Automatic advertising content generation system based on big data model

Abstract

The present invention discloses an automatic advertising content generation system based on a big data model, relates to the technical field of advertising production, and includes a data acquisition module, a preliminary analysis module, a video processing module, an image analysis module, an advertising integration module and a local adjustment module: the data acquisition module is used to obtain product information provided by a user, and pre-process the obtained product information to obtain product positioning; its technical points are: it can comprehensively analyze the product information uploaded by a customer, disassemble the video into images, and judge all the images, select representative high-quality images, unify the background, and then further analyze the matching value of the high-quality images to obtain the images required for the advertisement. This processing makes the generated advertisement style unified, the overall layout more coordinated, and can fully highlight the characteristics of the product, improve the quality of the advertisement, and have a good use effect and a good prospect.

Description

Technical Field

[0001] The present invention relates to the technical field of advertisement production, and in particular to an automatic advertisement content generation system based on a big data model. Background Art

[0002] Advertising, as the name suggests, is to make something known to the general public.

[0003] With the development of mobile data networks, the form of advertising has also changed a lot. In addition to traditional forms of advertising through newspapers, radio, television, movies, etc., web pages, emails, and built-in applications are also becoming popular. These new forms of advertising have the advantages of low cost and simple and convenient publishing methods; with the acceleration of digital transformation, the advertising industry is also actively exploring the application of big data and artificial intelligence technologies. The combination of automated advertising content generation and AIGC is an important direction for the digital transformation of the advertising industry; AIGC technology can automatically generate advertising content that conforms to the brand tone. These contents are not only creative, but also can be personalized according to the characteristics of the target audience; the combination of automated advertising content generation based on big data models and AIGC is an important trend in the advertising industry and technology.

[0004] However, with the rapid increase in the amount of advertising, it is difficult for the same old ads to attract people's sustained attention when the target audience is unclear, and the cost of manual updating is very high. For this reason, people have invented some automatic advertising generation systems.

[0005] The existing application publication number is CN105976203B, and the invention patent named "Internet advertising creative automatic generation method and device" records that it includes: obtaining multiple different information elements to form a database; creating multiple advertising templates that meet the preset size, the advertising template includes multiple areas, each area corresponds to an information field, and each information field corresponds to at least one information element; receiving a creative request, the creative request includes the total number of creatives and a number of information elements, wherein each information element corresponds to an information field, and selecting an advertising template corresponding to the creative request according to a preset algorithm; judging whether any area in the advertising template has an information field corresponding to the area in the creative request. Without the need for any design capabilities, the customer can automatically generate multiple corresponding advertising creatives through the advertising template provided by the present invention, which is convenient for customers to select, simple and convenient to make, and easy to use.

[0006] However, based on the above content and in combination with the prior art, the core solution of the solution recorded in the above patent is mainly used to make promotional posters, divide the advertisements into multiple sections, and then directly fill the sections with materials from the Internet or provided by customers to form advertisements.

[0007] However, in actual use, most solutions are based on the requirements of customers through simple search of ordinary patterns, and then directly form advertisements. There are many types of screening, and the integrity of the advertisement is not considered. Therefore, the combined advertising background colors are different, and the chaotic colors make it easy for users to fail to grasp the characteristics of the product. In addition, many customers provide different materials when using it, including videos, texts, pictures, etc. Existing advertisements are usually just a simple combination of text and pictures in a fixed form, which is difficult to reflect the characteristics of the product. Moreover, it is easy to have insufficient materials, resulting in a picture being played repeatedly in an advertisement, and the long playback time causing visual fatigue. The generated advertisements are easy to make the audience feel bored, and often fail to achieve the advertising effect and do not meet people's requirements for advertising. For this reason, we have developed an automatic advertising content generation system based on big data models. Summary of the invention

[0008] 1. Technical issues to be resolved

[0009] In view of the deficiencies in the prior art, the present invention provides an automated advertising content generation system based on a big data model, which can comprehensively analyze the product information uploaded by customers, decompose videos into pictures, and judge all the pictures, select representative high-quality pictures, unify the backgrounds, and then further analyze the matching values ​​of the high-quality pictures to obtain the pictures required for the advertisement. This processing makes the generated advertisements uniform in style and more coordinated in overall layout, which can fully highlight the characteristics of the products, improve the quality of the advertisements, have good effects, and have good prospects for use.

