventory feature of the historical image feature library can accurately reflect the similarity between the image to be determined and the historical image. Specifically, the inventory feature of the historical image can be an image feature of a preset dimension, and the image feature of the image to be determined is an image feature of a preset dimension, which can facilitate subsequent calculation of the feature metric distance between the inventory feature and the image feature of the image to be determined, reduce the calculation amount, and improve the calculation efficiency and the detection efficiency of the false image.

[0127] Specifically, the association between the inventory feature corresponding to the historical image and the additional information corresponding to the historical image can be that the corresponding index information is generated according to the inventory feature of the historical image, and the additional information corresponding to the historical image is stored in the storage space pointed by the index information corresponding to the historical image, thereby realizing the association between the inventory feature corresponding to the historical image and the additional information corresponding to the historical image.

[0128] In specific embodiments, the Faiss (Facebook AI Similarity Search) library can be used to associate the inventory feature of the historical image with the additional information corresponding to the historical image, and establish the historical image feature library.

[0129] In some embodiments, the historical image feature library is established according to the associated inventory feature and the corresponding additional information, and the historical image does not need to be included in the historical image feature library, which can reduce the data amount in the historical image feature library, thereby improving the feature matching efficiency of obtaining the target inventory feature by matching the image feature of the image to be determined with the inventory feature of the historical image feature library, and further improving the detection efficiency of the false image.

[0130] Figure 8 The part of the steps of the offline implementation of some embodiments of the application are schematically shown. Figure 8 As shown, after the images in the Places365 dataset are preprocessed, they are input into the ResNet50 model for training, and the trained ResNet50 model is removed from the scene classification layer to obtain a pre-trained model. After the historical image is preprocessed, it is input into the pre-trained model to obtain the inventory feature corresponding to the historical image through the pre-trained model. The inventory feature corresponding to the historical image is associated with the additional information corresponding to the historical image, and the historical image feature library is established according to the associated inventory feature and the corresponding additional information. Figure 8The step flow in the method can be implemented offline, that is, after the model training and the historical image feature library are completed offline, the pre-trained model and the historical image feature library are put online to the server.

[0131] S230. Obtain additional information associated with the target inventory feature, the additional information including objects associated with the target inventory feature.

[0132] Figure 9 A schematic diagram of a step flow of obtaining additional information associated with the target inventory feature in some embodiments of the present application is shown. As shown in the above embodiments, the step S230 of obtaining additional information associated with the target inventory feature can further include the following steps S910-S930. Figure 9

[0133] S910. Obtain index information associated with the target inventory feature;

[0134] S920. Read additional information associated with the index information according to the index information to determine objects associated with the index information;

[0135] S930. When there are identical objects in the objects associated with the index information, merge the identical objects, and obtain a total number of the objects associated with the index information after the identical objects are merged.

[0136] The index information associated with the target inventory feature is obtained, and the additional information associated with the index information is read from the database according to the index information to determine the objects associated with the index information. When there are identical objects in the objects associated with the index information, the identical objects are merged, and a total number of the objects associated with the index information is obtained after the identical objects are merged. Thus, the same image uploaded by the same merchant on the platform can be avoided from being repeatedly calculated as multiple objects, thereby improving the accuracy of the calculation of the total number of objects and facilitating the improvement of the detection accuracy of the fake image.

[0137] In some embodiments, on the basis of the above embodiments, the additional information further includes an object state corresponding to the object, and the object state includes a blacklist state and a normal state. After the step S230 of obtaining the additional information associated with the target inventory feature, the following steps can be further included:

[0138] When the proportion of the object state corresponding to the object associated with the target inventory feature in the blacklist state is higher than a preset proportion, the image to be determined is determined as a fake image.

[0139] ​Therefore, the object state corresponding to the object associated with the target inventory feature can help subsequent judgment of the image authenticity of the to-be-judged image similar to the historical image corresponding to the target inventory feature. When the proportion of the object state corresponding to the object associated with the target inventory feature in the black list state is higher than the preset proportion, it can be understood that the to-be-judged image similar to the historical image corresponding to the target inventory feature is an image with black list aggregation, that is, the to-be-judged image is likely to be an image used by a black production group to apply for a large number of objects. At this time, the to-be-judged image is determined to be a false image, which can identify the black production at the object access stage, thereby avoiding the black production group from using false and fake images to access and apply for a large number of objects, thereby improving the security of the platform and improving the credibility of other objects on the platform.

