Advertisement putting content generation method and system based on large model

Abstract

The invention relates to the technical field of advertisement putting, in particular to an advertisement putting content generation method and system based on a large model. The method comprises the following steps: acquiring industrial data of a city; carrying out industrial correlation modeling on industrial data of the city by utilizing the large model, and generating an industrial-advertisement type mapping matrix; performing advertisement type dynamic priority ranking on the industry-advertisement type mapping matrix to generate an advertisement type recommendation list of the city; acquiring geographic information data of the city; advertising screen labeling is carried out on the geographic information data of the city, and labeled advertising screen position information is obtained; and carrying out candidate site grading output on the marked advertisement screen position information to obtain an advertisement site priority list. According to the method, through combination of the large model, geographic information analysis, real-time crowd feature extraction and closed-loop optimization, the accuracy and dynamism of advertisement putting content generation are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of advertisement delivery, and in particular to a method and system for generating advertisement delivery content based on a large model. Background Art

[0002] The background technology behind advertising content generation methods has evolved through several stages. Initially, companies relied on manually crafted advertising content, relying on designers and copywriters to manually create content. This lacked personalization and automation, resulting in low production efficiency. With the development of information technology, particularly the rise of big data and machine learning, the advertising industry has gradually begun to adopt data analysis to optimize advertising effectiveness. In the 21st century, breakthroughs in deep learning technology have ushered in a new era for advertising content generation. Technologies such as natural language processing (NLP) and generative adversarial networks (GANs) have begun to be widely used for automated advertising content generation. In recent years, large-scale model-based advertising content generation methods have not only been applied to copywriting but have also achieved breakthroughs in areas such as visual design and video ad creation. However, traditional advertising methods often overlook the impact of geographic location on advertising effectiveness, and the selection of advertising venues often relies on manual experience. Furthermore, many traditional advertising systems fail to dynamically analyze real-time demographics, resulting in a mismatch between advertising content and audience needs, and consequently, low accuracy and dynamism in advertising content generation. Summary of the Invention

[0003] Based on this, it is necessary to provide a method and system for generating advertising content based on a large model to solve at least one of the above technical problems.

[0004] To achieve the above-mentioned purpose, a method for generating advertisement content based on a large model is provided, the method comprising the following steps:

[0005] Step S1: Obtain industry data for the city; use the big model to model the industry correlation of the industry data for the city, and generate an industry-advertising type mapping matrix; dynamically prioritize the advertising types in the industry-advertising type mapping matrix, and generate a recommended list of advertising types for the city;

[0006] Step S2: Obtain geographic information data of the city; annotate the advertising screens in the geographic information data of the city to obtain the location information of the annotated advertising screens; output the candidate sites according to the annotated advertising screen location information to obtain a priority list of advertising sites;

[0007] Step S3: Extract real-time crowd features based on geographic information data of the city, and generate an attention heat map for the labeled advertisement screen location information according to the real-time crowd features, to obtain a screen area attention distribution heat map; inject physical constraint conditions into the screen area attention distribution heat map; construct a dynamic position optimizer according to the injected conditions to generate advertisement element coordinate suggestion data;

[0008] Step S4: Combine advertisement content based on the advertisement type recommendation list, the site priority list, and the advertisement element coordinate suggestion to generate an advertisement image-text combination scheme; perform closed-loop effect evaluation and content weight adjustment on the advertisement image-text combination scheme to execute advertisement content generation optimization tasks.

[0009] The present application can accurately place advertisement content at the right time and place through comprehensive analysis based on industrial data, advertisement type recommendation, and real-time crowd features, improving the reach and conversion rate of advertisements. Dynamic optimization of advertisement screen locations and priority sorting of advertisement types make advertisement placement more accurate and reasonable, avoiding resource waste. Through closed-loop effect evaluation and content weight adjustment, the advertisement strategy can be adjusted in real time to ensure that the advertisement placement effect reaches the best state. Since the placement of advertisements is more personalized and meets the needs of users, the content and presentation of advertisements will be more attractive, thereby enhancing user acceptance and participation. By using a large model to model the correlation of industrial data, deep rules and patterns can be extracted from a large amount of data, making advertisement placement more scientific and intelligent. Therefore, the present application improves the accuracy and dynamics of advertisement content generation by combining large models, geographic information analysis, real-time crowd feature extraction, and closed-loop optimization.

[0010] Preferably, step S1 includes the following steps:

[0011] Step S11: Obtain industrial data of the city;

[0012] Step S12: Use a large model to extract industrial keywords and consumption scenarios from the industrial data of the city, and model the industrial correlation of the industrial keywords and consumption scenarios to generate an industry-advertisement type mapping matrix;

[0013] Step S13: Perform seasonal demand analysis on the industry-advertisement type mapping matrix to generate seasonal demand analysis data;

[0014] Step S14: Perform dynamic priority sorting of advertisement types on the industry-advertisement type mapping matrix through the seasonal demand analysis data to generate an advertisement type recommendation list for the city.

[0015] The present application models the association between industry keywords and consumer scenarios based on the generation of an industry-advertising type mapping matrix. This precise association model can help better match the characteristics of the industry and the needs of the target audience for advertising, avoiding mismatched advertising content. Seasonal demand analysis provides data support for dynamic priority ranking of advertising types, ensuring real-time adjustment of advertising content during different seasons and demand fluctuations, which maximizes the effectiveness of advertising and avoids wasting advertising budgets during low demand periods. By considering the impact of seasonal demand, advertising placement is not only based on static data analysis, but also keeps up with market dynamics. The advertising type recommendation list adjusts with seasonal changes, making the advertising content more consistent with the audience's needs and improving user engagement and advertising effectiveness. The analysis of industry data using large models, combined with seasonal demand forecasting, makes advertising placement more scientific and predictable. The improvement of advertising effectiveness not only depends on the current market situation, but also can predict future trends based on historical data to make adjustments in advance.

[0016] Preferably, step S12 comprises the following steps:

[0017] Step S121: Collecting consumer scenario data from the industry data of the city where the user is located to obtain the consumer scenario data of the city where the user is located;

[0018] Step S122: Extracting industry keywords from the industry data of the city where the user is located based on the consumer scenario data of the city where the user is located to obtain the core keywords of the industry;

[0019] Step S123: Clustering the core keywords of the industry to generate the main industry categories of the city where the user is located, and performing descriptive analysis on the consumer scenario data based on the main industry categories of the city where the user is located through text analysis technology to generate core features of the consumer scenario and consumer demand data;

[0020] Step S124: Using a large model to calculate the correlation of the core features of the consumer scenario and the consumer demand data to generate industry-consumer scenario association data; and performing data matrixization on the industry-consumer scenario association data to generate an industry-advertising type mapping matrix.

[0021] The present application can help to accurately identify and describe the main industrial characteristics of the city and the corresponding consumer demand through the extraction of industrial core keywords and the analysis of core features of consumer scenarios. Through text analysis technology, the relationship between industry and consumer scenarios is deeply analyzed, thereby providing more accurate basic data for advertisement placement. The generation of industry-consumer scenario association data is related by a large model, making the association between industry and consumer scenario more scientific. This deep analysis helps advertisers to determine which types of advertisements are more suitable for matching with which consumer scenarios, thereby improving the relevance and targeting of advertisements. The industry-advertising type mapping matrix generated by data matrixing makes the process of advertisement placement more intelligent and automated. This matrix can serve as the core framework of advertisement placement strategy, helping advertisers to optimize placement according to industrial characteristics, consumer demand and scenario relevance. The core features of industry and consumer scenario and consumer demand data provide a clear basis for the priority ranking and placement location selection of advertising types. By optimizing the precise matching of advertisements, the content of the advertisements can better meet the actual needs of the target users, thereby improving the participation and conversion rate of the advertisements. Descriptive analysis and cluster analysis of consumer scenario data can help advertisers better understand the behavior and preferences of consumers in the city. By deeply mining the core features of consumer scenarios, advertisers can design advertising content that better meets the needs of the audience, thereby improving the personalization and appeal of the advertisements.

[0022] Preferably, step S2 comprises the following steps:

[0023] Step S21: obtaining geographic information data of the city;

[0024] Step S22: labeling the advertising screen location information of the geographic information data of the city, and performing scene semantic modeling on the geographic information data of the city based on the labeled advertising screen location information using a CV model to generate a scene label system;

[0025] Step S23: performing multi-dimensional matching calculation on the advertising type recommendation list of the city and the scene label system to exclude conflicting scenes and generate advertising scene matching data;

[0026] Step S24: performing candidate site hierarchical output on the labeled advertising screen location information through the advertising scene matching data to obtain an advertising site priority list.

[0027] The application can accurately associate geographic information data with scene characteristics by advertising screen location information labeling and CV model scene semantic modeling, providing a more detailed and intelligent spatial understanding for advertising placement. This modeling can ensure that the location selection and display content of the advertising screen are highly consistent with the surrounding environment, thereby improving the visual appeal and audience attention of the advertisement. Multi-dimensional matching calculation matches the advertising type recommendation list with the scene tag system, which can automatically exclude scenes that do not match or are not suitable for the advertising type, effectively avoiding mismatch of advertising content and ensuring the accuracy and efficiency of advertising placement. By generating an advertising site priority list through hierarchical output of the labeled advertising screen location information, the advertiser can allocate advertising resources reasonably according to the importance of the site, exposure, and matching degree of the advertising type. The generation of this priority list not only improves the advertising effect, but also optimizes the allocation of advertising budget. The combination of the scene tag system and the advertising type can more accurately design advertising content according to different geographic regions and scene characteristics. For example, promotional advertisements are displayed in business districts, and transportation tool advertisements are placed in transportation hubs, etc. The compatibility of the advertisement with the environment improves the acceptance and conversion rate of the advertisement. Through multi-dimensional matching, personalized advertisements can be designed according to the consumer demand and scene characteristics of different geographic regions. Scene-based advertising display not only improves the interactivity of the advertisement, but also improves the relevance and user engagement of the advertisement.

