Image processing method and device, electronic equipment and computer readable storage medium

Abstract

The embodiment of the invention discloses an image processing method and device, electronic equipment and a computer readable storage medium. According to the embodiment of the invention, after a to-be-processed image is acquired and a processed image library corresponding to the to-be-processed image is determined, the to-be-processed image is compared with an intra-cluster historical image in a historical image cluster, and at least one target image cluster to which the to-be-processed image belongs is determined based on a comparison result and the historical image cluster; acquiring cluster security information of the target image cluster, and processing the intra-cluster image in the target image cluster according to the cluster security information; according to the scheme, the image processing efficiency can be improved on the basis of ensuring the image processing accuracy; the embodiment of the invention can be applied to various scenes such as artificial intelligence and computer vision.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and more specifically to an image processing method, apparatus, electronic device, and computer-readable storage medium. Background Art

[0002] Image management in application software involves multiple aspects, such as user avatars and transmitted images. Traditional image management solutions typically involve manually reviewing image content based on preliminarily determined harmful or potentially harmful criteria by machine algorithms to determine if there are any violations, and then processing the violating images. This approach requires a high investment of manpower and is relatively inefficient.

[0003] Currently, different applications have their own standards and governance solutions for image security operations. For some applications with a high number of daily active users (DAU), the corresponding image governance solutions require higher operating costs and human resource investment. Moreover, the traditional method of relying on manual review cannot significantly improve the efficiency of image review even with more resources and costs. Summary of the Invention

[0004] This application provides an image processing method, apparatus, electronic device, and computer-readable storage medium, which can improve the efficiency of image processing while ensuring the accuracy of image processing.

[0005] This application provides an image processing method, including:

[0006] Obtain the image to be processed and determine the processed image library corresponding to the image to be processed, wherein the processed image library includes at least one historical image cluster;

[0007] The image to be processed is compared with the historical images within the historical image cluster;

[0008] Based on the comparison results and the historical image clusters, at least one target image cluster to which the image to be processed belongs is determined;

[0009] Obtain cluster security information of the target image cluster, wherein the cluster security information indicates the risk category of the images in the target image cluster;

[0010] Based on the cluster security information, the intra-cluster images in the target image cluster are processed.

[0011] Accordingly, embodiments of this application provide an image processing apparatus, including:

[0012] An image acquisition unit is used to acquire an image to be processed and determine the processed image library corresponding to the image to be processed, wherein the processed image library includes at least one historical image cluster;

[0013] An image clustering unit is used to compare the image to be processed with the historical images within the historical image cluster;

[0014] An image localization unit is used to determine at least one target image cluster to which the image to be processed belongs, based on the comparison results and the historical image clusters.

[0015] An attribute determination unit is used to obtain cluster security information of the target image cluster, wherein the cluster security information indicates the risk category of the images in the target image cluster;

[0016] The image processing unit is used to process the intra-cluster images in the target image cluster according to the cluster security information.

[0017] In some embodiments, the image clustering unit may be used to extract features from the image to be processed to obtain a first image feature; extract features from historical images within the cluster to obtain a second image feature; and calculate the feature similarity between the first image feature and the second image feature.

[0018] In some embodiments, the image localization unit may be specifically used to, if the feature similarity is greater than or equal to a preset first threshold, take the historical image cluster corresponding to the historical image within the cluster as the target image cluster; if the feature similarity is less than the preset first threshold, construct the target image cluster to which the image to be processed belongs in the processed image library.

[0019] In some embodiments, the attribute determination unit may be specifically used to find the historical cluster security information of the target image cluster; if the target image cluster has the historical cluster security information, the historical cluster security information is used as the cluster security information of the target image cluster; if the target image cluster does not have the historical cluster security information, the risk category of the target image cluster is initialized to obtain the cluster security information of the target image cluster.

[0020] In some embodiments, the image processing apparatus further includes a risk identification unit, which can be used to acquire all historical images in the processed image library; cluster the historical images to obtain at least one historical image cluster; identify the risk category of the historical image cluster to obtain historical cluster security information of the historical image cluster.

[0021] In some embodiments, the risk identification unit may be specifically used to acquire a risk-marked image, search for the first image cluster to which the risk-marked image belongs from the historical image cluster; if the first image cluster exists, update the risk category of the first image cluster to the first category, and obtain the historical cluster security information of the first image cluster.

[0022] In some embodiments, the risk identification unit may be specifically used to compare the risk-marked image with the historical image if the first image cluster does not exist; if the comparison result shows that the similarity between the risk-marked image and the historical image is greater than or equal to a preset second threshold, the historical image cluster where the historical image is located is taken as the second image cluster to which the risk-marked image belongs; the risk category of the second image cluster is updated to the first category to obtain the historical cluster security information of the second image cluster.

[0023] In some embodiments, the risk identification unit may be specifically used to construct a third image cluster to which the risk-marked image belongs in the processed image library if the comparison result shows that the similarity between the risk-marked image and the historical image is less than the preset second threshold; set the risk category of the third image cluster to the first category, and obtain the historical cluster security information of the third image cluster.

[0024] In some embodiments, the risk identification unit may be specifically used to filter historical image clusters in the processed image library whose risk category is the first category to obtain a first remaining image cluster; to select historical image clusters in the first remaining image cluster whose number of historical images within the cluster is greater than a preset third threshold as second category candidate image clusters; to review the historical images within the clusters of the second category candidate image clusters, and to select the second category candidate image clusters that pass the review as second category image clusters; to update the risk category of the second category image clusters to the second category, thereby obtaining the historical cluster security information of the second category image clusters.

[0025] In some embodiments, the risk identification unit may be specifically used to obtain the application account corresponding to each cluster of historical images in the first remaining image cluster in the target application, and the historical interaction records of the application account; perform anomaly detection on the historical interaction records; if there is no abnormal behavior in the historical interaction records, take the first remaining image cluster corresponding to the historical interaction records as the second category image cluster; update the risk category of the second category image cluster to the second category, and obtain the historical cluster security information of the second category image cluster.

[0026] In some embodiments, the risk identification unit may be specifically used to filter historical image clusters in the processed image library that have risk categories of the first category and the second category to obtain a second remaining image cluster; and to update the risk category of the second remaining image cluster to a third category to obtain the historical cluster security information of the second remaining image cluster.

[0027] In some embodiments, the attribute determination unit may be specifically used to set the risk category of the target image cluster to a third category.

[0028] In some embodiments, the image processing unit may be specifically configured to: retain the target image cluster if the cluster security information indicates that the risk category of the target image cluster is a second category; obtain the target account corresponding to the image to be processed in the target application and send a prompt message to the target account if the cluster security information indicates that the risk category of the target image cluster is a first category; and identify the image to be processed and process the image to be processed according to the identification result if the cluster security information indicates that the risk category of the target image cluster is a third category.

[0029] In some embodiments, the image processing unit may be specifically used to retain the cluster security information of the target image cluster if the recognition result shows that the image to be processed is a safe image; and to update the risk category of the target image cluster to the first category if the recognition result shows that the image to be processed is a risky image.

[0030] In some embodiments, the image processing unit may be specifically configured to, if the cluster security information indicates that the risk category of the target image cluster includes a first category and a second category, obtain a first target image cluster corresponding to the first category and a second target image cluster corresponding to the second category from the target image cluster; use the intra-cluster images in the first target image cluster and the second target image cluster as re-examination images, cluster the re-examination images to obtain at least one redirected image cluster; perform a second review on the risk category of the re-examination images, and determine the cluster security information of the redirected image cluster based on the second review result.

[0031] Furthermore, embodiments of this application also provide an electronic device, including a processor and a memory, wherein the memory stores an application program, and the processor is used to run the application program in the memory to execute the image processing method provided in embodiments of this application.

[0032] Furthermore, embodiments of this application also provide a computer program product, including a computer program or instructions, which, when executed by a processor, implement the steps in the image processing method provided in embodiments of this application.

[0033] Furthermore, embodiments of this application also provide a computer-readable storage medium storing a plurality of instructions adapted for loading by a processor to execute steps in any of the image processing methods provided in embodiments of this application.

[0034] This application embodiment acquires an image to be processed and determines the processed image library corresponding to the image to be processed, wherein the processed image library includes at least one historical image cluster; then, the image to be processed is compared with the historical images within the historical image cluster; and based on the comparison result and the historical image cluster, at least one target image cluster to which the image to be processed belongs is determined; then, the cluster security information of the target image cluster is acquired, the cluster security information indicating the risk category of the images in the target image cluster; and then, the images within the target image cluster are processed according to the cluster security information. This scheme can compare the image to be processed with images in the processed image library in real time. Since there are already multiple historical image clusters in the processed image library, the target image cluster to which the image to be processed belongs can be determined based on the similarity between the image to be processed and the images in the processed image library during the comparison process. This allows the image to be processed to be associated with other images in the target image cluster. In a target image cluster, the images within each cluster have a high degree of similarity and influence each other, so they can be treated as a whole. Based on this, if the cluster security information of the target image cluster indicates that the target image cluster is at risk, the image to be processed can be processed directly. Alternatively, if the image to be processed is subsequently detected to be at risk, other clusters within the target image cluster to which the image to be processed belongs can be processed immediately. This achieves a processing effect from point to surface and from surface to point, improving the efficiency of image processing while ensuring the accuracy of image processing. Attached Figure Description

[0035] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0036] Figure 1 This is a schematic diagram illustrating an application scenario of the image processing method provided in the embodiments of this application;

[0037] Figure 2 This is a schematic flowchart of the image processing method provided in the embodiments of this application;

[0038] Figure 3 This is a schematic diagram illustrating the comparison between the image to be processed and the historical images within the historical image cluster provided in this application embodiment;

[0039] Figure 4A This is a schematic diagram illustrating one method for determining target image clusters provided in an embodiment of this application;

[0040] Figure 4B This is a schematic diagram illustrating another method for determining target image clusters provided in an embodiment of this application;

[0041] Figure 5 This is a schematic diagram of the process for risk identification of historical image clusters provided in an embodiment of this application;

[0042] Figure 6 This is a schematic diagram of the processing chain in the image processing method provided in the embodiments of this application;

[0043] Figure 7 This is a schematic diagram illustrating the processing of target image clusters provided in an embodiment of this application;

[0044] Figure 8 This is another schematic flowchart of the image processing method provided in the embodiments of this application;

[0045] Figure 9 This is a schematic diagram of the structure of the image processing apparatus provided in the embodiments of this application;

[0046] Figure 10 This is a schematic diagram of the structure of the electronic device provided in the embodiments of this application. Detailed Implementation

[0047] The technical solutions described below, with reference to the accompanying drawings, will be clearly and completely described. Obviously, the described embodiments are merely some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0048] This application provides an image processing method, apparatus, and computer-readable storage medium. The image processing apparatus can be integrated into an electronic device, which may be a server or a user terminal, etc.

