E-commerce product category information processing method and system based on vector database

Abstract

The invention provides an e-commerce product category information processing method and system based on a vector database, and the method comprises the steps: receiving product image information and product description information of a to-be-shelved product uploaded by a user through terminal equipment, and displaying a compliance detection control on a product compliance detection interface; receiving a trigger operation on the compliance detection control, sending a compliance detection request message to a server, and indicating the server to perform a compliance detection operation; receiving a compliance detection response message from the server; when the compliance state of the product description information and the non-compliance state of the product image information are detected, risk prompt information is displayed on a product compliance detection interface, and the risk prompt information is used for representing that the to-be-put-on-shelf product belongs to the forbidden product category. Therefore, in the application, the image detection method can be used to accurately recall illegal products under the condition that the product described by the text is inconsistent with the product shown by the actual image, and the comprehensiveness and accuracy of compliance detection are improved.

Description

Technical Field

[0001] The present application relates to the field of information processing technology, and in particular to an e-commerce product category information processing method based on a vector database. Background Art

[0002] The existing compliance detection of e-commerce platforms is usually based on the text of the product, while ignoring the detection of product images. In order to circumvent compliance detection, some users may change the text content of the product to compliant content while retaining the actual product image of the product. Therefore, the comprehensiveness and accuracy of compliance detection of e-commerce platforms in the existing technology need to be improved urgently. Summary of the invention

[0003] The present application provides an e-commerce product category information processing method and system based on a vector database. By using an image detection method, even if the product described in text is inconsistent with the product shown in the actual image, it is still possible to accurately recall the illegal product, thereby improving the comprehensiveness and accuracy of compliance detection of the e-commerce platform.

[0004] The present application provides an e-commerce product category information processing method based on a vector database, which is applied to a terminal device of a first e-commerce platform, wherein the first e-commerce platform includes a server and the terminal device, and the method includes: receiving product image information and product description information of a product to be put on the shelf uploaded by a user, and displaying a compliance detection control on a product compliance detection interface; and, receiving a trigger operation on the compliance detection control, creating a compliance detection request message, wherein the compliance detection request message includes the product image information and the product description information; and, sending the compliance detection request message to the server, wherein the compliance detection request message is used to instruct the server to perform a compliance detection operation; receiving a compliance detection response message from the server, wherein the compliance detection response message includes the detection status of the product image information and the product description information, the detection status including a compliance status and a non-compliance status; when the compliance status of the product description information and the non-compliance status of the product image information are detected, risk warning information is displayed on the product compliance detection interface, and the risk warning information is used to characterize that the product to be listed belongs to a banned product category; wherein the non-compliant status of the product image information is used to characterize that the similarity between the product image information and the first image in the first category image sub-library in the full library of banned product images is greater than a first preset similarity, and the first category image sub-library is associated with an image set of first category products of at least one second e-commerce platform, and the first category products are products that are banned from sale by the first e-commerce platform and are not banned from sale by the at least one second e-commerce platform.

[0005] In some embodiments, the full library of banned product images also includes a second category image sub-library, and the second category image sub-library is associated with a collection of images of second category products of the Internet platform, and the second category products are products that are banned from sale by both the first e-commerce platform and the at least one second e-commerce platform; the non-compliant status of the product image information is also used to characterize that the similarity between the product image information and the second image in the second category image sub-library is greater than the first preset similarity.

[0006] In some embodiments, the compliance status of the product description information is used to characterize that the similarity between the product description information and the text in the first category text sub-library and the text in the second category text sub-library in the entire text library of banned products is less than a second preset similarity, the first category text sub-library is associated with the text set of the first category products of the at least one second e-commerce platform, and the second category text sub-library is associated with the text set of the second category products in the policy file.

[0007] In some embodiments, the compliance detection operation includes the following steps: obtaining the product image information, the product description information, the image vector data library of the banned product image library, and the text vector data library of the banned product text library; determining the text detection result based on the product description information and the text vector data library; determining the image detection result based on the product image information and the image vector data library; and creating the compliance detection response message based on the text detection result and the image detection result.

[0008] In some embodiments, determining the text detection result based on the product description information and the entire text vector data library includes: vectorizing the product description information to obtain a product text vector; searching the entire text vector data library based on the product text vector and a text vector retrieval algorithm to obtain a target text vector that is most similar to the product text vector, and a text similarity with the target text vector; and determining the text detection result based on the text similarity.

[0009] In some embodiments, determining the image detection result based on the product image information and the entire image vector data library includes: vectorizing the product image information to obtain a product image vector; searching the entire image vector data library based on the product image vector and an image vector retrieval algorithm to obtain a target image vector that is most similar to the product image vector, and a target image similarity with the target image vector, wherein the target image corresponding to the target image vector is the first image or the second image; and determining the image detection result based on the target image similarity.

[0010] In some embodiments, the image detection result is determined based on the product image information and the image vector data library, including: performing image semantic analysis of the product image information in the product category dimension to obtain a reference category identifier of a reference product category to which the product to be put on the shelf belongs; calling a first image vector data sub-library corresponding to the reference category identifier in the image vector data library; performing vectorization processing on the product image information to obtain a product image vector; using the product image vector as a search term for a first image vector search operation, calling the first image vector data sub-library to perform the first image vector search operation, and obtaining multiple first image vector distances; screening the first image vector distance representing the highest similarity among the multiple first image vector distances; if the detected If it is detected that the first image vector distance selected out is less than the preset image vector distance, the image detection result is determined according to the first image vector distance; if it is detected that the first image vector distance selected out is greater than or equal to the preset image vector distance, the product image vector is used as the search term for the second image vector retrieval operation, and the second image vector data sub-library is called to perform the second image vector retrieval operation to obtain multiple second image vector distances, and the second image vector data sub-library is the image vector data sub-library in the entire image vector data library except the first image vector data sub-library; the second image vector distance representing the highest similarity among the multiple second image vector distances is selected; and the image detection result is determined according to the second image vector distance.

