An intelligent advertising placement processing method and system based on user data

Abstract

The present invention discloses an intelligent advertising delivery processing method and system based on user data, which relates to the technical field of data processing, and includes: delivering advertisements in the advertisement pool based on edge devices, collecting video information during the advertisement playback through the edge devices, analyzing the stored video information and extracting user feature information, integrating the user feature information into a user feature vector, uploading the user feature vector stored in the edge device to the cloud platform, the cloud platform analyzes the user feature vectors uploaded by each edge device to generate user portrait information, based on the user portrait information generated by the cloud platform, delivering corresponding advertisements to user groups with similar user portrait information, collecting feedback data, and dynamically optimizing the advertisements delivered to the edge devices and user groups based on the feedback data. The advantages of the present invention are: reducing the data transmission delay and the cloud computing pressure through the method of cloud-edge collaboration.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and in particular to a method and system for intelligent advertising delivery processing based on user data. Background Art

[0002] Today, with the widespread popularity of the Internet and mobile devices, digital advertising has become the main means for companies to promote their products and services. The traditional advertising model mainly relies on centralized cloud computing platforms to collect and analyze a large amount of user data to achieve accurate advertising positioning and personalized recommendations. However, the existing centralized advertising model faces many challenges and limitations;

[0003] The existing centralized cloud computing advertising delivery model has significant data transmission delays and insufficient computing resources when faced with a large amount of uploaded user data. User data needs to be frequently uploaded to the cloud for processing and analysis, which not only increases the time cost of data transmission, but also leads to a slow response speed for advertising delivery, and is unable to achieve real-time tracking and dynamic adjustment of user behavior. In addition, with the popularization of IoT devices and mobile terminals, the amount of data generated by edge devices has exploded. The existing centralized cloud computing model is difficult to cope with such a large amount of data, resulting in huge pressure and cost for cloud computing resources. Summary of the invention

[0004] In order to solve the above technical problems, a method and system for intelligent advertising delivery processing based on user data are provided. This technical solution solves the problem that the above-mentioned advertising delivery model using centralized cloud computing has significant data transmission delays and insufficient computing resources when facing a large amount of uploaded user data.

[0005] In order to achieve the above purpose, the technical solution adopted by the present invention is:

[0006] A method for intelligent advertising delivery based on user data, comprising:

[0007] Deliver advertisements from the advertising pool based on edge devices;

[0008] Collecting video information during the advertisement playing period through the edge device, wherein the video information is a user screen of the user watching the advertisement during the advertisement playing period, and the video information is stored in the edge device for extracting user feature information;

[0009] Analyze the stored video information and extract user feature information through the computer system of the edge device, and integrate the user feature information into a user feature vector;

[0010] The user characteristic information includes the user's age, gender, facial expression, viewing time and viewing duration, wherein the viewing time and viewing duration are respectively the specific time and duration when the user watches the advertisement played by the edge device;

[0011] Upload the user feature vector stored in the edge device to the cloud platform;

[0012] Analyze the user feature vectors uploaded by each edge device based on the cloud platform to generate user portrait information;

[0013] Based on the user portrait information generated by the cloud platform, corresponding advertisements are delivered to user groups with similar user portrait information;

[0014] The feedback data of the user groups with similar user portrait information on the corresponding advertisements placed by the cloud platform are collected, and the advertisements placed by the edge devices and user groups are dynamically optimized based on the feedback data.

[0015] Preferably, the placing of advertisements in the advertisement pool based on edge devices specifically includes:

[0016] In the initial stage, the edge device randomly selects advertisements from the advertising pool for delivery to obtain video information of users when watching different advertisements. In the subsequent stage, based on the collected feedback data, the edge device dynamically screens and optimizes the advertisements in the advertising pool for precise delivery.

[0017] Preferably, the collecting of video information during the advertisement playing and storing it in the edge device specifically includes:

[0018] During the advertisement playback, video information including the user's physical features, facial expressions and line of sight direction is collected through the edge device, and the currently playing advertisement identification information is synchronously recorded, including the advertisement ID and the playback time. The playback time is the specific time when the advertisement is played. The collected video information is matched with the corresponding advertisement identification information to ensure that each piece of video information is accurately associated with the specific advertisement content played, and the matched video information and the corresponding advertisement identification information are stored in the local storage module of the edge device in an orderly manner.

[0019] Preferably, the analyzing the stored video information and extracting user characteristic information specifically includes:

[0020] Perform data preprocessing on the stored video information, including extracting continuous image frames from the stored video information in time sequence ,in Represents the timestamp for each frame of image Perform resizing and grayscale normalization to generate a standard input format image ;

[0021] Using Haar feature based cascade classifier to classify images Perform face detection and obtain the location of the face area , based on the detected face position , crop the face area and perform posture correction to obtain the aligned face image ;

[0022] Using the pre-trained deep convolutional neural network model VGGNet, face images Perform feature extraction and extract high-dimensional feature vectors through the convolution layer and pooling layer of the deep convolutional neural network model , using the fully connected layer and Softmax activation function, the feature vector Conduct sex and age group Classification, calculate the output probability distribution:

[0023] ;

[0024] ;

[0025] In the formula, is the weight of the classifier, and is the bias of the classifier;

[0026] Use the CNN+LSTM model optimized for expression recognition to transform the face image Taking into account the temporal features of Category probability distribution:

[0027] ;

[0028] In the formula, is the weight of the classifier, is the bias of the classifier;

[0029] From face image Locate the eye area and extract eye feature points , use the eye feature points to calculate the sight direction vector , estimate the sight direction angle through the regression model:

[0030] ;

[0031] According to the sight direction angle Determine whether the user is looking at the screen;

[0032] Generate the user's time at each time stamp based on the sight direction and face detection results. Watch status badge ,in Indicates that you are watching. Indicates that the video has not been watched, and records the timestamp of when the user started watching the video and the timestamp of the end of viewing , watch time calculate:

[0033] ;

[0034] Integrate user feature information including gender, age, facial expression, viewing time and viewing duration into a user feature vector :

[0035] ;

[0036] User feature vector After being associated with the corresponding advertisement identification information, it is stored in the storage module of the edge device.

[0037] Preferably, uploading the user feature vector stored in the edge device to the cloud platform specifically includes:

[0038] Use a secure communication protocol to transfer the feature vector Transmitted from the edge device to the cloud platform, while the user is watching an advertisement, the edge device collects and analyzes video information in real time, and uploads the user feature vector obtained after analysis and integration to the cloud platform. When network conditions are poor, the edge device can cache the user feature vector locally, and upload it to the cloud platform in batches when conditions permit. The cloud platform receives and stores the user feature vector and adds a timestamp and advertising identification index to the uploaded user feature vector.

[0039] Preferably, the cloud platform analyzes the user feature vectors uploaded by each edge device to generate user portrait information, specifically including:

[0040] The cloud platform receives user feature vectors uploaded by each edge device and advertisement identification information associated with each feature vector;

[0041] The cloud platform conducts in-depth analysis of the associated user feature vectors to assess the user's interest in different ads. By analyzing the user's viewing time and facial expressions when watching a specific ad, the cloud platform can obtain the user's reaction data to the ad content, and the platform converts this reaction data into a quantitative interest score.

[0042] The cloud platform uses cluster analysis technology to identify user preference trends. Based on the comprehensive viewing behavior of different users for different types of advertisements, the cloud platform identifies users' viewing preferences in different time periods and for different types of advertisements.

[0043] Based on the user feature vector and the interest scores and preference trends analyzed by the cloud platform, the cloud platform builds detailed user profile information. Each user profile includes the user's basic attributes: gender, age, and viewing tendency;

[0044] The generated user portrait information is stored in the database of the cloud platform. The cloud platform establishes a complete indexing and labeling mechanism, so that the user portrait information can be quickly retrieved and called;

[0045] The cloud platform continuously receives and analyzes new user feature vectors from edge devices and continuously updates user portrait information.

[0046] Preferably, the placing of corresponding advertisements to user groups having similar user portrait information based on the user portrait information generated by the cloud platform specifically includes:

[0047] The cloud platform obtains user portrait information of a large number of user groups by collecting Internet data sources, and matches corresponding advertising content to user groups with similar user portrait information based on the user portrait information generated by the cloud platform. The Internet data sources include: personal information and behavior data of users obtained by social platforms or shopping platforms with the permission of users;

[0048] Determine the advertising content that matches the target user group, and the cloud platform formulates a specific advertising delivery strategy, including determining the delivery channel, delivery time and frequency.

[0049] Preferably, feedback data of the user groups with similar user portrait information on the corresponding advertisements placed by the cloud platform is collected, and the advertisements placed by the edge devices and the Internet groups are dynamically optimized based on the feedback data, specifically including:

[0050] After the advertisement is placed, the cloud platform will collect the user group's feedback data on the advertisement, including the user group's click-through rate, conversion rate, interaction rate and browsing time;

[0051] The cloud platform aggregates and organizes the collected feedback data, classifies them according to user groups and advertising identifiers, and conducts in-depth analysis of feedback data from different user groups;

[0052] Based on the analysis results, the cloud platform comprehensively evaluates the advertising delivery effect. By comparing the performance of each indicator, the cloud platform identifies efficient and inefficient advertising content and delivery strategies, further analyzes the degree of user groups' interest in the advertisements, thereby optimizing the user profile information. According to the evaluation results, the platform formulates dynamic optimization strategies to dynamically optimize the advertisements delivered to edge devices and the Internet population. The dynamic optimization strategies include adjusting the advertising content, optimizing the delivery time and frequency, and reallocating advertising resources.