[0028] Preferably, based on the labeled advertising screen location information, the geographic information data of the city is modeled using a CV model to include:

[0029] Extracting the planning features of the geographic information data of the city, and dividing the geographic information data of the city into spatial region levels to generate spatial region level data;

[0030] Based on the labeled advertising screen location information, the spatial region level data is positioned in the space where the advertising screen is located to generate advertising screen space positioning data; the image of the advertising screen space positioning data is retrieved by the regional camera, and the image is detected by the CV model to generate advertising scene visual features;

[0031] Extracting the scene elements of the advertising scene visual features, and using the regional generation network to model the scene semantics of the image of the advertising screen space positioning data to generate advertising screen environment scene modeling data;

[0032] Semantically associating and labeling the advertising screen environment scene modeling data to generate environment label association data; using the environment label association data and the spatial region level data to construct a scene tag system for the advertising screen environment scene modeling data to obtain the scene tag system.

[0033] The present application can accurately parse the location of the advertising screen and the surrounding geographical environment by generating spatial area level data and advertising screen spatial positioning data. Through planning feature extraction and spatial level division of urban geographical information, the advertiser can better understand the specific location of the advertising screen and its relative position in the urban space, thereby providing a clear framework for advertising placement selection. The extraction of advertising scene visual features combined with the target detection model can identify specific environmental features (such as traffic flow, building type, pedestrian density, etc.) around the advertising screen, and further provide detailed data on the advertising placement effect in this environment. This process can ensure that the advertising content is highly consistent with the surrounding environment, improving the relevance of the advertisement and the attention of the audience. Through scene element extraction and area generation network, the environmental features of the advertising screen are accurately converted into data that can be used for decision-making, further establishing advertising screen environment scene modeling data. This modeling not only considers the geographical location, but also covers the visual features and actual use of the scene, making advertising placement decisions more comprehensive and accurate. Using semantic association labels, specific semantic labels (such as business district, residential area, transportation hub, etc.) can be provided for advertising scene modeling data. Such labels not only help advertisers understand the characteristics of different scenes, but also enable targeted optimization of advertising content and placement strategies. Through the label system, advertisers can identify the special needs of the environment where each advertising screen is located, further enhancing the personalization and customization of advertising. The construction of the scene label system provides a powerful tool for content adaptation of advertising placement in different scenarios. For example, promotional advertisements in business districts and cultural advertisements in educational areas. Such precise matching can effectively improve the acceptance and effectiveness of advertising, avoiding ineffective exposure.

[0034] Preferably, in step S3, real-time crowd features are extracted based on geographical information data of the city, and attention heat map generation is performed on the labeled advertising screen location information according to the real-time crowd features, including:

[0035] Based on the geographical information data of the city, the regional camera of the region is called;

[0036] The regional camera of the region is subjected to environment image screening to obtain regional environment images; human body contours are identified from the regional environment images, and crowd attribute analysis is performed on the human body contours to generate real-time crowd features;

[0037] According to the real-time crowd features, a visual range analysis is performed on the labeled advertising screen location information to generate visual range data; an eye movement tracking prediction is performed on the visual range data using a pre-trained attention prediction model to generate eye movement tracking prediction data;

[0038] The eye movement tracking prediction data is converted into an attention heat map to generate a screen area attention distribution heat map.

[0039] The present application can efficiently obtain crowd feature data by calling regional cameras in the region and combining real-time image processing technology. The recognition of human body contours and crowd attribute analysis enable the system to obtain detailed information about the crowd in real time, such as the number of people, age range, gender, and behavior habits, thereby providing advertisers with more accurate audience analysis. Real-time crowd feature data can directly affect the accurate matching of advertising content. Through analysis of the location information of the advertising screen and the behavior of the crowd, it can be ensured that the content of the advertisement is highly consistent with the needs of the surrounding audience. For example, advertisements can be individually adjusted according to the age, interests, or consumption habits of the crowd to improve the acceptance and effectiveness of the advertisements. Based on the visual range analysis, advertisers can more accurately understand the exposure effect of different advertising screens in a specific area. This analysis can reveal the distance, angle, and visual range between the advertising screen and the target audience, making the selection of advertising screen positions and the presentation of advertising content more consistent with the visual distance of actual audiences, thereby avoiding waste of advertising. Using an eye tracking prediction model, advertisers can predict the area where the audience's gaze is concentrated in a specific location and environment. Eye tracking can help identify the focus of the audience's attention, revealing the visual path and areas of concentrated attention of the audience when watching advertisements, providing data support for the design of advertising content to ensure that key elements of the advertisement are located in the area most likely to be focused on by the audience. The attention heat map converts eye tracking data into a visual image, allowing advertisers to intuitively understand the distribution of the audience's attention on the advertising screen. Through this heat map, advertisers can see the areas of focus of the audience's attention and optimize the layout and visual design of the advertisement to maximize appeal and audience engagement.

[0040] Preferably, the physical constraint condition injection on the screen area attention distribution heat map in step S3 comprises:

[0041] The labeled advertising screen location information is subjected to advertising screen parameter information collection to obtain advertising screen information parameters; the advertising screen information parameters are subjected to screen resolution and pixel density analysis to obtain the advertising screen resolution and pixel density;

[0042] Based on the advertising screen resolution and pixel density, an initial advertising boundary range is set, and the set initial advertising boundary range is subjected to visual range calculation to obtain the advertising screen visual range;

[0043] The screen area attention distribution heat map is subjected to minimum font size limit physical constraint on the advertising screen visual range to obtain a content layout safety boundary constraint condition, wherein the calculation formula of the minimum font size limit physical constraint is as follows:

[0044]

[0045] In the formula, B(x) is the content layout safety boundary constraint, Aunsafe (x) is the part of the area of the advertising content that exceeds the visible area of the screen, A total is the total area of the screen, and x is an input parameter.

[0046] The present application can provide accurate display range for advertisers by analyzing the resolution and pixel density of the advertising screen, ensuring that the advertising content does not exceed the actual visible area of the screen. This analysis can avoid the loss of part of the information due to the exceeding of the visible range of the screen, thereby optimizing the overall effect of the advertisement. Based on the screen resolution and pixel density of the advertising screen, the preliminary advertising boundary range can be set under actual display conditions, which helps advertisers avoid excessive or insufficient layout problems when displaying content, ensuring the consistency and standardization of advertisements on different devices. The range of sight and the minimum font size limit ensure that the advertising content is not only clearly visible within the visual range, but also meets the actual viewing distance and screen size requirements. By accurately calculating the minimum font size of the advertising content, the advertising content will not affect the reading experience of the audience due to the small font size, improving the readability and attractiveness of the advertisement. The injection of physical constraint conditions, especially the minimum font size limit, can effectively avoid small fonts and ensure that the audience can clearly see the advertising information at a normal viewing distance. By ensuring the readability and visibility of the advertisement, the user experience of the advertisement can be significantly improved, enhancing the audience's attention to the advertising content. By calculating the parameters and layout safety boundaries of the advertising screen, the design of the advertising content between different advertising screens can reach consistent standards and specifications. Whether on screens of different sizes or in different display environments, the advertising content can follow the same set of physical and design rules, thereby improving the uniformity of advertising placement.

[0047] Preferably, the dynamic position optimizer construction according to the injection conditions in step S3 includes:

[0048] Based on the content layout safety boundary constraint condition, a multi-objective optimization parameter is constructed, wherein the multi-objective optimization parameter includes a minimum exposure rate and a minimum information interference value, and the calculation formula of the minimum exposure rate is as follows:

[0049]

[0050] In the formula, E(x) is the minimum exposure rate, A i is the display area of the advertisement in the i-th region, A total is the total area of the screen, V x is the exposure of the advertisement in the i-th region under different viewing angles, n is the number of advertising regions, and x is an input parameter.

[0051] The calculation formula of the minimum information interference value is as follows:

[0052]

[0053] where D(x) is the minimum information interference value, L j (x) is the advertisement brightness under the environmental lighting condition j, C j (x) is the information visible contrast under the environmental lighting condition j, L max and C max are the maximum brightness and the maximum contrast of the screen under the given condition respectively, m is the number of different environmental conditions, and x is the input parameter;

[0054] A multi-objective optimization function is constructed according to the minimum exposure rate, the minimum information interference value, and the content layout safety boundary constraint condition, and the multi-objective optimization function is obtained, wherein the formula of the multi-objective optimization function is as follows:

[0055] f(x) = w1·E(x) + w2·D(x) + w3·B(x);

[0056] where f(x) is the multi-objective optimization function, E(x) is the minimum exposure rate, D(x) is the minimum information interference value, B(x) is the content layout safety boundary constraint, w1 is the minimum exposure rate weight, w2 is the minimum information interference weight, w3 is the layout safety boundary constraint weight, and x is the input parameter;

[0057] The multi-objective optimization function is used to optimize the dynamic position of the advertisement element based on the screen area attention distribution heat map, and advertisement element coordinate suggestion data is generated.

[0058] The present application optimizes the display position of advertising elements through minimum exposure rate calculation, ensuring that advertisements have sufficient exposure in various regions. This process maximizes the visibility of advertisements, enhances their dissemination effect, and ensures that more viewers can see the advertising content, thereby increasing the effectiveness and influence of the advertisement. The calculation of the minimum information interference value helps analyze the brightness and contrast of advertisements under different environmental conditions, ensuring that advertisements are clearly visible under various lighting and contrast conditions. By reducing visual interference with the surrounding environment, the effectiveness of the advertising content is improved, and the viewer's attention is focused on the advertising content itself rather than external interference factors. By considering the content layout safety boundary constraints, it is ensured that the advertisement does not exceed the visible range of the screen, avoiding the impact of layout problems on the display effect of the advertisement. This constraint ensures that the advertisement can always be displayed correctly and clearly in all devices and environments, improving the user's visual experience and the attractiveness of the advertisement. The construction of the dynamic position optimizer allows the advertising content to be adapted to different viewing angles and exposure conditions, ensuring that the advertising content always achieves the best display effect under different environments and screen angles. Through this optimization, the advertisement can dynamically adapt to various viewing conditions, enhancing its performance in different scenarios. Through the multi-objective optimization function, the exposure of the advertisement, information interference, and layout safety are considered comprehensively, allowing the generation of precise advertising element position recommendations. This precise layout not only optimizes the display effect of the advertisement, but also avoids ineffective or overly crowded advertising element arrangements, ensuring that the advertising content maximizes its value.