[0049] The server can be a standalone physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing basic cloud computing services such as cloud services, cloud-preset databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content delivery network (CDN) acceleration services, and big data and artificial intelligence platforms. The terminal can be a smartphone, tablet, laptop, desktop computer, smart speaker, smartwatch, etc., but is not limited to these. The terminal and server can be directly or indirectly connected via wired or wireless communication, which is not limited herein.

[0050] Figure 1 A schematic diagram illustrating an application scenario of the image processing method provided in an embodiment of this application is shown. For example... Figure 1 As shown, taking the integration of an image processing device into an electronic device as an example, the electronic device can acquire an image to be processed and determine the processed image library corresponding to the image to be processed, wherein the processed image library includes at least one historical image cluster; then, the image to be processed is compared with the historical images within the historical image cluster; and based on the comparison result and the historical image cluster, at least one target image cluster to which the image to be processed belongs is determined; then, the cluster security information of the target image cluster is acquired, the cluster security information indicating the risk category of the images in the target image cluster; and then, the images within the target image cluster are processed according to the cluster security information.

[0051] The image processing method provided in this application relates to Computer Vision (CV) technology within Artificial Intelligence (AI). This application embodiment can acquire an image to be processed and determine the corresponding processed image library; then, compare the image to be processed with historical images within a historical image cluster; and based on the comparison result and the historical image cluster, determine at least one target image cluster to which the image to be processed belongs.

[0052] Artificial intelligence (AI) encompasses the theories, methods, technologies, and application systems that utilize digital computers or computers-controlled machines to simulate, extend, and expand human intelligence, enabling them to perceive the environment, acquire knowledge, and use that knowledge to achieve optimal results. In other words, AI is a comprehensive technology within computer science that attempts to understand the essence of intelligence and produce new intelligent machines capable of reacting in a manner similar to human intelligence. AI studies the design principles and implementation methods of various intelligent machines, enabling them to possess perception, reasoning, and decision-making capabilities.

[0053] Artificial intelligence (AI) is a comprehensive discipline encompassing a wide range of fields, including both hardware and software technologies. Fundamental AI technologies generally include sensors, dedicated AI chips, cloud computing, distributed storage, big data processing, pre-trained model technology, operating / interactive systems, and mechatronics. Pre-trained models, also known as large-scale models or foundational models, can be widely applied to downstream tasks across various AI fields after fine-tuning. AI software technologies primarily include computer vision, speech processing, natural language processing, and machine learning / deep learning.

[0054] Computer vision is the science that studies how to enable machines to "see." More specifically, it refers to machine vision, which uses cameras and computers to replace human eyes in recognizing, detecting, and measuring targets, and further processes images to create images more suitable for human observation or transmission to instruments. As a scientific discipline, computer vision studies related theories and technologies, attempting to build artificial intelligence systems capable of extracting information from images or multidimensional data. Large model technology has brought significant changes to the development of computer vision technology. Pre-trained models in the vision field, such as Swin-transformer, ViT, V-MOE, and MAE, can be quickly and widely applied to downstream tasks after fine-tuning. Computer vision technology typically includes image processing, image recognition, image semantic understanding, image retrieval, OCR, video processing, video semantic understanding, video content / behavior recognition, 3D object reconstruction, 3D technology, virtual reality, augmented reality, simultaneous localization and mapping (SLAM), and common biometric recognition technologies such as facial recognition and fingerprint recognition.

[0055] It is understood that, in the specific implementation of this application, data related to the image to be processed, historical images within the cluster, images within the cluster, risk-marked images, etc., are involved. When the following embodiments of this application are applied to specific products or technologies, permission or consent is required, and the collection, use and processing of related data must comply with the relevant laws, regulations and standards of the relevant countries and regions.

[0056] The following sections provide detailed descriptions of each example. It should be noted that the order in which the embodiments are described is not intended to limit the preferred order of the embodiments.

[0057] This embodiment will be described from the perspective of an image processing device, which can be integrated into an electronic device, such as a server or a terminal. The terminal can include tablet computers, laptops, personal computers (PCs), wearable devices, virtual reality devices, or other smart devices that can generate image files.

[0058] An image processing method includes: acquiring an image to be processed and determining a processed image library corresponding to the image to be processed, the processed image library including at least one historical image cluster; comparing the image to be processed with historical images within the historical image cluster; determining at least one target image cluster to which the image to be processed belongs based on the comparison result and the historical image cluster; acquiring cluster security information of the target image cluster, the cluster security information indicating the risk category of images in the target image cluster; and processing the images within the target image cluster according to the cluster security information.

[0059] Figure 2 A schematic flowchart of the image processing method provided in an embodiment of this application is shown. Figure 2 As shown, the specific process of this image processing method is as follows:

[0060] 101. Obtain the image to be processed and determine the processed image library corresponding to the image to be processed. The processed image library includes at least one historical image cluster.

[0061] The image to be processed can be a user's avatar in various applications. For example, in interactive applications, avatars are a fundamental business element, and almost every interactive application has its own avatar data, which is displayed during information interaction between users (such as chatting, commenting, posting on social media, publishing articles, etc.). The image to be processed can also be image content that users share through taking photos, uploading, transmitting, or other means while using various applications as information dissemination tools.

[0062] For ease of description, this application embodiment will use a user avatar in an application software as an example to illustrate the process. In application software, users can set their own avatars for normal use, and can also modify their avatars after setting them. For example, in chat software, users can set their own account avatars, which can reflect the user's image to a certain extent. Chat software generally has a very large user base, and some users may use inappropriate avatars to achieve various improper purposes. To create a healthy online interactive environment, it is necessary to review user avatars and take appropriate measures to deal with identified inappropriate avatars.

[0063] An image to be processed can be understood as a user's avatar in an application that is about to be reviewed; that is, the image has not yet been reviewed. For any user of the application, when that user initially uploads their avatar, that initial avatar can be used as the image to be processed. If the user's avatar has been used for some time, regardless of whether the current avatar has been reviewed, if the user replaces the current avatar with a new one, the new avatar can also be used as the image to be processed. Furthermore, if the user uses a previously used avatar as their latest avatar, this previously used avatar (historical avatar) may also be used as the image to be processed. In this case, if this historical avatar has already been reviewed and marked, the electronic device can directly recognize the relevant information of this historical image, and there is no need for it to be used as the image to be processed again; if this historical avatar has not yet been reviewed, it can be used as the image to be processed.

[0064] After acquiring the image to be processed, the dataset of user avatars in the application software that are located before the acquisition time of the image can be used as the boundary to form a processed image library. The processed image library includes multiple historical image clusters, and each historical image cluster includes at least one historical image within the cluster.

[0065] In this embodiment, at any point in time before acquiring the image to be processed, the existing user avatars in the application software can be clustered to divide the user avatars into multiple image subsets, each of which includes at least one user avatar. The historical image clusters in the processed image library are the image subsets, and the historical images within each historical image cluster are the user avatars within those subsets. Accordingly, the processed image library includes at least one historical image cluster, and each historical image cluster includes at least one historical image within that cluster.

[0066] 102. Compare the image to be processed with the historical images within the historical image cluster.

[0067] The process of comparing the image to be processed with the historical images within each historical image cluster can be understood as comparing the similarity of the image to be processed with the historical images within each historical image cluster in the processed image library, and calculating the similarity between the image to be processed and each historical image cluster. In the process of comparing the image to be processed with the historical images within each historical image cluster, a similarity clustering algorithm can be used to calculate the similarity between the image to be processed and each historical image cluster, or other similarity calculation methods can be used. This application embodiment does not impose any limitations on this approach.

[0068] Figure 3This illustration shows a diagram comparing the image to be processed with historical images within a historical image cluster, as provided in an embodiment of this application. Figure 3 As shown, the processed image library includes multiple historical image clusters, namely historical image cluster 1, historical image cluster 2, historical image cluster 3, historical image cluster 4, historical image cluster 5, historical image cluster 7, ..., historical image cluster n (n is a natural number greater than 7). At this point, the image to be processed can be compared with all historical images within historical image cluster 1 to obtain the similarity value between the image to be processed and historical image cluster 1; the image to be processed can be compared with all historical images within historical image cluster 2 to obtain the similarity value between the image to be processed and historical image cluster 2; and so on, the image to be processed can be compared with all historical images within other historical image clusters (including each historical image cluster from historical image cluster 3 to historical image cluster n) to obtain the corresponding similarity value.

[0069] The process of comparing the image to be processed with historical images within the same historical image cluster can include: extracting features from the image to be processed to obtain first image features; extracting features from the historical images within the cluster to obtain second image features; and calculating the feature similarity between the first image features and the second image features.

[0070] The following example illustrates the process of comparing the image to be processed with a historical image cluster (e.g., historical image cluster 1) in a processed image library. Electronic devices can use various neural network models (such as convolutional neural networks) to extract features from the image to be processed and the historical images within the cluster, obtaining their respective image features. The extracted image features can include geometric features, shape features, amplitude features, color features, etc.

