E-commerce supply chain AI collaborative integrated intelligent management method and system

Abstract

The invention discloses an E-commerce supply chain AI collaborative integrated intelligent management method and system, and the method comprises the following steps: S1, obtaining a plurality of to-be-put-on-shelf commodities from a supply chain database, obtaining the to-be-put-on-shelf commodities selected by a user, and obtaining the historical commodities of the same type as the to-be-put-on-shelf commodities from the supply chain database, the method comprises the following steps: acquiring an e-commerce selling platform link of a historical commodity, a commodity picture and a first content set name where the commodity picture is located, and acquiring main picture and detail picture data of the historical commodity from the e-commerce selling platform link as a reference picture set, obtaining a to-be-typeset target picture of the to-be-put-on-shelf commodity and a second content set name where each picture is located; according to the method, the commodities of the same type are obtained from the e-commerce supply chain, then the reference pictures in the e-commerce platform are obtained, the closest pictures are determined, and finally the pictures are automatically formed, typeset and uploaded to the e-commerce selling platform, so that the commodities are automatically put on the shelf, and the efficiency is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent management of e-commerce systems, and particularly to an AI collaborative integrated intelligent management method and system for e-commerce supply chains. Background Art

[0002] An e-commerce supply chain provides supply chain products to e-commerce operators through a supply chain system, and the e-commerce operators can list products on an e-commerce sales platform. When listing products, graphic content typesetting is required. In the existing process of e-commerce product display and graphic content typesetting, how to effectively organize the arrangement order and position layout of multiple pictures plays a key role in enhancing page aesthetics, improving user experience, and promoting conversion rates. Currently, most picture typesetting still relies on manual operations by designers or simple template rules, lacking intelligent understanding and adaptive capabilities for the internal content of pictures, text semantics, and typesetting layouts. Different personnel have different typesetting effects, and for those with insufficient experience, the effects may be poor, and at the same time, it will waste more time and be inefficient. Summary of the Invention

[0003] Therefore, there is a need to provide an AI collaborative integrated intelligent management method and system for e-commerce supply chains to solve the problem of a large amount of manual operations and relatively low efficiency between existing supply chain products and e-commerce sales platforms.

[0004] To achieve the above object, the present invention provides an AI collaborative integrated intelligent management method for e-commerce supply chains, including the following steps: Step S1: Obtain multiple products to be listed from a supply chain database, obtain the products to be listed selected by a user, and obtain historical products of the same category as the product to be listed from the supply chain database. Obtain the e-commerce sales platform link, product pictures, and the name of the first content set where the product pictures are located of the historical products. Obtain the main pictures and detailed picture data of the historical products from the e-commerce sales platform link as a reference picture set, and obtain the target pictures to be typeset of the product to be listed and the name of the second content set where each picture is located. Step S2: Sequentially obtain a reference picture and its first content set name according to the order of the reference picture set. Obtain all the in-set target pictures in the second content set with the same name as the first content set name. Perform image content recognition on the reference picture and the in-set target pictures respectively, and extract visual feature descriptions of the product main body, its position, and layout through an AI large language model. Step S3: Extract the text content and its spatial position information in the reference picture and the in-set target pictures through an AI large language model, and generate semantic feature descriptions. Step S4: Send the visual feature description and semantic feature description to the AI large language model for comparison and obtain the in-set target image closest to the one reference image; Step S5: Sequentially obtain the in-set target images corresponding to all reference images according to the order of the reference image set, sort them, and generate an automated layout result in combination with the layout size parameters of the reference images, and upload it to the e-commerce sales platform.

[0005] Further, it also includes Step S6: Receive the interactive adjustment operation of the user on the layout result, store the adjusted order and the historical product in an associated manner as feedback data, and obtaining the order of the historical product and its reference image set includes modifying the order of the reference image set based on the feedback data.

[0006] Further, Step S7: According to the product link title of the historical product, the name of the product to be listed selected by the user, and the text content obtained by OCR from the images in the layout result, mark this information respectively and generate a control instruction by name and send it to the AI large language model to generate the product link title of the product to be listed and upload it to the e-commerce sales platform.