[0059] Preferably, step S4 comprises the following steps:

[0060] Step S41: based on the advertisement type recommendation list, the site priority list, and the advertising element coordinate suggestion, the advertising content combination is recommended, and the advertising image-text combination scheme is generated;

[0061] Step S42: real-time click-through rate testing is performed on the advertising image-text combination scheme, and the real-time click-through rate testing result is obtained; based on the real-time click-through rate testing result, cross-scene style migration is performed on the advertising image-text combination scheme, and advertising style migration data is generated;

[0062] Step S43: using the advertising style migration data, the advertising image-text combination scheme is evaluated in a closed loop, and advertising placement effect evaluation data is generated; and based on the advertising placement effect evaluation data, the content weight of the advertising image-text combination scheme is adjusted to perform advertising placement content generation optimization.

[0063] The present application can optimize the advertisement graphic combination scheme more accurately for different venues and target user groups by combining the advertisement type recommendation list, the venue priority list, and the advertisement element coordinate suggestion. This process ensures a high degree of fit between the advertisement content and the audience's needs, improving the attractiveness and relevance of the advertisement. Real-time click-through rate testing provides immediate feedback for advertisement placement, allowing advertisers to adjust the advertisement content in a timely manner based on the audience's click behavior. This process makes the advertisement placement strategy more flexible and dynamic, effectively responding to changes in user preferences and behavior. Through cross-scene style transfer of the advertisement graphic combination scheme, the advertisement can adapt to the display requirements in different scenes. Whether on outdoor screens, mobile devices, or other types of advertisement carriers, the advertisement content can be adjusted adaptively according to environmental changes, thereby improving the universality and effectiveness of the advertisement. Advertisement style transfer can optimize the visual style of the advertisement according to different display environments and audience needs. Through style transfer, the advertisement content can adjust the layout, color, and style of the graphic and text while retaining the core information, making it more attractive to the target audience. Closed-loop effect evaluation provides precise feedback on the effectiveness of the advertisement by evaluating its actual performance after placement. This evaluation helps advertisers understand the actual conversion effect of the advertisement, including click-through rate, interaction, brand awareness, and other key indicators, further optimizing the effectiveness of advertisement placement.

[0064] In the present specification, a system for generating advertisement content based on a large model is provided for performing the above-mentioned method for generating advertisement content based on a large model, the system for generating advertisement content based on a large model comprising:

[0065] An advertisement type recommendation module is configured to obtain industrial data of a city; use a large model to model the industrial data of the city to generate an industry-advertisement type mapping matrix; and perform dynamic priority sorting on the industry-advertisement type mapping matrix to generate an advertisement type recommendation list for the city.

[0066] An advertisement venue recommendation module is configured to obtain geographic information data of a city; perform advertisement screen labeling on the geographic information data of the city to obtain labeled advertisement screen position information; and perform candidate venue hierarchical output on the labeled advertisement screen position information to obtain an advertisement venue priority list.

[0067] An advertisement content recommendation module is configured to extract real-time crowd features based on geographic information data of a city, and generate an attention heat map for the labeled advertisement screen position information based on the real-time crowd features to obtain a screen area attention distribution heat map; perform physical constraint condition injection on the screen area attention distribution heat map; construct a dynamic position optimizer based on the injected conditions to generate advertisement element coordinate suggestion data.

[0068] The advertising generation module is used to combine advertising content based on the advertising type recommendation list, venue priority list and advertising element coordinate suggestions, and generate an advertising image and text combination plan; conduct closed-loop effect evaluation and content weight adjustment on the advertising image and text combination plan to perform advertising content generation optimization operations.

[0069] The present invention provides the following benefits: the ad type recommendation module utilizes a large model to model industry relevance, accurately generating an industry-advertising type mapping matrix based on the city's industry data. This process ensures that advertising content closely matches industry demand, improving advertising relevance and effectiveness. Dynamic prioritization of ad types allows for real-time adjustments based on seasonality or market demand, thereby enhancing advertising effectiveness. The ad venue recommendation module obtains city geographic information and labels advertising screens to generate a venue priority list, ensuring that advertising is placed in areas with the greatest potential and exposure opportunities. This module provides advertisers with multiple venue options by grading candidate sites, helping them select the optimal placement and maximize advertising exposure and impact. The ad content recommendation module, through real-time crowd feature extraction and attention heat map generation, provides precise ad placement recommendations based on the audience attention level of the ad screen's location. By combining crowd feature analysis with viewing distance analysis, the placement and method of ad content can be further optimized to ensure that ads maximize the attention of the target audience. The ad generation module automatically generates the optimal ad image and text combination based on the ad type recommendation list, venue priority list, and ad element coordinate recommendations. Through closed-loop effect evaluation and content weight adjustment, the system can dynamically adjust the content according to the real-time feedback data of the advertisement, thereby continuously optimizing the effect of advertising delivery. Through real-time effect evaluation of the advertising image and text combination scheme, the advertising content can be adjusted in a timely manner according to data such as user click behavior, advertising exposure, and interaction rate. This closed-loop optimization process ensures that the advertising content always maintains efficient and accurate performance, avoiding invalid advertising and waste of resources. Therefore, the present invention improves the accuracy and dynamism of advertising content generation by combining large models, geographic information analysis, real-time crowd feature extraction and closed-loop optimization. BRIEF DESCRIPTION OF THE DRAWINGS

[0070] Figure 1 A flowchart of steps for a method for generating advertisement content based on a large model;

[0071] Figure 2 for Figure 1 Detailed implementation steps of step S2 in FIG.

[0072] Figure 3 for Figure 1 Detailed implementation steps of step S4 in FIG.

[0073] The objectives, functional characteristics and advantages of the present application will be further described with reference to the embodiments and in conjunction with the accompanying drawings. DETAILED DESCRIPTION

[0074] The technical method of the present application will be described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all the other embodiments obtained by those skilled in the art without creative work shall fall within the scope of protection of the present application.

[0075] In addition, the accompanying drawings are only schematic illustrations of the present application and are not necessarily drawn to scale. Identical reference numerals in the drawings represent identical or similar parts, and thus repeated descriptions thereof will be omitted. Some of the block diagrams shown in the drawings are functional entities, which do not necessarily have to correspond to physically or logically independent entities. The functional entities can be implemented in the form of software, or in one or more hardware modules or integrated circuits, or in different network and / or processor methods and / or microcontroller methods.

[0076] It should be understood that although the terms "first", "second" and the like can be used herein to describe various elements, these elements should not be limited by these terms. These terms are only used to distinguish one element from another. For example, without departing from the scope of the exemplary embodiments, a first element can be referred to as a second element, and similarly a second element can be referred to as a first element, and the term "and / or" as used herein includes any and all combinations of one or more of the associated associated items listed.

[0077] To achieve the above-mentioned object, please refer to Figures 1 to 3 A method for generating advertising content based on a large model, the method comprising the following steps:

[0078] Step S1: Obtain the industrial data of the city; use a large model to model the industrial correlation degree of the industrial data of the city, generate an industry-advertising type mapping matrix; perform dynamic priority sorting on the industry-advertising type mapping matrix to generate an advertising type recommendation list of the city;

[0079] Step S2: Obtain the geographic information data of the city; perform advertising screen annotation on the geographic information data of the city to obtain annotated advertising screen location information; perform candidate site hierarchical output on the annotated advertising screen location information to obtain an advertising site priority list;

[0080] Step S3: Extract real-time crowd features based on geographic information data of the city, and generate an attention heat map for the labeled advertising screen location information according to the real-time crowd features, to obtain a screen area attention distribution heat map; inject physical constraint conditions into the screen area attention distribution heat map; construct a dynamic position optimizer according to the injected conditions to generate advertising element coordinate suggestion data;

[0081] Step S4: Combine advertising content based on the advertising type recommendation list, the site priority list, and the advertising element coordinate suggestion to generate an advertising image-text combination scheme; perform closed-loop effect evaluation and content weight adjustment on the advertising image-text combination scheme to execute advertising content generation optimization tasks.

[0082] The present application can accurately place advertising content at the right time and place through comprehensive analysis based on industry data, advertising type recommendations, and real-time crowd characteristics, improving the reach and conversion rates of advertisements. Dynamic optimization of advertising screen locations and priority sorting of advertising types make advertising placement more accurate and reasonable, optimizing advertising budgets and avoiding resource waste. Through closed-loop effect evaluation and content weight adjustment, advertising strategies can be adjusted in real time to ensure optimal advertising placement. As advertising placement is more personalized and meets user needs, advertising content and presentation will be more attractive, enhancing user acceptance and engagement. By using a large model to model the correlation of industry data, deep rules and patterns can be extracted from a large amount of data, making advertising placement more scientific and intelligent. Therefore, the present application improves the accuracy and dynamics of advertising content generation by combining large models, geographic information analysis, real-time crowd feature extraction, and closed-loop optimization.

