Live broadcast e-commerce platform commodity content intelligent push management system based on big data

Abstract

The invention discloses a live broadcast e-commerce platform commodity content intelligent push management system based on big data. The big data-based live broadcast e-commerce platform commodity content intelligent push management system comprises a keyword searching and screening module, a commodity feature parameter division and extraction module, a user browsing image acquisition module, a target browsing image preprocessing module, an analysis server, a face angle detection module, a target element content information extraction module and a feature extraction module. An image coincidence coefficient analysis module; and an image association information pushing module. According to the invention, each commodity conforming to the keyword is screened from the plurality of commodities on the live broadcast e-commerce platform by inputting the commodity keyword, and the transaction information matched with the commodity keyword is screened from all the transaction information under the user login account through the input keyword; and the user deviation characteristics corresponding to the characteristic parameters of the commodity are analyzed from the transaction information, so that the shopping interest of the user is improved, and the deviation shopping demand of the user is met.

Description

Technical Field

[0001] This invention relates to the field of e-commerce platform management technology, specifically to a big data-based intelligent push management system for product content on live-streaming e-commerce platforms. Background Art

[0002] Currently, with the increasing penetration rate of mobile shopping, traditional e-commerce platforms are watching the shift from incremental growth to competition for existing customers, with overall traffic declining and customer acquisition costs rising. The emergence of live-streaming e-commerce has largely broken the status quo where consumers cannot see, touch, or feel the products. Compared to pictures and text, live-streaming presents richer information dimensions in video, allowing consumers to intuitively and comprehensively understand product and service information and immerse themselves in the shopping scenario. Therefore, live-streaming e-commerce is considered the future direction of e-commerce development.

[0003] When shopping on live-streaming e-commerce platforms, users enter product keywords into the search box, and the platform's backend pushes a series of products matching those keywords. However, existing live-streaming e-commerce platforms cannot recommend products based on user preferences, resulting in a low level of intelligence. This forces users to spend a significant amount of time searching through the recommended products to find those that match their interests, impacting their shopping interest and leading to a poor user experience. To enhance the user experience, there is an urgent need to design an intelligent product content recommendation and management system for live-streaming e-commerce platforms based on big data. Summary of the Invention

[0004] The purpose of this invention is to provide a big data-based intelligent push management system for product content on live-streaming e-commerce platforms, in order to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a big data-based intelligent push management system for product content on a live-streaming e-commerce platform, comprising:

[0006] The system includes a keyword search and filtering module, a product feature parameter segmentation and extraction module, a user browsing image acquisition module, a target browsing image preprocessing module, an analysis server, a face angle detection module, a target element content information extraction module, a feature extraction module, an image overlap coefficient analysis module, an image association information push module, a display terminal and storage database, a product comprehensive push order analysis module, and a background push module.

[0007] The keyword search and filtering module is used to filter products that match the keywords entered in the search box of the live e-commerce platform, and to number the filtered products in a preset order.

[0008] The product feature parameter segmentation and extraction module is used to segment the product categories represented by product keywords into various feature parameters, and extract the corresponding features of each feature parameter from the product live broadcast details for each numbered product according to the various feature parameters segmented by the product category. The extracted features of each product constitute the product feature set.

[0009] The user browsing image acquisition module is used to acquire images that the user browses on the target browsing platform and trigger target browsing image preprocessing instructions based on the user's dwell time on the browsing images.

[0010] The target browsing image preprocessing module is used to preprocess the user's target browsing image using image segmentation processing technology to obtain the foreground sub-image and background sub-image corresponding to the user's target browsing image;

[0011] The analysis server is connected to the face angle detection module, the target element content information extraction module, the feature extraction module, the feature analysis module, the image overlap coefficient analysis module, the image association information push module, the display terminal, and the storage database, respectively. The target browsing image preprocessing module is connected to the user browsing image acquisition module and the face angle detection module, respectively. The feature extraction module is connected to the feature analysis module, and the storage database is connected to the feature analysis module.

[0012] The face angle detection module includes a digital angle meter, which is used to receive the enhanced image number of each person sent by the image processing module. The digital angle meter measures the deflection angle of the face relative to the vertical direction and the pitch angle of the face relative to the horizontal plane in the enhanced image of each person. The module counts the deflection angle of the face relative to the vertical direction and the pitch angle of the face relative to the horizontal plane in the enhanced image of each person. The module then sends the set of deflection angles of the face relative to the vertical direction and the set of pitch angles of the face relative to the horizontal plane in the enhanced image of each person to the analysis server.