[0053] Furthermore, an intelligent advertising delivery processing system based on user data is proposed to implement the intelligent advertising delivery processing method based on user data as described above. The processing system consists of two parts: edge devices and a cloud platform, and realizes cloud-edge collaborative work through network connection. The edge devices specifically include:

[0054] A data acquisition module, which is used to acquire video information;

[0055] A data preprocessing module, which adjusts the size and performs grayscale normalization on the acquired video information, analyzes the video images using deep learning algorithms, extracts user feature information and user feature vectors, and associates the user feature vectors with the advertising identification information of the currently played advertisement;

[0056] A local storage module, which is used to store video information, user feature information, user feature vectors, and associated advertising identification information, and manage the read and write operations of the stored data;

[0057] An advertising delivery module, which is used to play the advertisements in the set advertising pool according to the settings, and perform advertising delivery according to the dynamic optimization strategy issued by the cloud platform;

[0058] A security protection module, which is used to encrypt the acquired and stored video information and user feature information, and only store them in the local storage module of the edge device. After the data preprocessing module extracts the user feature information and user feature vectors, the corresponding video information is deleted, the identities of the device and the user are verified, and the running state of the system is monitored to detect and defend against potential security threats and attacks.

[0059] Preferably, the cloud platform specifically includes:

[0060] A data storage module, which is used to store the user feature vectors, advertising identification information, and feedback data uploaded from the edge devices, regularly back up the stored data, provide a data recovery function, and manage the organizational structure and access rights of the data;

[0061] A data analysis module, which uses cluster analysis and association rule mining data analysis algorithms to conduct in-depth analysis of user feature vectors, generate user portrait information, analyze user group feedback data on advertisements, and evaluate the actual effect of advertisement delivery;

[0062] A model training module, wherein the model training module trains a machine learning model for predicting user interests and advertising effects based on the collected user feature vectors and feedback data;

[0063] System management module, which is used to manage the computing resources, storage resources and network resources of the cloud platform, coordinate and schedule various data processing and analysis tasks, optimize the system's resource utilization and processing efficiency, monitor the system's operating status in real time, and detect and handle faults.

[0064] Compared with the prior art, the present invention has the following beneficial effects:

[0065] The present invention proposes an intelligent advertising delivery processing solution based on user data, which collects and preliminarily processes video information of users watching advertisements through edge devices, and uploads the generated user feature vectors to the cloud platform for in-depth analysis and user portrait generation. The method greatly reduces the data transmission volume and delay by uploading only structured user feature vectors. At the same time, the present invention immediately deletes the original video information after local data processing and feature extraction on the edge device, effectively enhancing user privacy protection and reducing the transmission and storage risks of sensitive data. By sinking part of the data processing tasks to the edge device, the burden of the cloud computing platform is reduced, the operating cost is reduced, and the overall processing efficiency of the system is improved, ensuring that it can still operate efficiently and stably in a large-scale user environment. In addition, the cloud platform comprehensively analyzes the uploaded user feature vectors and feedback data, generates a dynamic optimization strategy, and continuously adjusts the advertising content, delivery time and frequency according to real-time feedback to form a closed-loop optimization mechanism, which significantly improves the effect of advertising delivery and user experience. The system adopts modular design and supports multiple communication protocols, has good scalability and adaptability, and can flexibly respond to different application scenarios and network environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0066] Figure 1 This is a flow chart of the intelligent advertising delivery processing method based on user data proposed by the present invention;

[0067] Figure 2 A flow chart of delivering advertisements in an advertisement pool based on edge devices in the present invention;

[0068] Figure 3 A flowchart of collecting video information during advertisement playing in the present invention;

[0069] Figure 4 A flowchart of analyzing stored video information and extracting user feature information in the present invention;

[0070] Figure 5 A flowchart of generating user portrait information in the present invention;

[0071] Figure 6 This is a structural block diagram of the intelligent advertisement delivery processing system based on user data proposed in the present invention. DETAILED DESCRIPTION

[0072] The following description is used to disclose the present invention so that those skilled in the art can implement the present invention. The preferred embodiments described below are only examples, and those skilled in the art may think of other obvious variations.

[0073] Reference Figure 1 As shown, a method for intelligent advertising delivery based on user data includes:

[0074] Placing advertisements in an advertisement pool based on an edge device, wherein the edge device is a smart terminal with a multi-functional module;

[0075] Collecting video information during the advertisement playing period through the edge device, wherein the video information is a user screen of the user watching the advertisement during the advertisement playing period, and the video information is stored in the edge device for extracting user feature information;

[0076] Analyze the stored video information and extract user feature information through the computer system of the edge device, and integrate the user feature information into a user feature vector;

[0077] The user characteristic information includes the user's age, gender, facial expression, viewing time and viewing duration, wherein the viewing time and viewing duration are respectively the specific time and duration when the user watches the advertisement played by the edge device;

[0078] Upload the user feature vector stored in the edge device to the cloud platform;

[0079] Analyze the user feature vectors uploaded by each edge device based on the cloud platform to generate user portrait information;

[0080] Based on the user portrait information generated by the cloud platform, corresponding advertisements are delivered to user groups with similar user portrait information;

[0081] The feedback data of the user groups with similar user portrait information on the corresponding advertisements placed by the cloud platform are collected, and the advertisements placed by the edge devices and user groups are dynamically optimized based on the feedback data.