[0071] Since a historical image cluster includes at least one intra-cluster historical image, the second image feature obtained by feature extraction from the intra-cluster historical image can be understood as a set of image features corresponding to each of the multiple intra-cluster historical images.

[0072] Furthermore, as mentioned earlier, since the user avatars already existing in the application software have been clustered at any point in time before the image to be processed is acquired, it means that every image in the processed image library (the historical images in each historical image cluster) has been clustered. Clustering generally includes a feature extraction step, so the image features of the historical images within the cluster can be directly obtained and used as the second image features.

[0073] After determining the first image feature corresponding to the image to be processed and the second image feature corresponding to each historical image cluster, the feature similarity between the first image feature and the second image feature can be calculated.

[0074] 103. Based on the comparison results and historical image clusters, determine at least one target image cluster to which the image to be processed belongs.

[0075] A target image cluster can be understood as the set of images containing the image to be processed. This target image cluster may be one or more historical image clusters in the processed image library, or it may be a newly created image cluster specifically for the image to be processed within the processed image library. The determination of the target image cluster depends on the comparison results between the image to be processed calculated in step 102 and the historical images within the historical image clusters. Furthermore, after determining the target image cluster, the image to be processed can be added to the target image cluster.

[0076] In this embodiment of the application, determining at least one target image cluster to which the image to be processed belongs based on comparison results and historical image clusters can include the following two cases:

[0077] The first case: If the feature similarity is greater than or equal to the preset first threshold, the historical image cluster corresponding to the historical image within the cluster is taken as the target image cluster;

[0078] The second scenario: If the feature similarity is less than the preset first threshold, construct the target image cluster to which the image to be processed belongs in the processed image library.

[0079] Figure 4A This illustration shows a schematic diagram of one method for determining target image clusters provided in an embodiment of this application. For example... Figure 4A As shown, for the first case: if, after comparison, it is determined that the feature similarity between the image to be processed and the historical images within the historical image cluster is greater than or equal to a preset first threshold, then the historical image cluster can be used as the target image cluster corresponding to the image to be processed. After determining the target image cluster corresponding to the image to be processed, the image to be processed can be added to the target image cluster.

[0080] For example, after comparison, if it is determined that the feature similarity between the image to be processed and the historical images within the historical image cluster 1 is greater than or equal to a preset first threshold, and the feature similarity between the image to be processed and the historical images within the historical image cluster 7 is also greater than or equal to a preset first threshold, then historical image cluster 1 and historical image cluster 7 can be used as target image clusters at the same time, that is, two target image clusters are obtained, and the image to be processed is added to historical image cluster 1 and historical image cluster 7 respectively.

[0081] It should be understood that Figure 4AThis is merely an example of how to determine a target image cluster. In real-world scenarios, the feature similarity between historical images within other historical image clusters and the image to be processed might also meet a preset first threshold. For example, the feature similarity between historical images within historical image cluster 2 and the image to be processed might be greater than or equal to the preset first threshold. In this case, historical image cluster 2 can be used as the target image cluster, and the image to be processed can be added to historical image cluster 2 accordingly. Other possible scenarios will not be listed in this embodiment.

[0082] Figure 4B This diagram illustrates another method for determining target image clusters provided in an embodiment of this application. For example... Figure 4B As shown, for the second case: if, after comparison, it is determined that the feature similarity between the image to be processed and the historical images within the historical image cluster is less than the preset first threshold, it indicates that there is no historical image cluster with high similarity to the image to be processed in the processed image library. At this time, a new image cluster can be created directly in the processed image library, and the image to be processed can be added to the new image cluster. The new image cluster is then used as the target image cluster to which the image to be processed belongs.

[0083] For example, a newly created image cluster is identified as "New Image Cluster 1," which is the target image cluster to which the image to be processed belongs. At this point, New Image Cluster 1 contains only one image from the image cluster to be processed. As time progresses, for subsequent new images to be processed, their corresponding processed image library includes New Image Cluster 1, and New Image Cluster 1 will then become a historical image cluster in the processed image library. If the similarity between a subsequent new image to be processed and the image to be processed in New Image Cluster 1 meets a preset condition, the subsequent new image to be processed can be added to New Image Cluster 1, thereby expanding the image size of New Image Cluster 1.

[0084] It should be understood that the specific value of the preset first threshold can be set according to the actual situation, and the embodiments of this application do not impose specific limitations here. Generally, the higher the preset first threshold, the higher the required feature similarity, and the fewer historical image clusters that the image to be processed may match in the processed image library.

[0085] 104. Obtain the cluster security information of the target image cluster. The cluster security information indicates the risk category of the images in the target image cluster.

[0086] As mentioned earlier, user avatars already existing in the application software can be clustered at any point in time before the image to be processed is acquired. Furthermore, after obtaining multiple historical image clusters through clustering, the risk category of these historical image clusters can be determined, thereby obtaining the historical cluster security information for each historical image cluster. Since the target image cluster to which the image to be processed belongs can originate from historical image clusters in the processed image library, after determining the target image cluster to which the image to be processed belongs, the cluster security information of the target image cluster can be determined based on the information of the target image cluster in the processed image library.

[0087] In this embodiment of the application, obtaining the cluster security information of the target image cluster may include: searching for the historical cluster security information of the target image cluster; if the target image cluster has historical cluster security information, using the historical cluster security information as the cluster security information of the target image cluster; if the target image cluster does not have historical cluster security information, initializing the risk category of the target image cluster to obtain the cluster security information of the target image cluster.

[0088] As mentioned earlier, the target image cluster may be one or more historical image clusters in the processed image library, or it may be a newly created image cluster in the processed image library specifically for the image to be processed. The cluster security information obtained for the target image cluster also differs for these two different scenarios.

[0089] For the first scenario: If the target image cluster originates from a historical image cluster in the processed image library, then the target image cluster possesses historical cluster security information. In this case, the historical cluster security information can be directly used as the cluster security information of the target image cluster. For example, if step 103 determines that the target image cluster to which the image to be processed belongs is historical image cluster 1 (referred to as target image cluster A for ease of distinction) and historical image cluster 7 (referred to as target image cluster B for ease of distinction), then the historical cluster security information of historical image cluster 1 can be used as the cluster security information of target image cluster A, and the historical cluster security information of historical image cluster 7 can be used as the cluster security information of target image cluster B.

[0090] In the second scenario: if the target image cluster is a newly created image cluster, there is not yet time to identify it. In this case, the target image cluster does not have historical cluster security information, and the risk category of the target image cluster can be directly initialized to obtain the cluster security information of the target image cluster. For example, if step 103 determines that the feature similarity between the image to be processed and the historical images within the historical image cluster is less than a preset first threshold, a new image cluster (new image cluster 1) is directly created in the processed image library as the target image cluster to which the image to be processed belongs (referred to as target image cluster C for easy distinction). At this time, the risk category of target image cluster C can be initialized to obtain the cluster security information of target image cluster C.

[0091] Initializing the risk category of the target image cluster may include setting the risk category of the target image cluster to the third category.

[0092] In this embodiment, the risk categories in cluster security information can be divided into three types: a first category, a second category, and a third category. The first category can be understood as a blacklist category, the second category as a whitelist category, and the third category as a general category that is neither a whitelist nor a blacklist category. Since both whitelist and blacklist categories require strict review, in the second case described above, where the feature similarity is less than a preset first threshold, and a target image cluster to which the image to be processed belongs is constructed in the processed image library, the risk category for this target image cluster can be directly set to the third category during initial risk category settings.

[0093] In this embodiment of the application, before step 104, the method may further include: acquiring all historical images in the processed image library; clustering the historical images to obtain at least one historical image cluster; identifying the risk category of the historical image cluster to obtain the historical cluster security information of the historical image cluster.

[0094] As mentioned earlier, the processed image library includes at least one historical image cluster, and each historical image cluster includes at least one intra-cluster historical image. Historical image clusters are obtained by clustering all historical images in the processed image library. During the clustering process, each historical image can be assigned to only one historical image cluster, or it can be assigned to two or more historical image clusters simultaneously. For example, taking one historical image in the processed image library (named historical image 1 for ease of explanation), after clustering, the processed image library includes historical image cluster 1, historical image cluster 2, historical image cluster 3, historical image cluster 4, historical image cluster 5, historical image cluster 7, ..., historical image cluster n. In this case, historical image 1 can be an intra-cluster historical image of one of these historical image clusters; for example, historical image 1 is located in historical image cluster 5. Furthermore, historical image 1 can also exist in two or more historical image clusters simultaneously; for example, historical image cluster 1 includes historical image 1, and historical image cluster 2 also includes historical image 1.

[0095] It should be noted that the number of historical image clusters that a historical image can exist in can be set according to the actual situation. For example, it can be set that each historical image can only be assigned to one historical image cluster when clustering historical images; or, it can be set that each historical image can be assigned to two or more historical image clusters when clustering historical images, provided that a preset similarity threshold is met. Even in this case, in the actual clustering process, a historical image may only be assigned to one historical image cluster due to similarity.

[0096] After clustering is completed, the risk category of each historical image cluster can be identified, and the corresponding historical cluster security information can be obtained. The risk category of historical image clusters can be identified based on the historical images within the cluster, or based on other information that is related to the historical images within the cluster.

[0097] Figure 5 This illustration shows a flowchart of risk identification for historical image clusters provided in an embodiment of this application. Figure 5 As shown, identifying the risk category of historical image clusters to obtain historical cluster security information can include: acquiring risk-marked images, searching for the first image cluster to which the risk-marked images belong from historical image clusters; if the first image cluster exists, updating the risk category of the first image cluster to the first category, and obtaining the historical cluster security information of the first image cluster.