[0083] In the embodiment of the present application, as shown in Figure 1 The method for generating advertising content based on a large model includes the following steps:

[0084] Step S1: Obtain industry data for the city; use a large model to model the industry correlation of the industry data for the city, generate an industry-advertising type mapping matrix, and sort the dynamic priority of the advertising types in the industry-advertising type mapping matrix to generate an advertising type recommendation list for the city;

[0085] In the embodiments of the present application, various industrial data of the city is obtained through city economy, industrial park, industry association and other ways. These data can include industry classification, industry scale, main enterprises, development trend, etc. The collected data is processed to remove noise data, and the data from different sources is standardized to make the format consistent, which is convenient for further analysis. A large pre-trained model suitable for analyzing industrial correlation (such as deep learning-based graph neural network, BERT, GPT, etc.) is selected and fine-tuned according to the industrial data. The cleaned industrial data is input to let the large model learn the correlation between different industries, for example, whether the growth of a certain industry will promote the development of related industries (such as advertising industry, manufacturing industry, etc.). The model outputs the correlation between industries, and quantifies the correlation of each pair of industries into a numerical value. Define the types of advertisements that can be used for advertising, such as video advertisements, social media advertisements, search engine advertisements, etc. Combine the results of industrial correlation to define the types of advertisements that are highly related to each industry. Based on the relationship between industries and advertisement types, generate an "industry-advertisement type mapping matrix", where each element represents the correlation between a certain industry and a certain advertisement type. According to the effect evaluation indicators of the advertisement types (such as click-through rate, conversion rate, market demand, etc.), the priority of the advertisement types is dynamically sorted. Use dynamic sorting algorithms (such as weighted sorting, machine learning sorting algorithms) to combine the real-time changes of industries and advertisement types to generate advertisement type priority adjusted according to the characteristics of the city and the current trends. According to the mapping matrix of industries and advertisement types, and the dynamic priority sorting of advertisement types, use collaborative filtering or content recommendation algorithms to generate an advertisement type recommendation list.

[0086] Step S2: Obtain geographic information data of the city; perform advertisement screen labeling on the geographic information data of the city to obtain labeled advertisement screen location information; perform candidate site hierarchical output on the labeled advertisement screen location information to obtain an advertisement site priority list;

[0087] In the embodiments of the present application, detailed geographic information data of the city where the user is located is obtained, including but not limited to city map, geographic coordinates, building information, road network, population density, traffic flow, etc. The data collected from different sources is uniformly formatted and standardized processed to ensure the consistency and availability of the data. The location data of all advertising screens (including LED screens, digital billboards, bus station advertising screens, etc.) in the city where the user is located is collected. The advertising screens can be obtained through the database provided by the city management system or the existing advertising platform. Detailed labeling is performed according to the physical location, size, shape, display characteristics, etc. of the advertising screens. The location of each advertising screen can be described by latitude and longitude coordinates, surrounding geographical environment (such as transportation hub, commercial area, etc.) and other related information. Define labeling rules to ensure the accuracy of the advertising screen data. For example, use a standard coordinate system to accurately calibrate the advertising screen, and add environmental information related to the advertising screen (such as whether it is located in a bustling commercial area, exposure during peak traffic hours, etc.). Establish a data table or database to organize the labeling information of each advertising screen (such as advertising screen number, latitude and longitude, size, display range, surrounding traffic, environment, etc.) into a structured data format. According to the labeled information, the location distribution of the advertising screens is analyzed to identify the density, distribution characteristics of the advertising screens in different regions, and the distance relationship with important geographical points (such as shopping centers, transportation hubs). Define the selection criteria for candidate sites, such as traffic flow, population density, commercial value, and time period exposure rate. According to the above criteria, a multi-dimensional weighted scoring system (such as linear weighting method or scoring model based on machine learning) is used to score the location information of all advertising screens, and they are divided into different priority categories. For example, it can be divided into "high priority", "medium priority" and "low priority". Output the priority list of advertising sites according to the scoring results for advertising placement decision-making. According to the scoring and grading results of the advertising screens, the priority of the candidate sites is sorted. Different advertising needs (such as brand exposure, promotion activities, etc.) can be considered to assign appropriate priorities to different sites.

[0088] Step S3: extracting real-time crowd features based on geographic information data of the city where the user is located, and generating an attention heat map for the labeled advertising screen location information according to the real-time crowd features to obtain a screen area attention distribution heat map; injecting physical constraint conditions into the screen area attention distribution heat map; constructing a dynamic position optimizer according to the injected conditions to generate advertising element coordinate suggestion data;

[0089] In an embodiment of the present invention, real-time crowd distribution is collected through smart surveillance cameras, public transportation systems, and location data (such as mobile phone positioning and Wi-Fi hotspot data) in a city. Geotagged content on social media platforms (such as geotags on Twitter and Instagram) is analyzed to assist in extracting crowd characteristics. Traffic flow information is obtained through a traffic management system, which can indirectly infer the characteristics of people in densely populated areas. Based on the crowd distribution data, crowd density characteristics are extracted for different time periods and geographical areas. A heat map analysis algorithm (e.g., density estimation based on a Gaussian kernel) is used to generate regional crowd distribution maps. Combining geographic data and sensor information, crowd behavior patterns are analyzed, including peak activity times, shopping hotspots, and workspace distribution. Combined with geographic information, demographic data (such as age, gender, and occupation) within the region is obtained to further refine crowd characteristics. Time series analysis, predictive modeling, and other methods are used to predict changes in crowd distribution over a period of time, ensuring that advertising placement decisions are based on real-time and future forecast data. Based on real-time crowd characteristic data and combined with the location information of advertising screens, the crowd density and activity patterns in the area surrounding each advertising screen are weighted to generate an attention heat map for the advertising screen. Taking into account factors such as crowd density, time of day, activity patterns, and traffic flow, a heat map is generated for each advertising screen's coverage area, representing the distribution of crowd attention. This heat map is modeled in multiple dimensions, both temporally and spatially, creating a dynamic, time-based heat map that adapts to the needs of different time periods. This multidimensional heat map, incorporating information such as crowd density, age distribution, and gender distribution, is used to analyze the differences in attention paid to the screens by different groups. A series of constraints are designed based on the physical characteristics and geographic location of the screens. These include the screen's viewing distance and angle of view. For example, some screens are located at intersections or in the corners of buildings, where viewing distance is limited. The actual display size and positioning of the screens are taken into account to ensure that the heat map generated matches the screen's physical display area, avoiding excessive attention to content outside the display area. The attention distribution reflected in the heat map is consistent with actual crowd flow, for example, preventing screens in high-traffic areas from being overlooked. Incorporating these physical constraints into the heat map data ensures that the heat map not only accurately reflects crowd distribution but also accounts for the practical limitations of the screens, such as viewing angle and distance. Construct an optimization objective function to ensure that the placement of advertising elements maximizes the distribution of attention. This objective function can take into account factors such as crowd density, time period, and attention heatmaps. Physical constraints are combined with the optimization objective to form a comprehensive optimization model. This model balances crowd attention with the physical characteristics of the advertising screen to maximize advertising effectiveness. Based on the optimization objective, select an appropriate algorithm, such as genetic algorithms, particle swarm optimization (PSO), simulated annealing, etc., to dynamically adjust the placement of advertising elements. Using real-time crowd data feedback, adjust the optimizer parameters to adapt to real-time changes in ad placement.For example, when the attention heat map of a certain advertising screen position changes, the optimizer can update the display position of the advertising elements in real time. When optimizing the position, multiple targets are considered, such as maximizing exposure, maximizing audience participation, and ensuring that physical conditions (such as sight distance and angle) are met. Based on the results of the optimization process, the optimizer outputs optimal advertising element coordinate recommendation data, which can be used as input to the advertising delivery system to accurately guide the selection of the position of the advertising content. According to different scenarios (such as traffic changes at different times, different regional behavior characteristics), the position of the advertising elements is dynamically adjusted to ensure the best delivery effect.

[0090] Step S4: Based on the advertising type recommendation list, the site priority list, and the advertising element coordinate suggestions, the advertising delivery content combination is performed to generate an advertising image-text combination scheme; the advertising image-text combination scheme is evaluated in a closed loop and the content weight is adjusted to perform an advertising delivery content generation optimization task.

[0091] In the embodiments of the present application, by recommending the types of advertisements in the list (such as promotion, brand promotion, event promotion, etc.) according to the types of advertisements, combined with the priority of the advertisement screen in the site priority list, the appropriate advertisement type is selected and matched with the corresponding advertisement screen. Different priorities are set for different types of advertisements, and brand promotion type advertisements are preferentially placed in high exposure and high traffic areas, and promotion advertisements can be selected in areas with high periodical traffic. The priority of the advertisement screen is combined with the optimized advertisement element coordinate suggestion to ensure that the advertisement elements can be accurately placed in the corresponding advertisement screen position. Considering the physical constraints of the advertisement screen (such as display size, viewing distance, etc.), the layout and size of the advertisement content are adjusted to adapt to the space of each advertisement screen, ensuring the optimal visual effect. According to the matched advertisement type, site priority and coordinate suggestion, an advertisement graphic-text combination scheme is generated. The scheme includes: designing graphic-text content according to the theme of the advertisement (brand, promotion, etc.), and the design of the graphic-text needs to meet the display requirements of the advertisement screen (such as size, display time, resolution, etc.). The specific time of advertisement placement is determined, including period, frequency, cycle, etc., and the advertisement exposure opportunity is optimized. The layout of the advertisement on the screen is designed, such as the proportion of image and text, the font size of the text, color matching, etc. Real-time tracking of the placement effect of the advertisement is realized by using monitoring tools and sensor data, and the performance of the advertisement graphic-text combination scheme is monitored through the feedback mechanism (such as audience dwell time, trigger behavior, etc.). A / B testing is performed, different advertisement graphic-text combination schemes are used for placement, the performance of each scheme is evaluated, and the optimal scheme is selected. Real-time data feedback (such as effect evaluation indicators) is fed back to the advertisement system for adjustment and optimization. Based on the analysis results, the system can adjust the advertisement placement strategy and content to continuously improve the advertisement effect. According to the advertisement type, advertisement screen position, period, audience group, etc., an advertisement content weight model is constructed. This model can dynamically adjust the weight of the content based on machine learning algorithms (such as weighted regression, support vector machine, etc.) or rule engines. Set the initial weight for each element (image, text, animation, etc.) in the advertisement graphic-text combination. According to the audience group characteristics of the advertisement screen and the advertisement effect evaluation results, dynamically adjust these weights. For example, in high traffic periods, the visual elements of the advertisement content will be given higher weights. Based on the advertisement effect evaluation results, the weights of the advertisement graphic-text content are fine-tuned. For example, if a certain type of advertisement graphic-text combination performs well in a specific period, the weight of that type of content can be increased to optimize the design and placement strategy of the advertisement content.