[0082] This solution uses edge devices to collect user images in real time during the advertising delivery process, and stores the collected video information in the edge devices for further extracting user feature information. The edge devices can be intelligent terminals with camera, display, audio and other functions and computer systems, such as mobile phones, tablets, intelligent station advertising screens, shopping mall advertising screens, elevator advertising screens and other intelligent terminals. As multi-functional intelligent terminals, edge devices use their computer systems to analyze the stored video information, extract user features such as age, gender, facial expressions, viewing time and viewing time, and integrate these feature information into user feature vectors. The generated user feature vectors are then uploaded to the cloud platform. The cloud platform conducts a comprehensive analysis of the feature vectors uploaded by each edge device to form detailed user portrait information. Based on this user portrait information, the system can deliver advertisements that match their interests to user groups with similar portrait features, and collect user feedback data on the delivered advertisements. The feedback data includes interactive information such as the click-through rate and viewing time of the user group. After analyzing these feedback data, the cloud platform dynamically optimizes the advertising content and delivery strategy, thereby improving the accuracy of advertising delivery and user experience.

[0083] Reference Figure 2 As shown, the advertisements in the advertisement pool based on edge devices specifically include:

[0084] In the initial stage, the edge device randomly selects advertisements from the advertising pool for delivery to obtain video information of users when watching different advertisements. In the subsequent stage, based on the collected feedback data, the edge device dynamically screens and optimizes the advertisements in the advertising pool for precise delivery.

[0085] Reference Figure 3 As shown, collecting video information during advertisement playback and storing it in the edge device specifically includes:

[0086] During the advertisement playback, video information including the user's physical features, facial expressions and line of sight direction is collected through the edge device, and the currently playing advertisement identification information is synchronously recorded, including the advertisement ID and the playback time. The playback time is the specific time when the advertisement is played. The collected video information is matched with the corresponding advertisement identification information to ensure that each piece of video information is accurately associated with the specific advertisement content played, and the matched video information and the corresponding advertisement identification information are stored in the local storage module of the edge device in an orderly manner.

[0087] It can be understood that during the advertising playback, in order to ensure that the video information corresponds one-to-one with the advertisement being played, the edge device collects video information including the user's physical features, facial expressions, and line of sight direction, and synchronously records the video information with the identification information of the currently playing advertisement, so as to accurately capture the user's behavioral data on watching advertisements. Specifically, the played advertisement content has a unique advertisement ID and playback time. The edge device matches this identification information with the video information one-to-one to ensure that each piece of user video information collected can be accurately associated with the corresponding advertisement content. In this process, all matched video information and advertisement identification information are stored in sequence in the local storage module of the edge device, thereby providing a comprehensive and coherent data foundation for subsequent user feature analysis and advertising delivery optimization.

[0088] Reference Figure 4 As shown, analyzing the stored video information and extracting user feature information specifically includes:

[0089] Perform data preprocessing on the stored video information, including extracting continuous image frames from the stored video information in time sequence ,in Represents the timestamp for each frame of image Perform resizing and grayscale normalization to generate a standard input format image ;

[0090] Using Haar feature based cascade classifier to classify images Perform face detection and obtain the location of the face area , based on the detected face position , crop the face area and perform posture correction to obtain the aligned face image ;

[0091] Using the pre-trained deep convolutional neural network model VGGNet, face images Perform feature extraction and extract high-dimensional feature vectors through the convolution layer and pooling layer of the deep convolutional neural network model , using the fully connected layer and Softmax activation function, the feature vector Conduct sex and age group Classification, calculate the output probability distribution:

[0092] ;

[0093] ;

[0094] In the formula, is the weight of the classifier, and is the bias of the classifier;

[0095] Use the CNN+LSTM model optimized for expression recognition to transform the face image Taking into account the temporal features of Category probability distribution:

[0096] ;

[0097] In the formula, is the weight of the classifier, is the bias of the classifier;

[0098] From face image Locate the eye area and extract eye feature points , use the eye feature points to calculate the sight direction vector , estimate the sight direction angle through the regression model:

[0099] ;

[0100] According to the sight direction angle Determine whether the user is looking at the screen;

[0101] Generate the user's time at each time stamp based on the sight direction and face detection results. Watch status badge ,in Indicates that you are watching. Indicates that the video has not been watched, and records the timestamp of when the user started watching the video and the timestamp of the end of viewing , watch time calculate:

[0102] ;

[0103] Integrate user feature information including gender, age, facial expression, viewing time and viewing duration into a user feature vector :

[0104] ;

[0105] User feature vector After being associated with the corresponding advertisement identification information, it is stored in the storage module of the edge device.

[0106] In one embodiment of the present invention, uploading the user feature vector stored in the edge device to the cloud platform specifically includes:

[0107] Use a secure communication protocol to transfer the feature vector Transmitted from the edge device to the cloud platform, while the user is watching an advertisement, the edge device collects and analyzes video information in real time, and uploads the user feature vector obtained after analysis and integration to the cloud platform. When network conditions are poor, the edge device can cache the user feature vector locally, and upload it to the cloud platform in batches when conditions permit. The cloud platform receives and stores the user feature vector and adds a timestamp and advertising identification index to the uploaded user feature vector.