[0098] Risk-marked images are highly risky pictures identified through various means (such as reports, image recognition models, and manual inspections). Risk-marked images can cover the most basic and common types of violations in the security industry, including but not limited to pornography, gambling, gore, and vulgarity, and can reflect various dimensions such as faces, objects, actions, and postures. In practical applications, risk-marked images can be categorized into information security risk-marked images, business security risk-marked images, or other types. Information security risk-marked images can include sensitive information related to leadership or other information that violates laws and regulations; business security risk-marked images can include pornography, vulgar content, gore, violence, malicious traffic redirection, or other types of information.

[0099] For the acquired risk-marked image, we can first determine whether the risk-marked image exists in the historical image cluster. If it does, the historical image library containing the risk-marked image can be designated as the first image cluster, and the risk category of the first image cluster can be updated to the first category, thus obtaining the historical cluster security information of the first image cluster. Before updating the risk category of the first image cluster to the first category, the risk category of the first image cluster may be one of the first, second, or third categories. In this case, regardless of the specific risk category of the first image cluster, it is updated to the first category, i.e., the blacklist category. After updating the risk category of the first image cluster to the first category, it means that all historical images within the first image cluster have the same level of risk as the risk-marked image.

[0100] After finding the first image cluster to which the risk-marked image belongs from the historical image clusters, the process may further include: if the first image cluster does not exist, comparing the risk-marked image with the historical images; if the comparison result shows that the similarity between the risk-marked image and the historical images is greater than or equal to a preset second threshold, taking the historical image cluster to which the historical image belongs as the second image cluster to which the risk-marked image belongs; updating the risk category of the second image cluster to the first category, and obtaining the historical cluster safety information of the second image cluster.

[0101] In determining whether a risk-marked image exists in a historical image cluster, if the result shows that the risk-marked image does not exist in the historical image cluster, a similarity comparison can be performed between the risk-marked image and the historical images. Based on the similarity comparison result, it can be determined whether there exists a historical image cluster with a preset second threshold that is similar to the risk-marked image. In this step, if the similarity comparison result shows that the similarity between the risk-marked image and the intra-cluster historical images in one or more historical image clusters is greater than or equal to the second preset similarity, the corresponding historical image cluster can be designated as the second image cluster (i.e., the risk-marked image is added to the corresponding historical image cluster, and the historical image cluster with the added risk-marked image is the second image cluster), and the risk category of the second image cluster is updated to the first category, thereby obtaining the historical cluster safety information of the second image cluster. For example, if the similarity between the risk-marked image and the intra-cluster historical images in historical image cluster 1 is greater than or equal to the preset second threshold, the risk category of historical image cluster 1 can be updated to the first category. Alternatively, if the similarity between the risk-marked image and the intra-cluster historical images in historical image cluster 1 is greater than or equal to a preset second threshold, and the similarity between the risk-marked image and the intra-cluster historical images in historical image cluster 7 is also greater than or equal to a preset second threshold, then the risk category of historical image cluster 1 can be updated to the first category, and at the same time, the risk category of historical image cluster 7 can be updated to the first category.

[0102] Before updating the risk category of the second image cluster to the first category, the risk category of the second image cluster may be one of the first, second, or third categories. In this case, regardless of the specific risk category of the second image cluster, it is updated to the first category, i.e., the blacklist category. After updating the risk category of the second image cluster to the first category, it means that all historical images within the second image cluster have the same level of risk as the risk-marked images.

[0103] It should be understood that the specific value of the preset second threshold can be set according to the actual situation. The preset second threshold and the preset first threshold can be the same or different. This application embodiment does not impose specific restrictions here.

[0104] After comparing the risk-marked image with historical images, the process may further include: if the comparison result shows that the similarity between the risk-marked image and the historical image is less than a preset second threshold, constructing a third image cluster to which the risk-marked image belongs in the processed image library; setting the risk category of the third image cluster to the first category, and obtaining the historical cluster security information of the third image cluster.

[0105] During the similarity comparison between the risk-marked image and historical images, if the comparison result shows that the similarity between the risk-marked image and historical images is less than a preset second threshold (i.e., there is no historical image cluster with a preset similarity to the risk-marked image in the processed image library), then a new image cluster can be created directly in the processed image library, and the risk-marked image can be added to the new image cluster. For easy differentiation, the newly created image cluster used to store the risk-marked image is called the third image cluster. For the newly created third image cluster, the risk category of the third image cluster can be directly set to the first category.

[0106] It should be noted that if multiple risk-marked images are processed simultaneously, and the processed image library does not contain a first image cluster or a second image cluster that meets the conditions, these risk-marked images can be clustered, and the number of new image clusters to be created can be determined based on the clustering results. For example, if there are 5 risk-marked images, namely risk-marked image 1, risk-marked image 2, risk-marked image 3, risk-marked image 4, and risk-marked image 5, clustering these 5 risk-marked images will result in three sets. The first set includes risk-marked image 1 and risk-marked image 3, the second set includes risk-marked image 2 and risk-marked image 5, and the third set includes risk-marked image 4. In this case, a new image cluster needs to be created for each set, resulting in 3 new third image clusters: third image cluster 1, third image cluster 2, and third image cluster 3. Correspondingly, third image cluster 1 includes risk-marked image 1 and risk-marked image 3, third image cluster 2 includes risk-marked image 2 and risk-marked image 5, and third image cluster 3 includes risk-marked image 4. At the same time, the risk categories of the third image cluster 1, the third image cluster 2, and the third image cluster 3 can all be set to the first category, thereby obtaining their respective historical cluster security information.

[0107] It should be noted that for historical image clusters with a risk level of Category 1, the images within the cluster (including risk-marked images) can be processed. For example, they can be marked as risk images, and risk warnings can be issued to the user accounts corresponding to the images to remind the relevant users to change their profile pictures. Alternatively, other methods can be used to process the images in historical image clusters with a risk level of Category 1, and this application does not impose any restrictions on this.

[0108] After determining the historical image clusters corresponding to the first category, the historical image clusters corresponding to the second category can be determined. Therefore, in this embodiment, identifying the risk category of historical image clusters to obtain historical cluster security information may further include: filtering historical image clusters in the processed image library whose risk category is the first category to obtain a first remaining image cluster; selecting historical image clusters in the first remaining image clusters whose number of intra-cluster historical images is greater than a preset third threshold as candidate image clusters for the second category; reviewing the intra-cluster historical images in the candidate image clusters for the second category, and selecting the candidate image clusters for the second category that pass the review as the second category image clusters; updating the risk category of the second category image clusters to the second category to obtain the historical cluster security information of the second category image clusters.

[0109] The first remaining image cluster can be understood as the historical image clusters in the processed image library excluding those in the first risk category. Since the second category can be understood as a whitelist category, historical images within a cluster of historical images in the second risk category are required to have extremely high security. Consequently, the workload for reviewing the second category is significantly greater. In practical applications, especially for products with a large user base, reviewing every image within the first remaining image cluster for the second category would be extremely resource-intensive. Considering that clustering can group commonly used and highly similar avatars together, resulting in some historical image clusters having a significantly larger number of avatars than others, reviewing these large-scale, frequently used avatars for the second category is essential. If one or more historical images within such a cluster do not meet the review requirements for the second category, the corresponding historical image cluster can be determined to have a certain level of risk and cannot be classified as the second category.

[0110] In this embodiment, historical image clusters with a high number of historical images within a cluster can be directly submitted for manual review. Specifically, the cluster can be filtered based on the number of historical images within it. Historical image clusters with a number of historical images within a cluster greater than a preset third threshold (which can be set according to actual conditions) are designated as second-category candidate image clusters and submitted for manual review. If all historical images within a second-category candidate image cluster pass the review, the risk category of the second-category candidate image cluster can be updated to the second category. Conversely, if one or more historical images within a second-category candidate image cluster fail the review, the risk category of the second-category candidate image cluster can be determined as either the first or third category based on the specific circumstances of the historical images that failed the review.

[0111] The process, after filtering historical image clusters in the processed image library that belong to the first risk category to obtain the first remaining image cluster, may further include: obtaining the application account corresponding to each cluster of historical images in the target application, and the historical interaction records of the application account; performing anomaly detection on the historical interaction records; if there is no abnormal behavior in the historical interaction records, taking the first remaining image cluster corresponding to the historical interaction records as the second category image cluster; updating the risk category of the second category image cluster to the second category, and obtaining the historical cluster security information of the second category image cluster.

[0112] The target application can be any application software in a real-world application scenario. Users can register an account (application account) in the target application and upload an account avatar (historical image). Therefore, there is a correlation between the application account and the account avatar, and they influence each other. In determining whether the risk category of the first remaining image cluster is the second category, this application embodiment can consider the historical interaction records of the application account corresponding to the historical images in each cluster of the first remaining image cluster.

[0113] Historical interaction records can include whether the application account has been penalized by the target application platform for violations during its use, and whether it has been reported or complained about by other users. Since users can change their account avatars at any time, the historical interaction records can also include whether any previously changed avatars of the application account have used prohibited avatars, and so on. Anomaly detection is performed on the historical interaction records exemplified above. The detection results can reflect the risk level of the historical images within the corresponding cluster, thereby determining the risk category of the first remaining image cluster corresponding to the historical images within the cluster. For example, if all the application accounts corresponding to the historical images within a certain first remaining image cluster have no abnormal behavior in their historical interaction records, the risk category of that first remaining image cluster can be updated to the second category, and that first remaining image cluster becomes the second category image cluster.

[0114] Before updating the risk category of the first remaining image cluster to the second category, the risk category of the first remaining image cluster could be either the second or third category. In this case, regardless of the specific risk category of the first remaining image cluster, it is updated to the second category, i.e., the whitelist category. After updating the risk category of the first remaining image cluster to the second category, it means that all historical images within that cluster are whitelisted images, thus possessing a higher level of security.

[0115] It should be noted that for second-category candidate image clusters where the number of historical images within a cluster exceeds a preset third threshold, the review process for these clusters can be conducted entirely through manual review. In this case, the review scope corresponding to the historical interaction record anomaly detection method is the historical image clusters within the first remaining image clusters, excluding the second-category candidate image clusters.