[0092] Preferably, step S1 comprises the following steps:

[0093] Step S11: Obtain the industrial data of the city;

[0094] Step S12: Use a large model to extract industry keywords and consumption scenarios from the industry data of the city, and model the industry correlation degree of the industry keywords and consumption scenarios to generate an industry-advertising type mapping matrix;

[0095] Step S13: Perform seasonal demand analysis on the industry-advertising type mapping matrix to generate seasonal demand analysis data;

[0096] Step S14: Perform advertising type dynamic priority sorting on the industry-advertising type mapping matrix through the seasonal demand analysis data to generate an advertising type recommendation list for the city.

[0097] In the embodiments of the present application, the industrial data of the city is collected through the use of public data sources, industry reports, and economic data released by the department. These data include the production scale, growth trend, market demand, and industrial distribution of various industries. The collected industrial data is preprocessed, such as cleaning, deduplication, and missing value filling, to ensure the accuracy and completeness of the data. Data processing tools such as Python's Pandas and SQL databases are used for data cleaning and standardization to ensure the uniform format of various data. Natural language processing techniques such as TF-IDF, Word2Vec, or BERT models are used to extract keywords from industrial-related text data such as industry reports, news, and social media comments. These industrial keywords may include industry names, product categories, consumer trends, and market hotspots. Based on the consumer behavior, purchase patterns, and consumption places in the industrial data, combined with scene modeling techniques such as scene perception models and user behavior analysis, consumption scenarios are extracted, including online / offline shopping, holiday consumption, and shopping behavior in specific areas. A large model such as a deep learning model or clustering algorithm is used to model the correlation between industrial keywords and consumption scenarios. By calculating the correlation between each industrial keyword and different consumption scenarios, an industrial-consumption scenario correlation matrix is generated. The model can be trained based on statistical methods such as collaborative filtering and similarity analysis or deep learning methods such as neural networks and graph neural networks to further improve the accuracy of the model. According to the results of industrial keyword and consumption scenario extraction, combined with the types of advertisements such as brand advertisements, promotional advertisements, and event advertisements, an industrial-advertising type mapping matrix is generated. This matrix reflects the correlation between different industries and advertising types, which can help understand which advertising types match better with which industries or consumption scenarios. Combined with historical sales data, seasonal commodity demand data, and climate change factors, the seasonal demand characteristics of advertising types are analyzed. For example, some advertising types have higher demand during holidays or seasonal changes. Time series analysis such as ARIMA, LSTM, or seasonal decomposition methods such as STL decomposition are used for seasonal demand analysis. Machine learning algorithms such as clustering analysis and trend analysis are used to identify the seasonal demand patterns of different industrial-advertising types. Through data mining techniques, the demand fluctuation patterns of each advertising type throughout the year are found. Based on the seasonal demand analysis, specific seasonal demand analysis data is generated. This data will include the demand intensity and fluctuation of each advertising type in different seasons, helping decision-makers understand the best timing for advertising types. Based on the seasonal demand analysis data, the priority of advertising types in the industrial-advertising type mapping matrix is adjusted. For advertising types with strong demand, higher priority is given; for advertising types with weak demand, their priority is reduced.A weighting algorithm can be used to set dynamic priorities for each ad type based on seasonal demand intensity, ad type matching degree with industry, and other factors. Based on the adjusted ad type priorities, a recommended list of ad types is generated for the city. The recommended list dynamically generates the most suitable ad type delivery order based on different seasons, industry demand, and ad effectiveness.

[0098] Preferably, step S12 comprises the following steps:

[0099] Step S121: Collect consumption scenario data of the city's industry data to obtain the city's consumption scenario data;

[0100] Step S122: Extract industry keywords from the city's industry data based on the city's consumption scenario data to obtain industry core keywords;

[0101] Step S123: Cluster the industry core keywords to generate the city's main industry categories, and perform descriptive analysis on the consumption scenario data based on the city's main industry categories through text analysis technology to generate core features of the consumption scenario and consumer demand data;

[0102] Step S124: Use a large model to calculate the correlation of the core features of the consumption scenario and the consumer demand data to generate industry-consumption scenario association data; and perform data matrixization on the industry-consumption scenario association data to generate an industry-ad type mapping matrix.

[0103] In the embodiments of the present application, data related to consumption scenarios is collected through e-commerce platforms (such as Taobao, Jingdong), offline mall data, industry reports, social media (such as Weibo, Douyin), and other channels. These data include consumers' shopping behavior, participation in activities, holiday consumption, etc. Consumption scenarios can be divided into multiple categories, such as online shopping, offline shopping, specific activities (such as festival promotions, Black Friday shopping), etc. According to the characteristics of different industries, the content and characteristics of each consumption scenario are clearly defined. Use text analysis tools (such as natural language processing technology) to label the collected consumption scenario data and extract industry-related consumption scenarios. Classify consumption scenarios and establish a standard classification system for consumption scenarios. For example: e-commerce shopping, offline experience, festival promotion, etc. Organize the collected and labeled consumption scenario data into a structured data format for subsequent analysis and processing. Use data processing tools (such as Pandas, SQL databases, etc.) to clean, remove duplicates, and standardize the data to ensure data quality. Perform text analysis on the industry data in the city and use natural language processing (NLP) techniques (such as TF-IDF, Word2Vec, BERT, etc.) to extract keywords from the industry data. These keywords may include industry names, product types, market demand, etc. According to the industry background and market dynamics of the city, extract keywords that represent the core characteristics of the industry, which involve industry fields, technology trends, consumer trends, etc. Use clustering algorithms (such as K-means, hierarchical clustering, etc.) to cluster the extracted keywords and filter out the core keywords for each industry field. Refine and optimize the extracted keywords to ensure only the most representative and highly correlated keywords are retained. Create a list of industry core keywords, which will serve as the basis for subsequent analysis. Use clustering algorithms (such as K-means, DBSCAN, hierarchical clustering, etc.) to perform clustering analysis on the industry core keywords and identify the main industry categories. According to the clustering results, classify multiple industry categories, such as retail, catering, automobiles, electronic products, etc., and determine the main representative keywords for each industry category. Use text analysis techniques to perform descriptive analysis on the main industry categories in the city. Through the correlation analysis of industry data and consumption scenario data, identify the corresponding consumption scenarios for each industry category. Combine data mining techniques to extract consumer demand characteristics in each industry category, such as purchase behavior, consumption frequency, preferred products, etc. Quantify the analysis results to generate core feature data and consumer demand data for consumption scenarios. Core feature data includes consumption time, purchase method, product type, etc.; consumer demand data includes price sensitivity, purchasing power, promotion responsiveness, etc. Organize the core features and consumer demand data of consumption scenarios to ensure consistent data formats and perform missing value processing, normalization, etc. Use tools such as Pandas to integrate these data into a clear data set for model training.The core features of the consumption scene and consumer demand data are trained using large models such as deep learning models, graph neural networks, etc. to calculate the correlation between each industry and the consumption scene. Through multi-layer neural networks, attention mechanisms, and other technologies, the potential complex correlation between industries and consumption scenes is captured. According to the results output by the model, the correlation data between industries and consumption scenes are generated, which reflect the correlation strength between different industry categories and different consumption scenes, helping to understand which industries are more suitable for which consumption scenes. The industry-consumption scene correlation data is converted into a matrix form. In this matrix, each row represents an industry category, each column represents a consumption scene, and each value in the matrix represents the correlation between the industry category and the consumption scene. According to the correlation between industries and consumption scenes, combined with the definition of ad types (such as brand ads, promotional ads, event ads, etc.), appropriate ad types are assigned to each industry category corresponding to the consumption scene. For example, if a certain industry has a high correlation in a specific consumption scene, the corresponding ad type will be evaluated as a priority ad type. The results of the above steps are integrated to generate the final industry-ad type mapping matrix. This matrix not only reflects the correlation between industries and consumption scenes, but also provides the most suitable ad type recommendation for each industry category.

[0104] As an example of the present application, reference is made to Fig. 1, which shows the steps S1-S4 of the method according to the present application. Figure 2 In this example, the step S2 includes:

[0105] Step S21: Obtain geographic information data of the city where the user is located;

[0106] Step S22: Perform ad screen location information annotation on the geographic information data of the city where the user is located, and perform scene semantic modeling on the geographic information data of the city where the user is located based on the annotated ad screen location information using a CV model to generate a scene label system;

[0107] Step S23: Perform multi-dimensional matching calculation on the ad type recommendation list of the city where the user is located and the scene label system, exclude conflicting scenes, and generate ad scene matching data;

[0108] Step S24: Perform candidate site hierarchical output on the annotated ad screen location information based on the ad scene matching data to obtain an ad site priority list.