[0007] Further, it also includes the steps: Obtain the online listing feedback result of the e-commerce sales platform and record the corresponding relationship between the link address and the product in this feedback result.

[0008] Further, obtain the historical products of the same category as the product to be listed from the supply chain database: Obtain multiple historical products of the same category as the product to be listed from the supply chain database, Obtain the e-commerce sales platform link of each historical product and obtain the sales volume, Obtain the historical product with the largest sales volume as the historical product for subsequent reference.

[0009] Further, it also includes the steps: Generate a preview image after layout, obtain the adjustment of the user on this preview image, and upload the adjusted image to the e-commerce sales platform.

[0010] Further, in Step S4, sending the visual feature description and semantic feature description to the AI large language model for comparison is as follows: Convert the image name to an abbreviation, and send the corresponding visual feature description, semantic feature description, and comparison instruction to the AI large language model for comparison.

[0011] Further, it also includes the steps: Obtain the thumbnail of the historical product and its sorting, use the corresponding current image to be typeset as the thumbnail and sort it in the same order.

[0012] Further, the AI large language model is Deepseek or OpenAI.

[0013] The present invention provides an AI collaborative integrated intelligent management system for an e-commerce supply chain, including a memory and a processor. A computer program is stored on the memory, and when the computer program is executed by the processor, the steps of the method described in any one of the present invention are implemented.

[0014] Different from the prior art, the above technical solution obtains commodities of the same category from the e-commerce supply chain, then obtains reference pictures in the e-commerce platform, and differentiates the photos by the first content set name, and then compares the photos in the same content set. When comparing, it combines the picture summary and text recognition of the AI (Artificial Intelligence) large language model, extracts feature descriptions, and finally the AI large language model summarizes and compares the text to determine the closest picture and finally automatically forms a picture layout and uploads it to the e-commerce sales platform. In this way, the collaborative integration operation of e-commerce and the supply chain can be realized on the system, the automatic shelving of commodities can be realized, and the efficiency is improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 It is a flowchart of the AI collaborative integrated intelligent management method for the e-commerce supply chain of the present invention; Figure 2 It is a flowchart of the method of another embodiment of the present invention; Figure 3 It is a schematic diagram of the system interface of an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0016] To describe in detail the technical content, structural features, achieved objectives and effects of the technical solution, the following is described in detail in combination with specific embodiments and with reference to the accompanying drawings.

[0017] Please refer to Figures 1 to 3 , the present invention provides an AI collaborative integrated intelligent management method for an e-commerce supply chain, and the system interface is as Figure 3 shown. The method includes the following steps: Step S1: Obtain multiple commodities to be shelved from the supply chain database, such as Figure 3Supply chain products can be selected on the left side of the middle part. Supply chain products are obtained from the supply chain database. In the supply chain, suppliers mainly upload some products that can be sold, including product names, product categories, product pictures, product prices, product quantities, etc. Users of the management platform can select a product to be listed from them. The system obtains the product to be listed selected by the user and historical products of the same category as the product to be listed from the supply chain database. Here, the same category refers to the same product category, such as all being range hoods, washing machines, etc. Obtain the e-commerce sales platform link, product picture, and the name of the first content set where the product picture is located for the historical product. Here, the e-commerce sales platform link is the link recorded after the e-commerce sales platform returns the product link each time the product is released to the e-commerce sales platform. The content set name is used to store different pictures in different folder names in the supply chain database. For example, there are product main display picture sets, product detail display picture sets, product comparison picture sets, evaluation picture sets, scene picture sets, text description picture sets, manufacturer introduction picture sets, etc. Suppliers store multiple pictures related to each in the corresponding picture set folder respectively. Then, the system obtains the main picture and detail picture data of the historical product from the e-commerce sales platform link in the background as a reference picture set. Here, the main picture is the picture that rotates at the top of the product page, and the detail picture is the picture for the detailed introduction of the product page. These pictures are cached in the background in the order they appear on the product page. At the same time, obtain the target pictures to be typeset for the product to be listed and the name of the second content set where each picture is located.

