Search

CN-121999309-A - AI image monitoring evidence obtaining platform and electronic equipment

CN121999309ACN 121999309 ACN121999309 ACN 121999309ACN-121999309-A

Abstract

The application discloses an AI image monitoring evidence obtaining platform and electronic equipment, wherein the AI image monitoring evidence obtaining platform is provided with a plurality of AI large models, and after the AI image monitoring evidence obtaining platform is started, the processing is carried out, namely, image information of a prior right is input into an image feature extraction large model to generate an image feature set, the image feature expansion large model is combined with image display information, sales operation and transaction feature information of known infringement commodities, the image display information is corrected and supplemented to obtain an expanded image feature set, potential infringement commodity information is searched and recalled in an electronic commerce platform based on the expanded image feature set, the image information of the prior right and the potential infringement commodity information are input into an infringement comparison large model to be compared to obtain an infringement analysis result, if infringement is judged, evidence obtaining is carried out, and the evidence obtaining result is sent to a timestamp authentication service system to be added with a timestamp. The method and the system can improve the efficiency and accuracy of online store infringement evidence collection.

Inventors

  • ZHANG CHANGLI
  • Diao Chunfei

Assignees

  • 北京联合信任技术服务有限公司

Dates

Publication Date
20260508
Application Date
20251230

Claims (19)

