Advertisement recommendation system and method based on crowd dressing appearance and body information

Abstract

This invention discloses an advertising recommendation system and method based on people's clothing, appearance, and body posture information, belonging to the field of advertising recommendation technology. The system includes: identifying a target group; acquiring initial images of the target group using a data acquisition device; generating target group images after preprocessing; extracting body posture features from the target group images; retrieving preset classification labels based on a crowd classification algorithm and determining their correspondence with body posture features; dividing the target group into multiple target sub-groups, each sub-group corresponding to a preset classification label; retrieving a preset advertising library based on an advertising recommendation algorithm; matching target advertisements in the preset advertising library corresponding to the preset classification labels and recommending them to the target sub-groups; collecting recommendation feedback data and optimizing the crowd classification algorithm and advertising recommendation algorithm; thereby enabling accurate classification based on the clothing style of outdoor people, recommending suitable advertising content, improving the accuracy of advertising recommendations, and enhancing user experience.

Description

Technical Field

[0001] This invention relates to the field of advertising recommendation technology, specifically to an advertising recommendation system and method based on information about people's clothing, appearance, and body shape. Background Technology

[0002] With the rapid development of internet technology, the advertising industry is gradually moving towards precision and intelligence. Advertising recommendation systems, as a bridge connecting consumers and brands, are becoming increasingly important. With the rapid development of big data, cloud computing, and artificial intelligence technologies, advertising recommendation systems have undergone an evolution from basic to complex, and from simple to diverse.

[0003] Traditional advertising recommendation methods often rely on user browsing history and purchase records for recommendations. However, in the out-of-home advertising sector, this method cannot directly access users' online behavior data, resulting in significant limitations. Existing technologies recommend ads by recognizing users' facial features, gender, and age. However, users' clothing, appearance, and physique also significantly influence ad preferences. For example, people dressed formally may be more interested in high-end brand ads, while those dressed casually may prefer travel or entertainment ads. Therefore, existing advertising recommendation methods often struggle to accurately identify target audiences and provide precise ad recommendations based on demographic characteristics, leading to wasted advertising resources and a degraded user experience.

[0004] Therefore, how to accurately categorize outdoor users based on their clothing styles and recommend suitable advertising content to improve the accuracy of ad recommendations and enhance user experience has become an urgent problem to be solved. Summary of the Invention

[0005] To address the aforementioned issues, this invention proposes an advertising recommendation system and method based on information about people's clothing, appearance, and body shape. This system enables precise classification of people's clothing, appearance, and body shape, improving the accuracy and effectiveness of advertising recommendations and reducing information waste. Furthermore, by combining user profiles and advertising content, it achieves personalized advertising recommendations, enhancing user experience and advertising conversion rates, avoiding interference from ineffective advertisements. Moreover, by collecting user feedback data in real time and continuously optimizing the recommendation algorithm, the system's accuracy and efficiency can be improved, thereby increasing advertising conversion rates.

[0006] To achieve the above objectives, the present invention is implemented through the following technical solution:

[0007] The present invention relates to an advertising recommendation system based on population clothing, appearance, and body shape information, the system comprising:

[0008] The acquisition module is used to identify the target population and acquire the initial population image corresponding to the target population through the acquisition device;

[0009] The preprocessing module is used to perform preprocessing operations on the initial crowd image to generate the target crowd image;

[0010] The extraction module is used to extract the body features corresponding to the target population in the target population image;

[0011] The classification module is used to retrieve preset classification labels based on a population classification algorithm, and divide the target population into multiple target sub-populations according to the correspondence between the preset classification labels and body features, with each target sub-population corresponding to a preset classification label;

[0012] The recommendation module is used to retrieve a preset advertising library based on an advertising recommendation algorithm, match the target advertisement in the preset advertising library corresponding to the preset category tag, determine the target sub-group corresponding to the preset category tag, and recommend the target advertisement to the target sub-group.

[0013] The feedback module is used to collect recommendation feedback data from the target sub-group and optimize the audience classification algorithm and the advertisement recommendation algorithm based on the recommendation feedback data.