[0011] In some embodiments, the performing of image semantic analysis on the product image information in the product category dimension to obtain a reference category identifier of a reference product category to which the product to be put on the shelves belongs includes: calling a pre-trained product category recognition model that is locally deployed; and using the product category recognition model to process the product image information to obtain a reference category identifier of the reference product category to which the product to be put on the shelves belongs.

[0012] In some embodiments, the process of creating the full library of banned product images includes the following steps: obtaining a first compliance policy of the first e-commerce platform and a second compliance policy of the at least one second e-commerce platform; determining the first category products and the second category products according to the first compliance policy and the second compliance policy; for the first category products, performing the following operations: obtaining the first product category to which the first category products belong; and, searching the at least one second e-commerce platform according to the first product category, storing the images obtained by the search as the first category image sub-library; for the second category products, performing the following operations: obtaining the second product category to which the second category products belong; and, searching the Internet platform according to the second product category, storing the images obtained by the search as the second category image sub-library.

[0013] The present application provides an e-commerce platform system, which includes a terminal device and a server, and the terminal device is used to execute the step instructions in any method in the above method embodiments.

[0014] It can be seen that in the embodiment of the present application, the terminal device receives the product image information and product description information of the product to be put on the shelf uploaded by the user, and displays the compliance detection control on the product compliance detection interface; receives the trigger operation of the compliance detection control, creates a compliance detection request message, and the compliance detection request message includes the product image information and the product description information; sends a compliance detection request message to the server, and the compliance detection request message is used to instruct the server to perform a compliance detection operation; receives a compliance detection response message from the server, and the compliance detection response message includes the detection status of the product image information and the product description information, and the detection status includes a compliance status and a non-compliance status; detects the compliance status of the product description information and the non-compliance status of the product image information, and displays risk warning information on the product compliance detection interface, and the risk warning information is used to indicate that the product to be put on the shelf belongs to the category of banned products; wherein the non-compliant status of the product image information is used to indicate that the similarity between the product image information and the first image in the first category image sub-library in the full library of banned product images is greater than the first preset similarity, and the first category image sub-library is associated with an image set of the first category products of at least one second e-commerce platform, and the first category products are products that are banned from sale by the first e-commerce platform and are not banned from sale by at least one second e-commerce platform. Therefore, in this application, the image detection method can be used, and even if the product described in the text is inconsistent with the product shown in the actual image, the illegal products can still be accurately recalled, thereby improving the comprehensiveness and accuracy of compliance detection of the e-commerce platform. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0016] Figure 1 A schematic diagram of the structure of an e-commerce platform provided in an embodiment of the present application; Figure 2 A flowchart of a method for processing e-commerce product category information based on a vector database provided in an embodiment of the present application; Figure 3 One of the schematic diagrams of the compliance detection interface provided in the embodiment of the present application; Figure 4 A schematic diagram of a waiting interface provided in an embodiment of the present application; Figure 5 A second schematic diagram of a compliance detection interface provided in an embodiment of the present application; Figure 6 The third schematic diagram of the compliance detection interface provided in the embodiment of the present application; Figure 7 A fourth schematic diagram of a compliance detection interface provided in an embodiment of the present application; Figure 8 A fifth schematic diagram of a compliance detection interface provided in an embodiment of the present application; Fig. 9 A schematic diagram of the structure of a terminal device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0017] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0018] The terms "first", "second", etc. in the specification and claims of this application and the above-mentioned drawings are used to distinguish different objects, rather than to describe a specific order. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or units is not limited to the listed steps or units, but optionally includes steps or units that are not listed, or optionally includes other steps or units inherent to these processes, methods, products or devices.

[0019] Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present application. The appearance of the phrase in various locations in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0020] In the embodiments of the present application, "and / or" describes the association relationship of the associated objects, indicating that three relationships may exist. For example, A and / or B can represent the following three situations: A exists alone; A and B exist at the same time; B exists alone. Among them, A and B can be singular or plural.

[0021] In the embodiment of the present application, the symbol " / " can indicate that the objects associated with each other are in an "or" relationship. In addition, the symbol " / " can also indicate a division sign, that is, performing a division operation. For example, A / B can indicate A divided by B.

[0022] In the embodiments of the present application, "at least one item" or similar expressions refer to any combination of these items, including any combination of single items or plural items, and refer to one or more, and multiple refers to two or more. For example, at least one item of a, b, or c can represent the following seven situations: a, b, c, a and b, a and c, b and c, a, b, and c. Among them, each of a, b, and c can be an element or a set containing one or more elements.

[0023] In the embodiments of the present application, "equal to" can be used in conjunction with greater than, and is applicable to the technical solution adopted when greater than, and can also be used in conjunction with less than, and is applicable to the technical solution adopted when less than. When equal to is used in conjunction with greater than, it is not used in conjunction with less than; when equal to is used in conjunction with less than, it is not used in conjunction with greater than. The existing compliance detection of e-commerce platforms is usually based on the text of the product, but ignores the detection of product images. In order to circumvent compliance detection, some users may change the text content of the product to compliant content while retaining the actual product image.

[0024] In order to solve the above problems, the present application provides an e-commerce product category information processing method and system based on a vector database. Using an image detection method, even if the product described in text is inconsistent with the product shown in the actual image, it is still possible to accurately recall the illegal products, thereby improving the comprehensiveness and accuracy of compliance detection of the e-commerce platform.

[0025] The technical solution of the present application and how the technical solution of the present application solves the above-mentioned technical problems are described in detail below with specific embodiments. The following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be repeated in some embodiments. The embodiments of the present application will be described below in conjunction with the accompanying drawings.

[0026] See also Figure 1 and Figure 2 , Figure 1 A schematic diagram of the structure of an e-commerce platform provided in an embodiment of the present application, Figure 2 A flowchart of a method for processing e-commerce product category information based on a vector database is provided in an embodiment of the present application.

[0027] like Figure 1 As shown, the e-commerce platform 1 includes a terminal device 10 and a server 20, and the terminal device 10 and the server 20 are communicatively connected via wired or wireless means.