  1. 1. The AI image monitoring evidence obtaining platform for discovering potential infringement behaviors of online stores is characterized in that the AI image monitoring evidence obtaining platform is provided with a plurality of pre-trained AI large models, the AI image monitoring evidence obtaining platform is in communication connection with a preset timestamp authentication service system, and the AI image monitoring evidence obtaining platform performs the following processing after being started: inputting the image information of the previous right into a pre-trained image feature extraction large model, wherein the image feature extraction large model is used for generating an image feature set of the previous right according to the image information of the previous right; Inputting the generated image feature set into a pre-trained image feature expansion large model, wherein the image feature expansion large model is used for correcting or supplementing the image feature set by utilizing image display information, sales operation and transaction feature information of one or more known infringement online stores on the known infringement commodities to obtain an expansion image feature set for searching the potential infringement commodities; searching and recalling potentially infringing commodity information in an electronic commerce platform based on the generated extended image feature set, wherein the potentially infringing commodity information comprises commodity display pictures, commodity packaging pictures and/or commodity use state pictures of potentially infringing commodities; Inputting the image information of the prior right and each piece of potentially infringing commodity information into a pre-trained infringing comparison large model, wherein the infringing comparison large model is used for analyzing the image information of the prior right to obtain a plurality of image features for defining the protection range of the prior right; if the infringement analysis result is that the potentially infringing commodity is judged to be the infringing commodity, evidence obtaining is carried out on the infringing commodity, and the evidence obtaining result is sent to a time stamp authentication service system so that the time stamp authentication service system can add a time stamp to the evidence obtaining result.
  2. 2. The AI image monitoring forensic platform according to claim 1 wherein the preceding claim comprises a patented design patent and the image information of the preceding claim comprises at least one of a six view and a perspective view of the design patent; the image feature extraction large model is used to perform the following processing: And (3) carrying out image analysis on each view image of the appearance design patent, and extracting overall contour features, key part shape features, pattern layout features, color collocation features and/or decoration element features as an image feature set of the appearance design patent.
  3. 3. The AI image monitoring forensic platform according to claim 2 wherein the image display information of the known infringing article comprises an infringing article picture of the known infringing article, the infringing article picture comprising at least one of an angle display picture, a pattern detail picture, and a color display picture of the known infringing article; the image feature extension large model is specifically used for executing the following processes: If the visual angle of the infringement commodity picture is inconsistent with the visual angles of the views of the design patent, extracting the image characteristics of the infringement commodity picture as parallel image characteristics to be supplemented into an expanded image characteristic set and/or And extracting a common background template, an interference light shadow or a coverage watermark for confusing the whole visual impression of the product from the infringement commodity picture as parallel image features, and supplementing the parallel image features into an extended image feature set.
  4. 4. The AI image monitoring and evidence obtaining platform according to claim 2 or 3, wherein the sales call of the known infringing online store comprises a text description of the appearance of the infringing commodity in a commodity detail page, a matched appearance description in an advertisement document and/or a description text of the commodity form in a user consultation question and answer; the image feature extension large model is specifically used for executing the following processes: extracting word description corresponding to key shape characteristics of the design patent from sales call operation; Extracting a fuzzification description of core design features adopted for avoiding design infringement from sales call operation; And converting the text description and/or the blurred description into image features, and supplementing the image features into the extended image feature set.
  5. 5. The AI image monitoring evidence obtaining platform of claim 2 or 3, wherein the transaction characteristic information of the known infringing on-line store includes a selling price interval and/or a shipping place of the known infringing on-line commodity sold by the known infringing on-line store; The image feature expansion large model is used for performing the process of supplementing the sales price interval and/or the delivery site of the known infringing commodity into an expansion feature word set as auxiliary screening features.
  6. 6. The AI image monitoring forensic platform according to claim 2 or 3 is characterized in that the infringement contrast large model is specifically used to perform the following processing: integrating image features of each view of the appearance design patent, determining a core design key point of the appearance design, decomposing the core design key point into a plurality of minimum visual units, wherein each minimum visual unit is an image feature, and the image feature comprises an integral modeling feature, a key part proportion feature, a pattern element combination feature or a color hierarchy feature; mapping and comparing the potentially infringing commodity information with the plurality of image features obtained by deconstructing one by one; If the potentially infringing commodity information contains all image features corresponding to core design points of the design patent, judging that the same or similar infringement is formed, and obtaining an infringement analysis result that the potentially infringing commodity is an infringement commodity.
  7. 7. The AI image monitoring forensic platform of claim 1 wherein searching and recalling potentially infringing merchandise information in an e-commerce platform based on the generated expanded image feature set comprises: decomposing each feature dimension in the extended image feature set into a search keyword, and synchronously generating corresponding feature vectors; In a commodity image library and a commodity description text library of the electronic commerce platform, respectively carrying out feature matching retrieval and semantic retrieval based on the feature vector and the retrieval keyword to obtain a potential infringement commodity list; And recalling the potentially infringing commodity information of each potentially infringing commodity in the potentially infringing commodity list.
  8. 8. The AI image monitoring forensic platform according to claim 1 in which the image feature set contains only part of the core image features defining the scope of the preceding claim or the image feature set contains all of the image features defining the scope of the preceding claim.
  9. 9. The AI image monitoring forensics platform according to claim 1 wherein the prior rights comprise a graphic-like trademark or a stereoscopic-like trademark, and the image information of the prior rights comprises a registered pattern of the graphic-like trademark or the stereoscopic-like trademark and goods or services specified for use; the image feature extraction large model is specifically used for executing the following processing: carrying out three-dimensional modeling analysis on the three-dimensional modeling pattern of the stereoscopic trademark to extract the overall stereoscopic profile, the three-dimensional morphology of the key part and/or the combination relation characteristics of all the components; And associating the extracted features with the properties of the goods or services by combining the approval registration category of the trademark and the appointed goods or services to form an image feature set of the graphic or stereo trademark.
  10. 10. The AI image monitoring and forensics platform according to claim 9 wherein the image presentation information of the known infringing article comprises an image used after mirror-image flipping, scaling, deformation or color adjustment of the graphic-like trademark or a three-dimensional presentation image used after scaling, local detail modification of the stereoscopic-like trademark; the image characteristic expansion large model is specifically used for executing the following processing: If the image display information of the known infringement commodity is identified to contain image features for mirror image overturning, zooming, deformation or color adjustment of the graphic trademark, taking the deformed or adjusted image features as parallel image features of the corresponding features of the graphic trademark, and supplementing the parallel image features into the expanded image feature set; If the image display information of the known infringed commodity is identified to contain the image features with the proportion adjustment or the local detail modification of the stereoscopic trademark, taking the adjusted or modified image features as the parallel image features of the corresponding features of the stereoscopic trademark, and supplementing the parallel image features into the extended image feature set.