[0010] (II) Technical solution

[0011] To achieve the above objectives, the present invention is implemented through the following technical solutions:

[0012] The automated advertising content generation system based on the big data model includes a data collection module, a preliminary analysis module, a video processing module, an image analysis module, an advertising integration module, and a local adjustment module:

[0013] Data collection module: used to obtain product information provided by users and pre-process the obtained product information to obtain product positioning. Product information includes key product introduction words, product videos, product pictures and product prices;

[0014] Preliminary analysis module: used to use big data models to perform channel analysis and attribute analysis on product positioning data, obtain advertising channels and the average advertising duration corresponding to the advertising channels, and select advertising production templates from the advertising template library according to the product type, advertising duration and attribute analysis results to obtain the number of photos L required for the advertisement;

[0015] Video processing module: used to perform frame-by-frame restoration on product videos, and then classify pictures according to key product introduction words, and make comprehensive evaluations on the classified pictures, calculate the evaluation values, and then extract the N pictures with the highest evaluation values ​​in each category, N < 5, and then uniformly extract the background colors of the pictures;

[0016] Image analysis module: used to analyze the extracted images, select the image with the highest evaluation value, and then calculate the matching value between the image in each category and the image with the highest evaluation value, and retrieve the K images with the highest matching value in each category, summarize and add them to obtain the L photos required for the template, and obtain the product introduction picture set;

[0017] Advertisement integration module: used to insert product key introduction words into the pictures in the product introduction picture set, and calculate the corresponding playback time of each picture according to the evaluation value, insert the product introduction picture set into the advertisement production template according to the calculated playback time, and input the advertising slogan generated by the big data model for the product key introduction words to generate advertisements;

[0018] Local adjustment module: used to obtain user feedback, calculate the user's modified value, compare the modified value with the set standard value, and execute the corresponding adjustment strategy based on the comparison result.

[0019] Preferably, the steps of preprocessing the acquired product information to obtain product positioning are as follows:

[0020] Get the name and price of the advertisement product;

[0021] Get the average price of the same product on the market, and calculate the price difference ratio between the advertised product price and the average price of the same product on the market. The specific calculation formula is as follows:

[0022]

[0023] In the formula, is the price difference ratio between the advertised product price and the average price of the same product on the market, For the advertised product price, is the average price of the same product in the market;

[0024] Compare the price difference ratio CY with the preset positioning range to obtain the product positioning;

[0025] The preset positioning ranges include the low-end range (-100%, -15%), the affordable range [-15%, 15%], the high-end range (15%, 100%) and the top range [100%, +∞).

[0026] Preferably, when performing frame-by-frame restoration processing on the product video, P pictures are randomly extracted per second, 3<P<6;

[0027] The steps to categorize images according to product key introduction words are as follows:

[0028] Extract the key parts of the product from the key words of the product introduction;

[0029] The key parts of the product, product packaging and overall shape of the product are used as labels to classify the extracted P pictures and the obtained product pictures.

[0030] Preferably, the formula for calculating the evaluation value is as follows:

[0031]

[0032] In the formula, is the calculated judgment value, is the area of ​​the product in the picture, is the total area of ​​the image, The ratio of product area to image area in the set advertising image. , For the clarity of the picture, This is the standard definition of the advertising image. is the average color difference between the product edge color and the background color in the image, The color difference coefficient for the preset advertising. ;

[0033] The calculation formula for the average color difference between the product edge color and the image background color in the image is as follows:

[0034]

[0035] In the formula, is the brightness difference between the edge color of the product in the picture and the background color of the picture. is the chroma difference between the edge color of the product in the picture and the background color of the picture. is the hue difference between the edge color of the product in the picture and the background color of the picture. is the brightness weight factor, is the chroma weight factor, 1≤j / c≤2, where , and The saturation of the image.

[0036] Preferably, the steps of unifying the extracted image backgrounds are as follows:

[0037] Filter advertisements of similar products by product type;

[0038] Further screen the advertisements of similar products according to product positioning and extract advertisements of similar products with the same positioning;

[0039] Get the background color of the pictures in the advertisements of the same type of products with the same positioning, select the majority background color as the background color of this product advertisement, and get the RGB value of the majority background color;

[0040] Adjust the RGB value of the extracted image background to the RGB value of the majority background color.

[0041] Preferably, the formula for calculating the matching value between the image in each category and the image with the highest evaluation value is as follows:

[0042]

[0043] In the formula, is the matching value between the image and the image with the highest judgment value, It is the color difference between the product in the picture with the highest judgment value in each category.

[0044] Preferably, K pictures with the highest matching values ​​in each category are retrieved, summarized and added to obtain L photos required for the template, and when the product introduction picture set is obtained, if L=K×μ, where μ is the number of categories, then the number of pictures in each category in the product introduction picture set is K, and no addition is required; if L=K×μ+A, A<K, A is the number of pictures to be added, and A is the A pictures with the highest matching values ​​extracted from the remaining N×K-K×μ pictures with the same background color.

[0045] Preferably, the steps of inserting the key product introduction words into the photo are as follows:

[0046] Adjust the font size of key introductory words so that the font size of the key introductory words occupies 1 / B of the image background area in the advertisement, 5≤B≤10;

[0047] Adjust the resized font of the key words to the background of the picture;

[0048] Adjust the font color so that the color difference between the font color and the image background color is greater than F, F>20.

[0049] Preferably, the formula for calculating the playback time corresponding to each picture according to the evaluation value is as follows:

[0050]

[0051] In the formula, is the playing time of the i-th picture in the advertisement, is the judgment value of the i-th picture, is the sum of the evaluation values ​​of L pictures, The duration of the advertisement.