[0028] The terminal device 10 may specifically include a front-end device applied to the user side and capable of realizing functions such as data collection and data transmission, and may be a user equipment (UE) such as a mobile phone, a smart phone, a laptop, a digital broadcast receiver, a personal digital assistant (PDA), a tablet computer (PAD), a handheld device, a vehicle-mounted device, a wearable device, a computing device or other processing equipment connected to a wireless modem, a mobile station (MS), a mobile terminal, etc. Alternatively, the terminal device 10 may also be a software application that can be run in the above electronic devices. For example, it may be an APP running on a mobile phone.

[0029] The server 20 may specifically include a server that is applied to one side of the network platform and is responsible for data processing in the background, which can realize functions such as data transmission and data processing. It may be a physical server, or a server cluster or distributed system composed of multiple physical servers. In this embodiment, the number of servers is not specifically limited. Alternatively, it may also be a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, content delivery networks (CDN), and big data and artificial intelligence platforms.

[0030] The terminal device 10 is Figure 2 The execution body of the e-commerce product category information processing method based on the vector database shown in the figure includes the following steps S201 to S204: Step S201: The terminal device receives product image information and product description information of the product to be put on the shelf uploaded by the user, and displays a compliance detection control on the product compliance detection interface.

[0031] For specific implementation, please refer to Figure 3 , Figure 3 One of the schematic diagrams of the compliance detection interface provided in the embodiment of the present application, such as Figure 3 As shown, the compliance detection interface 3 includes a file upload area 31, which includes a product description or manual input / file upload area 311 and a product image file upload area 312. The user can click the corresponding upload control to upload the corresponding file.

[0032] Step S202: The terminal device receives a trigger operation on the compliance detection control, creates a compliance detection request message; and sends the compliance detection request message to the server.

[0033] Among them, after the user clicks on the compliance detection control, the terminal device creates a compliance detection request message for the product to be listed based on the click event. The compliance detection request message includes product image information and product description information. The compliance detection request message is used to instruct the server to perform a compliance detection operation; the terminal device sends the compliance detection request message to the server; after receiving the compliance detection request message, the server performs a compliance detection operation and generates a compliance detection response message; finally, the server sends the compliance detection response message to the terminal device.

[0034] In the specific implementation, after the user uploads the product description information and product image information, click the compliance detection control to submit the product description information and product image information to the server, and the server performs product compliance detection and enters Figure 4 The waiting interface 4 is shown.

[0035] Step S203: The terminal device receives a compliance detection response message from the server.

[0036] The compliance detection response message includes the detection status of the product image information and the product description information, and the detection status includes a compliance status and a non-compliance status.

[0037] Among them, in this application, the test results include the following three forms: ① the non-compliant status of product description information; ② the compliant status of product description information, the compliant status of product image information; ③ the compliant status of product description information, the non-compliant status of product image information. Among them, form ① and form ③ both indicate that the product to be put on the shelf is unqualified and needs to be recalled, and form ② indicates that the product to be put on the shelf is qualified and can be released to the current e-commerce platform, and the release business needs to be executed.

[0038] Step S204: the terminal device detects the compliance status of the product description information and the non-compliance status of the product image information, and displays risk warning information on the product compliance detection interface.

[0039] Among them, the risk warning information is used to indicate that the products to be listed belong to the prohibited product category, which refers to the product category that is prohibited from sale by the current first e-commerce platform.

[0040] Among them, the non-compliant status of the product image information is used to characterize that the similarity between the product image information and the first image in the first category image sub-library in the full library of banned product images is greater than a first preset similarity, and the first category image sub-library is associated with an image set of first category products of at least one second e-commerce platform, and the first category products are products that are banned from sale on the first e-commerce platform and are not banned from sale on at least one second e-commerce platform.

[0041] In specific implementation, when the service detection is completed, a compliance detection response message is sent to the terminal device, and the detection results of the product description information and product image information are displayed on the compliance detection interface 3, and the detection results can be prompted to the user by marking methods such as color or text, see Figure 5 .

[0042] It can be seen that in the embodiment of the present application, the terminal device receives the product image information and product description information of the product to be put on the shelf uploaded by the user, and displays the compliance detection control on the product compliance detection interface; receives the trigger operation of the compliance detection control, creates a compliance detection request message, and the compliance detection request message includes the product image information and the product description information; sends a compliance detection request message to the server, and the compliance detection request message is used to instruct the server to perform a compliance detection operation; receives a compliance detection response message from the server, and the compliance detection response message includes the detection status of the product image information and the product description information, and the detection status includes a compliance status and a non-compliance status; detects the compliance status of the product description information and the non-compliance status of the product image information, and displays risk warning information on the product compliance detection interface, and the risk warning information is used to indicate that the product to be put on the shelf belongs to the category of banned products; wherein the non-compliant status of the product image information is used to indicate that the similarity between the product image information and the first image in the first category image sub-library in the full library of banned product images is greater than the first preset similarity, and the first category image sub-library is associated with an image set of the first category products of at least one second e-commerce platform, and the first category products are products that are banned from sale by the first e-commerce platform and are not banned from sale by at least one second e-commerce platform. Therefore, in this application, the image detection method can be used, and even if the product described in the text is inconsistent with the product shown in the actual image, the illegal products can still be accurately recalled, thereby improving the comprehensiveness and accuracy of compliance detection of the e-commerce platform.

[0043] In some embodiments, the full library of banned product images also includes a second category image sub-library, and the second category image sub-library is associated with a collection of images of second category products of the Internet platform, and the second category products are products that are banned from sale by both the first e-commerce platform and the at least one second e-commerce platform; the non-compliant status of the product image information is also used to characterize that the similarity between the product image information and the second image in the second category image sub-library is greater than the first preset similarity.

[0044] In this embodiment, the entire library of banned product images includes not only the first category image sub-library but also the second category image sub-library. The first category image sub-library includes a collection of images of first category products, where the first category refers to the product category that is currently banned from sale on the first e-commerce platform but can be legally sold on other second e-commerce platforms; the second category image sub-library includes a collection of images of second category products, where the second product category refers to the product category that is banned from sale on all platforms. The images in the first image sub-library are obtained from at least one second e-commerce platform, and the images in the second image sub-library are obtained from an Internet platform.