[0014] A further improvement of the present invention is that the acquisition module includes:

[0015] An installation unit is used to determine a target area and install the acquisition device in the target area;

[0016] The acquisition unit is used to determine the target population and acquire the initial population image corresponding to the target population through the acquisition device;

[0017] A storage unit is used to store the acquired initial crowd images into an initial crowd image database.

[0018] A further improvement of the present invention is that the preprocessing module includes:

[0019] The retrieval unit is used to retrieve the initial crowd images stored in the initial crowd image database;

[0020] The generation unit is used to perform denoising and enhancement processing on the initial crowd image to generate the target crowd image.

[0021] A further improvement of the present invention is that the extraction module extracts the body features corresponding to the target population in the target population image using the following feature extraction algorithm:

[0022]

[0023] In the formula, Indicates a spatial location index on a physical feature; Describe physical characteristics; Indicating the spatial location index of body features The value at; and Indicates the spatial location index on the convolution kernel; Represents images of the target audience; Indicates the spatial location index of the target crowd image. The value at; Represents the convolution kernel; Indicates the spatial index of the convolution kernel The weight value at the location; The term ΣmΣn represents the summation of all elements in the convolution kernel; Σ represents the summation sign.

[0024] Based on the feature extraction algorithm, feature extraction is performed on the target crowd image to obtain the body posture features corresponding to the target crowd.

[0025] A further improvement of the present invention is that: the classification module retrieves preset classification labels using the following population classification algorithm, and divides the target population into multiple target sub-populations based on the correspondence between the preset classification labels and body shape features, with each target sub-population corresponding to one preset classification label:

[0026] Calculate the information gain corresponding to the input body features:

[0027]

[0028] In the formula, This represents the target population dataset; Represents the input body posture features; Indicates information gain; Represents entropy; Represents the target population dataset Entropy; Represents the set of all values ​​for the input body posture features; Represents any value of the input body posture feature; This represents any value of the input body posture feature. Segmented target sub-group dataset; Represents the target sub-group dataset The number of users; Represents the target population dataset The number of users; Represents the body features of the input The corresponding gain coefficient; Σ represents the summation symbol;

[0029] in

[0030]

[0031] In the formula, This represents the target population dataset; This represents the entropy of the target population dataset; Indicates the total number of preset category labels; Indicates the first The proportion of the target sub-group dataset corresponding to each preset category label in the target group dataset; Represents a logarithmic function; Σ represents the summation symbol;

[0032] For the target population dataset Calculate the body features The corresponding information gain Select the body features with the highest information gain. The segmentation feature is used as the segmentation feature for the current node; based on the different values ​​of the segmentation feature, the target population dataset is... The dataset is divided into multiple target sub-groups. For each of the target sub-group datasets Repeat the above steps until the stopping condition is met, i.e., the target sub-group dataset is reached. All users belong to the same preset category label or the target sub-group dataset. The number of users in the system is less than a preset threshold;

[0033] When the above recursive process stops, the current node can be marked as a leaf node, i.e., the target sub-group dataset. and determine the target subgroup dataset. The corresponding preset category label.

[0034] A further improvement of the present invention is that: the recommendation module, using the following advertising recommendation algorithm, retrieves a preset advertising library, matches the target advertisement in the preset advertising library corresponding to the preset category tag, determines the target sub-group corresponding to the preset category tag, and recommends the target advertisement to the target sub-group:

[0035] Calculate the first matching degree function between the preset category tag and the target advertisement:

[0036]

[0037] In the formula, This function represents the first degree of match between the preset category tags and the target advertisement. This represents a predefined category label vector; Represents the target ad vector; This represents the dot product of the preset category label vector and the target ad vector; This represents the modulus of the preset category label vector; The modulus of the target ad vector;

[0038] Calculate the second matching degree function between the preset category label and the target subgroup:

[0039]

[0040] In the formula, This function represents the second degree of matching between the preset category labels and the target subgroup. This represents a predefined category label vector; Represents the target subgroup population vector; This represents the dot product of the preset category label vector and the target subgroup vector; This represents the modulus of the preset category label vector; The modulus of the target subgroup's population vector;