[0052] Preferably, the formula for calculating the user's modified value is as follows:

[0053]

[0054] In the formula, is the user's modified value. The number of categories modified for the user, is the classification difficulty coefficient, , The number of images modified for the specified purpose;

[0055] According to the comparison results, the corresponding adjustment strategies are as follows:

[0056] If the calculated If it is less than the standard value STD, the advertisement content will be partially modified according to the user's requirements;

[0057] If the calculated If it is greater than or equal to the standard value STD, the image will be reclassified and the advertisement will be regenerated according to customer requirements and key product introduction words.

[0058] (III) Beneficial effects

[0059] The present invention provides an automated advertising content generation system based on a big data model, which has the following beneficial effects:

[0060] 1. The present invention records an automated advertising content generation system based on a big data model, which can comprehensively analyze the product information uploaded by customers, disassemble videos into pictures, and judge all pictures, select representative high-quality pictures, unify the background, and then further analyze the matching value of the high-quality pictures, so as to obtain pictures required for advertisements. This processing makes the generated advertisements uniform in style and more coordinated in overall layout, which can fully highlight the characteristics of the products, improve the quality of advertisements, and have good use effects and good prospects for use;

[0061] 2. The present invention records an automated advertising content generation system based on a big data model, which can comprehensively analyze the product information uploaded by customers, obtain the customer's positioning of the product, and combine the big data model to perform channel analysis and attribute analysis on the product positioning data, accurately obtain the advertising channels and the average advertising duration corresponding to the advertising channels, more accurately and reasonably grasp the needs of users, and can generate advertisements that meet customer needs at one time, with good use effect and good use prospects;

[0062] 3. The present invention records an automated advertising content generation system based on a big data model, which can automatically screen out high-quality pictures, reasonably divide the picture time, and combine the pictures into advertising videos, which can more clearly and completely show the characteristics of the product. The advertisement generation process does not require manual participation from customers, and can be automatically generated, which is easy to use. In addition, it can subsequently further feedback and analyze customer requirements, and fine-tune or regenerate the advertisements. The whole process is relatively simple, and can quickly generate advertisements that meet customer requirements. It has a good effect and has a good prospect for use. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] Figure 1 It is a flowchart of the automatic advertising content generation system based on the big data model of the present invention;

[0064] Figure 2 A flowchart of video processing steps in the automatic advertising content generation system based on a big data model of the present invention;

[0065] Figure 3 It is a flow chart of the picture analysis steps in the automatic advertising content generation system based on the big data model of the present invention;

[0066] Figure 4 A basic template diagram for generating video advertisements in the automatic advertisement content generation system based on the big data model of the present invention;

[0067] Figure 5 It is a basic template diagram for generating print advertisements in the automatic advertisement content generation system based on the big data model of the present invention. DETAILED DESCRIPTION

[0068] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0069] Research reasons

[0070] Most existing systems divide advertisements into multiple sections, and then directly fill the sections with materials from the Internet or provided by customers to form advertisements. During use, they mainly perform simple assembly of pictures for the design of print advertisements.

[0071] However, in actual use, most solutions are based on customer requirements through simple search of ordinary patterns, and then directly form advertisements. In the process of advertising production, the advertisements require many types of pictures (generally multiple angles of the product combined with key areas of focus). Generally, the pictures with these characteristics are simply combined directly without considering the integrity of the advertisement. Therefore, the combined advertisement background colors are different (there are often situations where the product backgrounds are different, the pictures are not connected, and there is no connection between them). The chaotic colors make it easy for users to fail to grasp the product characteristics.

[0072] In addition, many customers provide different materials when using the product, including videos, texts, pictures, etc. Existing advertisements usually simply combine texts and pictures in a fixed form, which makes it difficult to reflect the characteristics of the product. Moreover, it is easy to have insufficient materials, resulting in some pictures being played repeatedly in an advertisement. The long playback time causes visual fatigue, and the generated advertisements easily make the audience feel bored and often fail to achieve the desired advertising effect.

[0073] There are also some video advertisements that directly combine different segments of videos, or edit short videos from long videos as advertisements. This method has extremely high requirements for video materials, and in real scenarios, most customers are unable to provide such high-demand videos. Therefore, the overall quality of advertisements prepared by this method is relatively poor.

[0074] Design ideas

[0075] During the research and development process, the initial ideas were consistent with the existing solutions, which mainly involved making full use of the data provided by customers, processing the data provided by customers reasonably, screening out customer needs as much as possible, and then generating corresponding advertisements based on customer needs.

[0076] However, in actual use, it was found that customers would not extract all the images corresponding to their needs. They would usually send a video combined with images and text, and manpower was required to select suitable images from them for processing. This method made the production of the entire advertisement take a long time and required high labor costs, which did not meet people's requirements.