[0013] The target element content information extraction module is used to obtain each target element corresponding to the foreground sub-image in the user's target browsing image and extract the content information of each target element corresponding to the foreground sub-image in the user's target browsing image;

[0014] The feature extraction module is used to receive the set of image numbers of each person watching the multimedia advertisement sent by the analysis server, extract features from each image of each person watching the multimedia advertisement, extract facial features and body features from each image, and send the extracted facial features and body features from each image to the feature analysis module.

[0015] The image overlap coefficient analysis module is used to analyze the overlap coefficient between the user's target browsing image and each similar stored image, and to filter the similar stored images with the highest overlap coefficient.

[0016] The image association information push module is used to extract the associated videos corresponding to each stored image in the platform database and push the most similar stored images and their corresponding associated videos with the highest overlap coefficient.

[0017] The display terminal is used to receive and display the interest index of people of different genders towards the pushed advertisements sent by the analysis server;

[0018] The storage database is used to store the preset range of facial deflection angles relative to the vertical direction and the preset range of facial tilt angles relative to the horizontal plane when people watch multimedia advertisements. It also stores the standard facial features and standard body features of men and women, and the total time of pushing male and female gender-type advertisements.

[0019] The product comprehensive push order analysis module receives the total correlation degree between each product and user preference sent by the feature parameter association matching module, receives the monthly sales of each product sent by the monthly sales extraction module, and receives the positive review rate of each product sent by the positive review rate statistics analysis module. It performs comprehensive push coefficient statistics for each product and arranges each product in descending order of comprehensive push coefficient according to the size of the comprehensive push coefficient.

[0020] The backend push module is used to push each product, which has been sorted by the product comprehensive push order analysis module, to the live e-commerce platform interface in the sorted order.

[0021] Preferably, the intelligent product content push management system for live-streaming e-commerce platforms based on big data also includes:

[0022] The user historical data collection module is connected to the central control module and is used to collect users' historical transaction data and other information.

[0023] The product information collection module is connected to the central control module and is used to collect product information sold on the e-commerce platform.

[0024] The keyword extraction module, connected to the central control module, is used to extract product keywords based on the collected product information;

[0025] The product classification module, connected to the central control module, is used to classify products based on extracted product keywords.

[0026] Preferably, the intelligent product content push management system for live-streaming e-commerce platforms based on big data also includes:

[0027] Obtain extracted product keywords, features, parameter information, user preference features, and demand information;

[0028] Filter products based on user needs and keywords; match the features and parameters of different products based on user preferences and keywords.

[0029] Based on the matching results of demand and keywords, user preference characteristics and product characteristics, and parameter matching results, the relevance between products and users is determined, and products are sorted according to the degree of relevance.

[0030] Preferably, the specific steps corresponding to the target element feature data extraction module include:

[0031] Extract the corresponding feature data based on the central target element corresponding to the user's target browsing image, and mark the feature data of the central target element corresponding to the user's target browsing image as, where v is an uncertain number in the feature sequence number of the central target element corresponding to the user's target browsing image;

[0032] Based on the analysis steps of the central target element corresponding to the user's target browsing image, the central target elements corresponding to each similar stored image in the platform database are screened to obtain the standard feature data corresponding to the central target elements of each similar stored image. The standard feature data corresponding to the central target elements of each similar stored image are marked as, where;

[0033] The feature data of the central target element corresponding to the user's target browsing image are compared with the standard feature data of the central target element corresponding to each similar stored image to obtain the difference of the feature data of the central target element between the user's target browsing image and each similar stored image.

[0034] Preferably, the image segmentation processing techniques in the target browsing image preprocessing module include:

[0035] The grayscale transformation process is performed on the user's target browsing image to obtain the grayscale value corresponding to each pixel in the user's target browsing image, and a grayscale histogram of the user's target browsing image is constructed.

[0036] The maximum and minimum gray values ​​of the user's target browsing image are obtained from the gray-level histogram of the image, and are denoted as and , respectively, and an initial threshold is set.

[0037] Collect all gray values ​​in the user's target browsing image that are less than the initial threshold and all gray values ​​that are greater than the initial threshold, and calculate the mean value of all gray values ​​less than the initial threshold and the mean value of all gray values ​​greater than the initial threshold in the user's target browsing image, respectively.