[0140] Figure 10 The association diagram of the target inventory feature and the additional information in the historical image feature library of some embodiments of the present application is schematically shown. In specific examples, please refer to Figure 10The image features of the image to be determined are matched with the inventory features of the historical image feature library to obtain target inventory feature 1, target inventory feature 2, target inventory feature 3, target inventory feature 4, …, and target inventory feature 9. The additional information includes the object and the subject associated with the object. In the historical image feature library, the object associated with the target inventory feature 1 is object a, the object state of the object a is a normal state, and the subject associated with the object a is subject A; the object associated with the target inventory feature 2 is object a, the object state of the object a is a normal state, and the subject associated with the object a is subject A; the object associated with the target inventory feature 3 is object b, the object state of the object b is a blacklist state, and the subject associated with the object b is subject B; the object associated with the target inventory feature 4 is object c, the object state of the object c is a blacklist state, and the subject associated with the object c is subject B; the object associated with the target inventory feature 5 is object c, the object state of the object c is a blacklist state, and the subject associated with the object c is subject B; the object associated with the target inventory feature 6 is object d, the object state of the object d is a normal state, and the subject associated with the object d is subject C; the object associated with the target inventory feature 7 is object d, the object state of the object d is a normal state, and the subject associated with the object d is subject C; the object associated with the target inventory feature 8 is object d, the object state of the object d is a normal state, and the subject associated with the object d is subject C; the object associated with the target inventory feature 9 is object e, the object state of the object e is a blacklist state, and the subject associated with the object e is subject D. In some embodiments, the preset proportion is 50%, and by calculation, the proportion of the object state corresponding to the object associated with the target inventory feature that is a blacklist state is (3 / 5)*100%=60%, which is higher than the preset proportion of 50%, and the image to be determined is determined to be a false image.

[0141] In some embodiments, on the basis of the above embodiments, after obtaining the additional information associated with the target inventory feature in step S230, the following steps can be further included:

[0142] When the proportion of the object state corresponding to the object associated with the target inventory feature that is a blacklist state is higher than zero but less than or equal to the preset proportion, the image to be determined is pushed to the artificial review for artificial review of the image to be determined. Thus, when there are a small number of blacklist images in the historical images corresponding to the target inventory feature, the image to be determined can be further determined to be a false image or a real image through artificial review, so as to realize accurate identification of false images and avoid misidentification of real images as false images.

[0143] In some embodiments, on the basis of the above examples, the additional information further includes the subjects associated with the main body; after obtaining the additional information associated with the target inventory characteristics in step S230, the following steps can be further included:

[0144] When the total number of the main body associated with the subjects associated with the target inventory characteristics is greater than the second preset number, it is determined that the to-be-judged image is a false image.

[0145] When the total number of the main body associated with the subjects associated with the target inventory characteristics is greater than the second preset number, it indicates that the aggregation scale of the main body of the subjects associated with the historical images similar to the current to-be-judged image is large. That is, the to-be-judged image is likely to be an access image of the object applied by a black production group through a plurality of main bodies. At this time, the to-be-judged image is determined to be a false image, the black production can be identified at the object access stage, so that the black production group can be prevented from applying a large number of merchants through false and fake images, thereby improving the security of the platform and improving the credibility of other merchants on the platform.

[0146] In some embodiments, the second preset number can be 3, please continue to refer to Figure 10 The image features of the to-be-judged image are matched with the inventory characteristics of the historical image feature library to obtain target inventory characteristics 1, target inventory characteristics 2, target inventory characteristics 3, target inventory characteristics 4, …, target inventory characteristics 9. Through statistical calculation, in Figure 10 The total number of the main body associated with the subjects associated with the target inventory characteristics is 4, which is greater than the second preset number 3, so the to-be-judged image is determined to be a false image.

[0147] In some embodiments, when the total number of the main body associated with the subjects associated with the target inventory characteristics is greater than the second preset number, after determining that the to-be-judged image is a false image, the main body associated with the subjects associated with the target inventory characteristics can be packaged and reported to the background for security analysis, thereby providing strong clues for identifying the black production group behind the objects and the main body.