[0045] Therefore, this embodiment expands the image coverage of the entire library of banned product images, making the results of searching for product image information in the entire library of banned product images more comprehensive and accurate, reducing false detections and missed detections, thereby ensuring that more banned products can be effectively identified and intercepted.

[0046] In some embodiments, the compliance status of the product description information is used to characterize that the similarity between the product description information and the text in the first category text sub-library and the text in the second category text sub-library in the entire text library of banned products is less than a second preset similarity, the first category text sub-library is associated with the text set of the first category products of the at least one second e-commerce platform, and the second category text sub-library is associated with the text set of the second category products in the policy file.

[0047] In this embodiment, the entire text library of banned products includes a first-category text sub-library and a second-category text sub-library. The first-category text sub-library is associated with a text set of first-category products in at least one second e-commerce platform, and the second-category text sub-library is associated with a text set of second-category products in a policy document.

[0048] It can be seen that in this embodiment, the text coverage of the banned product text library is expanded, making the search results for product text information in the banned product text library more comprehensive and accurate, reducing false detections and missed detections, thereby ensuring that more banned products can be effectively identified and intercepted.

[0049] In some embodiments, the specific process of the server performing the compliance detection operation includes the following steps a to d: Step a, the server obtains the product image information, the product description information, the image vector data library of the banned product image library, and the text vector data library of the banned product text library; In a specific implementation, the product images uploaded by users for the products to be put on the shelves include the marketing images of the products, which contain the ontological visualization information of the products and other image information in addition to the ontological visualization information. The other image information can be, for example, marketing-related graphic information. The server needs to first process the ontological visualization information of the product images to obtain ontological image feature information, and then further process the marketing-related graphic information to obtain non-ontological image feature information, and finally merge the ontological image feature information and non-ontological image feature information into the product image information of the product image uploaded this time. This can avoid missing the multi-dimensional information contained in the product images and improve the comprehensiveness and accuracy of subsequent vector retrieval.

[0050] For example, taking the gun barrel, one of the firearms parts banned from sale on the first platform as an example, the product to be listed may include the image information of the gun barrel itself and the brand and model information engraved on the outer surface of the gun barrel. The server can first process the image information of the gun barrel itself to obtain the main image feature information such as contour, structure, shape, material, color, etc., and then analyze the brand and model information on the outer surface of the gun barrel to obtain the brand feature information and model feature information, and finally merge the main image feature information such as contour, structure, shape, material, color, the brand feature information, and the model feature information into the product image information of the product picture uploaded currently.

[0051] Step b, the server determines the text detection result according to the product description information and the entire text vector data library; Step c, the server determines the image detection result according to the product image information and the image vector data library; Step d: The server creates the compliance detection response message according to the text detection result and the image detection result.

[0052] Among them, before step a "the server obtains the product image information, the product description information, the image vector data library of the banned product image library, and the text vector data library of the banned product text library", the method also includes: obtaining the banned product image library and the banned product text library; vectorizing the banned product image library and the banned product text library to obtain the image vector data library and the text vector data library.

[0053] Among them, for step b "the server determines the text detection result based on the product description information and the entire text vector data library", it specifically includes: vectorizing the product description information to obtain the product text vector; searching the entire text vector data library based on the product text vector and the text vector retrieval algorithm to obtain the target text vector that is most similar to the product text vector, and the text similarity with the target text vector; determining the text detection result based on the text similarity.

[0054] The server performs vectorization processing on the product description information to obtain a product text vector, which specifically includes the following steps: Step b101, text preprocessing: remove the punctuation marks from the product description text and perform word segmentation processing; for example, the description "This is an example product" will be segmented into ['this', 'is', 'a', 'example', 'product']; Step b102, using the Word2vec model to obtain word vectors: using the Word2vec model to obtain the vector of each word; Step b103, text vectorization: calculate the average of the word vectors in the product description to obtain the product text vector of a single product description content; assuming that there are N words in the description, and their word vectors are v1, v2, …, vN, then the description vector vdesc can be calculated by the following formula (1): , where vdesc is the vector representation of the product description, vi is the word vector of the i-th word, and N is the total number of words.

[0055] It should be noted that the above-mentioned process of vectorizing product description information is only an example provided in this application. Other methods such as the TF-IDF model can also be used to obtain the product text vector of a single product description information. This does not constitute a limitation on the technical solution of this application. The same text vectorization method in this embodiment can be used for the vectorization process of the entire text library of banned products.

[0056] Among them, the text vector retrieval algorithm includes but is not limited to cosine similarity, Euclidean distance, Pearson correlation coefficient, etc., which are all existing technologies and will not be repeated in this application.

[0057] Among them, in the text detection result determined according to the text similarity, the text detection result includes two states: compliance and non-compliance. The higher the text similarity, the greater the possibility of non-compliance. The specific judgment rule is: when the text similarity is greater than or equal to the second preset similarity, the text detection result is in a non-compliant state, and when the text similarity is less than the second preset similarity, the text detection result is in a qualified state. The second preset similarity is determined according to the detection training.

[0058] It can be seen that in this embodiment, by vectorizing the product description information and comparing it with the entire text vector database, high-precision text detection can be achieved. This method can more accurately identify content that matches the banned product information by calculating the similarity between vectors, thereby reducing the possibility of misjudgment and missed judgment.

[0059] Among them, for step c "the server determines the image detection result according to the product image information and the image vector data library", this application provides the following two specific embodiments of method 1 and method 2 for determination, which are as follows: Method 1: In some embodiments, step c "the server determines the image detection result according to the product image information and the image vector data library" specifically includes the following steps c11 to c13: Step c11, the server vectorizes the product image information to obtain a product image vector; Step c12: the server searches the entire image vector database according to the product image vector and the image vector retrieval algorithm to obtain a target image vector that is most similar to the product image vector and a target image similarity to the target image vector; the target image corresponding to the target image vector is the first image or the second image; Step c13: The server determines the image detection result according to the target image similarity.