[0038] A new threshold is calculated for the user's target browsing image. All pixels in the user's target browsing image with gray values ​​less than the new threshold are segmented and recombined to obtain the background sub-image corresponding to the user's target browsing image. All pixels in the user's target browsing image with gray values ​​greater than or equal to the new threshold are segmented and recombined to obtain the foreground sub-image corresponding to the user's target browsing image.

[0039] Preferably, the specific steps for a user to browse the image acquisition module include:

[0040] Define the browsing platform that the user is currently logged into as the target browsing platform, and obtain the images that the user is browsing on the target browsing platform;

[0041] Record the duration of time a user spends browsing images on the target browsing platform. If the duration of time a user spends browsing images on the target browsing platform exceeds the set duration threshold, then the image browsed by the user on the target browsing platform is recorded as the user's target browsing image, and the target browsing image preprocessing instruction is triggered.

[0042] The present invention proposes a big data-based intelligent product content push management system for live-streaming e-commerce platforms, which has the following advantages:

[0043] 1. This invention filters products from a pool of goods on a live-streaming e-commerce platform that match the input keywords. Simultaneously, it filters transaction information matching the keywords from all transaction records under the user's logged-in account. Furthermore, it analyzes the user preference characteristics corresponding to various feature parameters of the products from the transaction information. Meanwhile, the platform backend correlates and matches the features of each product selected by the input keywords with the user preference characteristics. This enables the products pushed by the live-streaming e-commerce platform to match the user's preferences, exhibiting a high level of intelligence, enhancing the user experience, avoiding users spending excessive time searching for products that match their preferences, increasing user shopping interest, and satisfying users' biased shopping needs.

[0044] 2. This invention analyzes and calculates the interest index of different genders towards pushed advertisements through a server, and displays it through a display module. This can intuitively show the interest of different genders in pushed advertisements, providing guidance for future advertisement pushes, facilitating the improvement of advertising resource utilization, and increasing the audience's interest in advertising content. Attached Figure Description

[0045] Figure 1 This is a schematic diagram of the module connection of the present invention. DETAILED DESCRIPTION

[0046] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0047] Please see Figure 1 This invention provides a technical solution: a big data-based intelligent push management system for product content on a live-streaming e-commerce platform, comprising:

[0048] The system includes a keyword search and filtering module, a product feature parameter segmentation and extraction module, a user browsing image acquisition module, a target browsing image preprocessing module, an analysis server, a face angle detection module, a target element content information extraction module, a feature extraction module, an image overlap coefficient analysis module, an image association information push module, a display terminal and storage database, a product comprehensive push order analysis module, and a background push module.

[0049] The specific steps corresponding to the user browsing image acquisition module include: denoting the browsing platform that the user is currently logging into as the target browsing platform, and acquiring the images that the user is browsing on the target browsing platform.

[0050] Record the duration of time a user spends browsing images on the target browsing platform. If the duration of time a user spends browsing images on the target browsing platform exceeds the set duration threshold, then the image browsed by the user on the target browsing platform is recorded as the user's target browsing image, and the target browsing image preprocessing instruction is triggered.

[0051] The keyword search and filtering module is used to filter products that match the keywords entered in the search box of the live e-commerce platform, and to number the filtered products in a preset order.

[0052] The product feature parameter segmentation and extraction module is used to segment the product categories represented by product keywords into various feature parameters, and extract the corresponding features of each feature parameter from the product live broadcast details for each numbered product according to the various feature parameters segmented by the product category. The extracted features of each product constitute the product feature set.

[0053] The user browsing image acquisition module is used to acquire images that the user browses on the target browsing platform and trigger target browsing image preprocessing instructions based on the user's dwell time on the browsing images.

[0054] The target browsing image preprocessing module is used to preprocess the user's target browsing image using image segmentation processing technology to obtain the foreground sub-image and background sub-image corresponding to the user's target browsing image.

[0055] The image segmentation processing technology in the target browsing image preprocessing module includes: performing grayscale transformation processing on the user target browsing image to obtain the grayscale value corresponding to each pixel in the user target browsing image, and constructing a grayscale histogram of the user target browsing image.

[0056] The maximum and minimum gray values ​​of the target image are obtained from the gray-level histogram of the target image and denoted as and , respectively, and an initial threshold is set.