[0148] In some embodiments, on the basis of the above examples, the additional information further includes the main body state corresponding to the main body associated with the object; after obtaining the additional information associated with the target inventory characteristics in step S230, the following steps can be further included:

[0149] When the proportion of the main body state corresponding to the main body associated with the subjects associated with the target inventory characteristics is higher than the preset proportion, it is determined that the to-be-judged image is a false image.

[0150] S240. When the total number of objects associated with the target inventory feature is greater than the first preset number, it is determined that the image to be determined is a false image.

[0151] When the total number of objects associated with the target inventory feature is greater than the first preset number, it indicates that the aggregation scale of the objects associated with the historical image similar to the current image to be determined is large. That is, the image to be determined is likely to be a similar object store image tampered by a black production group, and the black production group applies a large number of objects on the platform through the tampered similar object store image. At this time, the image to be determined is determined to be a false image, which can identify the black production at the object access stage, thereby avoiding the black production group from applying a large number of objects through false and counterfeit images at the object access stage, thereby improving the security of the platform and improving the credibility of other objects on the platform.

[0152] In some embodiments, the first preset number can be 4, please continue to refer to Figure 10 The image features of the image to be determined are matched with the inventory features of the historical image feature library to obtain target inventory feature 1, target inventory feature 2, target inventory feature 3, target inventory feature 4, …, and target inventory feature 9. Through statistical calculation, the total number of objects associated with the target inventory feature is 5 in Figure 10 , which is greater than the first preset number 4, and it is determined that the image to be determined is a false image.

[0153] In some embodiments, when the total number of objects associated with the target inventory feature is greater than the first preset number, after determining that the image to be determined is a false image, the objects associated with the target inventory feature can be packaged and reported to the background for security analysis, thereby providing strong clues for identifying the black production group behind the objects.

[0154] In some embodiments, after obtaining the additional information associated with the target inventory feature in step S230, the following steps can be further included:

[0155] When the total number of objects associated with the target inventory feature is less than or equal to the first preset number, it is determined that the image to be determined is a real image.

[0156] When the total number of objects associated with the target inventory feature is less than or equal to the first preset number, it indicates that the aggregation scale of the objects associated with the historical image similar to the current image to be determined is not large, and therefore, the image to be determined is determined to be a real image, which can achieve the purpose of releasing the merchant corresponding to the image to be determined with a small aggregation scale, and can avoid excessive strict review to limit the object access.

[0157] In some embodiments, after obtaining the additional information associated with the target inventory feature in step S230, the method further comprises the following steps:

[0158] When the total number of objects associated with the target inventory feature is less than or equal to a third preset number, it is determined that the image to be determined is a real image, wherein the third preset number is less than the first preset number.

[0159] When the total number of objects associated with the target inventory feature is greater than the third preset number but less than the first preset number, the image to be determined is pushed to the artificial review for artificial review of the image to be determined.

[0160] When the total number of objects associated with the target inventory feature is greater than the third preset number but less than the first preset number, the image to be determined is pushed to the artificial review for artificial review of the image to be determined. Thus, when the aggregation scale of the objects associated with the historical images corresponding to the target inventory feature has a certain scale but not large, the image to be determined can be further determined to be a false image or a real image through artificial review, so as to realize accurate identification of false images and avoid misidentification of real images as false images.

[0161] In some embodiments, the first preset number can be 6 and the third preset number can be 3. Please refer to Figure 10 The image features of the image to be determined are matched with the inventory features of the historical image feature library to obtain target inventory features 1, 2, 3, 4, …, and 9. Through statistical calculation, the total number of objects associated with the target inventory feature is 5, which is greater than the third preset number 3 but less than the first preset number 6, so the image to be determined is pushed to the artificial review for artificial review of the image to be determined. Figure 10

[0162] Figure 11 Part of the steps of the online implementation of some embodiments of the present application are schematically shown. As Figure 11 After the image to be determined is preprocessed and input into the pre-trained model, 512-dimensional image features of the image to be determined are obtained. The image features of the image to be determined are matched with the 512-dimensional inventory features of the historical image feature library to obtain target inventory features. After obtaining the additional information associated with the target inventory feature, the image to be determined is determined to be a false image or a real image according to the above embodiments. Figure 11 The step flow in Figure 11 The step flow in Figure 8 ​The part of the steps of the embodiment of the present application shown online can save the online resources of the server and improve the implementation efficiency of the server Figure 11 The identification efficiency of the false image.