[0060] Among them, for step c11 "the server vectorizes the product image information to obtain a product image vector", the main body image information and the non-main body image information in the product image information are vectorized according to the following steps c111 to c113 respectively to obtain the main body image vector and the non-main body image vector: Step c111, image preprocessing: resize the image to the size of the input required by the model; extract the RGB of each pixel of the image to form a three-dimensional array, where two dimensions identify the height and width of the image, and the third dimension represents the color channel (red, green, blue); normalize the image pixel values; Step c112, feature extraction: use a deep learning model that has been trained on a large-scale dataset (such as ImageNet); input the preprocessed image into the model, and the model will output a feature vector, which is a high-dimensional feature representation of the image; Step c113, feature vector processing: processing the multi-dimensional feature vector into a one-dimensional vector; normalizing the one-dimensional feature vector to obtain a single image vector.

[0061] It should be noted that the above-mentioned process of vectorizing product image information is only an example provided in this application. Other methods can also be used to obtain the product image vector of a single product image information. This does not constitute a limitation on the technical solution of this application. The same image vectorization method in this embodiment can be used for the vectorization process of the entire library of banned product images.

[0062] The image vector retrieval algorithm can be calculated using a brute force algorithm, an inverted index algorithm, a graph algorithm, and other algorithms. Exemplarily, when the image vector retrieval algorithm is a brute force algorithm, step c12 "the server searches the entire image vector database according to the product image vector and the image vector retrieval algorithm to obtain the target image vector that is most similar to the product image vector, and the target image similarity with the target image vector" specifically includes the following steps c121-c122: Step c121, the server calculates the distance between the product image vector and each image vector in the entire image vector database to obtain multiple vector distances, and a single vector distance is negatively correlated with the image similarity; In a specific implementation, since the reference role ratios of the main image information and the non-main image information in the image information of the product images of products of different product categories are different, the reference role ratio refers to the weight value artificially calibrated by the developer based on experience. The larger the weight value, the greater the role of the image information corresponding to the weight value in the prohibited product identification business. Through the design of the weight value, the server can dynamically adjust the weight values ​​of the vector distance of the main image vector and the vector distance of the non-main image vector according to the weight configuration information corresponding to the product category to which the image vector in the currently processed image vector data library belongs during the process of calculating the distance between the product image vector and each image vector in the image vector data library. Then, the vector distance between the currently processed image vector and the product image vector is weightedly calculated according to the weight value in the weight configuration information, and the calculation result of the vector distance will be more accurate than the calculation result using a uniform ratio for image vectors of all prohibited product categories.

[0063] For example, taking the gun barrel, one of the gun parts banned from sale on the first platform as an example, developers can use big data analysis to determine that the product images of firearms components must contain brand and exclusive model information, and the brand and model information are significantly different from other product categories, while the main body image information of the gun barrel is not significantly different from products such as toy guns. Therefore, the weight value of the main body image information of the gun barrel product can be set to be smaller than the weight value of the non-main body image information, for example, 0.2 (main body weight value) and 0.8 (non-main body weight value).

[0064] Step c122: The server obtains an image vector corresponding to the minimum vector distance among the multiple vector distances as the target image vector, and determines the target image similarity according to the vector distance between the target image vector and the product image vector.

[0065] Among them, for step c13 "the server determines the image detection result based on the target image similarity", the image detection result includes a compliant state and a non-compliant state. When the target image similarity is greater than or equal to the first preset similarity, the image detection result is a non-compliant state; when the target image similarity is less than the first preset similarity, the image detection result is a compliant state.

[0066] It can be seen that in this embodiment, the image vector retrieval algorithm is used to search the entire image vector database and calculate the similarity, which can effectively detect and match the images of banned products.

[0067] Method 2: In some embodiments, step c "the server determines the image detection result according to the product image information and the image vector data library" specifically includes the following steps c21-26: Step c21, the server performs image semantic analysis on the product image information in terms of product category, and obtains a reference category identifier of a reference product category to which the product to be put on the shelf belongs; Step c22, the server calls the first image vector data sub-library corresponding to the reference category identifier in the full image vector data library; Among them, the full image vector data library includes the correspondence between category identifiers and image vector data sub-libraries, that is, each product category has its corresponding image vector data sub-library, and multiple image vector data sub-libraries corresponding to multiple product categories that are banned from sale constitute the full image vector data library.

[0068] Step c23, the server performs vectorization processing on the product image information to obtain a product image vector; Among them, the product image information is consistent with the product image information of method 1. The product image information of products of different product categories includes entity image information and non-entity image information. The corresponding product entity image vector and product non-entity image vector are generated according to the entity image information and product non-entity image information in the product image. For example, for the barrel of a firearm part, the entity image information includes contour, structure, shape, material, color, etc., and the non-entity image information includes characteristic information such as brand and model.

[0069] Step c24, the server uses the product image vector as a search term for the first image vector search operation, calls the first image vector database sub-library to perform the first image vector search operation, and obtains a plurality of first image vector distances; and selects the first image vector distance representing the highest similarity among the plurality of first image vector distances; Among them, the reference role ratios of the ontological image information and the non-ontological image information in the product image information are different. The reference role ratio refers to the weight value artificially calibrated by the developer based on experience. The larger the weight value, the greater the role of the image information corresponding to the weight value in the prohibited product identification business. Through the design of the weight value, the server can dynamically adjust the weight values ​​of the vector distance of the ontological image vector and the vector distance of the non-ontological image vector according to the weight configuration information corresponding to the product category described by the image vector in the first image vector data sub-library currently being processed in the process of calculating the distance between the product image vector and each image vector in the first image vector data sub-library, and then calculate the vector distance between the currently processed image vector and the product image vector by weight according to the weight value in the weight configuration information. And the calculation result of the vector distance will be more accurate than the calculation result of using a uniform ratio for all image vectors of prohibited product categories.