[0108] Reference Figure 5 As shown in the figure, the cloud platform analyzes the user feature vectors uploaded by each edge device and generates user portrait information, including:

[0109] The cloud platform conducts in-depth analysis of the associated user feature vectors to assess the user's interest in different ads. By analyzing the user's viewing time and facial expressions when watching a specific ad, the cloud platform can obtain the user's reaction data to the ad content, and the platform converts this reaction data into a quantitative interest score.

[0110] The cloud platform uses cluster analysis technology to identify user preference trends. Based on the comprehensive viewing behavior of different users for different types of advertisements, the cloud platform identifies users' viewing preferences in different time periods and for different types of advertisements.

[0111] Based on the user feature vector and the interest scores and preference trends analyzed by the cloud platform, the cloud platform builds detailed user profile information. Each user profile includes the user's basic attributes: gender, age, and viewing tendency;

[0112] The generated user portrait information is stored in the database of the cloud platform. The cloud platform establishes a complete indexing and labeling mechanism, so that the user portrait information can be quickly retrieved and called;

[0113] The cloud platform continuously receives and analyzes new user feature vectors from edge devices and continuously updates user portrait information.

[0114] It is understandable that the cloud platform combines the user's behavioral data and emotional response data in the process of analyzing the user's feature vector, which can achieve a more accurate user interest assessment. By conducting in-depth analysis of the user's viewing time and physiological data such as facial expressions when watching advertisements, the cloud platform can quantify the user's interest in different advertising content and generate specific interest scores. For example, when a woman between the ages of 25 and 30 watches an advertisement about cosmetics, her facial expression is a positive emotion such as a smile or curiosity, and the viewing time is more than half of the advertisement duration, then her interest score for this advertisement is high. At the same time, the platform further identifies the user's preference trend through cluster analysis technology. For example, the interest score of men between the ages of 40 and 45 in watching alcohol advertisements near holidays is significantly higher than that of daily life. The user's viewing behavior of advertisements, especially the response in different time periods and different types of advertisements, is combined to construct a user's viewing preference model. Based on the above analysis results, the cloud platform can generate detailed user portrait information. User portraits not only cover the basic attributes of users, such as gender and age, but also include viewing tendencies based on interest ratings and viewing preference analysis. These portrait information is stored in the platform's database and managed through a complete indexing and labeling mechanism, so that user portrait information can be quickly retrieved and called. In addition, the cloud platform does not generate user portraits once, but dynamically updates and improves user portraits by continuously receiving new user feature vectors from edge devices. This continuous updating mechanism enables the platform to adjust its portrait information in real time according to changes in user behavior, ensuring the accuracy and timeliness of user portraits.

[0115] In one embodiment of the present invention, based on the user portrait information generated by the cloud platform, placing corresponding advertisements to user groups having similar user portrait information specifically includes:

[0116] The cloud platform obtains user portrait information of a large number of user groups by collecting Internet data sources, and matches corresponding advertising content to user groups with similar user portrait information based on the user portrait information generated by the cloud platform. The Internet data sources include: personal information and behavior data about users obtained by social platforms or shopping platforms with the permission of users;

[0117] Determine the advertising content that matches the target user group, and the cloud platform formulates a specific advertising delivery strategy, including determining the delivery channel, delivery time and frequency.

[0118] In one embodiment of the present invention, feedback data of user groups with similar user portrait information on corresponding advertisements placed by the cloud platform is collected, and the advertisements placed by edge devices and Internet groups are dynamically optimized based on the feedback data, specifically including:

[0119] After the advertisement is placed, the cloud platform will collect the user group's feedback data on the advertisement, including the user group's click-through rate, conversion rate, interaction rate and browsing time;

[0120] The cloud platform aggregates and organizes the collected feedback data, classifies them according to user groups and advertising identifiers, and conducts in-depth analysis of feedback data from different user groups;

[0121] Based on the analysis results, the cloud platform conducts a comprehensive evaluation of the effectiveness of advertising delivery. By comparing the performance of various indicators, the cloud platform identifies efficient and inefficient advertising content and delivery strategies, and further analyzes the user group's interest in advertising, thereby optimizing user portrait information. According to the evaluation results, the platform formulates dynamic optimization strategies to dynamically optimize the advertisements delivered by edge devices and Internet groups. The dynamic optimization strategies include adjusting advertising content, optimizing delivery time and frequency, and reallocating advertising resources.

[0122] In the specific implementation process, after the advertisement is delivered, the cloud platform will automatically collect the user group's feedback data on the advertisement. The feedback data includes key indicators such as the user group's click-through rate, conversion rate, interaction rate and browsing time. The collected feedback data will be classified and managed according to the user group and advertisement identification. The cloud platform will summarize and organize these classified data and conduct in-depth analysis on the feedback data of different user groups. The data analysis module of the cloud platform can dig out the behavioral differences of different user groups when facing the same or different advertisements, and identify the key factors affecting the advertising effect. For example, for a specific advertisement, the platform can analyze the different genders and age groups. Or data such as the interaction rate and click-through rate of interest groups, to find out which groups are more interested in the advertisement and which groups have a lower conversion rate. By comparing the performance of various feedback indicators, the cloud platform can identify the efficient and inefficient parts of advertising. For example, some advertising content may show higher click-through rates and conversion rates in specific groups, but poor results in other groups. Based on these analyses, the platform can further determine whether the advertising content is in line with the user's interest preferences and adjust the advertising delivery strategy. For example, if a certain type of advertisement placed on the Internet receives good feedback from a certain user group, this type of advertisement can be placed on edge devices involving more such user groups.