[0163] In some embodiments, on the basis of the above embodiments, after determining that the to-be-determined image is a real image or after determining that the to-be-determined image is a false image, the following steps can be further included:

[0164] The image features of the to-be-determined image and the additional information corresponding to the to-be-determined image are added to the historical image feature library.

[0165] Therefore, after determining whether the to-be-determined image is a real image or a false image, the image features of the to-be-determined image and the additional information corresponding to the to-be-determined image can be added to the historical image feature library, so as to realize a real-time updating mechanism of the historical image feature library, avoid the problem that the identification of a large number of false images submitted in a concentrated manner is not timely due to the delay of data updating, improve the comprehensiveness of the identification of false images, and thus improve the accuracy of the identification of false images, so as to comprehensively avoid the black production group from applying for a large number of merchants by using false images, thereby further improving the security of the platform and further improving the credibility of other merchants on the platform.

[0166] In some embodiments, on the basis of the above embodiments, the to-be-determined image can be a to-be-audited store image of an object submitted by the object. In some embodiments, the to-be-determined image can also be other business access images, such as a food image of the object, a product image of the object, an environment image of the object, and the like.

[0167] Specifically, the to-be-determined image can be a to-be-audited store image of a merchant submitted by the merchant. In some embodiments, the to-be-determined image can also be other business access images, such as a food image of the merchant, a product image of the merchant, an environment image of the merchant, and the like.

[0168] In summary, the embodiment of the present application realizes intelligent detection of the to-be-determined image by matching the image features of the to-be-determined image with the inventory features of the historical image feature library to obtain target inventory features, and determining that the to-be-determined image is a false image when the total number of objects associated with the target inventory features is greater than a first preset number, so as to identify the false image in the to-be-determined image, thereby saving the auditing manpower for the false image and improving the detection efficiency of the false image.

[0169] It should be noted that although the various steps of the methods of the present application are described in a particular order in the accompanying drawings, this is not required or implied that the steps must be performed in that particular order, or that all of the steps shown must be performed to achieve the desired result. Additionally or alternatively, certain steps can be omitted, multiple steps can be combined into a single step, and / or a single step can be broken up into multiple steps, etc.

[0170] The device embodiments of the present application are introduced below, which can be used to execute the false image determination method in the above-mentioned embodiments of the present application. Figure 12 The structural block diagram of the false image determination device provided by the embodiments of the present application is schematically shown. As shown in Figure 12 The false image determination device 1200 includes:

[0171] The feature extraction module 1210 is configured to perform feature extraction on the to-be-determined image to obtain the image feature of the to-be-determined image.

[0172] The feature matching module 1220 is configured to perform feature matching between the image feature of the to-be-determined image and the inventory feature of the historical image feature library to obtain a target inventory feature, wherein the inventory feature is the image feature of a historical image stored in the historical image feature library.

[0173] The additional information acquisition module 1230 is configured to acquire additional information associated with the target inventory feature, and the additional information includes an object associated with the target inventory feature.

[0174] The first false image determination module 1240 is configured to determine that the to-be-determined image is a false image when the total number of the objects associated with the target inventory feature is greater than a first preset number.

[0175] In some embodiments of the present application, based on the above embodiments, the feature matching module includes:

[0176] The feature matching unit is configured to perform feature matching between the image feature of the to-be-determined image and each inventory feature in the historical image feature library respectively to obtain a corresponding feature similarity.

[0177] The target inventory feature determination unit is configured to determine the inventory feature corresponding to the feature similarity as the target inventory feature when the feature similarity is greater than a preset threshold.

[0178] In some embodiments of the present application, based on the above embodiments, the feature matching unit includes:

[0179] The feature metric distance calculation sub-unit is configured to calculate the feature metric distance between the image feature of the to-be-determined image and each inventory feature of the historical image feature library.

[0180] The feature similarity calculation subunit is configured to calculate a corresponding feature similarity according to the feature metric distance, wherein the feature similarity is inversely proportional to the feature metric distance.

[0181] In some embodiments of the present application, based on the above embodiments, the feature extraction module comprises:

[0182] The image to be judged pre-processing unit is configured to pre-process the image to be judged, wherein the pre-processing comprises processing the image to obtain the image to be judged with a fixed image size and a fixed pixel value range.

[0183] The image feature acquisition unit is configured to input the pre-processed image to be judged into the pre-trained model to acquire the image feature of the image to be judged by the pre-trained model.