[0057] Collect all gray values ​​in the user's target browsing image that are less than the initial threshold and all gray values ​​that are greater than the initial threshold, and calculate the mean value of each gray value in the target browsing image.

[0058] A new threshold is calculated for the user's target browsing image. All pixels in the user's target browsing image with gray values ​​less than the new threshold are segmented and recombined to obtain the background sub-image corresponding to the user's target browsing image. All pixels in the user's target browsing image with gray values ​​greater than or equal to the new threshold are segmented and recombined to obtain the foreground sub-image corresponding to the user's target browsing image.

[0059] The target browsing image preprocessing module is used to preprocess the user's target browsing image using image segmentation processing technology to obtain the foreground sub-image and background sub-image corresponding to the user's target browsing image;

[0060] The analysis server is connected to the face angle detection module, the target element content information extraction module, the feature extraction module, the feature analysis module, the image overlap coefficient analysis module, the image association information push module, the display terminal, and the storage database, respectively. The target browsing image preprocessing module is connected to the user browsing image acquisition module and the face angle detection module, respectively. The feature extraction module is connected to the feature analysis module, and the storage database is connected to the feature analysis module.

[0061] The face angle detection module includes a digital angle meter, which is used to receive the enhanced image number of each person sent by the image processing module. The digital angle meter measures the deflection angle of the face relative to the vertical direction and the pitch angle of the face relative to the horizontal plane in the enhanced image of each person. The module counts the deflection angle of the face relative to the vertical direction and the pitch angle of the face relative to the horizontal plane in the enhanced image of each person. The module then sends the set of deflection angles of the face relative to the vertical direction and the set of pitch angles of the face relative to the horizontal plane in the enhanced image of each person to the analysis server.

[0062] The target element content information extraction module is used to obtain each target element corresponding to the foreground sub-image in the user's target browsing image and extract the content information of each target element corresponding to the foreground sub-image in the user's target browsing image;

[0063] The feature extraction module is used to receive the set of image numbers of each person watching the multimedia advertisement sent by the analysis server, extract features from each image of each person watching the multimedia advertisement, extract facial features and body features from each image, and send the extracted facial features and body features from each image to the feature analysis module.

[0064] The feature analysis module receives facial and body features from images of individuals sent by the feature extraction module. It extracts standard facial and body features for men and women stored in the database. The module compares the facial features in the received images with these standard features, calculates the similarity between the facial features and the standard male or female features, and selects the gender corresponding to the facial features with the highest similarity. It also calculates the gender corresponding to the facial features in the images of individuals of the corresponding gender. Simultaneously, it compares the body features of the images corresponding to the gender with the standard body features of that gender. The system compares the similarity between the body features of individuals with facial features and the standard body features of their respective genders in their images. The statistical similarity is then compared with a set similarity threshold. If the statistical similarity is less than the set threshold, feature extraction and analysis need to be performed again until the statistical similarity is greater than or equal to the set threshold. At this point, the gender corresponding to each individual's image is output, thus avoiding a lack of targeting in advertising. This lays the foundation for later calculation of the male-to-female ratio coefficient among individuals watching multimedia advertisements. The gender corresponding to each individual's image is then sent to the analysis server.

[0065] The image overlap coefficient analysis module is used to analyze the overlap coefficient between the user's target browsing image and each similar stored image, and to filter the similar stored images with the highest overlap coefficient.

[0066] The image association information push module is used to extract the associated videos corresponding to each stored image in the platform database and push the most similar stored images and their corresponding associated videos with the highest overlap coefficient.

[0067] The display terminal is used to receive and display the interest index of different genders towards the pushed advertisements sent by the analysis server. It can intuitively show the interest of different genders in the pushed advertisements, provide guidance for future advertisement pushes, facilitate the improvement of advertising resource utilization, and increase the audience's interest in the advertising content.

[0068] The storage database is used to store the preset range of facial deflection angles relative to the vertical direction and the preset range of facial tilt angles relative to the horizontal plane when people watch multimedia advertisements. It also stores the standard facial features and standard body features of men and women, and the total time of pushing male and female gender-type advertisements.

[0069] The product comprehensive push order analysis module receives the total correlation degree between each product and user preference sent by the feature parameter association matching module, receives the monthly sales of each product sent by the monthly sales extraction module, and receives the positive review rate of each product sent by the positive review rate statistics analysis module. It performs comprehensive push coefficient statistics for each product and arranges each product in descending order of comprehensive push coefficient according to the size of the comprehensive push coefficient.