[0070] In specific implementation, the product ontology image vector Vprimary and the product non-ontology image vector Vnon-primary in the product image vector Vcurrent are obtained, and the coordinates of the product ontology image vector Vprimary and the product non-ontology image vector Vnon-primary in the n-dimensional space are (v11, v12, ..., v1n) and (v21, v22, ..., v2n) respectively; the ontology weight value Wprimary and the non-ontology weight value Wnon-primary are determined according to the product category; for the image vector Vi in the first image vector data sub-library, the coordinates of the image vector Vi in the n-dimensional space are (vi1, vi2, ..., vin), where i refers to the i-th image vector in the first image vector data sub-library, and the first vector distance D (Vi, Vcurrent) between the image vector Vi and the product image vector Vcurrent is calculated, including the following steps ①-③: ① Calculate the vector distance D (Vi, Vprimary) between the image vector Vi and the product body image vector Vprimary: D (Vi,Vprimary)= ; ② Calculate the vector distance D (Vi, Vnon-primary) between the image vector Vi and the product non-primary image vector Vnon-primary: D (Vi,Vnon-primary)= ; ③According to the ontology weight value Wprimary and the non-ontology weight value Wnon-primary, the weighted distance calculation of D (Vi, Vprimary) and D (Vi, Vnon-primary) is performed to obtain the first image vector distance D (Vi, Vcurrent) between the image vector Vi and the product image vector Vcurrent: D(Vi,Vcurrent)=Wprimary· D (Vi,Vprimary)+ Wnon-primary·D (Vi,Vnon-primary)= Wprimary· +Wnon-primary .

[0071] Step c25: if the server detects that the first image vector distance selected is less than a preset image vector distance, the server determines the image detection result according to the first image vector distance; Step c26, if the server detects that the first image vector distance selected is greater than or equal to the preset image vector distance, the server uses the product image vector as the search term for the second image vector search operation, calls the second image vector data sub-library to perform the second image vector search operation, and obtains multiple second image vector distances, where the second image vector data sub-library is the image vector data sub-library in the entire image vector data library except the first image vector data sub-library; and selects the second image vector distance representing the highest similarity among the multiple second image vector distances; and determines the image detection result according to the second image vector distance; Among them, the specific process of the server calculating the distance between the second image vector based on the product image vector and each image vector in the second image vector data sub-library is consistent with the specific process of calculating the distance between the first image vector in the above step c24. The specific process is detailed in the above steps ①-③ and will not be repeated again.

[0072] Among them, for step c21, the reference category identifier refers to the product category identifier corresponding to the product to be put on the shelf, and "the server performs image semantic analysis of the product image information in the product category dimension to obtain the reference category identifier of the reference product category to which the product to be put on the shelf belongs", which specifically includes: calling a pre-trained product category recognition model deployed locally; using the product category recognition model to process the product image information to obtain the reference category identifier of the reference product category to which the product to be put on the shelf belongs.

[0073] The first image vector data sub-library refers to a collection of image vectors corresponding to reference category identifiers in the entire image vector data library.

[0074] The vectorization process of step c23 is the same as that of step c11 of method 1, and will not be described again here.

[0075] Wherein, for step c24, the first image vector retrieval operation performed by the server means that the server calculates the vector distance between the product image vector and each image vector in the first image vector data sub-library, and the vector distance is negatively correlated with the similarity, that is, the larger the vector distance, the smaller the similarity, and the smaller the vector distance, the greater the similarity. Therefore, the smallest first image vector distance among the multiple first image vector distances is determined, and the preset image vector distance corresponding to the first preset similarity is calculated according to the correlation between the similarity and the vector distance; If the minimum first image vector distance is less than or equal to the preset image vector distance, it indicates that there is an image in the first image vector data sub-library that is highly similar to the product image vector. In this case, the image detection result is in a non-compliant state; If the minimum first image vector distance is greater than the preset image vector distance, it indicates that there is no image in the first image vector data sub-library that is highly similar to the product image vector. At this time, it is necessary to perform vector retrieval again from the second image vector data sub-library; the image detection result is determined according to the relationship between the second image vector distance between the product image vector and the image vector in the second image vector data sub-library and the preset image vector distance; If the smallest second image vector distance is less than or equal to the preset image vector distance, the image detection result is in a non-compliant state; If the smallest second image vector distance is greater than the preset image vector distance, the image detection result is in compliance state.

[0076] It can be seen that in this embodiment, the server first searches the first image vector data sub-library based on the reference category. If the first image vector data sub-library is found to have a vector distance less than or equal to the preset image vector distance between the product image vector and the server, the image detection result is directly determined to be in a non-compliant state. In this case, the server can directly give a detection result of a non-compliant state without searching the entire library, thereby improving detection efficiency. When the first image vector data sub-library does not have a vector distance less than or equal to the preset image vector distance between the product image vector and the server, the second image vector data sub-library other than the first image vector data sub-library in the entire image vector data library is searched to obtain the detection result, thereby ensuring the comprehensiveness and accuracy of the search.

[0077] In some embodiments, the process of creating the entire library of banned product images by the server includes the following steps e-h: Step e: The server obtains a first compliance policy of the first e-commerce platform and a second compliance policy of the at least one second e-commerce platform; Step f, the server determines the first category products and the second category products according to the first compliance policy and the second compliance policy; Step g: the server performs the following operations for the first category products: obtaining the first product category to which the first category products belong; and searching the at least one second e-commerce platform according to the first product category, and storing the images obtained by the search as the first category image sub-library; In step h, the server performs the following operations for the second category products: obtaining the second product category to which the second category products belong; and searching the Internet platform according to the second product category, and storing the images obtained by the search as the second category image sub-library.

[0078] Among them, the compliance policy of the e-commerce platform includes multiple product categories that can be legally sold on the platform and multiple product categories that are prohibited from being sold. Therefore, the server can determine the first category of products that are prohibited from being sold on the first e-commerce platform but not on the second e-commerce platform according to the first compliance strategy and the second compliance strategy, and determine the second category of products that are prohibited from being sold on both the first and second e-commerce platforms. For the first category of products, since they can be legally sold on the second e-commerce platform, they can be searched on the second e-commerce platform according to their corresponding category information, and the searched picture collection is stored as the first category image sub-library; for the second category of products, since neither the first nor the second e-commerce platform can legally sell them, they can only be searched on the Internet platform according to their corresponding category information, and the searched picture collection is stored as the second category image sub-library.

[0079] It can be seen that in this embodiment, by selecting image sources based on legal sales and prohibited sales, an image sub-library related to the product category can be constructed more accurately, thereby reducing unnecessary image data, thereby improving the quality and relevance of the entire library of prohibited product images.