[0123] For further information, see Figure 6 As shown, based on the same inventive concept as the above-mentioned intelligent advertisement delivery processing method based on user data, this solution proposes an intelligent advertisement delivery processing system based on user data, including:

[0124] Edge devices include:

[0125] A data acquisition module, wherein the data acquisition module is used to acquire video information;

[0126] A data preprocessing module, which performs size adjustment and grayscale normalization processing on the collected video information, analyzes the video image using a deep learning algorithm, extracts user feature information and user feature vectors, and associates the user feature vectors with the identification information of the currently playing advertisement;

[0127] A local storage module, the local storage module is used to store video information, user feature information, user feature vectors and associated advertisement identification information, and manage read and write operations of the stored data;

[0128] An advertisement delivery module, which is used to play advertisements in the advertisement pool according to settings and deliver advertisements according to the dynamic optimization strategy issued by the cloud platform;

[0129] The security protection module is used to encrypt the collected and stored video information and user feature information, and only store them in the local storage module of the edge device. After the data preprocessing module extracts the user feature information and user feature vector, the corresponding video information is deleted, the identity of the device and the user is verified, and the operating status of the system is monitored to detect and defend against potential security threats and attacks.

[0130] The cloud platform specifically includes:

[0131] A data storage module, which is used to store user feature vectors, advertising identification information and feedback data uploaded from edge devices, regularly back up stored data, provide data recovery functions, and manage the organizational structure and access rights of data;

[0132] A data analysis module, which uses cluster analysis and association rule mining data analysis algorithms to conduct in-depth analysis of user feature vectors, generate user portrait information, analyze user group feedback data on advertisements, and evaluate the actual effect of advertisement delivery;

[0133] A model training module, wherein the model training module trains a machine learning model for predicting user interests and advertising effects based on the collected user feature vectors and feedback data;

[0134] System management module, which is used to manage the computing resources, storage resources and network resources of the cloud platform, coordinate and schedule various data processing and analysis tasks, optimize the system's resource utilization and processing efficiency, monitor the system's operating status in real time, and detect and handle faults.

[0135] Specifically, the data acquisition module in the edge device is used to obtain the user's real-time video information and pass the video information to the data preprocessing module. After receiving the video information, the data preprocessing module first performs preprocessing operations such as resizing and grayscale normalization on the video to ensure that the format and quality of the video data are suitable for further analysis. After the preprocessing is completed, the data preprocessing module uses a deep learning algorithm to analyze the video image and extract the user's feature information and feature vectors. At the same time, the system will associate these extracted user feature vectors with the currently playing advertisement identification information to ensure that the user feature information matches the advertisement delivery content. The local storage module is used to store all collected video information, user feature information, user feature vectors, and associated advertisement identification information. This module can effectively manage storage The data reading and writing operations ensure the safe storage and efficient call of data locally. The advertising delivery module automatically selects suitable advertisements from the advertising pool for playback according to the system settings. At the same time, according to the dynamic optimization strategy issued by the cloud platform, the content, playback time and frequency of the advertisements are adjusted in real time to improve the effect of advertising delivery. The security protection module ensures that all collected and stored video information and user feature information are encrypted and only stored in the local storage module of the edge device. After the data preprocessing module successfully extracts the user feature information and feature vectors, the system will automatically delete the corresponding video information to ensure user privacy. At the same time, the security protection module will verify the identity of the device and the user, and monitor the operating status of the system in real time, detect and defend against potential security threats and attacks, and ensure the security and stability of the edge device.

[0136] The data storage module of the cloud platform is mainly responsible for storing user feature vectors, advertising identification information and feedback data of advertising delivery uploaded from edge devices. The module also has the functions of regular backup and data recovery to ensure the security and integrity of the data, and reasonably manage the organizational structure and access rights of the data to prevent unauthorized access. The data analysis module uses a clustering analysis algorithm to deeply analyze user feature vectors, generate user portraits, and evaluate the actual advertising delivery effect based on advertising feedback data. At the same time, the model training module continuously trains and optimizes the machine learning model that predicts user interests and advertising effects based on the collected user feature vectors and feedback data to improve the accuracy of advertising delivery. The system management module is used to manage the computing resources, storage resources and network resources of the cloud platform, ensure the coordinated scheduling of various data processing and analysis tasks, and optimize the system's resource utilization and processing efficiency. At the same time, the module monitors the system's operating status in real time, quickly detects and handles any possible faults, and ensures the stable operation of the cloud platform.