[0116] In addition, during the review of candidate image clusters in the second category, a combination of manual review and historical interaction record anomaly detection can be used for judgment. If the historical images within a cluster of candidate images in the second category have passed manual review and there are no anomalies in the historical interaction records, the risk category of the candidate image cluster in the second category can be updated to the second category. If the historical images within a cluster of candidate images in the second category have only passed manual review but there are anomalies in the historical interaction records, the risk category of the candidate image cluster in the second category cannot be updated to the second category. If the historical images within a cluster of candidate images in the second category have not passed manual review, the risk category of the candidate image cluster in the second category cannot be updated to the second category regardless of whether there are anomalies in its historical interaction records.

[0117] After identifying the historical image clusters corresponding to the first category and the second category, the remaining historical image clusters are the historical image clusters corresponding to the third category. Accordingly, identifying the risk category of the historical image clusters to obtain their cluster security information may further include: filtering the historical image clusters in the processed image library that have risk categories one and two to obtain a second remaining image cluster; updating the risk category of the second remaining image clusters to the third category to obtain the historical cluster security information of the second remaining image clusters.

[0118] The second remaining image cluster can be understood as the historical image clusters in the processed image library other than those with risk categories one and two. Before updating the risk category of the second remaining image cluster to category three, the risk category of the second remaining image cluster may be category two or category three. At this time, regardless of the specific risk category of the second remaining image cluster, it is updated to category three, i.e., the ordinary category.

[0119] It should be understood that the above steps “obtain all historical images in the processed image library; cluster the historical images to obtain at least one historical image cluster; identify the risk category of the historical image cluster to obtain the historical cluster security information of the historical image cluster” and their sub-steps can also be located before steps 103, 102, or 101. Figure 6 A schematic diagram of the processing chain in the image processing method provided in the embodiments of this application is shown. For example... Figure 6 As shown, in steps 101-105 of this application, the image processing process with the image to be processed as the core can be understood as a real-time processing link. However, the above steps "obtain all historical images in the processed image library; cluster the historical images to obtain at least one historical image cluster; identify the risk category of the historical image cluster to obtain the historical cluster security information of the historical image cluster" and their sub-steps can be understood as an asynchronous link with a processing timeline different from the real-time link.

[0120] 105. Based on the cluster security information, process the intra-cluster images in the target image cluster.

[0121] After obtaining the cluster security information of each target image cluster (i.e., determining the risk category of each target image cluster), the intra-cluster images within the target image cluster can be processed based on the cluster security information. In this embodiment, processing the intra-cluster images within the target image cluster based on the cluster security information can include the following methods:

[0122] Method 1: If the cluster security information indicates that the risk category of the target image cluster is Category 2, retain the target image cluster;

[0123] Method 2: If the cluster security information indicates that the risk category of the target image cluster is Category 1, obtain the target account corresponding to the image to be processed in the target application and send a prompt message to the target account;

[0124] Method 3: If the cluster security information indicates that the risk category of the target image cluster is Category 3, the image to be processed is identified, and the image to be processed is processed according to the identification result.

[0125] Figure 7 This illustration shows a schematic diagram of processing a target image cluster according to an embodiment of this application. For example... Figure 7 As shown, different risk categories of target image clusters require different processing methods.

[0126] In Method 1, if the cluster security information indicates that the risk category of the target image cluster is Category 2, it means that the target image cluster is a whitelisted cluster. Images in the whitelisted cluster have a high level of security and can be retained.

[0127] In Method Two, if the cluster security information indicates that the risk category of the target image cluster is Category 1, it means that the target image cluster is a blacklisted cluster, and the images within the cluster pose a security risk. Since the image to be processed is a member of the target image cluster, the risk handling process for the image to be processed can be triggered. For example, the target account corresponding to the image to be processed in the target application can be obtained, and a prompt message can be sent to the target account to inform the user that the current profile picture has not passed the review and needs to be changed to another profile picture.

[0128] In Method 3, if the cluster security information indicates that the risk category of the target image cluster is Category 3, it means that the target image cluster is a normal cluster. In this case, the image to be processed can be identified separately, and then processed according to the identification result. There are various ways to identify the image to be processed. For example, an image recognition model can be used. Specifically, the image recognition model can extract features from various dimensions of the image to be processed (such as pornography, gambling, gore, vulgarity, etc., based on dimensions such as faces, objects, actions, and postures), and predict the risk level of the image to be processed based on the extracted features, thus obtaining the identification result of the image to be processed. Afterwards, the image to be processed according to the identification result can include the following two cases:

[0129] (1) If the identification result shows that the image to be processed is a safe image, the cluster safety information of the target image cluster is retained. If the image to be processed is a safe image, considering that there may be other historical images of other clusters in the target image cluster to which it belongs, and the safety of other historical images of other clusters has not yet been identified, the cluster safety information of the target image cluster can be retained, that is, the risk category of the target image cluster to which the image to be processed belongs is kept as the third category.

[0130] (2) If the recognition result shows that the image to be processed is a risky image, the risk category of the target image cluster is updated to the first category. If the image to be processed is a risky image, it means that the image to be processed is risky. Since the target image cluster to which the image to be processed belongs is determined after similarity comparison, it means that the historical images of the cluster in the target image cluster have a high similarity with the image to be processed. Therefore, if the image to be processed is risky, the historical images of the cluster in the target image cluster are also risky. At this time, the risk category of the target image cluster can be updated from the third category to the first category, so as to achieve the effect of retrospectively processing other images in the target image cluster to which the image to be processed belongs based on the recognition result of the image to be processed.

[0131] As mentioned above, the number of target image clusters determined in step 103 is at least one. If the number of target image clusters is one, in step 105, the corresponding processing method can be directly applied based on the cluster security information of the target image cluster. If the number of target image clusters is two or more, generally, in step 105, the corresponding processing method can be applied based on the cluster security information of each target image cluster. For example, if there are three target image clusters determined in step 103, namely target image cluster A, target image cluster B, and target image cluster C, where the risk category of target image cluster A is the second category, the risk category of target image cluster B is the third category, and the risk category of target image cluster C is the third category, then in step 105, the risk category of target image cluster A remains unchanged and is still the second category. Since both target image clusters B and C fall under the third risk category, further identification of the images to be processed is required. If the images to be processed are identified as safe, the risk categories of both target image clusters B and C remain unchanged at the third category. If the images to be processed are identified as risky, the risk category of target image cluster B needs to be updated to the first category, and the risk category of target image cluster C also needs to be updated to the first category. Using the processing method given in step 105, with the risk category indicated by the cluster safety information as the filtering standard, if a target image cluster hits the first or second category, there is no need to review the images to be processed within that cluster, reducing review costs and improving image processing efficiency. If a target image cluster hits the first or second category, only the images to be processed within that cluster need to be reviewed, and the traceability of the review results extends only to the hit target image clusters, which can control the scope of influence of the images to be processed to a certain extent and improve the accuracy of image processing.

[0132] When an image to be processed, after similarity clustering, belongs to two or more target image clusters, and these target image clusters have both a first category and a second category of risk, for example, if there are two target image clusters determined in step 103, namely target image cluster A and target image cluster B, and if the risk category of target image cluster A is the first category (indicating that the image to be processed has a high risk) and the risk category of target image cluster B is the second category (indicating that the image to be processed is a safe image), then the safety attributes of the image to be processed in different target image clusters are completely opposite. This may be due to an error in the similarity clustering process in step 102. This result can be fed back to optimize and update the similarity clustering method used in step 102. Therefore, in this embodiment of the application, processing the intra-cluster images in the target image cluster according to the cluster security information may further include: if the cluster security information indicates that the risk category of the target image cluster includes a first category and a second category, obtaining the first target image cluster corresponding to the first category and the second target image cluster corresponding to the second category from the target image cluster; using the intra-cluster images in the first target image cluster and the second target image cluster as re-examination images, clustering the re-examination images to obtain at least one redirected image cluster; performing a second review on the risk category of the re-examination images, and determining the cluster security information of the redirected image cluster based on the second review result.

[0133] When the target image cluster exhibits the aforementioned situation, the target image cluster corresponding to the first category can be designated as the first image cluster, and the target image cluster corresponding to the second category can be designated as the second image cluster. All images within the first image cluster (including the image to be processed) can be designated as re-examination images, and all images within the second image cluster (including the image to be processed) can also be designated as re-examination images. For example, target image cluster A can be designated as the first target image cluster, and target image cluster B as the second target image cluster. All images within the clusters of both target image clusters A and B are re-examination images. Then, all re-examination images are re-clustered to obtain at least one redirected image cluster. A second review of the risk category of the re-examination images is then performed. The method for this second review can refer to the aforementioned method of "clustering historical images to obtain at least one historical image cluster and identifying the risk category of the historical image cluster," which will be elaborated upon here. Through re-clustering and second review, the aforementioned process can be retrospectively checked, thereby continuously optimizing the image processing accuracy of each stage in the image processing method.

[0134] As can be seen from the above, the embodiments of this application obtain an image to be processed and determine the processed image library corresponding to the image to be processed, wherein the processed image library includes at least one historical image cluster; then, the image to be processed is compared with the historical images within the historical image cluster; and based on the comparison result and the historical image cluster, at least one target image cluster to which the image to be processed belongs is determined; then, the cluster security information of the target image cluster is obtained, the cluster security information indicating the risk category of the images in the target image cluster; and then, the images within the target image cluster are processed according to the cluster security information. This scheme can compare the image to be processed with images in the processed image library in real time. Since there are already multiple historical image clusters in the processed image library, the target image cluster to which the image to be processed belongs can be determined based on the similarity between the image to be processed and the images in the processed image library during the comparison process. This allows the image to be processed to be associated with other images in the target image cluster. In a target image cluster, the images within each cluster have a high degree of similarity and influence each other, so they can be treated as a whole. Based on this, if the cluster security information of the target image cluster indicates that the target image cluster is at risk, the image to be processed can be processed directly. Alternatively, if the image to be processed is subsequently detected to be at risk, other clusters within the target image cluster to which the image to be processed belongs can be processed immediately. This achieves a processing effect from point to surface and from surface to point, improving the efficiency of image processing while ensuring the accuracy of image processing.