[0077] Therefore, in the middle of the research and development, we changed our thinking and studied how to extract the required pictures from the video based on the processed video. Therefore, we developed a solution to divide the video into frames, restore the processed pictures, and then judge the pictures to screen out suitable pictures.

[0078] However, in actual use, it was found that although this operation can screen out high-quality pictures, the overall backgrounds of the pictures are different. Therefore, the advertising backgrounds made with these pictures are relatively messy, and the colors of the advertising backgrounds are complicated. The resulting advertisements are incoherent and cannot well display the characteristics of the product.

[0079] Therefore, in the later stage of research and development, we fully combined the previous plans to study the combination plan of pictures and how to further screen and process the pictures, so as to obtain more suitable pictures with higher matching values. The specific plan is as follows.

[0080] Study plan

[0081] The automated advertising content generation system based on big data models includes data collection module, preliminary analysis module, video processing module, image analysis module, advertising integration module and local adjustment module: the big data model used in the solution can automatically generate advertising content that meets the brand tone through AIGC technology; at the same time, for the generation of advertising images and videos, the generative adversarial network (GAN) and deep learning technology can also be used to automatically generate high-quality advertising images and video content, which not only reduces the workload of designers, but also improves the creation efficiency;

[0082] Product information and user requirements acquisition

[0083] In order to generate advertisements that meet customer requirements, you must understand customer needs and fully understand product information. To do this, you need to collect relevant data, and the process of collecting relevant data is based on the data collection module.

[0084] Data collection module: used to obtain product information provided by users and pre-process the obtained product information to obtain product positioning. Product information includes key product introduction words, product videos, product pictures and product prices;

[0085] The data collection module is set up as an existing questionnaire. Customers fill in relevant content according to the prompts, and the key product introduction words, pictures, videos and product positioning (price and other data) are classified and displayed, so as to obtain product information and customer requirements, grasp the user's thinking, and design advertisements that meet user requirements.

[0086] The steps for preprocessing the acquired product information and obtaining product positioning are as follows:

[0087] Get the name and price of the advertisement product;

[0088] Get the average price of the same product on the market, and calculate the price difference ratio between the advertised product price and the average price of the same product on the market. The specific calculation formula is as follows:

[0089]

[0090] In the formula, is the price difference ratio between the advertised product price and the average price of the same product on the market, For the advertised product price, is the average price of the same product in the market;

[0091] Compare the price difference ratio CY with the preset positioning range to obtain the product positioning;

[0092] The preset positioning ranges include the low-end range (-100%, -15%), the affordable range [-15%, 15%], the high-end range (15%, 100%) and the top range [100%, +∞).

[0093] Different channels have different detailed requirements for advertising, different requirements for the number of pictures and advertising duration, and different product positioning (the same product may be used for different groups, and the requirements of different groups are different, so precise positioning is required). Only by understanding the product positioning can you develop advertisements that meet the requirements.

[0094] For example, shoes at different price ranges have completely different introductions. Low-end shoes often attract customers from the perspective of color, while high-end shoes mainly attract customers through shape, materials and special designs. Therefore, it is necessary to fully understand the positioning of the product in the early stage of advertising design.

[0095] The present invention records an automated advertising content generation system based on a big data model, which can conduct a comprehensive analysis of product information uploaded by customers, obtain customers' positioning of products, and perform channel analysis and attribute analysis on product positioning data in combination with the big data model, accurately obtain advertising channels and the average advertising duration corresponding to the advertising channels, grasp user needs more accurately and reasonably, and generate advertisements that meet customer needs at one time, with good use effect and good prospects for use.

[0096] Further analysis of advertising demand

[0097] After obtaining product information and user requirements, in order to better advertise the product, it is necessary to conduct corresponding analysis based on existing advertising channels and create advertisements that are more suitable for the product. This process relies on the preliminary analysis module.

[0098] Preliminary analysis module: used to use big data models to perform channel analysis and attribute analysis on product positioning data, obtain advertising channels and the average advertising duration corresponding to the advertising channels, and select advertising production templates from the advertising template library according to the product type, advertising duration and attribute analysis results to obtain the number of photos L required for the advertisement;

[0099] When the big data model performs channel analysis on product positioning data, it analyzes the visit sources of users of the same product with the same positioning, measures the channel quality by evaluating basic indicators such as the number of visiting users, number of visits, visit duration, and bounce rate, and customizes conversion goals to comprehensively measure the conversion effect of the channel. The channel analysis model helps companies optimize marketing strategies, improve conversion rates and user experience, and can analyze user behavior in different channels, discover pain points and room for improvement in user experience, and thus optimize products and services.

[0100] Through channel analysis, we can get the advertising channels, and once we get the advertising channels, we can naturally get the corresponding average advertising duration.

[0101] When the big data model performs attribute analysis on product positioning data, it can classify users according to their attributes and view the statistics of users in different groups at the same time. Attribute analysis can help developers grasp the characteristics of a specified player group from multiple angles and in all directions, and grasp the composition and preferences of the overall players at a macro level, thus providing a basis for refined operations.