[0070] The backend push module is used to push each product, which has been sorted by the product comprehensive push order analysis module, to the live e-commerce platform interface in the sorted order.

[0071] As a preferred solution, a further intelligent product content push management system for live-streaming e-commerce platforms based on big data also includes:

[0072] The user historical data collection module is connected to the central control module and is used to collect users' historical transaction data and other information.

[0073] The product information collection module is connected to the central control module and is used to collect product information sold on the e-commerce platform.

[0074] The keyword extraction module, connected to the central control module, is used to extract product keywords based on the collected product information;

[0075] The product classification module, connected to the central control module, is used to classify products based on extracted product keywords.

[0076] As a preferred solution, a further intelligent product content push management system for live-streaming e-commerce platforms based on big data also includes:

[0077] Obtain extracted product keywords, features, parameter information, user preference features, and demand information;

[0078] Filter products based on user needs and keywords; match the features and parameters of different products based on user preferences and keywords.

[0079] Based on the matching results of demand and keywords, user preference characteristics and product characteristics, and parameter matching results, the relevance between products and users is determined, and products are sorted according to the degree of relevance.

[0080] S1 collects users' historical transaction data and other information through the user historical data collection module; and collects product information sold on the e-commerce platform through the product information collection module.

[0081] S2, through the central control module, uses a microcontroller or controller to control the keyword extraction module to extract product keywords based on the collected product information; and through the product classification module, performs product classification based on the extracted product keywords;

[0082] S3 extracts product features and parameters based on the collected product information through the product feature parameter extraction module; and extracts user preferences based on the collected user historical transaction data through the user preference extraction module.

[0083] S4: The user demand collection module obtains the input information entered by the user into the search box of the search engine of the live e-commerce platform; the feature parameter association matching module matches the most relevant products based on the extracted user preferences and user demand.

[0084] S5 pushes products based on the matching results through the background push module.

[0085] As a preferred option, the specific steps corresponding to the target element feature data extraction module further include:

[0086] Extract the corresponding feature data based on the central target element corresponding to the user's target browsing image, and mark the feature data of the central target element corresponding to the user's target browsing image as, where v is an uncertain number in the feature sequence number of the central target element corresponding to the user's target browsing image;

[0087] Based on the analysis steps of the central target element corresponding to the user's target browsing image, the central target elements corresponding to each similar stored image in the platform database are screened to obtain the standard feature data corresponding to the central target elements of each similar stored image. The standard feature data corresponding to the central target elements of each similar stored image are marked as, where;

[0088] The feature data of the central target element corresponding to the user's target browsing image are compared with the standard feature data of the central target element corresponding to each similar stored image to obtain the difference of the feature data of the central target element between the user's target browsing image and each similar stored image.

[0089] As a preferred option, the image segmentation processing techniques in the target browsing image preprocessing module further include:

[0090] The grayscale transformation process is performed on the user's target browsing image to obtain the grayscale value corresponding to each pixel in the user's target browsing image, and a grayscale histogram of the user's target browsing image is constructed.

[0091] The maximum and minimum gray values ​​of the user's target browsing image are obtained from the gray-level histogram of the image, and are denoted as and , respectively, and an initial threshold is set.

[0092] Collect all gray values ​​in the user's target browsing image that are less than the initial threshold and all gray values ​​that are greater than the initial threshold, and calculate the mean value of all gray values ​​less than the initial threshold and the mean value of all gray values ​​greater than the initial threshold in the user's target browsing image, respectively.

[0093] A new threshold is calculated for the user's target browsing image. All pixels in the user's target browsing image with gray values ​​less than the new threshold are segmented and recombined to obtain the background sub-image corresponding to the user's target browsing image. All pixels in the user's target browsing image with gray values ​​greater than or equal to the new threshold are segmented and recombined to obtain the foreground sub-image corresponding to the user's target browsing image.

[0094] As a preferred option, the specific steps for users to browse the image acquisition module further include:

[0095] Define the browsing platform that the user is currently logged into as the target browsing platform, and obtain the images that the user is browsing on the target browsing platform;

[0096] Record the duration of time a user spends browsing images on the target browsing platform. If the duration of time a user spends browsing images on the target browsing platform exceeds the set duration threshold, then the image browsed by the user on the target browsing platform is recorded as the user's target browsing image, and the target browsing image preprocessing instruction is triggered.