[0018] Then enter step S2: According to the order of the reference picture set, the main pictures and detail pictures themselves have sorting. Then, they are combined and sorted with the main pictures first and the detail pictures later, and a reference picture and its first content set name are obtained in sequence. Here, the name of the first content set is specifically obtained by obtaining the original picture of the reference picture in the supply chain database, and then obtaining the name of the first content set of the original picture. Since the reference picture may have been adjusted, an image similarity comparison algorithm can be used for comparison to find the original picture. Then, all the in-set target pictures in the second content set name that is the same as the first content set name are obtained. Image content recognition is performed on the reference picture and the in-set target pictures respectively, and the visual feature descriptions of the product main body, its position, and layout are extracted through the AI large language model. In this step, the picture can be sent to the AI large language model, informing it to identify the product main body and its position in the picture, and extracting the position, size, and layout information of the product in the picture.

[0019] Taking a historical product as an example of a range hood, the first reference picture is Figure A (a distant front view with text descriptions such as "upgrade" at the top of the picture). The original picture of Figure A and the corresponding name of the first content set, "Product Main Body Display Atlas", are found. Then it is sent to the AI large language model, and at the same time, the instruction is sent: "Please give a visual feature description of the product main body and its position and layout." The text description extracted by the AI large language model is: "The product main body is a black top-mounted range hood, located above the center of the picture, symmetrically arranged, occupying about 60% of the overall image." Then, two pictures are found in the second content set named "Product Main Body Display Atlas", namely Figure B (a close-up side stereo view with text descriptions such as "upgrade" at the bottom of the picture) and Figure C (a front view with text descriptions such as "suction power" at the bottom of the picture), and they are sent to the AI large language model, along with the same instruction. The text descriptions extracted by the AI large language model are: for Figure B, "The product main body is a side-mounted range hood, located above and to the left of the center of the picture, accounting for about 40%, showing a three-dimensional effect from a top-down perspective"; for Figure C, "The product main body is a side-mounted range hood, located in the center of the picture, vertically symmetrically arranged, accounting for about 55%, with a clear and simple structure." The pictures and their corresponding descriptions are stored together.

[0020] Then enter step S3: Extract the text content and its spatial position information in the one reference picture and the target pictures in the set through the AI large language model to generate semantic feature descriptions. Taking the above three pictures as an example, by sending the pictures to the AI large language model and giving the instruction for extracting semantic feature descriptions, such as "Please give the text theme content in the picture and its position description in the picture." The AI large language model gives the description results. Taking the above three pictures as an example, the AI large language model gives the following descriptions respectively: for Figure A, "Brand + Model in the center, selling points listed above"; for Figure B, "Function slogan in the center, brand name repeated at the bottom"; for Figure C, "Model + Core selling point in the center, brand at the top". By extracting feature descriptions, it can be avoided that when directly using the AI large language model for comparison, due to the too many uncertain understandings of the language by the large language model, the layout of the pictures and the layout of the text will be ignored, and then mostly text content comparison is carried out. The comparison effect of the direct comparison on the picture layout is poor, and the error rate is large during the comparison.

[0021] Then enter step S4: Send the visual feature description and the semantic feature description to the AI large language model for comparison and obtain the target picture in the set that is closest to the one reference picture. Send the above pictures and their corresponding visual feature descriptions and semantic feature descriptions to the AI large language model, along with the similarity control instruction "According to the description, find the picture similar to Figure A in Figure B or Figure C" and send it to the AI large language model together. The content returned by the system obtained from the AI large language model is "According to the picture layout and visual feature description, Figure C is more similar to Figure A." Then Figure C is used as the same layout reference as Figure A.