[0109] In the embodiments of the present application, detailed geographic information data of the city is obtained by using open data platforms (such as OpenStreetMap, government public data platforms) and third-party data providers (such as Gaode Map, Baidu Map). Including city block division, building location, road network, public facilities, commercial area, traffic flow and other data. Through GPS positioning and map labeling tools (such as ArcGIS, QGIS), the precise position of the advertising screen in the city is labeled. The labeled data includes the latitude and longitude of the advertising screen, the street block, the surrounding environment, the street facing the advertising screen, the traffic flow and other information. The accuracy of the labeling can be improved by combining manual labeling and automated tools. The labeled advertising screen position information is identified by using a computer vision (CV) model. First, high-resolution satellite images or street view data are used to process images of different areas of the city. Deep learning models (such as convolutional neural networks CNN, object detection models YOLO or Faster R-CNN) are used to extract features in the images and identify different scene types (such as commercial areas, residential areas, transportation hubs, etc.). Combined with geographic information system (GIS) data, spatial information fusion is performed to accurately locate the scene type of each advertising screen. Based on the scene recognition results and geographic information data, a scene label system is constructed. For example, the scene label can be divided into multiple dimensions, such as "traffic-intensive area", "shopping center surrounding area", "tourist attraction area", etc. Each advertising screen is assigned one or more labels to represent the specific scene type of the advertising screen. The advertising type recommendation list generated in step S12 is matched with the scene label system generated in step S22 in multiple dimensions. Each advertising type will be matched and calculated between different scene labels according to its characteristics. Multiple factors are considered in the calculation, such as the target audience of the advertisement, the timeliness of the advertisement, the consumer attributes of the scene, etc. A weight scoring algorithm is used to score the matching degree between the advertising type and the scene label. The higher the matching degree, the better the effect of the advertising type in the scene. The advertising type and scene label in the matching result are detected for conflicts. According to business needs, scenes that are not suitable for advertising are excluded. For example, some advertising types are not suitable for certain scenes, such as children's product advertising in educational or medical scenes. Use predefined rules (such as matching of advertising content, scene type and consumer interest) to filter out inappropriate scenes. Finally, an advertising scene matching data is generated, which contains the pairing information of each advertising type and its best scene label, and the matching degree of each pair is evaluated.

[0110] Preferably, based on the labeled advertising screen position information, the CV model is used to model the scene semantics of the geographic information data of the city, including:

[0111] extract planning features of geographic information data of the city where the user is located, and perform spatial region level division on the geographic information data of the city where the user is located to generate spatial region level data;

[0112] based on the labeled advertisement screen location information, the spatial region level data is located in the space of the advertisement screen to generate advertisement screen spatial positioning data; the image of the advertisement screen spatial positioning data is retrieved by using the regional camera, and the target is detected by using the CV model to generate the advertisement scene visual feature;

[0113] extract the scene elements of the advertisement scene visual feature, combine the regional generation network to model the scene semantics of the image of the advertisement screen spatial positioning data, and generate the advertisement screen environment scene modeling data;

[0114] the advertisement screen environment scene modeling data is labeled by semantic association label, and environment label association data is generated; the advertisement screen environment scene modeling data is constructed by using the environment label association data and the spatial region level data to obtain the scene label system.

[0115] In the embodiments of the present application, the spatial features of the city are extracted by using urban planning data such as road network, building distribution, green space and public area, etc. These features help to understand the city's infrastructure, traffic flow, commercial district distribution, etc. Using GIS tools (such as ArcGIS, QGIS) for spatial analysis, the planning features of different regions are identified, such as traffic-intensive areas, commercial center areas, residential areas, school areas, etc. According to the extracted planning features, the city space is divided into different levels of regions. For example, it can be divided into multiple levels such as block level, community level, commercial center level, etc. For each region, different weights are given according to the characteristics of the region (such as traffic flow, commercial activity, etc.), and corresponding spatial region level data is generated for each region. The hierarchical clustering algorithm (such as K-means, DBSCAN) can be used for automatic division of regions, and expert knowledge can also be combined for manual adjustment to ensure that the hierarchical data accurately reflects the spatial structure of the city. Combined with the labeled advertising screen location information (latitude, longitude, block location, etc.), the spatial region of the advertising screen is located. Using geographic information system (GIS) technology, the advertising screen coordinates are mapped to the corresponding region in the spatial region level data according to the labeled advertising screen coordinates. According to the location information of the advertising screen, combined with the spatial region level data, the specific spatial positioning data of each advertising screen is generated, which clearly indicates the location of the advertising screen in the city and the region level it is in. Through the cameras deployed in the city (such as traffic monitoring cameras, public facility cameras, etc.), real-time image or video data of the region where the advertising screen is located is obtained. The camera position can be selected according to the spatial region level data to select the cameras located around the advertising screen to obtain the image. Using target detection technology in deep learning (such as YOLO, Faster R-CNN) to detect the image, the key elements in the image (such as pedestrians, traffic flow, buildings, shops, etc.) are identified. Through the target detection model, the visual features around the advertising screen are extracted, including the crowd distribution, traffic density, building type, etc. Based on the target detection results, the visual features of the advertising scene are extracted, such as scene type (commercial area, residential area, transportation hub, etc.), crowd density, traffic flow, etc. Using these visual features to describe the current state of the environment where the advertising screen is located provides data support for subsequent advertising content optimization. The visual features of the advertising scene are subdivided to extract key scene elements such as building type, road width, crowd mobility, shop density, etc. Using image segmentation technology in computer vision (such as Mask R-CNN) to perform more detailed region segmentation on the image, different types of scene elements are distinguished. The extracted scene elements are combined with the region generation network (such as convolutional neural network CNN, generative adversarial network GAN) to model the image in the region where the advertising screen is located semantically.The model is used to assign semantic labels to elements in the image, such as "traffic-intensive", "commercial district", or "leisure area", etc. to generate semantic data for each advertising screen scene, and map it to the corresponding advertising screen spatial positioning data. The advertising screen environment scene modeling data is labeled, combined with regional characteristics and scene element information, to generate scene labels related to advertising screen location. Using natural language processing (NLP) technology, the scene elements are classified to generate environment labels such as "high traffic", "commercial shopping area", "transportation hub", etc. The generated labels are associated with the spatial positioning data of the advertising screen to form the label relationship between the advertising screen and its environment. Through label association data, the characteristics of the area where the advertising screen is located are analyzed to provide personalized suggestions for advertising placement. Based on the environment label association data and spatial region hierarchical data, a complete scene label system is constructed. The scene label system includes specific environmental labels of the advertising screen and various information related to the environment (such as crowd flow, traffic density, commercial area attributes, etc.).

[0116] Preferably, in step S3, real-time crowd characteristics are extracted based on the geographical information data of the city, and attention heat map generation is performed on the labeled advertising screen location information according to the real-time crowd characteristics, including:

[0117] Based on the geographical information data of the city, the regional camera of the region is called;

[0118] The regional camera of the region is subjected to environment image screening to obtain regional environment images; human body contours are identified from the regional environment images, and crowd attribute analysis is performed on the human body contours to generate real-time crowd characteristics;

[0119] According to the real-time crowd characteristics, the visibility analysis is performed on the labeled advertising screen location information to generate visibility data; the eye movement tracking prediction is performed on the visibility data using the pre-trained attention prediction model to generate eye movement tracking prediction data;

[0120] The eye movement tracking prediction data is converted into an attention heat map to generate a screen area attention distribution heat map.

[0121] In the embodiments of the present application, the location information of the cameras in the city area is obtained through geographic information system (GIS) data. The cameras deployed in the region are called, mainly including cameras in public places, commercial areas, transportation hubs and other regional cameras. Through the integration of city monitoring systems or other related data platforms, the camera video stream in the region is called. From the called regional camera images, images containing the environment around the advertising screen are selected. Image recognition methods in computer vision can be used to remove irrelevant backgrounds or other interference factors, ensuring that the image content is related to the location of the advertising screen and the crowd flow. Scene classification models (such as ResNet, VGG) can be used for image screening to ensure that the image meets the requirements of environment and advertising screen positioning. Human pose detection algorithms (such as OpenPose, HRNet, etc.) are used to process the regional environment images to identify the human contours in the images. This process identifies the body contours, poses, and other information of each person in the crowd, providing basic data for subsequent crowd analysis. Based on the identified human contours, attribute analysis of the crowd is performed, including the number of people, gender, age group, behavior patterns (such as walking, staying, gathering, etc.). Convolutional neural networks (CNN) or deep learning models (such as YOLO, FasterR-CNN) are used to further analyze the images, identify and classify the features of the crowd, such as crowd density, active areas, etc., to generate real-time crowd feature data, including the number, location, and activity status of the crowd. Based on the location of the advertising screen, the surrounding environment, and the real-time crowd features, the viewing distance of the advertising screen is calculated. The viewing distance analysis considers the effects of line-of-sight obstruction, environmental interference, and crowd behavior on the visibility range of the advertising screen. The visibility distance, angle, and audience distance of the advertising screen are evaluated through three-dimensional modeling and simulation (e.g., deep learning models based on scene reconstruction). The viewing distance data is calculated and generated to describe the viewing distance relationship between the advertising screen and the crowd in different regions. The viewing distance data provides a basis for subsequent advertising placement and content optimization. A pre-trained eye tracking prediction model (such as the EYE-Track model based on convolutional neural networks) is used to analyze the viewing distance data to predict the eye movement trajectories of the crowd in the advertising screen area. This model will predict which areas attract more attention based on the viewing distance data and the behavior patterns of the crowd, simulate the movement of the user's line of sight, and identify the focus points in the advertising screen. The eye tracking data output by the model includes the areas that the audience is looking at at each time point and the duration of the focus. These data can be quantified to describe the attention intensity of each area in the advertising screen. Based on the eye tracking prediction data, a heat map generation algorithm (such as the Gaussian smoothing or heat map fitting method) is used to convert the line-of-sight distribution of the focus areas into a heat map. The color depth in the heat map represents the strength of the area's focus, and color gradients (such as red for high focus and blue for low focus) are often used to represent it. The eye tracking data is converted into a focus distribution heat map of the screen area, showing the focus of different areas.

[0122] Preferably, the physical constraint injection on the screen area attention distribution heat map in step S3 comprises:

[0123] The annotated advertising screen location information is subjected to advertising screen parameter information collection to obtain advertising screen information parameters; the advertising screen information parameters are subjected to screen resolution and pixel density analysis to obtain advertising screen resolution and pixel density;

[0124] Based on the advertising screen resolution and pixel density, initial advertising boundary range setting is performed, and the set initial advertising boundary range is subjected to visual range calculation to obtain advertising screen visual range;

[0125] The advertising screen visual range is subjected to minimum font size limit physical constraint through the screen area attention distribution heat map to obtain content layout safety boundary constraint condition, wherein the calculation formula of the minimum font size limit physical constraint is as follows:

[0126]

[0127] In the formula, B(x) is the content layout safety boundary constraint, A unsafe (x) is the partial area of the advertising content beyond the screen visual area, A total is the total area of the screen, and x is an input parameter.