[0041] Based on the first matching degree function and the second matching degree function, calculate the recommendation matching degree function between the target advertisement and the target sub-group:

[0042]

[0043] In the formula, This represents a predefined set of category tags; A function representing the recommendation matching degree between the target advertisement and the target sub-audience; This parameter represents the weighting between the target ad and the target sub-audience. This function represents the first degree of match between the preset category tags and the target advertisement. This represents the second matching degree function between the preset category labels and the target subgroup; Σ represents the summation symbol;

[0044] Based on the recommendation matching degree function, the recommendation matching degree between the target sub-group and the target advertisement is calculated, the target advertisement corresponding to the maximum recommendation matching degree is determined, and recommended to the target sub-group.

[0045] The present invention provides an advertising recommendation method based on population clothing, appearance, and body shape information, the method comprising:

[0046] Identify the target population and acquire initial population images corresponding to the target population using acquisition devices;

[0047] The initial crowd image is preprocessed to generate the target crowd image;

[0048] Extract the body posture features corresponding to the target population from the target population image;

[0049] Based on the crowd classification algorithm, preset classification labels are retrieved. According to the correspondence between the preset classification labels and body features, the target crowd is divided into multiple target sub-groups, and each target sub-group corresponds to one preset classification label.

[0050] Based on the advertising recommendation algorithm, a preset advertising library is retrieved, and the target advertisement in the preset advertising library corresponding to the preset category tag is matched to determine the target sub-group corresponding to the preset category tag, and the target advertisement is recommended to the target sub-group.

[0051] Collect recommendation feedback data from the target sub-groups, and optimize the audience classification algorithm and the ad recommendation algorithm based on the recommendation feedback data.

[0052] The beneficial effects of this invention are as follows: This invention proposes an advertising recommendation system and method based on people's clothing, appearance, and body posture information. Through a collection module, a preprocessing module, an extraction module, a classification module, a recommendation module, and a feedback module, the target group can be determined. An initial group image corresponding to the target group is acquired through a collection device; the initial group image is preprocessed to generate a target group image; body posture features corresponding to the target group are extracted from the target group image; based on a group classification algorithm, preset classification labels are retrieved; according to the correspondence between the preset classification labels and body posture features, the target group is divided into multiple target subgroups, and each target subgroup corresponds to a preset classification label; based on an advertising recommendation algorithm, a preset advertising database is retrieved. The system matches target ads in a pre-defined ad library with pre-defined category tags, identifies target sub-groups corresponding to these tags, and recommends the target ads to these sub-groups. It collects recommendation feedback data from these sub-groups and optimizes both the audience classification and ad recommendation algorithms based on this data. This enables precise classification of audiences by clothing, appearance, and body type, improving the accuracy and effectiveness of ad recommendations, reducing information waste, and combining user profiles with ad content to achieve personalized ad recommendations. This enhances user experience and ad conversion rates, avoids interference from ineffective ads, and continuously optimizes the recommendation algorithm by collecting user feedback data in real time, thereby improving the accuracy and efficiency of the recommendation system and ultimately increasing ad conversion rates. Attached Figure Description

[0053] Figure 1 This is a schematic diagram of the structure of an advertising recommendation system based on people's clothing, appearance, and body shape information provided in an embodiment of the present invention; Figure 2 This is a schematic diagram of the structure of the data collection module of the advertising recommendation system based on people's clothing, appearance and body posture information provided in an embodiment of the present invention; Figure 3 This is a schematic diagram of the preprocessing module of an advertising recommendation system based on people's clothing, appearance, and body shape information provided in an embodiment of the present invention. Figure 4 A flowchart illustrating the advertising recommendation method based on people's clothing, appearance, and body shape information provided in an embodiment of the present invention; Figure 5 This is a schematic diagram of the hardware structure of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0054] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0055] See Figure 1 This is a schematic diagram of the structure of an advertising recommendation system based on people's clothing, appearance, and body shape information provided in an embodiment of the present invention. The recommendation system includes:

[0056] The acquisition module is used to identify the target population and acquire the initial population image corresponding to the target population through the acquisition device;

[0057] The preprocessing module is used to perform preprocessing operations on the initial crowd image to generate the target crowd image;

[0058] The extraction module is used to extract the body features corresponding to the target population in the target population image;

[0059] The classification module is used to retrieve preset classification labels based on a population classification algorithm, and divide the target population into multiple target sub-populations according to the correspondence between the preset classification labels and body features, with each target sub-population corresponding to a preset classification label;

[0060] The recommendation module is used to retrieve a preset advertising library based on an advertising recommendation algorithm, match the target advertisement in the preset advertising library corresponding to the preset category tag, determine the target sub-group corresponding to the preset category tag, and recommend the target advertisement to the target sub-group.