  11. 11. The AI image monitoring forensic platform according to claim 9 wherein sales of the known infringing online store includes textual interpretation of graphic or stereoscopic type brands used in advertising literature or product detail description, appearance descriptive text bundled with known product models; the image characteristic expansion large model is specifically used for executing the following processing: extracting text interpretation of core elements of the graphic trademark from sales call; Extracting the text description of the three-dimensional form of the three-dimensional trademark from the sales call; Extracting appearance description text of a known product model which is bundled with a graphic type or a stereo type trademark from sales operation; Converting the literal interpretation, the literal description and/or the appearance description text into image features, supplementing into the expanded image feature set.
  12. 12. The AI image monitoring forensic platform according to claim 9 is characterized in that the infringement contrast large model is specifically configured to: Carrying out feature analysis on image information of graphic or stereo trademarks to obtain a plurality of image features for defining the protection range of the graphic or stereo trademarks, wherein the plurality of image features comprise core graphic elements, overall visual effects, obvious identification parts and stable presentation patterns; analyzing whether the potentially infringing commodity information contains core graphic elements of the trademark, and respectively calculating the similarity between the potentially infringing commodity information and the whole visual effect, the obvious identification part and the stable presentation style of the trademark; If the potentially infringing commodity information contains core graphic elements of the trademark, and the similarity between the core graphic elements and the overall visual effect, the obvious identification part and the stable presentation style of the trademark is higher than a preset threshold value, and the potentially infringing commodity information is used on the same or similar commodity or service, judging that the trademark infringes, and obtaining an infringing analysis result that the potentially infringing commodity is the infringing commodity.
  13. 13. The AI image monitoring forensic platform according to claim 1 wherein the preceding rights comprise a copyrighted photographic work or artwork, and the image information of the preceding rights comprises an original scan of the photographic work or artwork, a work name, author information, and at least a portion of a core image segment; The image feature extraction large model is used for executing the following processing: carrying out feature analysis on the image information of the photographic work or the fine works to extract core composition features, color key features, main element morphological features, light and shadow processing features and/or image features corresponding to original expression techniques of the work; and taking the image characteristics corresponding to the name, the author information, the core composition characteristics, the color key characteristics, the main element morphological characteristics, the light and shadow processing characteristics and/or the original expression technique of the work as the image characteristic set of the photographic work or the fine works.
  14. 14. The AI image monitoring evidence obtaining platform according to claim 13, wherein the image display information of the known infringing article includes a cut-out section, a flip section or a splice section of a camera work or a artwork used on the known infringing article, and the sales art of the known infringing online store includes an article detail description of the known infringing article and a descriptive text of the infringing image used in the advertisement document; the image feature extension large model is specifically used for executing the following processes: extracting image fragment characteristics formed by intercepting, overturning or splicing the camera work or the artwork from image display information of known infringed commodities; Extracting a description text of an infringement image from sales call operation of a known infringement online store, and converting the description text into corresponding image characteristics; and supplementing the image fragment characteristics and the converted image characteristics into the extended image characteristic set.
  15. 15. The AI image monitoring forensic platform according to claim 13 wherein the infringement contrast large model is specifically configured to perform the following: performing thought and expression binary filtering on the image information of the photographic works or the fine works, and filtering out public domain image elements, general composition modes and common light and shadow effects which are not protected by copyright; Image block segmentation and feature extraction are carried out on the residual original expression part after filtering, so that a plurality of image features are generated, and each image feature corresponds to one original image expression unit; giving weight coefficients determined based on originality height, core degree in works and creation difficulty to different image features; respectively calculating the image similarity between the potentially infringing commodity information and each image feature, and carrying out weighted calculation on the image similarity according to the weight coefficient to obtain a weighted similarity score; And if the weighted similarity score is greater than or equal to a preset weighted threshold, judging that the physical similarity is formed, and obtaining an infringement analysis result that the potential infringement commodity is an infringement commodity.
  16. 16. The AI image monitoring forensic platform according to any one of claims 6, 12 or 15 in which the infringement analysis results further comprise at least one of an infringement type, a feature alignment table, a confidence level, and an infringement evidence guide; The infringement type comprises appearance design infringement, trademark infringement, copyright infringement or non-infringement, the feature comparison table comprises similarity comparison details between a plurality of image features of the prior right and corresponding image features in the potentially infringement commodity information, the confidence comprises a certainty score of the infringement comparison large model to an infringement type judgment conclusion, and the infringement evidence guide comprises an image region for feature comparison and a corresponding position of the image region in the prior right image.
  17. 17. The AI image monitoring forensic platform of claim 16 further for, in performing forensics, performing the following: Based on the image area pointed in the infringement evidence guide and the corresponding position information thereof, locating and extracting the corresponding image area of the infringement commodity image and the prior right image, performing differential comparison and labeling, and generating an image comparison report showing the same part, the substantially similar part and the differential part; Based on the differential comparison result, statistics is carried out, and quantized data is generated, wherein the quantized data comprises at least one of pixel occupation ratio of similar image areas, number of core similar features and area proportion of similar parts occupying previous right images; the image comparison report is used with the quantized data as part of the evidence obtaining result.
  18. 18. The AI image monitoring forensic platform of claim 1 wherein forensic for infringing articles comprises: For infringed commodities identified in infringement analysis results, capturing a complete page containing commodity display pictures, commodity package pictures, online store names and prices of the infringed commodities through screen capturing, and synchronously recording environment information of a client during evidence obtaining operation, wherein the environment information comprises at least one of an access URL, a client IP address, a time stamp, an operating system version and a browser user agent; And/or starting a recording function, recording from a page entering an online store, gradually recording the online store name, commodity details, commodity picture switching and commodity description rolling viewing process, completely recording the display condition of infringed commodities, and synchronously recording the environmental information of the client side during evidence obtaining operation.
  19. 19. An electronic device deployed with the AI image monitoring forensic platform of any one of claims 1-18.