[0080] In some embodiments, the server's creation process for the banned product text library includes the following steps: obtaining a first compliance policy of the first e-commerce platform and a second compliance policy of the at least one second e-commerce platform; determining the first category products and the second category products based on the first compliance policy and the second compliance policy; for the first category products, performing the following operations: obtaining the first product category to which the first category products belong; and, searching the at least one second e-commerce platform based on the first product category, storing the text description information obtained from the search as a first category text sub-library; for the second category products, performing the following operations: obtaining the second product category to which the second category products belong; and, searching policy documents based on the second product category, storing the text description information obtained from the search as a second category text sub-library.

[0081] It can be seen that in this embodiment, by selecting the source of text description information based on legal sales and prohibited sales, a text sub-library related to the product category can be constructed more accurately, thereby reducing unnecessary text description information, thereby improving the quality and relevance of the entire text library of prohibited products.

[0082] In some embodiments, the risk warning information includes the target image and the product recall control. After the risk warning information is displayed on the product compliance detection interface, the method further includes: after receiving a trigger operation from the user on the product recall control, creating a product recall request message; sending the product recall request message to the server, wherein the product recall request message is used to instruct the server to perform a product recall business operation.

[0083] Among them, after the terminal device displays the risk warning information, it prompts the user to confirm the risk warning information. If the product to be put on the shelf is consistent with the target image, the user is prompted to click the recall control to recall the product to be put on the shelf.

[0084] Among them, the risk warning information also includes anomaly detection controls and detail viewing controls. If the user determines that the product to be listed is inconsistent with the target image, he can click the detail viewing control and choose to view multiple other images with higher similarity. If there is an image consistent with the product to be listed, the user is prompted to click the recall control; if there is no consistent image, the user clicks the anomaly control and sends the detection anomaly information to the server.

[0085] When implementing it specifically, Figure 5 As shown in the figure, if the product description information is compliant and the product image information is not compliant, the image most similar to the image of the product to be listed will be displayed in the product image area, and a recall control will be displayed in the compliance detection interface; the user determines whether the most similar image is substantially similar to the image of the product to be listed. If they are substantially similar, the user clicks the recall control to cancel the release of the product to be listed; if they are substantially not similar, the user clicks the detail view control to view multiple images with high similarity to the image of the product to be listed, such as Figure 6 As shown, the user determines whether there is a product that is substantially the same as the image of the product to be put on the shelf. If so, click the recall control. If not, click the detect anomaly control to re-detect. If the number of detections exceeds 3 times, the server sends an abnormality prompt to the administrator.

[0086] It can be seen that in this embodiment, when the user determines that the target image of the risk warning information is consistent with the product to be put on the shelf, the user can directly click the recall control to recall the product to be put on the shelf, which helps the user to quickly and accurately identify potential violations and reduce the difficulty of assessing the risk of product-triggered e-commerce platform policy violations.

[0087] In a possible embodiment, after receiving the compliance detection response message from the server, the method further includes: detecting the non-compliant status of the product description information, and displaying second risk warning information on the product compliance detection interface, the second risk warning information being used to characterize that the product to be listed belongs to a banned product category, and the second risk warning information includes product information corresponding to the target description information and the product recall control.

[0088] In specific implementation, if the product description information is not compliant, such as Figure 7 As shown, the description information most similar to the description information of the product to be put on the shelf is displayed in the product description area, and the recall control is displayed in the compliance detection interface, and the undetected status is displayed in the product image area.

[0089] In a possible embodiment, after receiving the compliance detection response message from the server, the method further includes: detecting the compliance status of the product description information and the compliance status of the product image information, and displaying compliance prompt information on the product compliance detection interface, wherein the compliance prompt information is used to indicate that the product to be listed does not belong to the prohibited product category.

[0090] In a possible embodiment, the compliance prompt information includes a listing control. After the compliance prompt information is displayed on the product compliance detection interface, the method further includes: receiving a trigger operation on the listing control, creating a product listing request message; sending the product listing request message to the server, and the product listing request message is used to instruct the server to perform the listing business.

[0091] In specific implementation, if the product description information is compliant and the product image information is compliant, such as Figure 8 As shown, a listing control is displayed on the compliance detection interface, and the user clicks the listing control to publish the product to be listed on the current e-commerce platform.

[0092] also, Figure 3 , Figure 4-Figure 8 The compliance detection interface is also used to display the current detection status, including unsubmitted and completed.

[0093] It should be noted that the specific implementation process of this embodiment can refer to the specific implementation process described in the above method embodiment, which will not be described here.

[0094] With the above Figure 2 For details on the embodiments shown in the drawings, please refer to Fig. 9 , Fig. 9 A schematic diagram of the structure of a terminal device provided in an embodiment of the present application is shown in FIG. Fig. 9As shown, the terminal device 10 includes a processor 91, a memory 93, a communication interface 92, and one or more programs 931. The one or more programs 931 are stored in the memory 93 and are configured to be executed by the processor 91. The above-mentioned programs include methods for executing the methods described in the above-mentioned embodiments.

[0095] An embodiment of the present application also provides a computer storage medium, wherein the computer storage medium stores a computer program for electronic data exchange, wherein the computer program enables a computer to execute part or all of the steps of any method described in the above method embodiments, and the above computer includes an electronic device.

[0096] It should be noted that, for the aforementioned method embodiments, for the sake of simplicity, they are all expressed as a series of action combinations, but those skilled in the art should be aware that the present application is not limited by the described order of actions, because according to the present application, certain steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily required by the present application.

[0097] In the above embodiments, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0098] In the several embodiments provided in the present application, it should be understood that the disclosed device can be implemented in other ways. For example, the device embodiments described above are only schematic, such as the division of the above-mentioned units, which is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, and the indirect coupling or communication connection of devices or units can be electrical or other forms.

[0099] The units described above as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0100] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0101] If the above-mentioned integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable memory. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a memory, including a number of instructions to enable a computer device (which can be a personal computer, server or network device, etc.) to execute all or part of the steps of the above-mentioned methods in each embodiment of the present application. The aforementioned memory includes: U disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), mobile hard disk, disk or optical disk and other media that can store program codes.