[0135] Based on the method described in the above embodiments, the following examples will provide further detailed explanations.

[0136] In this embodiment, the image processing device will be specifically integrated into an electronic device, which will be a server, as an example for explanation.

[0137] Figure 8 Another schematic flowchart of the image processing method provided in an embodiment of this application is shown. Figure 8 As shown, an image processing method has the following specific steps:

[0138] 201. The server retrieves all historical images from the processed image library, clusters the historical images, and obtains at least one historical image cluster.

[0139] For example, during the process of clustering historical images, each historical image may be assigned to only one historical image cluster, or it may be assigned to two or more historical image clusters at the same time.

[0140] 202. The server identifies the risk category of historical image clusters and obtains the historical cluster security information of the historical image clusters.

[0141] For example, the server can obtain risk-marked images and find the first image cluster to which the risk-marked images belong from the historical image clusters; if the first image cluster exists, the risk category of the first image cluster is updated to the first category, and the historical cluster security information of the first image cluster is obtained.

[0142] For example, if the first image cluster does not exist, the server can compare the risk-marked image with historical images; if the comparison result shows that the similarity between the risk-marked image and the historical image is greater than or equal to a preset second threshold, the historical image cluster to which the historical image belongs is taken as the second image cluster to which the risk-marked image belongs; the risk category of the second image cluster is updated to the first category, and the historical cluster security information of the second image cluster is obtained.

[0143] For example, if the first image cluster does not exist, the server can compare the risk-marked image with historical images; if the comparison result shows that the similarity between the risk-marked image and the historical image is greater than or equal to a preset second threshold, the server can take the historical image cluster where the historical image is located as the second image cluster to which the risk-marked image belongs; update the risk category of the second image cluster to the first category, and obtain the historical cluster security information of the second image cluster.

[0144] For example, the server can filter historical image clusters in the processed image library that are classified as risk category 1 to obtain a first remaining image cluster; the historical image clusters in the first remaining image clusters with a number of historical images within the cluster greater than a preset third threshold are selected as candidate image clusters for category 2; the historical images within the candidate image clusters for category 2 are reviewed, and the candidate image clusters for category 2 that pass the review are selected as the candidate image clusters for category 2; the risk category of the candidate image clusters for category 2 is updated to category 2 to obtain the historical cluster security information of the candidate image clusters for category 2.

[0145] For example, the server can obtain the application account corresponding to each cluster of historical images in the first remaining image cluster in the target application, as well as the historical interaction records of the application account; perform anomaly detection on the historical interaction records; if there is no abnormal behavior in the historical interaction records, the first remaining image cluster corresponding to the historical interaction records is taken as the second category image cluster; update the risk category of the second category image cluster to the second category, and obtain the historical cluster security information of the second category image cluster.

[0146] For example, the server can filter historical image clusters in the processed image library that are classified as risk category 1 and 2 to obtain a second remaining image cluster; then update the risk category of the second remaining image cluster to category 3 to obtain the historical cluster security information of the second remaining image cluster.

[0147] 203. The server obtains the image to be processed and determines the processed image library corresponding to the image to be processed. The processed image library includes at least one historical image cluster.

[0148] For example, after acquiring the image to be processed, the server can use the acquisition time of the image as a boundary to create a processed image library consisting of user avatars from the application software that occurred before that acquisition time. The processed image library includes multiple historical image clusters, and each historical image cluster includes at least one historical image from within that cluster.

[0149] 204. The server compares the image to be processed with the historical images within the historical image cluster.

[0150] For example, the server can extract features from the image to be processed to obtain the first image features; extract features from historical images within the cluster to obtain the second image features; and calculate the feature similarity between the first image features and the second image features.

[0151] 205. Based on the comparison results and historical image clusters, the server determines at least one target image cluster to which the image to be processed belongs.

[0152] For example, if the feature similarity is greater than or equal to a preset first threshold, the server can use the historical image cluster corresponding to the historical image within the cluster as the target image cluster; if the feature similarity is less than the preset first threshold, the server can construct the target image cluster to which the image to be processed belongs in the processed image library.

[0153] 206. The server obtains the cluster security information of the target image cluster, which indicates the risk category of the images in the target image cluster.

[0154] For example, the server can look up the historical cluster security information of the target image cluster; if the target image cluster has historical cluster security information, the historical cluster security information is used as the cluster security information of the target image cluster; if the target image cluster does not have historical cluster security information, the risk category of the target image cluster is initialized to obtain the cluster security information of the target image cluster.

[0155] For example, the server can set the risk category of the target image cluster to the third category.

[0156] 207. The server processes the intra-cluster images in the target image cluster based on the cluster security information.

[0157] For example, the server can process images within a target image cluster in the following ways: Method 1: If the cluster security information indicates that the risk category of the target image cluster is Category 2, retain the target image cluster; Method 2: If the cluster security information indicates that the risk category of the target image cluster is Category 1, obtain the target account corresponding to the image to be processed in the target application, and send a prompt message to the target account; Method 3: If the cluster security information indicates that the risk category of the target image cluster is Category 3, identify the image to be processed, and process the image to be processed according to the identification result.

[0158] For example, if the recognition result shows that the image to be processed is a safe image, the server can retain the cluster security information of the target image cluster; if the recognition result shows that the image to be processed is a risky image, the server can update the risk category of the target image cluster to the first category.

[0159] For example, if the cluster security information indicates that the risk category of the target image cluster includes a first category and a second category, the server can obtain the first target image cluster corresponding to the first category and the second target image cluster corresponding to the second category from the target image cluster; take the intra-cluster images in the first target image cluster and the second target image cluster as re-examination images, cluster the re-examination images to obtain at least one redirected image cluster; perform a second review on the risk category of the re-examination images, and determine the cluster security information of the redirected image cluster based on the second review result.

[0160] As described above, the server acquires all historical images from the processed image library, clusters these historical images to obtain at least one historical image cluster; then, it identifies the risk category of each historical image cluster to obtain historical cluster security information; next, it acquires the image to be processed and determines the processed image library corresponding to the image to be processed, wherein the processed image library includes at least one historical image cluster; then, it compares the image to be processed with the historical images within the historical image cluster; and based on the comparison result and the historical image cluster, it determines at least one target image cluster to which the image to be processed belongs; then, it acquires the cluster security information of the target image cluster, which indicates the risk category of the images in the target image cluster; and finally, it processes the images within the target image cluster according to the cluster security information. This scheme can compare the image to be processed with images in the processed image library in real time. Since there are already multiple historical image clusters in the processed image library, the target image cluster to which the image to be processed belongs can be determined based on the similarity between the image to be processed and the images in the processed image library during the comparison process. This allows the image to be processed to be associated with other images in the target image cluster. In a target image cluster, the images within each cluster have a high degree of similarity and influence each other, so they can be treated as a whole. Based on this, if the cluster security information of the target image cluster indicates that the target image cluster is at risk, the image to be processed can be processed directly. Alternatively, if the image to be processed is subsequently detected to be at risk, other clusters within the target image cluster to which the image to be processed belongs can be processed immediately. This achieves a processing effect from point to surface and from surface to point, improving the efficiency of image processing while ensuring the accuracy of image processing.

[0161] To better implement the above methods, this application also provides an image processing apparatus that can be integrated into a network device, such as a server or terminal. The terminal may include a tablet computer, a laptop computer, and / or a personal computer.

[0162] Figure 9 A schematic diagram of the structure of the image processing apparatus provided in an embodiment of this application is shown. Figure 9 As shown, the image processing device may include an image acquisition unit 301, an image clustering unit 302, an image localization unit 303, an attribute determination unit 304, an image processing unit 305, and a risk identification unit 306, as follows:

[0163] (1) Image acquisition unit 301;

[0164] The image acquisition unit 301 is used to acquire the image to be processed and determine the processed image library corresponding to the image to be processed. The processed image library includes at least one historical image cluster.

[0165] For example, the image acquisition unit 301 can be used to, after acquiring the image to be processed, use the acquisition time of the image to be processed as the boundary to take the dataset of user avatars in the application software that are located before the acquisition time as the processed image library.

[0166] (2) Image clustering unit 302;

[0167] Image clustering unit 302 is used to compare the image to be processed with historical images within the historical image cluster.

[0168] For example, the image clustering unit 302 can be used to extract features from the image to be processed to obtain first image features; extract features from historical images within the cluster to obtain second image features; and calculate the feature similarity between the first image features and the second image features.

[0169] (3) Image positioning unit 303;

[0170] Image localization unit 303 is used to determine at least one target image cluster to which the image to be processed belongs based on comparison results and historical image clusters.

[0171] For example, the image localization unit 303 can be used to take the historical image cluster corresponding to the historical image within the cluster as the target image cluster if the feature similarity is greater than or equal to a preset first threshold; and to construct the target image cluster to which the image to be processed belongs in the processed image library if the feature similarity is less than the preset first threshold.

[0172] (4) Attribute determination unit 304;

[0173] The attribute determination unit 304 is used to obtain cluster security information of the target image cluster, which indicates the risk category of the images in the target image cluster.

[0174] For example, the attribute determination unit 304 can be used to find the historical cluster security information of the target image cluster; if the target image cluster has historical cluster security information, the historical cluster security information is used as the cluster security information of the target image cluster; if the target image cluster does not have historical cluster security information, the risk category of the target image cluster is initialized to obtain the cluster security information of the target image cluster.

[0175] For example, the attribute determination unit 304 can be used to set the risk category of the target image cluster to the third category.