[0184] In some embodiments of the present application, based on the above embodiments, the false image judgment device further comprises:

[0185] The data set image pre-processing unit is configured to pre-process the images in the scene recognition image data set, wherein the scene recognition image data set comprises the images and the scene category labels corresponding to the images.

[0186] The model training unit is configured to train the neural network model for scene classification by using the scene recognition image data set, wherein the neural network model comprises a feature layer and a scene classification layer, the feature layer is used to extract features of the images to obtain image features with a preset dimension, and the scene classification layer is used to determine the scene categories corresponding to the images according to the image features of the images.

[0187] The scene classification layer removal unit is configured to remove the scene classification layer from the trained neural network model to obtain the pre-trained model.

[0188] In some embodiments of the present application, based on the above embodiments, the false image judgment device further comprises:

[0189] The historical image pre-processing unit is configured to pre-process the historical image, wherein the historical image is a historical received image, and the historical image is associated with additional information corresponding to the historical image.

[0190] The stock feature acquisition unit is configured to input the pre-processed historical image into the pre-trained model to acquire the stock feature corresponding to the historical image by the pre-trained model.

[0191] The historical image feature library establishment unit is configured to associate the stock feature corresponding to the historical image with the additional information corresponding to the historical image, and establish a historical image feature library according to the associated stock feature and the corresponding additional information.

[0192] In some embodiments of the present application, based on the above embodiments, the false image determination device further comprises:

[0193] The real image determination unit is configured to determine the image to be determined as a real image when the total number of objects associated with the target inventory feature is less than or equal to a first preset number.

[0194] In some embodiments of the present application, based on the above embodiments, the false image determination device further comprises:

[0195] The historical image feature library updating unit is configured to add the image features of the image to be determined and the additional information corresponding to the image to be determined into the historical image feature library.

[0196] In some embodiments of the present application, based on the above embodiments, the additional information acquisition module comprises:

[0197] The index information acquisition unit is configured to acquire index information associated with the target inventory feature;

[0198] The additional information reading unit is configured to read the additional information associated with the index information according to the index information to determine the object associated with the index information.

[0199] The total number of objects calculation unit is configured to merge the same objects when there are same objects among the objects associated with the index information, and to acquire the total number of objects associated with the index information after the same objects are merged.

[0200] In some embodiments of the present application, based on the above embodiments, the additional information further comprises an object state corresponding to the object, and the object state comprises a blacklist state and a normal state. The false image determination device further comprises:

[0201] The second false image determination unit is configured to determine the image to be determined as a false image when the proportion of the object state corresponding to the object associated with the target inventory feature is higher than a preset proportion.

[0202] In some embodiments of the present application, based on the above embodiments, the additional information further comprises a subject associated with the object. The false image determination device further comprises:

[0203] The third false image determination unit is configured to determine the image to be determined as a false image when the total number of subjects associated with the object associated with the target inventory feature is greater than a second preset number.

[0204] In some embodiments of the present application, based on the above embodiments, the image to be determined is a store image of an object to be audited.

[0205] The specific details of the false image determination apparatus provided in the embodiments of the present application have been described in detail in the corresponding method embodiments, and will not be repeated here.

[0206] Figure 13 A computer system structure block diagram of an electronic device for implementing the embodiments of the present application is schematically shown.

[0207] It should be noted that, Figure 13 The computer system 1300 of the electronic device shown is only an example, and should not impose any limitation on the functions and use range of the embodiments of the present application.

[0208] As Figure 13 shown, the computer system 1300 includes a central processing unit 1301 (CPU), which can perform various appropriate actions and processes according to programs stored in a read-only memory 1302 (ROM) or loaded from a storage portion 1308 into a random access memory 1303 (RAM). Various programs and data required for system operation are also stored in the random access memory 1303. The central processing unit 1301, the read-only memory 1302, and the random access memory 1303 are connected to each other through a bus 1304. An input / output interface 1305 (I / O interface) is also connected to the bus 1304.

[0209] The following components are connected to the input / output interface 1305: an input portion 1306 including a keyboard, a mouse, and the like; an output portion 1307 including a cathode ray tube (CRT), a liquid crystal display (LCD), and the like, and a speaker, and the like; a storage portion 1308 including a hard disk, and the like; and a communication portion 1309 including a network interface card such as a local area network card, a modem, and the like. The communication portion 1309 performs communication processing via a network such as the Internet. A drive 1310 is also connected to the input / output interface 1305 as necessary. A removable recording medium 1311 such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, and the like is attached to the drive 1310 as necessary, so that a computer program read therefrom is installed in the storage portion 1308 as necessary.