[0137] To sum up, the advantages of the present invention are: the edge device extracts user features through its own computer system, and only transmits key information such as user feature vectors to the cloud platform, which effectively reduces the delay in data transmission and alleviates the computing pressure of the cloud platform when facing huge amounts of data. In addition, the collaborative work of the edge device and the cloud platform not only ensures the accuracy and personalization of advertising, but also improves the advertising effect through dynamic optimization strategies. At the same time, the present invention ensures user privacy and data security through strict data encryption and security protection mechanisms, and ensures the efficient operation of the system and the rational use of resources.

[0138] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The above embodiments and descriptions only describe the principles of the present invention. The present invention may be subject to various changes and improvements without departing from the spirit and scope of the present invention. These changes and improvements fall within the scope of the present invention. The scope of protection claimed by the present invention is defined by the attached claims and their equivalents.

Claims

1. A method for intelligent advertising delivery based on user data, characterized in that: include: Deliver advertisements from the advertising pool based on edge devices; Collecting video information during the advertisement playing period through the edge device, wherein the video information is a user screen of the user watching the advertisement during the advertisement playing period, and the video information is stored in the edge device for extracting user feature information; Analyze the stored video information and extract user feature information through the computer system of the edge device, and integrate the user feature information into a user feature vector; The user characteristic information includes the user's age, gender, facial expression, viewing time and viewing duration, wherein the viewing time and viewing duration are respectively the specific time and duration when the user watches the advertisement played by the edge device; Upload the user feature vector stored in the edge device to the cloud platform; Analyze the user feature vectors uploaded by each edge device based on the cloud platform to generate user portrait information; Based on the user portrait information generated by the cloud platform, corresponding advertisements are delivered to user groups with similar user portrait information; Collect feedback data of the user groups with similar user portrait information on corresponding advertisements placed by the cloud platform, and dynamically optimize the advertisements placed by the edge devices and user groups based on the feedback data; The cloud platform analyzes the user feature vectors uploaded by each edge device to generate user portrait information, specifically including: The cloud platform receives the user feature vectors and the advertising identification information associated with each feature vector uploaded by each edge device, and adds a timestamp and advertising identification index to the uploaded user feature vector; The cloud platform conducts in-depth analysis of the associated user feature vectors to assess the user's interest in different ads. By analyzing the user's viewing time and facial expressions when watching a specific ad, the cloud platform can obtain the user's reaction data to the ad content, and the platform converts this reaction data into a quantitative interest score. The cloud platform uses cluster analysis technology to identify user preference trends. Based on the comprehensive viewing behavior of different users for different types of advertisements, the cloud platform identifies users' viewing preferences in different time periods and for different types of advertisements. Based on the user feature vector and the interest scores and preference trends analyzed by the cloud platform, the cloud platform builds detailed user profile information. Each user profile includes the user's basic attributes: gender, age, and viewing tendency; The generated user portrait information is stored in the database of the cloud platform. The cloud platform establishes a complete indexing and labeling mechanism, so that the user portrait information can be quickly retrieved and called; The cloud platform continuously receives and analyzes new user feature vectors from edge devices and continuously updates user portrait information.

2. The method for intelligent advertising delivery based on user data according to claim 1, characterized in that: The advertisement in the advertisement pool based on the edge device specifically includes: In the initial stage, the edge device randomly selects advertisements from the advertising pool for delivery to obtain video information of users when watching different advertisements. In the subsequent stage, based on the collected feedback data, the edge device dynamically screens and optimizes the advertisements in the advertising pool for precise delivery.

3. The method for intelligent advertising delivery based on user data according to claim 1, characterized in that: The collecting of video information during the advertisement playing and storing it in the edge device specifically includes: During the advertisement playback, video information including the user's physical features, facial expressions and line of sight direction is collected through the edge device, and the currently playing advertisement identification information is synchronously recorded, including the advertisement ID and the playback time. The playback time is the specific time when the advertisement is played. The collected video information is matched with the corresponding advertisement identification information to ensure that each piece of video information is accurately associated with the specific advertisement content played, and the matched video information and the corresponding advertisement identification information are stored in the local storage module of the edge device in an orderly manner.

4. The method for intelligent advertising delivery based on user data according to claim 1, characterized in that: The analyzing the stored video information and extracting user characteristic information specifically includes: Perform data preprocessing on the stored video information, including extracting continuous image frames from the stored video information in time sequence ,in Represents the timestamp for each frame of image Perform resizing and grayscale normalization to generate a standard input format image ; Using Haar feature based cascade classifier to classify images Perform face detection and obtain the location of the face area , based on the detected face position , crop the face area and perform posture correction to obtain the aligned face image ; Using the pre-trained deep convolutional neural network model VGGNet, face images Perform feature extraction and extract high-dimensional feature vectors through the convolution layer and pooling layer of the deep convolutional neural network model , using the fully connected layer and Softmax activation function, the feature vector Conduct sex and age group Classification, calculate the output probability distribution: ； ； In the formula, is the weight of the classifier, and is the bias of the classifier; Use the CNN+LSTM model optimized for expression recognition to transform the face image Taking into account the temporal features of Category probability distribution: ； In the formula, is the weight of the classifier, is the bias of the classifier; From face image Locate the eye area and extract eye feature points , use the eye feature points to calculate the sight direction vector , estimate the sight direction angle through the regression model: ； According to the sight direction angle Determine whether the user is looking at the screen; Generate the user's time at each time stamp based on the sight direction and face detection results. Watch status badge ,in Indicates that you are watching. Indicates that the video has not been watched, and records the timestamp of when the user started watching the video and the timestamp of the end of viewing , watch time calculate: ； Integrate user feature information including gender, age, facial expression, viewing time and viewing duration into a user feature vector : ； User feature vector After being associated with the corresponding advertisement identification information, it is stored in the storage module of the edge device.