[0097] Extract the constituent images corresponding to the foreground sub-images in the user's target browsing image and record them as target elements. Obtain the target elements corresponding to the foreground sub-images in the user's target browsing image and the target elements corresponding to the foreground sub-images in each scene matching storage image. Compare the target elements corresponding to the foreground sub-images in the user's target browsing image with the target elements corresponding to the foreground sub-images in each scene matching storage image. Filter out scene matching storage images that are similar to the target elements corresponding to the foreground sub-images in the user's target browsing image and record them as similar storage images. Number the similar storage images sequentially.

[0098] The above-mentioned filtering of stored images that match the scene corresponding to the background sub-image in the user's target browsing image includes: obtaining the matching degree between the scene corresponding to the background sub-image in the user's target browsing image and the scene corresponding to the background sub-image in each stored image based on comparison; if the matching degree between the scene corresponding to the background sub-image in the user's target browsing image and the scene corresponding to the background sub-image in a certain stored image is greater than or equal to a set matching degree threshold, it indicates that the scene corresponding to the background sub-image in the stored image matches the scene corresponding to the background sub-image in the user's target browsing image, and filtering the stored images that match the scene corresponding to the background sub-image in the user's target browsing image.

[0099] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A big data-based intelligent push management system for product content on a live-streaming e-commerce platform, characterized in that: The big data-based intelligent product content push management system for live-streaming e-commerce platforms includes: The system includes a keyword search and filtering module, a product feature parameter segmentation and extraction module, a user browsing image acquisition module, a target browsing image preprocessing module, an analysis server, a face angle detection module, a target element content information extraction module, a feature extraction module, an image overlap coefficient analysis module, an image association information push module, a display terminal and storage database, a product comprehensive push order analysis module, and a background push module. The keyword search and filtering module is used to filter products that match the keywords entered in the search box of the live e-commerce platform, and to number the filtered products in a preset order. The product feature parameter segmentation and extraction module is used to segment the product categories represented by product keywords into various feature parameters, and extract the corresponding features of each feature parameter from the product live broadcast details for each numbered product according to the various feature parameters segmented by the product category. The extracted features of each product constitute the product feature set. The user browsing image acquisition module is used to acquire images that the user browses on the target browsing platform and trigger target browsing image preprocessing instructions based on the user's dwell time on the browsing images. The target browsing image preprocessing module is used to preprocess the user's target browsing image using image segmentation processing technology to obtain the foreground sub-image and background sub-image corresponding to the user's target browsing image; The analysis server is connected to the face angle detection module, the target element content information extraction module, the feature extraction module, the feature analysis module, the image overlap coefficient analysis module, the image association information push module, the display terminal, and the storage database, respectively. The target browsing image preprocessing module is connected to the user browsing image acquisition module and the face angle detection module, respectively. The feature extraction module is connected to the feature analysis module, and the storage database is connected to the feature analysis module. The face angle detection module includes a digital angle meter, which is used to receive the enhanced image number of each person sent by the image processing module. The digital angle meter measures the deflection angle of the face relative to the vertical direction and the pitch angle of the face relative to the horizontal plane in the enhanced image of each person. The module counts the deflection angle of the face relative to the vertical direction and the pitch angle of the face relative to the horizontal plane in the enhanced image of each person. The module then sends the set of deflection angles of the face relative to the vertical direction and the set of pitch angles of the face relative to the horizontal plane in the enhanced image of each person to the analysis server. The target element content information extraction module is used to obtain each target element corresponding to the foreground sub-image in the user's target browsing image and extract the content information of each target element corresponding to the foreground sub-image in the user's target browsing image; The feature extraction module is used to receive the set of image numbers of each person watching the multimedia advertisement sent by the analysis server, extract features from each image of each person watching the multimedia advertisement, extract facial features and body features from each image, and send the extracted facial features and body features from each image to the feature analysis module. The image overlap coefficient analysis module is used to analyze the overlap coefficient between the user's target browsing image and each similar stored image, and to filter the similar stored images with the highest overlap coefficient. The image association information push module is used to extract the associated videos corresponding to each stored image in the platform database and push the most similar stored images and their corresponding associated videos with the highest overlap coefficient. The display terminal is used to receive and display the interest index of people of different genders towards the pushed advertisements sent by the analysis server; The storage database is used to store the preset range of facial deflection angles relative to the vertical direction and the preset range of facial tilt angles relative to the horizontal plane when people watch multimedia advertisements. It also stores the standard facial features and standard body features of men and women, and the total time of pushing male and female gender-type advertisements. The product comprehensive push order analysis module receives the total correlation degree between each product and user preference sent by the feature parameter association matching module, receives the monthly sales of each product sent by the monthly sales extraction module, and receives the positive review rate of each product sent by the positive review rate statistics analysis module. It performs comprehensive push coefficient statistics for each product and arranges each product in descending order of comprehensive push coefficient according to the size of the comprehensive push coefficient. The backend push module is used to push each product, which has been sorted by the product comprehensive push order analysis module, to the live e-commerce platform interface in the sorted order.