[0061] The feedback module is used to collect recommendation feedback data from the target sub-group and optimize the audience classification algorithm and the advertisement recommendation algorithm based on the recommendation feedback data.

[0062] Understandably, the data acquisition module uses image acquisition devices, such as high-definition cameras or infrared sensors, to accurately capture dynamic images of the target crowd and generate an initial set of crowd images. This acquisition device boasts high sensitivity and accuracy, and can protect user privacy, preventing information leakage.

[0063] The preprocessing module uses image processing techniques, such as noise reduction, contrast enhancement, and image segmentation, to perform in-depth optimization on the initial crowd image, remove redundant information, and improve image quality, thereby generating a clear and accurate image of the target crowd.

[0064] The extraction module can accurately extract the physical characteristics of the target population from the optimized target population image, including physical features such as height, body shape, and gait, as well as clothing features such as clothing style and accessory selection. This facilitates subsequent population classification.

[0065] The classification module, based on a population classification algorithm and combined with a pre-defined classification label system (such as age group, gender, and interests), can divide the target population into multiple target subgroups with distinct characteristics by determining the correspondence between physical features and pre-defined classification labels. This population classification algorithm possesses high accuracy and generalization ability and can be continuously optimized to adapt to market changes.

[0066] The recommendation module can use advertising recommendation algorithms to intelligently match target ads that best match the characteristics of the target sub-group from a preset ad library, thereby improving the accuracy and effectiveness of ad recommendations, reducing information waste, and combining user profiles and ad content to achieve personalized ad recommendations, improve user experience and ad conversion rates, and avoid interference from invalid ads.

[0067] The feedback module collects and analyzes feedback data from target sub-groups regarding recommended ads, including click-through rates, conversion rates, and satisfaction surveys. This data allows for precise evaluation of ad recommendation effectiveness, enabling continuous optimization of both the audience segmentation and ad recommendation algorithms. Through iterative optimization, the accuracy and efficiency of the recommendation system can be improved, user satisfaction increased, and ultimately, ad conversion rates enhanced, resulting in more efficient and precise personalized marketing.

[0068] See Figure 2 The acquisition module 2 includes an installation unit 21, an acquisition unit 22, and a storage unit 23.

[0069] An installation unit is used to determine a target area and install the acquisition device in the target area;

[0070] The acquisition unit is used to determine the target population and acquire the initial population image corresponding to the target population through the acquisition device;

[0071] A storage unit is used to store the acquired initial crowd images into an initial crowd image database.

[0072] It should be noted that the installation unit can accurately locate and delineate specific target areas. The unit possesses a high degree of spatial awareness and environmental adaptability, ensuring that the selected area effectively covers monitoring needs while maximizing resource utilization. By integrating Geographic Information Technology (GIS) with automated deployment algorithms, the unit can automatically assess factors such as terrain, lighting conditions, and pedestrian density to precisely plan the optimal installation locations for the data acquisition devices. These devices can be high-definition cameras, infrared sensors, etc., capable of comprehensively capturing dynamic information within the target area.

[0073] The acquisition unit can define the target population range based on pre-set filtering conditions, such as age, gender, and behavioral characteristics, combined with artificial intelligence technologies such as facial recognition and behavior recognition, enabling efficient identification and tracking of specific groups. The acquisition device can capture and generate initial images of the target population in real time, while ensuring image quality and preventing distortion or occlusion.