[0022] Final step S5: Sequentially obtain the in-set target pictures corresponding to all reference pictures according to the order of the reference picture set and sort them. Combine the layout size parameters of the reference pictures to generate an automated layout result. Here, operations such as enlarging and reducing the pictures are performed according to the layout size parameters, and then preview can be carried out. Among them, the main picture and the detail pictures are at different preview positions. The preview is as shown in Figure 3 the right side of the middle part in []. Picture preview can be carried out, and then, after editing and confirmation together with some product parameter data, it can be uploaded to the e-commerce sales platform. The main picture and the detail pictures are uploaded to different positions on the e-commerce sales platform according to the main picture and the detail pictures in the reference pictures respectively.

[0023] In the present invention, similar products are obtained from the e-commerce supply chain, and then the reference pictures in the e-commerce platform are obtained. The photos are distinguished by the first content set name, and then the photos are compared within the same content set. When comparing, the picture summary and text recognition of the AI large language model are combined to extract feature descriptions. Finally, the AI large language model summarizes and compares the texts to determine the closest pictures and finally automatically form a picture layout and upload it to the e-commerce sales platform. In this way, the collaborative integration operation of e-commerce and the supply chain can be realized on the system, the automatic listing of supply chain products can be achieved, and the efficiency is improved.

[0024] To achieve the adjustment of pictures, as shown in Figure 2 Figure [], it further includes step S6: Receive the interactive adjustment operation of the user on the layout result, store the adjusted order associated with the historical product as feedback data. Obtaining the order of the historical product and its reference picture set includes modifying the order of the reference picture set based on the feedback data. In this way, when obtaining the reference picture set in the e-commerce sales platform link, the original layout of the e-commerce sales platform link is not changed, but the order of the reference picture set is changed through the feedback data. In this way, when referring to this e-commerce sales platform link later, a new picture layout order will be formed. By introducing the user interaction adjustment mechanism, the system can collect the subjective optimization opinions of users on the automatic sorting result, so as to obtain a sorting adjustment sample that is more in line with the actual operation strategy. Associate the sorting adjustment result with the historical product data, and then provide improved data for subsequent sorting optimization, which is convenient for the next product listing.

[0025] Furthermore, as shown in Figure 2As shown, it further includes step S7: Based on the product link title of the historical product, the name of the product to be listed selected by the user, and the text content obtained by OCR (Optical Character Recognition) from the pictures in the layout result, mark this information respectively (marking means using text to explain that the above text content is the reference product link title, the product name to be output, or the picture recognition text content) and generate a control instruction through the name (such as "generate the title name of the new product based on the reference product link title, the new product name, and its performance content") and send it to the AI large language model to generate the product link title of the product to be listed and upload it to the e-commerce sales platform. For example, the product link title of the historical product is "Brand A top and side dual-suction range hood, 21 cubic, 7-shaped range hood, small size 750mm, large suction, self-cleaning, wave control KL71". Then the name of the product to be listed is "Brand B Little Black D1P", and the text content obtained by OCR from the pictures in the layout result is such as "26m / min hurricane speed suction, top and side dual-net suction, infrared intelligent control to understand instructions instantly, new upgraded form, double-chamber little black wing, black and white two-color, versatile kitchen, easy to disassemble and clean". After sending it to the AI large language model, we get "Brand B Little Black D1P top and side dual-suction range hood, 26 cubic, hurricane speed suction, double-chamber little black wing, infrared intelligent control, easy to disassemble and clean, black and white two-color, household range hood". Then this title can be filled into Figure 3 the product name as the link title and uploaded to the e-commerce sales platform. This can achieve the automatic generation of the title by the AI large language model and improve the listing efficiency.

[0026] In some embodiments, in order to obtain the listing result, it further includes the steps of: obtaining the listing feedback result of the e-commerce sales platform and recording the corresponding relationship between the link address and the product in this feedback result. By recording the feedback data after the product pictures and texts are listed (such as product links, platform return status, listing results, etc.) and establishing a one-to-one mapping relationship with the original product data, the corresponding relationship between the automatic generation of the product and the real release on the platform can be achieved. It can be used as a standard for subsequent reference, facilitating the next automated listing. As Figure 3 shown, after the listing is completed, the upload time, the corresponding e-commerce platform, the link website, etc. are displayed in the upper left corner. By clicking to view, you can jump to the link address to view the e-commerce upload result.