5. The method for intelligent advertising delivery based on user data according to claim 1, characterized in that: The uploading of the user feature vector stored in the edge device to the cloud platform specifically includes: A secure communication protocol is used to transmit feature vectors from edge devices to the cloud platform. When users watch advertisements, the edge devices collect and analyze video information in real time, and upload the user feature vectors obtained after analysis and integration to the cloud platform. When network conditions are poor, the edge devices can cache user feature vectors locally and upload them to the cloud platform in batches when conditions permit. The cloud platform receives and stores user feature vectors and adds timestamps and advertising identification indexes to the uploaded user feature vectors.

6. The method for intelligent advertising delivery based on user data according to claim 1, characterized in that: The method of placing corresponding advertisements to user groups having similar user portrait information based on the user portrait information generated by the cloud platform specifically includes: The cloud platform obtains user portrait information of a large number of user groups by collecting Internet data sources, and matches corresponding advertising content to user groups with similar user portrait information based on the user portrait information generated by the cloud platform. The Internet data sources include: personal information and behavior data about users obtained by social platforms or shopping platforms with the permission of users; Determine the advertising content that matches the target user group, and the cloud platform formulates a specific advertising delivery strategy, including determining the delivery channel, delivery time and frequency.

7. The method for intelligent advertising delivery based on user data according to claim 1, characterized in that: Collect feedback data of the user groups with similar user portrait information on the corresponding advertisements placed by the cloud platform, and dynamically optimize the advertisements placed by the edge devices and Internet groups based on the feedback data, specifically including: After the advertisement is placed, the cloud platform will collect the user group's feedback data on the advertisement, including the user group's click-through rate, conversion rate, interaction rate and browsing time; The cloud platform aggregates and organizes the collected feedback data, classifies them according to user groups and advertising identifiers, and conducts in-depth analysis of feedback data from different user groups; Based on the analysis results, the cloud platform conducts a comprehensive evaluation of the effectiveness of advertising delivery. By comparing the performance of various indicators, the cloud platform identifies efficient and inefficient advertising content and delivery strategies, and further analyzes the user group's interest in advertising, thereby optimizing user portrait information. According to the evaluation results, the platform formulates dynamic optimization strategies to dynamically optimize the advertisements delivered by edge devices and Internet groups. The dynamic optimization strategies include adjusting advertising content, optimizing delivery time and frequency, and reallocating advertising resources.

8. An intelligent advertising delivery processing system based on user data, characterized in that: Used to implement the intelligent advertising delivery processing method based on user data as described in any one of claims 1 to 7, the processing system consists of an edge device and a cloud platform, and cloud-edge collaborative work is achieved through a network connection. The edge device specifically includes: A data acquisition module, wherein the data acquisition module is used to acquire video information; A data preprocessing module, which performs size adjustment and grayscale normalization processing on the collected video information, analyzes the video image using a deep learning algorithm, extracts user feature information and user feature vectors, and associates the user feature vectors with the identification information of the currently playing advertisement; A local storage module, the local storage module is used to store video information, user feature information, user feature vectors and associated advertisement identification information, and manage read and write operations of the stored data; An advertisement delivery module, which is used to play advertisements in the advertisement pool according to settings and deliver advertisements according to the dynamic optimization strategy issued by the cloud platform; The security protection module is used to encrypt the collected and stored video information and user feature information, and only store them in the local storage module of the edge device. After the data preprocessing module extracts the user feature information and user feature vector, the corresponding video information is deleted, the identity of the device and the user is verified, and the operating status of the system is monitored to detect and defend against potential security threats and attacks.

9. The intelligent advertisement delivery processing system based on user data according to claim 8, characterized in that: The cloud platform specifically includes: A data storage module, which is used to store user feature vectors, advertising identification information and feedback data uploaded from edge devices, regularly back up stored data, provide data recovery functions, and manage the organizational structure and access rights of data; A data analysis module, which uses cluster analysis and association rule mining data analysis algorithms to conduct in-depth analysis of user feature vectors, generate user portrait information, analyze user group feedback data on advertisements, and evaluate the actual effect of advertisement delivery; A model training module, wherein the model training module trains a machine learning model for predicting user interests and advertising effects based on the collected user feature vectors and feedback data; System management module, which is used to manage the computing resources, storage resources and network resources of the cloud platform, coordinate and schedule various data processing and analysis tasks, optimize the system's resource utilization and processing efficiency, monitor the system's operating status in real time, and detect and handle faults.

Images