[0074] The storage unit securely and efficiently stores the initial crowd images transmitted by the acquisition unit into the initial crowd image database. This storage unit possesses massive data storage capacity, as well as high-speed read / write, data redundancy backup, and encryption protection functions, ensuring data security and integrity. The storage unit often employs a distributed storage architecture, combined with cloud storage technology, enabling flexible expansion and efficient management of data resources. Furthermore, through regular data maintenance and optimization strategies, such as data cleaning, compression, and index rebuilding, the database's query efficiency and data quality can be further improved.

[0075] Through the above methods, the installation unit, acquisition unit, and storage unit work in close coordination to achieve effective monitoring of the target area, accurate identification of the target population, and real-time storage of image data, which facilitates subsequent classification and recommendation processing.

[0076] See Figure 3 The preprocessing module 3 includes a retrieval unit 31 and a generation unit 32:

[0077] The retrieval unit is used to retrieve the initial crowd images stored in the initial crowd image database;

[0078] The generation unit is used to perform denoising and enhancement processing on the initial crowd image to generate the target crowd image.

[0079] The retrieval unit can accurately retrieve and extract specified images from the initial crowd image database. Utilizing an efficient indexing mechanism and query algorithm, this retrieval unit can quickly locate and load the initial crowd image data stored in the database.

[0080] The generation unit can receive initial crowd images from the retrieval unit and perform image preprocessing operations on these images. First, the generation unit uses denoising techniques to effectively filter out random noise and interference information in the images, preserving key crowd feature details. Then, through image enhancement algorithms, such as contrast adjustment, sharpening, and color correction, it further enhances the visual effect and clarity of the images, generating high-quality target crowd images. This improves the visual quality of the images and ensures the accuracy of subsequent classification and recognition.

[0081] The extraction module extracts the body features corresponding to the target population in the target population image using the following feature extraction algorithm:

[0082]

[0083] In the formula, Indicates a spatial location index on a physical feature; Describe physical characteristics; Indicating the spatial location index of body features The value at; and Indicates the spatial location index on the convolution kernel; Represents images of the target audience; Indicates the spatial location index of the target crowd image. The value at; Represents the convolution kernel; Indicates the spatial index of the convolution kernel The weight value at the location; The term ΣmΣn represents the summation of all elements in the convolution kernel; Σ represents the summation sign.

[0084] Based on the feature extraction algorithm, feature extraction is performed on the target crowd image to obtain the body posture features corresponding to the target crowd.

[0085] The classification module uses the following crowd classification algorithm to retrieve preset classification labels and, based on the correspondence between the preset classification labels and body features, divides the target crowd into multiple target sub-groups, with each target sub-group corresponding to one preset classification label:

[0086] Calculate the information gain corresponding to the input body features:

[0087]

[0088] In the formula, This represents the target population dataset; Represents the input body posture features; Indicates information gain; Represents entropy; Represents the target population dataset Entropy; Represents the set of all values ​​for the input body posture features; Represents any value of the input body posture feature; This represents any value of the input body posture feature. Segmented target sub-group dataset; Represents the target sub-group dataset The number of users; Represents the target population dataset The number of users; Represents the body features of the input The corresponding gain coefficient; Σ represents the summation symbol;

[0089] in

[0090]

[0091] In the formula, This represents the target population dataset; This represents the entropy of the target population dataset; Indicates the total number of preset category labels; Indicates the first The proportion of the target sub-group dataset corresponding to each preset category label in the target group dataset; Represents a logarithmic function; Σ represents the summation symbol;

[0092] For the target population dataset Calculate the body features The corresponding information gain Select the body features with the highest information gain. The segmentation feature is used as the segmentation feature for the current node; based on the different values ​​of the segmentation feature, the target population dataset is... The dataset is divided into multiple target sub-groups. For each of the target sub-group datasets Repeat the above steps until the stopping condition is met, i.e., the target sub-group dataset is reached. All users belong to the same preset category label or the target sub-group dataset. The number of users in the system is less than a preset threshold;

[0093] When the above recursive process stops, the current node can be marked as a leaf node, i.e., the target sub-group dataset. and determine the target subgroup dataset. The corresponding preset category label.