[0027] Furthermore, obtain historical products of the same category as the product to be listed from the supply chain database: Obtain multiple historical products of the same category as the product to be listed from the supply chain database, obtain the e-commerce sales platform links of each historical product and obtain the sales volume, and use the historical product with the largest sales volume as the historical product for subsequent reference. By differentiating through the sales volume, it is convenient to adopt the layout result with a high sales volume, which helps to increase the sales volume of the product to be listed.

[0028] In some embodiments, it further includes the steps of: generating a typeset preview picture, obtaining the adjustments made by the user to the preview picture, and uploading the adjusted picture to an e-commerce sales platform. By showing the typeset preview picture to the user and allowing the user to make fine-tuning, the trust and acceptance of the user for the automated results of the system can be improved, and the interactive friendliness can be enhanced. The system records the user's adjustment results and then supports one-click uploading to the target e-commerce sales platform, realizing the full-process automation from automatic generation to user verification and then to automatic release, effectively reducing the manual operation steps and improving the efficiency of product listing.

[0029] Further, in step S4, sending the visual feature description and the semantic feature description to the AI large language model for comparison is as follows: converting the picture name into an abbreviation and then sending the corresponding visual feature description, semantic feature description, and the comparison instruction to the AI large language model for comparison. Through abbreviations, such as the above-mentioned Figure A, Figure B, and Figure C, or Figure 1 、 Figure 2 for differentiation, which can make the input information of the large language model less and avoid the long picture name information from affecting the judgment of the large language model.

[0030] Further, it further includes the steps of: obtaining the thumbnail of the historical product and its sorting, using the corresponding current picture to be typeset as the thumbnail and sorting it in the same order, and then uploading it to the e-commerce platform, so as to realize quick thumbnail browsing. By referring to the thumbnail sorting template of the historical product pictures, a recommended typesetting structure for the current product to be listed can be quickly generated. This strategy can not only continue the effective past typesetting styles but also provide a default sorting scheme in scenarios where data is insufficient or there is no reference picture, enhancing the system's adaptability. The comparison method here can refer to the comparison method of the main picture and the detail picture above.

[0031] Further, the AI large language model is Deepseek or OpenAI. In the present invention, using a general large language model (LLM) such as Deepseek or OpenAI as the core natural language and graphic integration processing engine can significantly enhance the system's ability to understand complex text information, picture content, user intent, and context semantics. By invoking the above advanced AI model, the system can achieve more accurate intelligent typesetting, personalized recommendation, and multi-round interactive feedback processing, improving the overall algorithm performance and user experience.

[0032] The present invention provides an AI collaborative integrated intelligent management system for e-commerce supply chains, including a memory and a processor. A computer program is stored on the memory, and when the computer program is executed by the processor, the steps of the method described in any one of the present invention are implemented. The storage medium of this embodiment can be a storage medium provided in an electronic device. The electronic device can read the content of the storage medium and achieve the effects of the present invention. The storage medium can also be a separate storage medium. When this storage medium is connected to the electronic device, the electronic device can read the content in the storage medium and implement the method steps of the present invention. This system obtains commodities of the same category in the e-commerce supply chain, then obtains reference pictures in the e-commerce platform, and differentiates the pictures by the first content set name. Then, the pictures are compared in the same content set. When comparing, the picture summary and text recognition of the AI large language model are combined to extract feature descriptions. Finally, the AI large language model summarizes and compares the texts to determine the closest picture and finally automatically forms a picture layout and uploads it to the e-commerce sales platform, realizing the automatic listing of commodities and improving the efficiency.

[0033] It should be noted that although the above embodiments have been described in this article, the patent protection scope of the present invention is not limited thereby. Therefore, based on the innovative concept of the present invention, any changes and modifications made to the embodiments described in this article, or equivalent structural or equivalent process transformations made using the content of the specification and drawings of the present invention, directly or indirectly applying the above technical solutions to other related technical fields, are all included in the patent protection scope of the present invention.