2. The intelligent push management system for product content on a live-streaming e-commerce platform based on big data, as described in claim 1, is characterized in that: The big data-based intelligent product content push management system for live-streaming e-commerce platforms also includes: The user historical data collection module is connected to the central control module and is used to collect users' historical transaction data and other information. The product information collection module is connected to the central control module and is used to collect product information sold on the e-commerce platform. The keyword extraction module, connected to the central control module, is used to extract product keywords based on the collected product information; The product classification module, connected to the central control module, is used to classify products based on extracted product keywords.

3. The intelligent push management system for product content on a live-streaming e-commerce platform based on big data, as described in claim 1, is characterized in that: The big data-based intelligent product content push management system for live-streaming e-commerce platforms also includes: Obtain extracted product keywords, features, parameter information, user preference features, and demand information; Filter products based on user needs and keywords; match the features and parameters of different products based on user preferences and keywords. Based on the matching results of demand and keywords, user preference characteristics and product characteristics, and parameter matching results, the relevance between products and users is determined, and products are sorted according to the degree of relevance.

4. A smart product content push management system for a live-streaming e-commerce platform based on big data, as described in claim 1, is characterized in that: The specific steps corresponding to the target feature data extraction module include: Based on the central target element corresponding to the user's target browsing image, extract the corresponding feature data and mark the feature data of the central target element corresponding to the user's target browsing image as, where v is an uncertain number in the feature sequence number of the central target element corresponding to the user's target browsing image; Based on the analysis steps of the central target element corresponding to the user's target browsing image, the central target elements corresponding to each similar stored image in the platform database are screened to obtain the standard feature data corresponding to the central target elements of each similar stored image. The standard feature data corresponding to the central target elements of each similar stored image are marked as, where; The feature data of the central target element corresponding to the user's target browsing image are compared with the standard feature data of the central target element corresponding to each similar stored image to obtain the difference of the feature data of the central target element between the user's target browsing image and each similar stored image.

5. A smart product content push management system for a live-streaming e-commerce platform based on big data, as described in claim 1, is characterized in that: The image segmentation techniques in the target browsing image preprocessing module include: The grayscale transformation process is performed on the user's target browsing image to obtain the grayscale value corresponding to each pixel in the user's target browsing image, and a grayscale histogram of the user's target browsing image is constructed. The maximum and minimum gray values ​​of the user's target browsing image are obtained from the gray-level histogram of the image, and are denoted as and , respectively, and an initial threshold is set. Collect all gray values ​​in the user's target browsing image that are less than the initial threshold and all gray values ​​that are greater than the initial threshold, and calculate the mean value of all gray values ​​less than the initial threshold and the mean value of all gray values ​​greater than the initial threshold in the user's target browsing image, respectively. A new threshold is calculated for the user's target browsing image. All pixels in the user's target browsing image with gray values ​​less than the new threshold are segmented and recombined to obtain the background sub-image corresponding to the user's target browsing image. All pixels in the user's target browsing image with gray values ​​greater than or equal to the new threshold are segmented and recombined to obtain the foreground sub-image corresponding to the user's target browsing image.

6. A smart product content push management system for a live-streaming e-commerce platform based on big data, as described in claim 1, is characterized in that: The specific steps for users to browse the image acquisition module include: Define the browsing platform that the user is currently logged into as the target browsing platform, and obtain the images that the user is browsing on the target browsing platform; Record the duration of time a user spends browsing images on the target browsing platform. If the duration of time a user spends browsing images on the target browsing platform exceeds the set duration threshold, then the image browsed by the user on the target browsing platform is recorded as the user's target browsing image, and the target browsing image preprocessing instruction is triggered.

Images