[0094] The recommendation module uses the following advertising recommendation algorithm to retrieve a preset advertising library, match the target advertisement in the preset advertising library corresponding to the preset category tag, determine the target sub-group corresponding to the preset category tag, and recommend the target advertisement to the target sub-group:

[0095] Calculate the first matching degree function between the preset category tag and the target advertisement:

[0096]

[0097] In the formula, This function represents the first degree of match between the preset category tags and the target advertisement. This represents a predefined category label vector; Represents the target ad vector; This represents the dot product of the preset category label vector and the target ad vector; This represents the modulus of the preset category label vector; The modulus of the target ad vector;

[0098] Calculate the second matching degree function between the preset category label and the target subgroup:

[0099]

[0100] In the formula, This function represents the second degree of matching between the preset category labels and the target subgroup. This represents a predefined category label vector; Represents the target subgroup population vector; This represents the dot product of the preset category label vector and the target subgroup vector; This represents the modulus of the preset category label vector; The modulus of the target subgroup's population vector;

[0101] Based on the first matching degree function and the second matching degree function, calculate the recommendation matching degree function between the target advertisement and the target sub-group:

[0102]

[0103] In the formula, This represents a predefined set of category tags; A function representing the recommendation matching degree between the target advertisement and the target sub-audience; This parameter represents the weighting between the target ad and the target sub-audience. This function represents the first degree of match between the preset category tags and the target advertisement. This represents the second matching degree function between the preset category labels and the target subgroup; Σ represents the summation symbol;

[0104] Based on the recommendation matching degree function, the recommendation matching degree between the target sub-group and the target advertisement is calculated, the target advertisement corresponding to the maximum recommendation matching degree is determined, and recommended to the target sub-group.

[0105] See Figure 4 This is a flowchart illustrating an advertising recommendation method based on clothing, appearance, and body shape information of a group of people, provided in an embodiment of the present invention. The recommendation method includes:

[0106] S1, determine the target population, and acquire the initial population image corresponding to the target population through the acquisition device;

[0107] S2, perform preprocessing operations on the initial crowd image to generate the target crowd image;

[0108] S3, extract the body features corresponding to the target population in the target population image;

[0109] S4. Based on the crowd classification algorithm, retrieve the preset classification labels, and divide the target crowd into multiple target sub-groups according to the correspondence between the preset classification labels and body features, with each target sub-group corresponding to one preset classification label;

[0110] S5, based on the advertising recommendation algorithm, retrieve the preset advertising library, match the target advertisement in the preset advertising library corresponding to the preset category tag, determine the target sub-group corresponding to the preset category tag, and recommend the target advertisement to the target sub-group;

[0111] S6. Collect recommendation feedback data from the target sub-group, and optimize the group classification algorithm and the advertisement recommendation algorithm based on the recommendation feedback data.

[0112] It should be noted that, Figure 4The execution entity of the method shown can be a software and / or hardware device. The execution entity of this application can include, but is not limited to, at least one of the following: user equipment, network equipment, etc. User equipment can include, but is not limited to, computers, smartphones, personal digital assistants (PDAs), and the aforementioned electronic devices. Network equipment can include, but is not limited to, a single network server, a server group consisting of multiple network servers, or a cloud based on cloud computing consisting of a large number of computers or network servers. Cloud computing is a type of distributed computing, consisting of a super virtual computer composed of a group of loosely coupled computers. This embodiment does not impose any limitations on this.

[0113] See Figure 5 This is a schematic diagram of the hardware structure of an electronic device provided in an embodiment of the present invention. The electronic device 50 includes: a processor 51, a memory 52, and a computer program; wherein:

[0114] The memory 52 is used to store the computer program, and the memory may also be flash memory. The computer program is, for example, an application program or functional module that implements the above method.

[0115] The processor 51 is configured to execute the computer program stored in the memory to implement the various steps performed by the device in the above method. For details, please refer to the relevant descriptions in the preceding method embodiments.

[0116] Alternatively, the memory 52 can be either standalone or integrated with the processor 51.

[0117] When the memory 52 is a device independent of the processor 51, the device may further include:

[0118] Bus 53 is used to connect the memory 52 and the processor 51.

[0119] The present invention also provides a readable storage medium storing a computer program, which, when executed by a processor, is used to implement the methods provided in the various embodiments described above.