[0128] In the embodiment of the application, the basic parameters of the advertising screen are collected, including the size of the advertising screen, the display technology (such as LED or LCD), the screen resolution, the pixel density (PPI), and the installation angle of the advertising screen (relative to the ground or the viewing area). It is ensured that the information source is accurate, and the actual installation environment, size and configuration of the advertising screen need to be synchronized with the database or real-time monitoring system. The horizontal and vertical resolutions of the advertising screen are confirmed, and common formats are 1920x1080, 3840x2160, etc. The pixel density (PPI) is calculated based on the physical size and resolution of the advertising screen. Based on the size, resolution and visual range analysis of the advertising screen, the initial boundary range of the advertising is determined. The boundary range defines the "display area" of the advertising, that is, all advertising content should be kept within the area to ensure complete visibility within the viewing range of the viewer. Through visual range analysis, the visual range of the advertising screen is calculated. The visual range calculation considers the distribution of the crowd, the viewing angle, and the installation position of the advertising screen. A visual range model can be used, such as based on the camera angle and scene depth to calculate the area where the crowd can clearly see the advertising content. Using the generated screen area attention distribution heat map, the attention hotspots of the audience to the advertising screen are mapped to the display area of the advertising screen to determine the areas with high and low attention. In order to ensure the readability and safety of the advertising content, the minimum font size needs to be set to avoid the advertising content being unable to be clearly visible in the case of the audience being far away or at a large angle. The partial area Aunsafe (x) If a part of the ad content is out of the viewable area (e.g. not visible at a far distance), the area of this part needs to be calculated and adjusted. The minimum font size restriction is used to adjust the physical constraint to ensure that the ad content is not out of the viewable range of the screen, which is calculated as follows: where B(x) is the content layout safety boundary constraint, A unsafe (x) is the area of the part of the ad content that is out of the viewable area of the screen, A total is the total area of the screen, and x is the input parameter. According to the calculation result, the layout of the ad content is adjusted to ensure that all the ad content is within the viewable range and meets the minimum font size restriction. By redefining the position of the ad content in the heat map, important information is ensured to be displayed in the focal point area of the viewer’s line of sight.

[0129] Preferably, the dynamic position optimizer construction according to the injection condition in step S3 comprises:

[0130] A multi-objective optimization parameter is constructed based on the content layout safety boundary constraint condition, wherein the multi-objective optimization parameter comprises a minimum exposure rate and a minimum information interference value, and the calculation formula of the minimum exposure rate is as follows:

[0131]

[0132] where E(x) is the minimum exposure rate, A i is the display area of the ad in the i-th region, A total is the total area of the screen, V x is the exposure of the ad in the i-th region under different viewing angles, n is the number of ad regions, and x is the input parameter.

[0133] The calculation formula of the minimum information interference value is as follows:

[0134]

[0135] where D(x) is the minimum information interference value, L j (x) is the ad brightness under the j-th environmental lighting condition, C j (x) is the information visible contrast under the j-th environmental lighting condition, L max and C max are the maximum brightness and the maximum contrast of the screen under the given condition, respectively, m is the number of different environmental conditions, and x is the input parameter.

[0136] A multi-objective optimization function is constructed according to the minimum exposure rate, the minimum information interference value, and the content layout safety boundary constraint condition, and the multi-objective optimization function is obtained, and the formula of the multi-objective optimization function is as follows:

[0137] f(x) = w1 * E(x) + w2 * D(x) + w3 * B(x);

[0138] In the formula, f(x) is a multi-objective optimization function, E(x) is a minimum exposure rate, D(x) is a minimum information interference value, B(x) is a content layout safety boundary constraint, w1 is a minimum exposure rate weight, w2 is a minimum information interference weight, w3 is a layout safety boundary constraint weight, and x is an input parameter.

[0139] The multi-objective optimization function is used for performing dynamic position optimization on the screen area attention distribution heat map of an advertisement element, and advertisement element coordinate suggestion data is generated.

[0140] In the embodiment of the application, the display area A i of the advertisement in each area is calculated according to the viewing distance and viewing angle of each area, the exposure V x is evaluated through a viewing angle model or through camera capture data, finally, the exposure rates of all areas are summed up to obtain the minimum exposure rate E(x). The brightness and contrast of the advertisement screen are monitored and calculated in real time, and the influence under different environmental light conditions is considered; the illumination data is acquired and the brightness L j (x) and contrast C j(x); normalize the values with the maximum brightness and maximum contrast, and sum them to obtain the minimum information interference value D(x). Through parameter tuning, the weights w1, w2, and w3 of each target are determined, which can be adjusted according to the actual advertising effect and priority. For example: if the exposure rate of the advertisement is prioritized, w1 will be relatively large; if the screen viewable range and content security are prioritized, the weight of w3 will be increased accordingly; if the environmental impact is large, w2 needs to be adjusted. By adjusting the weights, the balance of each target is optimized to build an optimization function suitable for the current advertising needs, and the formula of the multi-objective optimization function is as follows: f(x) = w1·E(x) + w2·D(x) + w3·B(x); in the formula, f(x) is a multi-objective optimization function, E(x) is the minimum exposure rate, D(x) is the minimum information interference value, B(x) is the content layout security boundary constraint, w1 is the minimum exposure rate weight, w2 is the minimum information interference weight, w3 is the layout security boundary constraint weight, and x is the input parameter; according to the generated content layout security boundary constraint, minimum exposure rate, and minimum information interference value, the dynamic position optimization process is started. An algorithm based on optimization theory (such as genetic algorithm, particle swarm optimization, simulated annealing, etc.) is used to optimize the position of the advertising elements on the screen. This optimization process will ensure that the position of the advertising elements can maximize the exposure within the viewable range, while minimizing information interference and meeting the layout security boundary constraint. After optimization, the obtained advertising element coordinates are the best display position of the advertisement on the screen, and the generated advertising element coordinate suggestion data is optimized according to the current advertising screen viewable range, exposure demand, environmental impact, etc. to maximize the advertising effect.

[0141] As an example of the present application, reference is made to Figure 3 In this example, the step S4 includes:

[0142] Step S41: based on the advertisement type recommendation list, the site priority list, and the advertising element coordinate suggestion, the advertising content combination is performed to generate an advertising graphic combination scheme;

[0143] Step S42: real-time click rate testing is performed on the advertising graphic combination scheme to obtain a real-time click rate test result; cross-scene style migration is performed on the advertising graphic combination scheme according to the real-time click rate test result to generate advertising style migration data;

[0144] Step S43: the advertising style migration data is used to perform closed-loop effect evaluation on the advertising graphic combination scheme to generate advertising placement effect evaluation data, and the content weight of the advertising graphic combination scheme is adjusted through the advertising placement effect evaluation data to perform advertising content generation optimization work.

[0145] In the embodiments of the present application, the advertising target types are analyzed and corresponding advertising type recommendation lists are generated according to advertising targets such as brand exposure, product promotion, and interactive guidance. The recommendation lists are dynamically adjusted considering factors such as audience behavior data, historical click data, and regional characteristics. Based on the actual situation of the advertising placement area (such as geographic location, audience density, and exposure period), a site priority list is generated. Sites with high priority are areas with dense human flow, transportation hubs, and commercial centers. Combining advertising placement environment data (such as geographic information and site characteristics) with advertising type recommendations, a dynamic optimizer generates advertising element coordinate suggestions. Through multi-objective optimization algorithms, the consistency of advertising display location with site priority and advertising type demand is ensured. The advertising images, text, and video elements are reasonably combined according to the coordinate suggestions to ensure that the advertising content visually meets the audience preferences of the site. Through content layout optimization, the best visual and interactive effect of the graphic advertising is ensured. In specific placement scenarios, the advertising graphic combination scheme is tested, and user click data is monitored in real time. A / B testing and other methods are used to compare the effects of different advertising combinations, and key data such as click-through rate and conversion rate are collected. According to the test results, the advertising combination is adjusted, and the advertising effect is optimized. According to the visual characteristics and audience needs of different scenarios, the style of the advertising graphic combination is migrated. Using a pre-trained style migration model, the image style, color tone, layout, and other elements of the advertising are adjusted to adapt to different styles. Through a deep learning model, the advertising style is automatically adjusted to ensure consistency across scenarios. Through style migration, adjustment data including advertising color tone, design elements, and text layout are generated. Using these data, the advertising team can further understand the performance of advertising content in different scenarios. Through user feedback, click-through rate, dwell time, and other data, the visual appeal and information transmission effect of the advertising are evaluated. Using multi-dimensional analysis methods, including the visual effect of the advertising, user interaction, and conversion rate, the advertising effect is evaluated. If the effect after style migration does not meet expectations, the advertising design and style migration parameters are adjusted according to the evaluation results. Real-time click-through rate, conversion rate, user dwell time, and exposure are summarized and analyzed to form a detailed advertising effect evaluation report. The performance of the advertising in different sites, different time periods, and different audience groups is analyzed to guide the placement strategy. According to the evaluation data, the weight of elements in the advertising graphic combination is adjusted. For example, if a certain advertising content type (such as text advertising or image advertising) achieves a higher click-through rate, the weight of that element is increased. Different elements in the advertising content (such as titles, pictures, and videos) are weighted and adjusted to ensure that their proportion in the advertising combination is more reasonable. After adjusting the weight of the advertising content, an optimized advertising graphic combination is generated. Using the advertising placement system, the optimized advertising scheme is pushed in real time for further testing and evaluation to complete the closed-loop optimization.

[0146] Therefore, the embodiments should be regarded, at any point, as being exemplary and not limiting, the scope of the application being defined by the appended claims and not by the above description, and all changes which come within the meaning and range of equivalency of the claims are therefore intended to be embraced therein.