[0176] (5) Image processing unit 305;

[0177] The image processing unit 305 is used to process intra-cluster images in the target image cluster based on cluster security information.

[0178] For example, the image processing unit 305 can be specifically used to retain the target image cluster if the cluster security information indicates that the risk category of the target image cluster is the second category; if the cluster security information indicates that the risk category of the target image cluster is the first category, obtain the target account corresponding to the image to be processed in the target application and send a prompt message to the target account; if the cluster security information indicates that the risk category of the target image cluster is the third category, identify the image to be processed and process the image to be processed according to the identification result.

[0179] For example, the image processing unit 305 can be used to retain the cluster security information of the target image cluster if the recognition result shows that the image to be processed is a safe image; and to update the risk category of the target image cluster to the first category if the recognition result shows that the image to be processed is a risky image.

[0180] For example, the image processing unit 305 can be specifically used to: if the cluster security information indicates that the risk category of the target image cluster includes a first category and a second category, obtain the first target image cluster corresponding to the first category and the second target image cluster corresponding to the second category from the target image cluster; take the intra-cluster images in the first target image cluster and the second target image cluster as re-examination images, cluster the re-examination images to obtain at least one redirected image cluster; perform a second review on the risk category of the re-examination images, and determine the cluster security information of the redirected image cluster based on the second review result.

[0181] (6) Risk identification unit 306;

[0182] The risk identification unit 306 is used to acquire all historical images in the processed image library; cluster the historical images to obtain at least one historical image cluster; identify the risk category of the historical image cluster to obtain the historical cluster security information of the historical image cluster.

[0183] For example, the risk identification unit 306 can be used to acquire risk-marked images, find the first image cluster to which the risk-marked image belongs from the historical image clusters; if the first image cluster exists, update the risk category of the first image cluster to the first category, and obtain the historical cluster security information of the first image cluster.

[0184] For example, the risk identification unit 306 can be used to compare the risk-marked image with historical images if the first image cluster does not exist; if the comparison result shows that the similarity between the risk-marked image and the historical image is greater than or equal to a preset second threshold, the historical image cluster where the historical image is located is taken as the second image cluster to which the risk-marked image belongs; the risk category of the second image cluster is updated to the first category, and the historical cluster security information of the second image cluster is obtained.

[0185] For example, the risk identification unit 306 can be used to construct a third image cluster to which the risk-marked image belongs in the processed image library if the comparison result shows that the similarity between the risk-marked image and the historical image is less than a preset second threshold; set the risk category of the third image cluster to the first category, and obtain the historical cluster security information of the third image cluster.

[0186] For example, the risk identification unit 306 can be used to filter historical image clusters with a risk category of the first category in the processed image library to obtain a first remaining image cluster; to select historical image clusters in the first remaining image clusters whose number of historical images within the cluster is greater than a preset third threshold as second category candidate image clusters; to review the historical images within the second category candidate image clusters, and to select second category candidate image clusters that pass the review as second category image clusters; and to update the risk category of the second category image clusters to the second category to obtain the historical cluster security information of the second category image clusters.

[0187] For example, the risk identification unit 306 can be used to obtain the application account corresponding to each cluster of historical images in the first remaining image cluster in the target application, as well as the historical interaction records of the application account; perform anomaly detection on the historical interaction records; if there is no abnormal behavior in the historical interaction records, take the first remaining image cluster corresponding to the historical interaction records as the second category image cluster; update the risk category of the second category image cluster to the second category, and obtain the historical cluster security information of the second category image cluster.

[0188] For example, the risk identification unit 306 can be used to filter historical image clusters in the processed image library that have risk categories of the first and second categories to obtain the second remaining image cluster; and update the risk category of the second remaining image cluster to the third category to obtain the historical cluster security information of the second remaining image cluster.

[0189] In practice, each of the above units can be implemented as an independent entity or can be arbitrarily combined to be implemented as the same or several entities. For the specific implementation of each of the above units, please refer to the previous method embodiments, which will not be repeated here.

[0190] As can be seen from the above, in this embodiment of the application, after the image acquisition unit 301 acquires the image to be processed and determines the processed image library corresponding to the image to be processed, the image clustering unit 302 compares the image to be processed with the historical images within the historical image cluster. Then, the image positioning unit 303 determines at least one target image cluster to which the image to be processed belongs based on the comparison result and the historical image cluster. After that, the attribute determination unit 304 acquires the cluster security information of the target image cluster, and then the image processing unit 305 processes the images within the target image cluster according to the cluster security information. This scheme can compare the image to be processed with images in the processed image library in real time. Since there are already multiple historical image clusters in the processed image library, the target image cluster to which the image to be processed belongs can be determined based on the similarity between the image to be processed and the images in the processed image library during the comparison process. This allows the image to be processed to be associated with other images in the target image cluster. In a target image cluster, the images within each cluster have a high degree of similarity and influence each other, so they can be treated as a whole. Based on this, if the cluster security information of the target image cluster indicates that the target image cluster is at risk, the image to be processed can be processed directly. Alternatively, if the image to be processed is subsequently detected to be at risk, other clusters within the target image cluster to which the image to be processed belongs can be processed immediately. This achieves a processing effect from point to surface and from surface to point, improving the efficiency of image processing while ensuring the accuracy of image processing.

[0191] This application also provides an electronic device, such as... Figure 10 As shown, it illustrates a structural schematic diagram of the electronic device involved in the embodiments of this application, specifically:

[0192] The electronic device may include components such as a processor 401 with one or more processing cores, a memory 402 with one or more computer-readable storage media, a power supply 403, and an input unit 404. Those skilled in the art will understand that... Figure 10The electronic device structure shown does not constitute a limitation on the electronic device and may include more or fewer components than shown, or combine certain components, or have different component arrangements. Wherein:

[0193] The processor 401 is the control center of the electronic device, connecting various parts of the device via various interfaces and lines. It executes software programs and / or modules stored in the memory 402, and calls data stored in the memory 402, to perform various functions and process data. Optionally, the processor 401 may include one or more processing cores; preferably, the processor 401 may integrate an application processor and a modem processor, wherein the application processor mainly handles the operating system, user interface, and applications, and the modem processor mainly handles wireless communication. It is understood that the modem processor may not be integrated into the processor 401.

[0194] The memory 402 can be used to store software programs and modules. The processor 401 executes various functional applications and data processing by running the software programs and modules stored in the memory 402. The memory 402 may mainly include a program storage area and a data storage area. The program storage area may store the operating system, application programs required for at least one function (such as sound playback function, image playback function, etc.), etc.; the data storage area may store data created according to the use of the electronic device, etc. In addition, the memory 402 may include high-speed random access memory, and may also include non-volatile memory, such as at least one disk storage device, flash memory device, or other volatile solid-state storage device. Accordingly, the memory 402 may also include a memory controller to provide the processor 401 with access to the memory 402.

[0195] The electronic device also includes a power supply 403 that supplies power to the various components. Preferably, the power supply 403 can be logically connected to the processor 401 through a power management system, thereby enabling functions such as charging, discharging, and power consumption management through the power management system. The power supply 403 may also include one or more DC or AC power supplies, recharging systems, power fault detection circuits, power converters or inverters, power status indicators, and other arbitrary components.

[0196] The electronic device may also include an input unit 404, which can be used to receive input digital or character information, and generate keyboard, mouse, joystick, optical or trackball signal inputs related to user settings and function control.

[0197] Although not shown, the electronic device may also include a display unit, etc., which will not be described in detail here. Specifically, in this embodiment, the processor 401 in the electronic device loads the executable files corresponding to the processes of one or more applications into the memory 402 according to the following instructions, and the processor 401 runs the applications stored in the memory 402 to realize various functions, as follows:

[0198] The process involves acquiring an image to be processed and determining the corresponding processed image library, which includes at least one historical image cluster; comparing the image to be processed with historical images within the historical image cluster; determining at least one target image cluster to which the image to be processed belongs based on the comparison results and the historical image cluster; acquiring cluster security information of the target image cluster, which indicates the risk category of the images in the target image cluster; and processing the images within the target image cluster according to the cluster security information.

[0199] For example, an electronic device can acquire all historical images in a processed image library, cluster the historical images to obtain at least one historical image cluster; then, identify the risk category of the historical image cluster to obtain historical cluster security information; then, acquire the image to be processed and determine the processed image library corresponding to the image to be processed, wherein the processed image library includes at least one historical image cluster; then, compare the image to be processed with the historical images within the historical image cluster; and based on the comparison result and the historical image cluster, determine at least one target image cluster to which the image to be processed belongs; then, acquire the cluster security information of the target image cluster, which indicates the risk category of the images in the target image cluster; and then process the images within the target image cluster according to the cluster security information, and so on.

[0200] For details on the implementation of each of the above operations, please refer to the previous examples, which will not be repeated here.

[0201] As can be seen from the above, the embodiments of this application obtain an image to be processed and determine the processed image library corresponding to the image to be processed, wherein the processed image library includes at least one historical image cluster; then, the image to be processed is compared with the historical images within the historical image cluster; and based on the comparison result and the historical image cluster, at least one target image cluster to which the image to be processed belongs is determined; then, the cluster security information of the target image cluster is obtained, the cluster security information indicating the risk category of the images in the target image cluster; and then, the images within the target image cluster are processed according to the cluster security information. This scheme can compare the image to be processed with images in the processed image library in real time. Since there are already multiple historical image clusters in the processed image library, the target image cluster to which the image to be processed belongs can be determined based on the similarity between the image to be processed and the images in the processed image library during the comparison process. This allows the image to be processed to be associated with other images in the target image cluster. In a target image cluster, the images within each cluster have a high degree of similarity and influence each other, so they can be treated as a whole. Based on this, if the cluster security information of the target image cluster indicates that the target image cluster is at risk, the image to be processed can be processed directly. Alternatively, if the image to be processed is subsequently detected to be at risk, other clusters within the target image cluster to which the image to be processed belongs can be processed immediately. This achieves a processing effect from point to surface and from surface to point, improving the efficiency of image processing while ensuring the accuracy of image processing.