[0120] The readable storage medium can be a computer storage medium or a communication medium. A communication medium includes any medium that facilitates the transfer of computer programs from one location to another. A computer storage medium can be any available medium accessible to a general-purpose or special-purpose computer. For example, a readable storage medium is coupled to a processor, enabling the processor to read information from and write information to the readable storage medium. Of course, the readable storage medium can also be a component of the processor. The processor and the readable storage medium can reside in an Application-Specific Integrated Circuit (ASIC). Alternatively, the ASIC can be located in a user equipment. Of course, the processor and the readable storage medium can also exist as discrete components in a communication device. The readable storage medium can be a read-only memory (ROM), random access memory (RAM), CD-ROM, magnetic tape, floppy disk, and optical data storage device, etc.

[0121] The present invention also provides a program product including executable instructions stored in a readable storage medium. At least one processor of the device can read the executable instructions from the readable storage medium, and the at least one processor executes the executable instructions to cause the device to implement the methods provided in the various embodiments described above.

[0122] In the embodiments of the above-described device, it should be understood that the processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), etc. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the method disclosed in this invention can be directly manifested as execution by a hardware processor, or execution by a combination of hardware and software modules within the processor.

[0123] Through the above embodiments, the present invention, through an advertising recommendation system and method based on people's clothing, appearance, and body posture information, utilizes a collection module, a preprocessing module, an extraction module, a classification module, a recommendation module, and a feedback module to determine the target audience. The system acquires an initial audience image corresponding to the target audience using a collection device; preprocesses the initial audience image to generate a target audience image; extracts the body posture features corresponding to the target audience from the target audience image; based on an audience classification algorithm, it retrieves preset classification labels and, according to the correspondence between the preset classification labels and body posture features, divides the target audience into multiple target sub-groups, with each target sub-group corresponding to a preset classification label; and based on an advertising recommendation algorithm, it retrieves a preset advertising database. The system matches target ads in a pre-defined ad library with pre-defined category tags, identifies target sub-groups corresponding to these tags, and recommends the target ads to these sub-groups. It collects recommendation feedback data from these sub-groups and optimizes both the audience classification and ad recommendation algorithms based on this data. This enables precise classification of audiences by clothing, appearance, and body type, improving the accuracy and effectiveness of ad recommendations, reducing information waste, and combining user profiles with ad content to achieve personalized ad recommendations. This enhances user experience and ad conversion rates, avoids interference from ineffective ads, and continuously optimizes the recommendation algorithm by collecting user feedback data in real time, thereby improving the accuracy and efficiency of the recommendation system and ultimately increasing ad conversion rates.

[0124] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. An advertising recommendation system based on people's clothing, appearance, and body shape information, characterized in that, The system includes: The acquisition module is used to identify the target population and acquire the initial population image corresponding to the target population through the acquisition device; The preprocessing module is used to perform preprocessing operations on the initial crowd image to generate the target crowd image; The extraction module is used to extract the body features corresponding to the target population in the target population image; The classification module is used to retrieve preset classification labels based on a population classification algorithm, and divide the target population into multiple target sub-populations according to the correspondence between the preset classification labels and body features, with each target sub-population corresponding to a preset classification label; The recommendation module is used to retrieve a preset advertising library based on an advertising recommendation algorithm, match the target advertisement in the preset advertising library corresponding to the preset category tag, determine the target sub-group corresponding to the preset category tag, and recommend the target advertisement to the target sub-group. The feedback module is used to collect recommendation feedback data from the target sub-group and optimize the audience classification algorithm and the advertisement recommendation algorithm based on the recommendation feedback data.

2. The advertising recommendation system based on people's clothing, appearance, and body shape information according to claim 1, characterized in that, The acquisition module includes: An installation unit is used to determine a target area and install the acquisition device in the target area; The acquisition unit is used to determine the target population and acquire the initial population image corresponding to the target population through the acquisition device; A storage unit is used to store the acquired initial crowd images into an initial crowd image database.