[0147] The foregoing is considered as illustrative only of the principles of the application. Numerous modifications and changes will readily occur to those skilled in the art, and the generic principles defined herein can be applied to other embodiments without departing from the spirit or scope of the application. Therefore, the application is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for generating advertising content based on a large model, characterized in that: The following steps are involved: Step S1: Obtain industry data for the city; use the big model to model the industry correlation of the industry data for the city, and generate an industry-advertising type mapping matrix; dynamically prioritize the advertising types in the industry-advertising type mapping matrix, and generate a recommended list of advertising types for the city; Step S2: Obtain geographic information data of the city; annotate the advertising screens in the geographic information data of the city to obtain the location information of the annotated advertising screens; output the candidate sites according to the annotated advertising screen location information to obtain a priority list of advertising sites; Step S3: extracting real-time crowd characteristics based on the geographic information data of the city, and generating an attention heat map for the marked advertising screen location information based on the real-time crowd characteristics to obtain a screen area attention distribution heat map; Inject physical constraints into the screen area attention distribution heat map; build a dynamic position optimizer based on the injected conditions to generate recommended coordinate data for advertising elements; Step S4: Based on the advertisement type recommendation list, the venue priority list, and the advertisement element coordinate suggestions, the advertisement content is combined to generate an advertisement image and text combination plan; the advertisement image and text combination plan is evaluated in a closed loop and the content weight is adjusted to perform the advertisement content generation optimization operation.

2. The method for generating advertisement content based on a large model according to claim 1, characterized in that: Step S1 includes the following steps: Step S11: Obtaining industrial data of the city; Step S12: Using the big model, extract industry keywords and consumption scenarios from the industry data of the city, and model the industry correlation between the industry keywords and consumption scenarios to generate an industry-advertising type mapping matrix; Step S13: performing seasonal demand analysis on the industry-advertising type mapping matrix to generate seasonal demand analysis data; Step S14: Dynamically prioritize the advertising types in the industry-advertising type mapping matrix using seasonal demand analysis data to generate a recommended advertising type list for the city.

3. The method for generating advertisement content based on a large model according to claim 2, characterized in that: Step S12 includes the following steps: Step S121: Collect consumption scenario data for the city's industrial data to obtain consumption scenario data for the city; Step S122: extracting industry keywords from the industry data of the city based on the consumption scenario data of the city to obtain core industry keywords; Step S123: Clustering the core industry keywords to generate the main industry categories of the city, and using text analysis technology to perform descriptive analysis on the consumption scenario data based on the main industry categories of the city to generate the core characteristics of the consumption scenario and consumer demand data; Step S124: Use the big model to calculate the correlation between the core features of the consumption scenario and the consumer demand data to generate industry-consumption scenario related data; matrix the industry-consumption scenario related data to generate an industry-advertising type mapping matrix.

4. The method for generating advertisement content based on a large model according to claim 1, characterized in that: Step S2 includes the following steps: Step S21: Obtaining geographic information data of the city; Step S22: labeling the location information of the advertising screens in the geographic information data of the city, and using the CV model to perform scene semantic modeling on the geographic information data of the city based on the labeled location information of the advertising screens to generate a scene label system; Step S23: Perform multi-dimensional matching calculations on the city's advertising type recommendation list and the scene label system, eliminate conflicting scenes, and generate advertising scene matching data; Step S24: The marked advertising screen location information is graded and output as candidate sites using the advertising scene matching data to obtain an advertising site priority list.

5. The method for generating advertisement content based on a large model according to claim 4, characterized in that: Based on the labeled advertising screen location information, the CV model is used to perform scene semantic modeling on the city's geographic information data, including: Extract planning features of the city's geographic information data, divide the city's geographic information data into spatial regional hierarchies, and generate spatial regional hierarchical data; Based on the labeled advertising screen location information, the spatial area hierarchical data is used to locate the advertising screen in the space, generating the advertising screen spatial positioning data; the regional camera is used to retrieve the image of the advertising screen spatial positioning data, and the CV model is used to perform target detection on the image to generate the visual features of the advertising scene; Extract scene elements of the visual features of the advertising scene, combine them with the region generation network to perform scene semantic modeling on the image of the advertising screen spatial positioning data, and generate advertising screen environment scene modeling data; The advertising screen environment scene modeling data is semantically associated with the label calibration to generate environment label associated data; the environment label associated data and spatial region hierarchical data are used to construct a scene label system for the advertising screen environment scene modeling data to obtain a scene label system.

6. The method for generating advertisement content based on a large model according to claim 1, characterized in that: The step S3 of extracting real-time crowd characteristics based on the geographic information data of the city and generating an attention heat map for the marked advertising screen location information according to the real-time crowd characteristics includes: Retrieve regional cameras in the area based on the geographic information data of the city; Filter the environmental images of the regional cameras in the area to obtain the regional environmental image; identify the human body contours in the regional environmental image, and perform crowd attribute analysis on the human body contours to generate real-time crowd features; Perform visual distance analysis on the marked advertising screen location information based on real-time crowd characteristics to generate visual distance data; use the pre-trained attention prediction model to perform eye tracking prediction on the visual distance data to generate eye tracking prediction data; The eye tracking prediction data is converted into an attention heat map to generate a screen area attention distribution heat map.

7. The method for generating advertisement content based on a large model according to claim 1, characterized in that: Injecting physical constraints into the screen area attention distribution heat map in step S3 includes: Collect advertising screen parameter information for the marked advertising screen location information to obtain advertising screen information parameters; perform screen resolution and pixel density analysis on the advertising screen information parameters to obtain the advertising screen resolution and pixel density; The initial advertising boundary range is set based on the screen resolution and pixel density of the advertising screen, and the viewing distance range is calculated for the set initial advertising boundary range to obtain the visible range of the advertising screen; The minimum font size limit physical constraint is imposed on the visible range of the advertising screen through the screen area attention distribution heat map, and the content layout safety boundary constraint condition is obtained. The calculation formula of the minimum font size limit physical constraint is as follows: Where B(x) is the content layout security boundary constraint, A unsafe (x) is the area of ​​the advertisement content that exceeds the visible area of ​​the screen, A total is the total area of ​​the screen, and x is the input parameter.

8. The method for generating advertisement content based on a large model according to claim 1, characterized in that: The construction of a dynamic location optimizer according to the injection conditions in step S3 includes: Based on the content layout safety boundary constraints, multi-objective optimization parameters are constructed. The multi-objective optimization parameters include minimum exposure rate and minimum information interference value. The calculation formula for the minimum exposure rate is as follows: Where E(x) is the minimum exposure rate, A i is the display area of ​​the advertisement in the i-th area, A total is the total area of ​​the screen, V x is the exposure of the advertisement in different viewing angles in the i-th area, n is the number of advertising areas, and x is the input parameter; The calculation formula of the minimum information interference value is as follows: Where D(x) is the minimum information interference value, L j (x) is the advertising brightness under the ambient light condition k, C j (x) is the visible contrast of the information under the ambient lighting condition k, L max and C max are the maximum brightness and maximum contrast of the screen under given conditions, m is the number of different environmental conditions, and x is the input parameter; A multi-objective optimization function is constructed based on the minimum exposure rate, minimum information interference value, and content layout safety boundary constraints to obtain the multi-objective optimization function. The formula of the multi-objective optimization function is as follows: f(x)=w1·E(x)+w2·D(x)+w3·B(x); Where f(x) is a multi-objective optimization function, E(x) is the minimum exposure rate, D(x) is the minimum information interference value, B(x) is the content layout safety boundary constraint, w1 is the minimum exposure rate weight, w2 is the minimum information interference weight, w3 is the layout safety boundary constraint weight, and x is an input parameter; The multi-objective optimization function is used to dynamically optimize the position of advertising elements on the screen area attention distribution heat map to generate advertising element coordinate recommendation data.

9. The method for generating advertisement content based on a large model according to claim 1, characterized in that: Step S4 includes the following steps: Step S41: Combining advertising content based on the advertising type recommendation list, the venue priority list, and the advertising element coordinate suggestions to generate an advertising image and text combination plan; Step S42: performing a real-time click-through rate test on the advertising image and text combination scheme to obtain a real-time click-through rate test result; performing cross-scene style migration on the advertising image and text combination scheme based on the real-time click-through rate test result to generate advertising style migration data; Step S43: Use the advertising style migration data to conduct a closed-loop effect evaluation on the advertising image and text combination plan, generate advertising delivery effect evaluation data, and adjust the content weight of the advertising image and text combination plan based on the advertising delivery effect evaluation data to perform advertising delivery content generation optimization operations.

10. A system for generating advertising content based on a large model, characterized in that: A method for generating advertisement content based on a large model according to claim 1, wherein the system comprises: The advertising type recommendation module is used to obtain the industry data of the city; use the big model to model the industry correlation of the industry data of the city and generate an industry-advertising type mapping matrix; dynamically prioritize the advertising types in the industry-advertising type mapping matrix and generate a list of recommended advertising types for the city; The advertising venue recommendation module is used to obtain the geographic information data of the city; mark the advertising screens in the geographic information data of the city to obtain the location information of the marked advertising screens; output the candidate venues according to the marked advertising screen location information, and obtain a priority list of advertising venues; The advertising content recommendation module is used to extract real-time crowd characteristics based on the city's geographic information data, and generate an attention heat map for the marked advertising screen location information based on the real-time crowd characteristics, obtaining a screen area attention distribution heat map; inject physical constraints into the screen area attention distribution heat map; and construct a dynamic location optimizer based on the injected conditions to generate advertising element coordinate recommendation data; The advertising generation module is used to combine advertising content based on the advertising type recommendation list, venue priority list and advertising element coordinate suggestions, and generate an advertising image and text combination plan; conduct closed-loop effect evaluation and content weight adjustment on the advertising image and text combination plan to perform advertising content generation optimization operations.

Images