[0102] Moreover, attribute analysis mainly marks various characteristics of users, understands the distribution structure of user attributes or attribute intersections, and then hierarchically marks users for subsequent product and operation actions. Attribute analysis can include basic attributes (such as name, age, family, marital status, gender, highest education level, etc.), social relationships (such as married or unmarried, whether there are children, etc.) and behavioral characteristics (such as registration time, source of the user's first visit channel, etc.).

[0103] For example, through attribute analysis, you can quickly view the distribution of users in various provinces, draw a user portrait of the overall user group, select the best templates for related products in the provinces with the largest number of people, and then further filter them based on the average advertising duration to obtain a template production template suitable for the advertisement.

[0104] Channel analysis and attribute analysis are two commonly used types of big data model analysis (event analysis, attribute analysis, channel analysis, session analysis, retention analysis, attribution analysis, funnel analysis, path analysis and distribution analysis). Therefore, the specific analysis process will not be discussed here. After searching, it is found that there is currently no application of big data model analysis in the field of advertising production.

[0105] The advertising production template recorded in the present invention is mainly a video production template. A template of a picture combination similar to that recorded in the existing patent document recorded in the background technology part of this application may also be used. However, the video production template can better reflect the characteristics of the product, and its advertising effect is better than other templates. Therefore, this solution is discussed based on the video production template.

[0106] Product information processing and analysis

[0107] After understanding the basic needs of advertising, in order to make the produced advertisements better and more in line with customer requirements, the information provided by the customer (voice, video, text and pictures) will be further sorted and analyzed. During the analysis process, the voice will be recognized into text to facilitate analysis. This process is based on the video processing module and the picture analysis module. The pictures are first screened and then analyzed and processed.

[0108] Video processing module: used to perform frame-by-frame restoration on product videos, and then classify pictures according to key product introduction words, and make comprehensive evaluations on the classified pictures, calculate the evaluation values, and then extract the N pictures with the highest evaluation values ​​in each category, N < 5, and then uniformly extract the background colors of the pictures;

[0109] In order to better reflect the characteristics of the product, we take the approach of frame-by-frame restoration of the product video to obtain more representative pictures.

[0110] When performing frame-by-frame restoration on a product video, P pictures are randomly extracted per second, where 3<P<6. The larger the value of P, the more pictures are obtained, but the amount of data for subsequent picture calculation and processing will increase. Therefore, if the product is positioned in the low-end and affordable range, 4 is selected, while if the product is positioned in the high-end and top range, 5 is selected.

[0111] The steps to categorize images according to product key introduction words are as follows:

[0112] Extract the key parts of the product from the key words of the product introduction;

[0113] For example, if the product is a pair of shoes, and it mentions that the upper is made of mesh material and the sole is made of high-elastic material, then the key parts of the product will be the upper introduction picture and the sole introduction picture.

[0114] The key parts of the product, product packaging and overall shape of the product are used as labels to classify the extracted P pictures and the obtained product pictures.

[0115] For example, if the product is a pair of shoes, the product packaging is a shoe box, and the overall shape of the product is a three-dimensional image and six-view image of the product. At this time, the categories for classifying the product images are upper introduction image, sole introduction image, product packaging image, three-dimensional image and six-view image of the product.

[0116] For products without special shapes, the partial classification in the six views can be removed and the three views can be used.

[0117] The formula for calculating the judgment value is as follows:

[0118]

[0119] In the formula, is the calculated judgment value, is the area of ​​the product in the picture, is the total area of ​​the image, The ratio of product area to image area in the set advertising image. , For the clarity of the picture, This is the standard definition of the advertising image. is the average color difference between the product edge color and the background color in the image, The color difference coefficient for the preset advertising. ;

[0120] The above is based on a comprehensive consideration of product proportion, picture clarity, and color difference between the product and the background.

[0121] The higher the proportion of the product, the more prominent it is. However, if the proportion reaches a certain level, the overall beauty of the picture will be reduced. Therefore, in general promotion, the product occupies 60-80% of the picture. Usually 75 is taken.

[0122] Image clarity can directly reflect the quality of the image.

[0123] The greater the color difference between the product and the background, the easier it is to adjust the image when the background is unified later. The image will also look better after adjustment, and white edges will not easily appear around the product.

[0124] The pictures judged by the above evaluation values ​​are easy to adjust, clear, and have comprehensive product display, and the resulting advertisements have good effects.

[0125] The calculation formula for the average color difference between the product edge color and the image background color in the image is as follows:

[0126]

[0127] In the formula, is the brightness difference between the edge color of the product in the picture and the background color of the picture. is the chroma difference between the edge color of the product in the picture and the background color of the picture. is the hue difference between the edge color of the product in the picture and the background color of the picture. is the brightness weight factor, is the chroma weight factor, 1≤j / c≤2, where , and The saturation of the image.

[0128] The color difference is calculated using the CMC color difference calculation formula.