[0202] Those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be performed by instructions, or by instructions controlling related hardware. These instructions can be stored in a computer-readable storage medium and loaded and executed by a processor.

[0203] Therefore, embodiments of this application provide a computer-readable storage medium storing a plurality of instructions that can be loaded by a processor to execute steps in any of the image processing methods provided in embodiments of this application. For example, the instructions can execute the following steps:

[0204] The process involves acquiring an image to be processed and determining the corresponding processed image library, which includes at least one historical image cluster; comparing the image to be processed with historical images within the historical image cluster; determining at least one target image cluster to which the image to be processed belongs based on the comparison results and the historical image cluster; acquiring cluster security information of the target image cluster, which indicates the risk category of the images in the target image cluster; and processing the images within the target image cluster according to the cluster security information.

[0205] For example, all historical images in the processed image library are acquired, and the historical images are clustered to obtain at least one historical image cluster. Then, the risk category of the historical image cluster is identified to obtain the historical cluster security information of the historical image cluster. Next, the image to be processed is acquired, and the processed image library corresponding to the image to be processed is determined, wherein the processed image library includes at least one historical image cluster. Then, the image to be processed is compared with the historical images within the historical image cluster. Based on the comparison result and the historical image cluster, at least one target image cluster to which the image to be processed belongs is determined. Then, the cluster security information of the target image cluster is obtained, and the cluster security information indicates the risk category of the images in the target image cluster. Then, the images within the target image cluster are processed according to the cluster security information, and so on.

[0206] For details on the implementation of each of the above operations, please refer to the previous examples, which will not be repeated here.

[0207] The computer-readable storage medium may include: read-only memory (ROM), random access memory (RAM), disk or optical disk, etc.

[0208] Since the instructions stored in the computer-readable storage medium can execute the steps of any of the image processing methods provided in the embodiments of this application, the beneficial effects that any of the image processing methods provided in the embodiments of this application can achieve can be realized, as detailed in the preceding embodiments, and will not be repeated here.

[0209] According to one aspect of this application, a computer program product or computer program is provided, comprising computer instructions stored in a computer-readable storage medium. A processor of an electronic device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the electronic device to perform the methods provided in the various alternative implementations of the data access aspect described above.

[0210] The foregoing has provided a detailed description of an image processing method, apparatus, and computer-readable storage medium provided in the embodiments of this application. Specific examples have been used to illustrate the principles and implementation methods of the present invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of the present invention. At the same time, those skilled in the art will recognize that there will be changes in the specific implementation methods and application scope based on the ideas of the present invention. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. An image processing method, characterized in that, include: Obtain the image to be processed and determine the processed image library corresponding to the image to be processed, wherein the processed image library includes at least one historical image cluster; The image to be processed is compared with the historical images within the historical image cluster; Based on the comparison results and the historical image clusters, at least one target image cluster to which the image to be processed belongs is determined; Obtain cluster security information of the target image cluster, wherein the cluster security information indicates the risk category of the images in the target image cluster; Based on the cluster security information, the intra-cluster images in the target image cluster are processed.

2. The image processing method as described in claim 1, characterized in that, The step of comparing the image to be processed with the historical images within the historical image cluster includes: Feature extraction is performed on the image to be processed to obtain the first image features; Feature extraction is performed on the historical images within the cluster to obtain the second image features; Calculate the feature similarity between the first image feature and the second image feature.

3. The image processing method as described in claim 2, characterized in that, The step of determining at least one target image cluster to which the image to be processed belongs, based on the comparison results and the historical image clusters, includes: If the feature similarity is greater than or equal to a preset first threshold, the historical image cluster corresponding to the historical image within the cluster is taken as the target image cluster. If the feature similarity is less than the preset first threshold, a target image cluster to which the image to be processed belongs is constructed in the processed image library.

4. The image processing method according to any one of claims 1-3, characterized in that, The step of obtaining the cluster security information of the target image cluster includes: Locate the historical cluster security information of the target image cluster; If the target image cluster contains the historical cluster security information, the historical cluster security information shall be used as the cluster security information of the target image cluster. If the target image cluster does not have the historical cluster security information, the risk category of the target image cluster is initialized to obtain the cluster security information of the target image cluster.

5. The image processing method as described in claim 4, characterized in that, Before obtaining the historical cluster security information of the target image cluster, the method further includes: Retrieve all historical images from the processed image library; The historical images are clustered to obtain at least one cluster of historical images; The risk category of the historical image cluster is identified to obtain the historical cluster security information of the historical image cluster.

6. The image processing method as described in claim 5, characterized in that, The process of identifying the risk category of the historical image cluster to obtain the historical cluster security information includes: Obtain the risk-marked image, and find the first image cluster to which the risk-marked image belongs from the historical image cluster; If the first image cluster exists, update the risk category of the first image cluster to the first category to obtain the historical cluster security information of the first image cluster.

7. The image processing method as described in claim 6, characterized in that, After finding the first image cluster to which the risk-marked image belongs from the historical image cluster, the method further includes: If the first image cluster does not exist, the risk-marked image is compared with the historical image; If the comparison results show that the similarity between the risk-marked image and the historical image is greater than or equal to a preset second threshold, the historical image cluster to which the historical image belongs is taken as the second image cluster to which the risk-marked image belongs; The risk category of the second image cluster is updated to the first category to obtain the historical cluster security information of the second image cluster.

8. The image processing method as described in claim 7, characterized in that, After comparing the risk-marked image with the historical image, the method further includes: If the comparison result shows that the similarity between the risk-marked image and the historical image is less than the preset second threshold, a third image cluster to which the risk-marked image belongs is constructed in the processed image library; The risk category of the third image cluster is set to the first category to obtain the historical cluster security information of the third image cluster.

9. The image processing method as described in claim 6, characterized in that, The step of identifying the risk category of the historical image cluster to obtain the historical cluster security information of the historical image cluster also includes: Filter the historical image clusters in the processed image library that belong to the first risk category to obtain the first remaining image cluster; Historical image clusters in the first remaining image cluster with a number of historical images within the cluster greater than a preset third threshold are designated as candidate image clusters of the second category. The historical images within the second category candidate image clusters are reviewed, and the second category candidate image clusters that pass the review are designated as the second category image clusters. The risk category of the second category image cluster is updated to the second category to obtain the historical cluster security information of the second category image cluster.

10. The image processing method as described in claim 9, characterized in that, After filtering the historical image clusters in the processed image library that belong to the first risk category to obtain the first remaining image cluster, the method further includes: Obtain the application account corresponding to each cluster of historical images in the first remaining image cluster in the target application, as well as the historical interaction records of the application account; Anomaly detection is performed on the historical interaction records; If there is no abnormal behavior in the historical interaction record, the first remaining image cluster corresponding to the historical interaction record is taken as the second category image cluster; The risk category of the second category image cluster is updated to the second category to obtain the historical cluster security information of the second category image cluster.

11. The image processing method as described in claim 9, characterized in that, The step of identifying the risk category of the historical image cluster to obtain the cluster security information of the historical image cluster also includes: The historical image clusters in the processed image library that fall under the first and second risk categories are filtered to obtain the second remaining image clusters; The risk category of the second remaining image cluster is updated to the third category to obtain the historical cluster security information of the second remaining image cluster.

12. The image processing method as described in claim 4, characterized in that, The initialization of the risk category for the target image cluster includes: The risk category of the target image cluster is set to the third category.

13. The image processing method as described in claim 1, characterized in that, The step of processing intra-cluster images in the target image cluster based on the cluster security information includes: If the cluster security information indicates that the risk category of the target image cluster is the second category, the target image cluster is retained; If the cluster security information indicates that the risk category of the target image cluster is the first category, obtain the target account corresponding to the image to be processed in the target application, and send a prompt message to the target account; If the cluster security information indicates that the risk category of the target image cluster is the third category, the image to be processed is identified, and the image to be processed is processed according to the identification result.

14. The image processing method as described in claim 13, characterized in that, The step of processing the image to be processed based on the recognition result includes: If the recognition result shows that the image to be processed is a safe image, the cluster security information of the target image cluster is retained; If the recognition result indicates that the image to be processed is a risky image, the risk category of the target image cluster is updated to the first category.

15. The image processing method as described in claim 1, characterized in that, The step of processing intra-cluster images in the target image cluster based on the cluster security information includes: If the cluster security information indicates that the risk category of the target image cluster includes a first category and a second category, obtain the first target image cluster corresponding to the first category and the second target image cluster corresponding to the second category from the target image cluster; The images within the first target image cluster and the second target image cluster are used as re-examined images. The re-examined images are clustered to obtain at least one redirected image cluster. The risk category of the re-examined image is subject to a second review, and the cluster security information of the redirected image cluster is determined based on the results of the second review.

16. An image processing apparatus, characterized in that, include: An image acquisition unit is used to acquire an image to be processed and determine the processed image library corresponding to the image to be processed, wherein the processed image library includes at least one historical image cluster; An image clustering unit is used to compare the image to be processed with the historical images within the historical image cluster; An image localization unit is used to determine at least one target image cluster to which the image to be processed belongs, based on the comparison results and the historical image clusters. An attribute determination unit is used to obtain cluster security information of the target image cluster, wherein the cluster security information indicates the risk category of the images in the target image cluster; The image processing unit is used to process the intra-cluster images in the target image cluster according to the cluster security information.

17. An electronic device, characterized in that, It includes a processor and a memory, the memory storing an application program, and the processor running the application program within the memory to perform the steps of the image processing method according to any one of claims 1 to 15.

18. A computer program product comprising a computer program / instructions, characterized in that, When the computer program / instructions are executed by the processor, they implement the steps of the image processing method according to any one of claims 1 to 15.

19. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a plurality of instructions adapted for loading by a processor to perform the steps of the image processing method according to any one of claims 1 to 15.

Images