[0102] Those skilled in the art will appreciate that all or part of the steps in the various methods of the above embodiments may be completed by instructing related hardware through a program, and the program may be stored in a computer-readable memory, which may include a flash drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, etc.

[0103] The embodiments of the present application are introduced in detail above. Specific examples are used in this article to illustrate the principles and implementation methods of the present application. The description of the above embodiments is only used to help understand the method and core idea of ​​the present application. At the same time, for general technical personnel in this field, according to the idea of ​​the present application, there will be changes in the specific implementation method and application scope. In summary, the content of this specification should not be understood as a limitation on the present application.

Claims

1. A method for processing e-commerce product category information based on a vector database, characterized in that: A terminal device applied to a first e-commerce platform, wherein the first e-commerce platform includes a server and the terminal device, and the method includes: receiving product image information and product description information of a product to be put on the shelf uploaded by a user, and displaying a compliance detection control on a product compliance detection interface; and, receiving a trigger operation on the compliance detection control, creating a compliance detection request message, the compliance detection request message including the product image information and the product description information; and, sending the compliance detection request message to the server, the compliance detection request message being used to instruct the server to perform a compliance detection operation; receiving a compliance detection response message from the server, wherein the compliance detection response message includes detection statuses of the product image information and the product description information, wherein the detection statuses include a compliance status and a non-compliance status; The compliance status of the product description information and the non-compliance status of the product image information are detected, and risk warning information is displayed on the product compliance detection interface, wherein the risk warning information is used to characterize that the product to be listed belongs to a banned product category; wherein the non-compliant status of the product image information is used to characterize that a similarity between the product image information and a first image in a first category image sub-library in a full library of banned product images is greater than a first preset similarity, and the first category image sub-library is associated with an image set of first category products of at least one second e-commerce platform, and the first category products are products that are banned from sale by the first e-commerce platform and are not banned from sale by the at least one second e-commerce platform.

2. The method according to claim 1, characterized in that The full library of banned product images also includes a second-category image sub-library, which is an image collection of second-category products associated with the Internet platform, and the second-category products are products that are banned from sale by both the first e-commerce platform and the at least one second e-commerce platform; the non-compliant status of the product image information is also used to characterize that the similarity between the product image information and the second image in the second-category image sub-library is greater than the first preset similarity.

3. The method according to claim 2, characterized in that The compliance status of the product description information is used to characterize that the similarity between the product description information and the text in the first category text sub-library and the text in the second category text sub-library in the entire text library of banned products is less than a second preset similarity, the first category text sub-library is associated with the text set of the first category products of the at least one second e-commerce platform, and the second category text sub-library is associated with the text set of the second category products in the policy file.

4. The method according to claim 3, characterized in that The compliance detection operation includes the following steps: Acquire the product image information, the product description information, the image vector data library of the banned product image library, and the text vector data library of the banned product text library; Determine a text detection result according to the product description information and the entire text vector data library; Determine an image detection result according to the product image information and the image vector data library; The compliance detection response message is created according to the text detection result and the image detection result.

5. The method according to claim 4, characterized in that The determining of the text detection result according to the product description information and the full text vector data library includes: Vectorizing the product description information to obtain a product text vector; Searching the entire text vector database according to the product text vector and the text vector search algorithm to obtain the target text vector that is most similar to the product text vector and the text similarity with the target text vector; The text detection result is determined according to the text similarity.

6. The method according to claim 4, characterized in that The determining of the image detection result according to the product image information and the image vector data library includes: Vectorizing the product image information to obtain a product image vector; Searching the entire image vector database according to the product image vector and the image vector retrieval algorithm to obtain a target image vector most similar to the product image vector and a target image similarity to the target image vector, wherein the target image corresponding to the target image vector is the first image or the second image; The image detection result is determined according to the target image similarity.

7. The method according to claim 4, characterized in that The determining of the image detection result according to the product image information and the image vector data library includes: Performing image semantic analysis of the product image information in the product category dimension to obtain a reference category identifier of the reference product category to which the product to be put on the shelf belongs; Calling a first image vector data sub-library corresponding to the reference category identifier in the full image vector data library; Performing vectorization processing on the product image information to obtain a product image vector; Using the product image vector as a search term for a first image vector search operation, calling the first image vector data sub-library to perform the first image vector search operation, and obtaining a plurality of first image vector distances; Selecting a first image vector distance representing the highest similarity among the plurality of first image vector distances; If it is detected that the first image vector distance selected is less than a preset image vector distance, determining the image detection result according to the first image vector distance; If it is detected that the first image vector distance selected is greater than or equal to the preset image vector distance, the product image vector is used as the search term for the second image vector search operation, and the second image vector data sub-library is called to perform the second image vector search operation to obtain multiple second image vector distances, and the second image vector data sub-library is an image vector data sub-library in the image vector data full library except the first image vector data sub-library; Selecting a second image vector distance representing the highest similarity among the plurality of second image vector distances; The image detection result is determined according to the second image vector distance.

8. The method according to claim 7, characterized in that The performing image semantic analysis of the product image information in the product category dimension to obtain a reference category identifier of a reference product category to which the product to be put on the shelf belongs includes: Call the locally deployed pre-trained product category recognition model; The product image information is processed using the product category recognition model to obtain a reference category identifier of a reference product category to which the product to be put on the shelf belongs.

9. The method according to claim 2, characterized in that: The process of creating the full library of banned product images includes the following steps: Obtaining a first compliance policy of the first e-commerce platform and a second compliance policy of the at least one second e-commerce platform; Determine the first category products and the second category products according to the first compliance strategy and the second compliance strategy; For the first category products, the following operations are performed: obtaining a first product category to which the first category products belong; and searching the at least one second e-commerce platform according to the first product category, and storing the images obtained by the search as the first category image sub-library; For the second category products, perform the following operations: obtain the second product category to which the second category products belong; and search the Internet platform according to the second product category, and store the images obtained from the search as the second category image sub-library.

10. An e-commerce platform system, characterized in that: The e-commerce platform system includes a terminal device and a server, and the terminal device is used to execute the step instructions in any method of claims 1-9.

Images