Claims

1. The e-commerce supply chain AI collaborative integrated intelligent management method is characterized by: The steps include: Step S1: obtaining multiple products to be put on the shelves from the supply chain database, obtaining the product to be put on the shelves selected by the user and obtaining historical products of the same category as the product to be put on the shelves from the supply chain database, obtaining the e-commerce sales platform link, product pictures and the name of the first content set where the product pictures are located of the historical products, obtaining the main picture and detail picture data of the historical products from the e-commerce sales platform link as a reference picture set, and obtaining the target pictures to be typeset of the product to be put on the shelves and the name of the second content set where each picture is located; Step S2: obtaining a reference image and its first content set name in sequence according to the order of the reference image set, obtaining all the target images in the second content set name in the second content set name that is the same as the first content set name, performing image content recognition on the reference image and the target image in the set respectively, and extracting the visual feature description of the commodity body and its position and layout through the AI ​​large language model; Step S3: extracting text content and spatial position information in the reference image and the target image in the set through the AI ​​large language model to generate a semantic feature description; Step S4: sending the visual feature description and the semantic feature description to the AI ​​large language model for comparison and obtaining the target image in the set that is closest to the reference image; Step S5: According to the order of the reference image set, the target images in the set corresponding to all reference images are obtained and sorted in turn, and the automatic layout results are generated and uploaded to the e-commerce sales platform in combination with the layout size parameters of the reference images.

2. The e-commerce supply chain AI collaborative integrated intelligent management method according to claim 1 is characterized in that: The method also includes step S6: receiving interactive adjustment operations of the user on the typesetting results, associating the adjusted order with the historical products and storing them as feedback data, and obtaining the order of the historical products and their reference picture sets, including modifying the order of the reference picture set based on the feedback data.

3. The e-commerce supply chain AI collaborative integrated intelligent management method according to claim 1 is characterized in that: Step S7: Based on the product link title of the historical product, the name of the product to be listed selected by the user, and the text content of the picture in the typesetting result obtained through OCR, this information is marked separately and sent to the AI ​​large language model through the name generation control instruction to generate the product link title of the product to be listed and upload it to the e-commerce sales platform.

4. The e-commerce supply chain AI collaborative integrated intelligent management method according to claim 1 is characterized in that: Also includes the steps: Obtain the listing feedback results of the e-commerce sales platform and record the corresponding relationship between the link address and the product in the feedback results.

5. The e-commerce supply chain AI collaborative integrated intelligent management method according to claim 1 is characterized in that: The historical products of the same category as the product to be put on the shelf obtained from the supply chain database include: Obtain multiple historical products of the same category as the product to be listed from the supply chain database, Get the e-commerce sales platform link of each historical product and obtain the sales volume. Get the historical products with the largest sales volume as a reference for subsequent use.

6. The e-commerce supply chain AI collaborative integrated intelligent management method according to claim 1 is characterized in that: Also includes the steps: Generate a preview image after typeset, obtain the user's adjustment to the preview image, and upload the adjusted image to the e-commerce sales platform.

7. The e-commerce supply chain AI collaborative integrated intelligent management method according to claim 1 is characterized in that: In step S4, the visual feature description and the semantic feature description are sent to the AI ​​large language model for comparison: The image name is converted into an abbreviation and the corresponding visual feature description and semantic feature description, as well as the comparison instruction are sent to the AI ​​large language model for comparison.

8. The e-commerce supply chain AI collaborative integrated intelligent management method according to claim 1 is characterized by: The method also includes the steps of obtaining thumbnails of historical products and their order, and using the corresponding current pictures to be typeset as thumbnails and ordering them in the same order.

9. The e-commerce supply chain AI collaborative integrated intelligent management method according to claim 1 is characterized by: The AI ​​large language model is Deepseek or OpenAI.

10. E-commerce supply chain AI collaborative integrated intelligent management system, characterized by: The method comprises a memory and a processor, wherein a computer program is stored in the memory, and when the computer program is executed by the processor, the steps of the method according to any one of claims 1 to 9 are implemented.

Images