[0129] After the pictures are screened out, they need to be further processed and analyzed. The steps of processing and analyzing the pictures are based on the picture analysis module: used to analyze the extracted pictures, select the pictures with the highest evaluation value, and then calculate the matching value between the pictures in each category and the pictures with the highest evaluation value, and retrieve the K pictures with the highest matching value in each category, summarize and add them to obtain the template requirement L pictures, and obtain the product introduction picture set;

[0130] The steps to unify the extracted image background are as follows:

[0131] Filter advertisements of similar products by product type;

[0132] Further screen the advertisements of similar products according to product positioning and extract advertisements of similar products with the same positioning;

[0133] For example, the background color is red. Red symbolizes enthusiasm, vitality, danger and warning. It can attract people's attention and is often used to convey a sense of urgency and vitality. In advertising design, red is often used to emphasize the characteristics of a brand or product, making it more prominent and eye-catching. It is suitable for industries such as catering, clubs, games and electronics.

[0134] The background color is blue, which represents professionalism, stability, and reliability. Using blue in advertising can give people an impression of thoughtfulness and enhance the credibility of the brand or product. In addition, blue can also convey a sense of calmness and peace, which is suitable for advertisements in the fields of finance, technology, etc.

[0135] The background color is yellow, which represents vitality, happiness and innovation, and can give people a relaxed and happy feeling. In advertising design, yellow is often used to present the highlights and features of products or services to attract consumers' attention. It is suitable for industries such as catering, electronics, home furnishings and logistics.

[0136] The background color is green, which can give consumers a sense of relaxation and comfort, and increase the credibility of products or brands. In some advertisements for health foods, eco-tourism, etc., green is often used to convey sustainable development and environmental awareness, and is often used in the service industry, health care industry and medical institutions.

[0137] The background color is orange, which can attract people's attention and give people a positive and cheerful feeling. In some advertisements for sports goods, beverages, etc., orange is often used to enhance the vitality and youthfulness of the brand.

[0138] The background color is purple, which is often used to attract consumers with unique tastes, individuality and uniqueness. Purple can also convey a mysterious and attractive charm, which is suitable for advertisements of some luxury goods, works of art, etc.

[0139] The above background color is used to make the advertisement less obtrusive and achieve good effect.

[0140] Get the background color of the pictures in the advertisements of similar products with the same positioning, select the majority background color (the color analyzed above) as the background color of this product advertisement, and get the RGB value of the majority background color;

[0141] Adjust the RGB value of the extracted image background to the RGB value of the majority background color.

[0142] The color adjustment is achieved by adjusting the RGB value. The RGB color mode is a color standard in the industry and is an existing technology, so no further description is given.

[0143] Because pictures are affected by light and other angles, they will not be completely consistent. Therefore, there will be a certain color difference in the products in the pictures. In order to make the advertisement more natural and harmonious, the pictures must be screened to select pictures with colors closer to the products. Therefore, the coordination value between the pictures needs to be calculated.

[0144] The formula for calculating the matching value between the image in each category and the image with the highest judgment value is as follows:

[0145]

[0146] In the formula, is the matching value between the image and the image with the highest judgment value, It is the color difference between the product in the picture with the highest judgment value in each category.

[0147] The higher the matching value, the closer the product colors in the picture are.

[0148] Retrieve the K pictures with the highest matching value in each category, summarize and add them to obtain L photos required for the template, and when obtaining the product introduction picture set, if L=K×μ, where μ is the number of categories (for example, there are 10 categories of shoes, and 20 photos are required, two photos for each category), then the number of pictures in each category in the product introduction picture set is K, and no addition is required. If L=K×μ+A, A<K, A is the number of pictures to be added, and A is the A pictures with the highest matching value extracted from the remaining N×K-K×μ pictures with the same background color. For example, 13 photos are needed, then one photo is taken from each category, and the remaining 3 photos are further screened from the remaining photos. The screened photos are summarized to obtain the picture set.

[0149] The present invention records an automated advertising content generation system based on a big data model, which can conduct a comprehensive analysis of product information uploaded by customers, disassemble videos into pictures, and judge all pictures, select representative high-quality pictures, unify the backgrounds, and then further analyze the matching values ​​of the high-quality pictures to obtain pictures required for advertisements. This processing makes the generated advertisements uniform in style and more coordinated in overall layout, which can fully highlight the characteristics of the products, improve the quality of advertisements, have good use effects, and have good prospects for use.

[0150] Ad generation

[0151] After the templates and images required for advertisement generation are processed, it is necessary to proceed to the advertisement generation step, which is based on the advertisement integration module.

[0152] Advertisement integration module: used to insert product key introduction words into the pictures in the product introduction picture set, and calculate the corresponding playback time of each picture according to the evaluation value, insert the product introduction picture set into the advertisement production template according to the calculated playback time, and input the advertising slogan generated by the big data model for the product key introduction words to generate advertisements;

[0153] The generation of advertising slogans is a basic application of big data technology and is very common in existing technologies. Therefore, the generation of advertising slogans and advertising voices will not be described in detail. The combination of audio and video is also a commonly used method in the advertising field, so it will not be described in detail.