Description

AI image monitoring evidence obtaining platform and electronic equipment Technical Field The application relates to the technical field of infringement evidence obtaining, in particular to an AI image monitoring evidence obtaining platform and electronic equipment for finding out potential infringement behaviors of online stores. Background With the vigorous development of the electronic commerce industry, the number of online stores is rapidly increased, and the problem of online store infringement is increasingly highlighted, wherein the image infringement behavior with appearance design, graphics/stereoscopic trademark, photographic works, fine works and the like as cores is particularly prominent, so that the legal rights and interests of the prior rights and interests are damaged. The traditional online store infringement evidence obtaining and processing mode has a plurality of defects, and the traditional online store infringement evidence obtaining and processing mode cannot adapt to the supervision requirement of image infringement. On the one hand, the traditional mode mainly relies on the manual work to carry out the investigation of online store one by one, commodity one by one, and not only inefficiency, omission appears moreover easily. For the infringement of images, the defects are more remarkable, namely, the mass of image information such as commodity display pictures, package pictures, use state pictures and the like on an electronic commerce platform is complicated, each image is manually compared in a visual mode, time and labor are consumed, the influence of subjective judgment differences is avoided, the infringement behaviors such as fine approximation of the visual design, deformation imitation of trademark patterns, interception and splicing of work images and the like are difficult to comprehensively and timely examine, and therefore many hidden image infringement behaviors cannot be timely found and prevented. On the other hand, traditional evidence-taking approaches lack effective time proof and authority. When infringement disputes occur, the validity and effectiveness of the evidence and the accuracy of the evidence obtaining time are important. The traditional evidence obtaining process is generally only simple operations such as screenshot, storing commodity image pages and the like, the image evidence is easy to tamper with and replace, and the legally proving force is relatively weak. Moreover, due to the lack of a unified time authentication mechanism, it is difficult to accurately prove the time when the image infringement occurs, which brings great difficulty to subsequent maintenance work. In addition, in order to evade supervision, an infringer continuously adopts a concealing means aiming at image infringement, so that not only can the naming, description and sales of commodities be changed, but also the infringement image can be modified deliberately, and therefore, the potential infringement commodities are difficult to accurately identify in a traditional searching mode based on fixed program codes (or fixed rules). For example, some infringers can perform view angle adjustment and partial occlusion on the image of the externally designed patent product, mirror image overturning and scaling deformation on the graphic trademark, and the like. The hidden operations make the traditional online store infringement searching evidence obtaining mode difficult to accurately identify the image infringement commodities, and further increase the difficulty and cost of evidence obtaining. Disclosure of Invention In view of the above, the embodiment of the application provides an AI image monitoring evidence obtaining platform and electronic equipment for discovering potential infringement behaviors of online stores, which are used for solving the technical problems of low online store infringement evidence obtaining efficiency and high difficulty. In a first aspect, an embodiment of the present application provides an AI image monitoring and evidence obtaining platform for discovering potential infringement behavior of an online store, where the AI image monitoring and evidence obtaining platform is mounted with a plurality of pre-trained AI large models, and the AI image monitoring and evidence obtaining platform is connected with a preset timestamp authentication service system in a communication manner, and performs the following processing after the AI image monitoring and evidence obtaining platform is started: inputting the image information of the previous right into a pre-trained image feature extraction large model, wherein the image feature extraction large model is used for generating an image feature set of the previous right according to the image information of the previous right; Inputting the generated image feature set into a pre-trained image feature expansion large model, wherein the image feature expansion large model is used for correcting or supplementin