[0210] In particular, according to embodiments of the present application, the processes described in the various method flowcharts can be implemented as a computer software program. For example, embodiments of the present application include a computer program product comprising a computer program carried on a computer readable medium, the computer program comprising program code for executing the methods illustrated by the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via the communication section 1309, and / or installed from the removable medium 1311. When the computer program is executed by the central processing unit 1301, various functions defined in the system of the present application are executed.

[0211] It should be noted that the computer readable medium shown in the embodiments of the present application can be a computer readable signal medium or a computer readable storage medium or any combination of the two. The computer readable storage medium may, for example, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or apparatus, or any combination of the above. More specific examples of the computer readable storage medium can include, but are not limited to, an electrical connection having one or more wires, a portable computer diskette, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM), a flash memory, an optical fiber, a portable compact disk read-only memory (Compact Disc Read-Only Memory, CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In the present application, the computer readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, device or apparatus. In the present application, the computer readable signal medium can include a data signal carried in a baseband or as a part of a carrier wave, which carries computer readable program code. Such a propagated data signal can take many forms, including but not limited to an electromagnetic signal, an optical signal or any suitable combination of the above. The computer readable signal medium can also be any computer readable medium other than the computer readable storage medium, which can send, propagate or transmit the program for use by or in conjunction with an instruction execution system, device or apparatus. The program code contained on the computer readable medium can be transmitted by any suitable medium, including but not limited to wireless, wired, or the like, or any suitable combination of the above.

[0212] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in a block diagram or flowchart, and combinations of blocks in a block diagram or flowchart, may be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0213] It should be noted that although several modules or units for the device used to perform actions have been mentioned in the detailed description above, this division is not mandatory. In fact, according to the embodiments of this application, the features and functions of two or more modules or units described above can be embodied in one module or unit. Conversely, the features and functions of one module or unit described above can be further divided and embodied by multiple modules or units.

[0214] Through the above description of the embodiments, those skilled in the art will readily understand that the exemplary embodiments described herein can be implemented by software or by combining software with necessary hardware. Therefore, the technical solutions according to the embodiments of this application can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (such as a CD-ROM, USB flash drive, external hard drive, etc.) or on a network, including several instructions to cause a computing device (such as a personal computer, server, touch terminal, or network device, etc.) to execute the method according to the embodiments of this application.

[0215] Other embodiments of this application will readily occur to those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. This application is intended to cover any variations, uses, or adaptations of this application that follow the general principles of this application and include common knowledge or customary techniques in the art not disclosed herein.

[0216] It should be understood that this application is not limited to the precise structure described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of this application is limited only by the appended claims.

Claims

1. A method for identifying fake images, characterized in that, include: Feature extraction is performed on the image to be judged to obtain the image features of the image to be judged; The image features of the image to be judged are matched with the existing features in the historical image feature library to obtain the target existing features. The existing features are the image features of historical images stored in the historical image feature library. The historical images are images received by the platform in the past for object admission. Obtain additional information associated with the target stock characteristics, the additional information including objects associated with the target stock characteristics; When the total number of objects associated with the target stock feature is greater than a first preset number, the image to be judged is determined to be a fake image; When the total number of objects associated with the target stock feature is less than or equal to the first preset number, the image to be judged is determined to be a real image; The image features of the image to be judged and the corresponding additional information of the image to be judged are added to the historical image feature library; The acquisition of additional information associated with the target stock characteristics includes: Obtain the index information associated with the target stock characteristics; Based on the index information, additional information associated with the index information is read to determine the object associated with the index information; When identical objects exist among the objects associated with the index information, the identical objects are merged, and after merging the identical objects, the total number of objects associated with the index information is obtained.

2. The method for determining fake images according to claim 1, characterized in that, The step of matching the image features of the image to be determined with the existing features in the historical image feature database to obtain the target existing features includes: The image features of the image to be judged are matched with each existing feature in the historical image feature library to obtain the corresponding feature similarity. When the feature similarity is greater than a preset threshold, the existing feature corresponding to the feature similarity is determined as the target existing feature.