[0154] The steps to insert product key words on photos are as follows:

[0155] Adjust the font size of key introductory words so that the font size of the key introductory words occupies 1 / B of the background area of ​​the image in the advertisement, 5≤B≤10; this step can effectively avoid the difference between fonts that are too large or too small.

[0156] Adjust the resized font of the key words to the background of the picture;

[0157] Adjust the font color so that the color difference between the font color and the image background color is greater than F, F>20.

[0158] This step makes the font clear and can show the characteristics of the product more clearly.

[0159] Users can also describe the font type and size during the collection step, so that the generated advertisements are more in line with customer requirements.

[0160] The formula for calculating the playback time of each picture based on the evaluation value is as follows:

[0161]

[0162] In the formula, is the playing time of the i-th picture in the advertisement, is the judgment value of the i-th picture, is the sum of the evaluation values ​​of L pictures, The duration of the advertisement.

[0163] The better the picture, the longer it will be displayed compared to poor quality pictures, which enables a more reasonable allocation of resources.

[0164] Post-production adjustment

[0165] After the advertisement is generated, not all customers will accept it and will make suggestions for modification. This process relies on the local adjustment module.

[0166] Local adjustment module: used to obtain user feedback, calculate the user's modified value, compare the modified value with the set standard value, and execute the corresponding adjustment strategy based on the comparison result.

[0167] The local adjustment module adjusts and regenerates the advertisement according to the customer's requirements until the generated advertisement meets the user's requirements.

[0168] The formula to calculate the user's modified value is as follows:

[0169]

[0170] In the formula, is the user's modified value. The number of categories modified for the user, is the classification difficulty coefficient, , The number of images modified for the specified purpose;

[0171] The classification difficulty coefficient is set according to the number of categories and the number of images that need to be classified.

[0172] When the user makes large changes, it is more efficient to choose the regeneration solution.

[0173] According to the comparison results, the corresponding adjustment strategies are as follows:

[0174] If the calculated If it is less than the standard value STD, the advertisement content will be partially modified according to the user's requirements;

[0175] If the calculated If it is greater than or equal to the standard value STD, the image will be reclassified and the advertisement will be regenerated according to customer requirements and key product introduction words.

[0176] For example, in a shoe advertisement, the previous advertisement was for white shoes, and each color of shoe has a separate advertisement, so it needs to be regenerated.

[0177] For example, if the fonts on some pictures need to be adjusted, only the pictures need to be modified individually. In this case, only partial modifications are required.

[0178] Because some customers tend to get entangled with certain images when generating advertisements, establishing this mechanism can increase the speed of advertisement generation, making advertisement production more efficient and the effect of use better.

[0179] The present invention records an automated advertising content generation system based on a big data model, which can automatically screen out high-quality pictures, reasonably divide picture time, and combine pictures into advertising videos, which can more clearly and completely show the characteristics of products. The advertisement generation process does not require manual participation of customers, and can be automatically generated, which is easy to use. In addition, it can subsequently further feedback and analyze customer requirements, and perform fine-tuning or regeneration of advertisements. The whole process is relatively simple, can quickly generate advertisements that meet customer requirements, has good use effects, and has good prospects for use.

[0180] The above formula is a formula obtained by collecting a large amount of data and performing software simulation to obtain the most recent real situation. The preset parameters in the formula are set by technicians in this field according to actual conditions.

[0181] The above embodiments may be implemented in whole or in part by software, hardware, firmware or any other combination thereof. When implemented using software, the above embodiments may be implemented in whole or in part in the form of a computer program product. A person of ordinary skill in the art may appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein may be implemented in electronic hardware, or in a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution.

[0182] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, and may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0183] The above description is only a specific implementation manner of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application.

Claims

1. An automated advertising content generation system based on a big data model, characterized in that: include: Data collection module: used to obtain product information provided by users and pre-process the obtained product information to obtain product positioning. Product information includes key product introduction words, product videos, product pictures and product prices; Preliminary analysis module: used to use big data models to perform channel analysis and attribute analysis on product positioning data, obtain advertising channels and the average advertising duration corresponding to the advertising channels, and select advertising production templates from the advertising template library according to the product type, advertising duration and attribute analysis results to obtain the number of photos L required for the advertisement; Video processing module: used to perform frame-by-frame restoration on product videos, and then classify pictures according to key product introduction words, and make comprehensive evaluations on the classified pictures, calculate the evaluation values, and then extract the N pictures with the highest evaluation values ​​in each category, N < 5, and then uniformly extract the background colors of the pictures; Image analysis module: used to analyze the extracted images, select the image with the highest evaluation value, and then calculate the matching value between the image in each category and the image with the highest evaluation value, and retrieve the K images with the highest matching value in each category, summarize and add them to obtain the L photos required for the template, and obtain the product introduction picture set; Advertisement integration module: used to insert product key introduction words into the pictures in the product introduction picture set, and calculate the corresponding playback time of each picture according to the evaluation value, insert the product introduction picture set into the advertisement production template according to the calculated playback time, and input the advertising slogan generated by the big data model for the product key introduction words to generate advertisements; Local adjustment module: used to obtain user feedback, calculate the user's modified value, compare the modified value with the set standard value, and execute the corresponding adjustment strategy based on the comparison result.