3. The method for determining fake images according to claim 2, characterized in that, The step of matching the image features of the image to be judged with each existing feature in the historical image feature database to obtain the corresponding feature similarity includes: Calculate the feature metric distance between the image features of the image to be judged and each existing feature in the historical image feature database; The corresponding feature similarity is calculated based on the feature metric distance, wherein the feature similarity is inversely proportional to the feature metric distance.

4. The method for determining fake images according to claim 1, characterized in that, The step of extracting features from the image to be judged to obtain the image features of the image to be judged includes: The image to be judged is preprocessed, wherein the preprocessing includes processing the image to obtain an image to be judged with a fixed image size and a fixed pixel value range; The preprocessed image to be judged is input into a pre-trained model to obtain the image features of the image to be judged through the pre-trained model.

5. The method for determining fake images according to claim 4, characterized in that, Before inputting the preprocessed image to be judged into the pre-trained model to obtain the image features of the image to be judged through the pre-trained model, the method further includes: The preprocessing is performed on the images in the scene recognition image dataset, wherein the scene recognition image dataset includes images and scene category labels corresponding to the images; The neural network model for scene classification is trained using the scene recognition image dataset. The neural network model includes a feature layer and a scene classification layer. The feature layer is used to extract image features of a preset dimension, and the scene classification layer is used to determine the scene category corresponding to the image based on the image features. The scene classification layer is removed from the trained neural network model to obtain the pre-trained model.

6. The method for determining fake images according to claim 5, characterized in that, Before performing feature matching between the image features of the image to be determined and the existing features in the historical image feature library to obtain the target existing features, the method further includes: The preprocessing is performed on historical images, wherein the historical images are historically received images and are associated with additional information corresponding to the historical images; The preprocessed historical image is input into the pre-trained model to obtain the existing features corresponding to the historical image through the pre-trained model; The existing features corresponding to the historical images are associated with the additional information corresponding to the historical images, and the historical image feature library is established based on the associated existing features and the corresponding additional information.

7. The method for determining fake images according to claim 1, characterized in that, The additional information also includes the object status corresponding to the object, which includes a blacklist status and a normal status; after obtaining the additional information associated with the target stock characteristics, the method further includes: When the proportion of objects whose status is blacklisted and associated with the target stock feature is higher than a preset proportion, the image to be judged is determined to be a fake image.

8. The method for determining fake images according to claim 1, characterized in that, The additional information also includes the subject associated with the object; after obtaining the additional information associated with the target stock characteristics, the method further includes: When the total number of subjects associated with the object associated with the target stock feature is greater than the second preset number, the image to be judged is determined to be a fake image.

9. The method for determining fake images according to any one of claims 1-8, characterized in that, The image to be judged is the store image of the object to be reviewed.

10. A device for determining fake images, characterized in that, include: The feature extraction module is configured to extract features from the image to be judged, thereby obtaining the image features of the image to be judged. The feature matching module is configured to perform feature matching between the image features of the image to be judged and the stored features of the historical image feature library to obtain the target stored features, wherein the stored features are the image features of historical images stored in the historical image feature library, and the historical images are images received by the platform in the past for object admission. The additional information acquisition module is configured to acquire additional information associated with the target stock feature, the additional information including objects associated with the target stock feature; The fake image determination module is configured to determine that the image to be determined is a fake image when the total number of objects associated with the target stock feature is greater than a first preset number. When the total number of objects associated with the target stock feature is less than or equal to the first preset number, the image to be judged is determined to be a real image; The image features of the image to be judged and the corresponding additional information of the image to be judged are added to the historical image feature library; The acquisition of additional information associated with the target stock characteristics includes: Obtain the index information associated with the target stock characteristics; Based on the index information, additional information associated with the index information is read to determine the object associated with the index information; When identical objects exist among the objects associated with the index information, the identical objects are merged, and after merging the identical objects, the total number of objects associated with the index information is obtained.

11. A computer-readable medium having a computer program stored thereon, which, when executed by a processor, implements the fake image determination method according to any one of claims 1 to 9.

12. An electronic device, characterized in that, include: processor; as well as Memory for storing the executable instructions of the processor; The processor is configured to execute the fake image determination method according to any one of claims 1 to 9 by executing the executable instructions.

13. A computer program product, characterized in that, The computer program product includes computer instructions stored in a computer-readable storage medium; the processor of the computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to perform the fake image determination method according to any one of claims 1 to 9.

Images