2. The automatic advertising content generation system based on the big data model according to claim 1, characterized in that: The steps for preprocessing the acquired product information and obtaining product positioning are as follows: Get the name and price of the advertisement product; Get the average price of the same product on the market, and calculate the price difference ratio between the advertised product price and the average price of the same product on the market. The specific calculation formula is as follows: ； In the formula, is the price difference ratio between the advertised product price and the average price of the same product on the market, For the advertised product price, is the average price of the same product in the market; Compare the price difference ratio CY with the preset positioning range to obtain the product positioning; The preset positioning ranges include the low-end range (-100%, -15%), the affordable range [-15%, 15%], the high-end range (15%, 100%) and the top range [100%, +∞).

3. The automatic advertising content generation system based on the big data model according to claim 2 is characterized by: When performing frame-by-frame restoration on a product video, P pictures are randomly extracted per second, 3<P<6; The steps to categorize images according to product key introduction words are as follows: Extract the key parts of the product from the key words of the product introduction; The key parts of the product, product packaging and overall shape of the product are used as labels to classify the extracted P pictures and the obtained product pictures.

4. The automatic advertising content generation system based on the big data model according to claim 3 is characterized by: The formula for calculating the judgment value is as follows: ； In the formula, is the calculated judgment value, is the area of ​​the product in the picture, is the total area of ​​the image, The ratio of product area to image area in the set advertising image. , For the clarity of the picture, This is the standard definition of the advertising image. is the average color difference between the product edge color and the background color in the image, The color difference coefficient for the preset advertising. ; The calculation formula for the average color difference between the product edge color and the image background color in the image is as follows: ； In the formula, is the brightness difference between the edge color of the product in the picture and the background color of the picture. is the chroma difference between the edge color of the product in the picture and the background color of the picture. is the hue difference between the edge color of the product in the picture and the background color of the picture. is the brightness weight factor, is the chroma weight factor, 1≤j / c≤2, where , and The saturation of the image.

5. The automatic advertising content generation system based on the big data model according to claim 4 is characterized by: The steps to unify the extracted image background are as follows: Filter advertisements of similar products by product type; Further screen the advertisements of similar products according to product positioning and extract advertisements of similar products with the same positioning; Get the background color of the pictures in the advertisements of the same type of products with the same positioning, select the majority background color as the background color of this product advertisement, and get the RGB value of the majority background color; Adjust the RGB value of the extracted image background to the RGB value of the majority background color.

6. The automatic advertising content generation system based on the big data model according to claim 5 is characterized by: The formula for calculating the matching value between the image in each category and the image with the highest judgment value is as follows: ； In the formula, is the matching value between the image and the image with the highest judgment value, It is the color difference between the product in the picture with the highest judgment value in each category.

7. The automatic advertising content generation system based on the big data model according to claim 6 is characterized by: Retrieve the K pictures with the highest matching value in each category, summarize and add them to obtain the L photos required for the template, and when obtaining the product introduction picture set, if L=K×μ, where μ is the number of categories, then the number of pictures in each category in the product introduction picture set is K, and no addition is needed. If L=K×μ+A, A<K, A is the number of pictures that need to be added, and A is the A pictures with the highest matching value extracted from the remaining N×K-K×μ pictures with the same background color.

8. The automatic advertising content generation system based on the big data model according to claim 7 is characterized by: The steps to insert product key words on photos are as follows: Adjust the font size of key introductory words so that the font size of the key introductory words occupies 1 / B of the image background area in the advertisement, 5≤B≤10; Adjust the resized font of the key words to the background of the picture; Adjust the font color so that the color difference between the font color and the image background color is greater than F, F>20.

9. The automatic advertising content generation system based on the big data model according to claim 8, characterized in that: The formula for calculating the playback time of each picture based on the evaluation value is as follows: ； In the formula, is the playing time of the i-th picture in the advertisement, is the judgment value of the i-th picture, is the sum of the evaluation values ​​of L pictures, The duration of the advertisement.

10. The automatic advertising content generation system based on the big data model according to claim 9, characterized in that: The formula to calculate the user's modified value is as follows: ； In the formula, is the user's modified value. The number of categories modified for the user, is the classification difficulty coefficient, , The number of images modified for the specified purpose; According to the comparison results, the corresponding adjustment strategies are as follows: If the calculated If it is less than the standard value STD, the advertisement content will be partially modified according to the user's requirements; If the calculated If it is greater than or equal to the standard value STD, the image will be reclassified and the advertisement will be regenerated according to customer requirements and key product introduction words.

Images