3. The advertising recommendation system based on people's clothing, appearance, and body shape information according to claim 1, characterized in that, The preprocessing module includes: The retrieval unit is used to retrieve the initial crowd images stored in the initial crowd image database; The generation unit is used to perform denoising and enhancement processing on the initial crowd image to generate the target crowd image.

4. The advertising recommendation system based on people's clothing, appearance, and body shape information according to claim 1, characterized in that, The extraction module extracts the body features corresponding to the target population in the target population image using the following feature extraction algorithm: ； In the formula, Indicates a spatial location index on a physical feature; Describe physical characteristics; Indicating the spatial location index of body features The value at; and Indicates the spatial location index on the convolution kernel; Represents images of the target audience; Indicates the spatial location index of the target crowd image. The value at; Represents the convolution kernel; Indicates the spatial index of the convolution kernel The weight value at the location; The term ΣmΣn represents the summation of all elements in the convolution kernel; Σ represents the summation sign. Based on the feature extraction algorithm, feature extraction is performed on the target crowd image to obtain the body posture features corresponding to the target crowd.

5. The advertising recommendation system based on people's clothing, appearance, and body shape information according to claim 1, characterized in that, The classification module uses the following crowd classification algorithm to retrieve preset classification labels and, based on the correspondence between the preset classification labels and body features, divides the target crowd into multiple target sub-groups, with each target sub-group corresponding to one preset classification label: Calculate the information gain corresponding to the input body features: ； In the formula, This represents the target population dataset; Represents the input body posture features; Indicates information gain; Represents entropy; Represents the target population dataset Entropy; Represents the set of all values ​​for the input body posture features; Represents any value of the input body posture feature; This represents any value of the input body posture feature. Segmented target sub-group dataset; Represents the target sub-group dataset The number of users; Represents the target population dataset The number of users; Represents the body features of the input The corresponding gain coefficient; Σ represents the summation symbol; in ； In the formula, This represents the target population dataset; This represents the entropy of the target population dataset; Indicates the total number of preset category labels; Indicates the first The proportion of the target sub-group dataset corresponding to each preset category label in the target group dataset; Represents a logarithmic function; Σ represents the summation symbol; For the target population dataset Calculate the body features The corresponding information gain Select the body features with the highest information gain. The segmentation feature is used as the segmentation feature for the current node; based on the different values ​​of the segmentation feature, the target population dataset is... The dataset is divided into multiple target sub-groups. For each of the target sub-group datasets Repeat the above steps until the stopping condition is met, i.e., the target sub-group dataset is reached. All users belong to the same preset category label or the target sub-group dataset. The number of users in the system is less than a preset threshold; When the above recursive process stops, the current node can be marked as a leaf node, i.e., the target sub-group dataset. and determine the target subgroup dataset. The corresponding preset category label.

6. The advertising recommendation system based on people's clothing, appearance, and body shape information according to claim 1, characterized in that, The recommendation module uses the following advertising recommendation algorithm to retrieve a preset advertising library, match the target advertisement in the preset advertising library corresponding to the preset category tag, determine the target sub-group corresponding to the preset category tag, and recommend the target advertisement to the target sub-group: Calculate the first matching degree function between the preset category tag and the target advertisement: ； In the formula, This function represents the first degree of match between the preset category tags and the target advertisement. This represents a predefined category label vector; Represents the target ad vector; This represents the dot product of the preset category label vector and the target ad vector; This represents the modulus of the preset category label vector; The modulus of the target ad vector; Calculate the second matching degree function between the preset category label and the target subgroup: ； In the formula, This function represents the second degree of matching between the preset category labels and the target subgroup. This represents a predefined category label vector; Represents the target subgroup population vector; This represents the dot product of the preset category label vector and the target subgroup vector; This represents the modulus of the preset category label vector; The modulus of the target subgroup's population vector; Based on the first matching degree function and the second matching degree function, calculate the recommendation matching degree function between the target advertisement and the target sub-group: ； In the formula, This represents a predefined set of category tags; A function representing the recommendation matching degree between the target advertisement and the target sub-audience; This parameter represents the weighting between the target ad and the target sub-audience. This function represents the first degree of match between the preset category tags and the target advertisement. This represents the second matching degree function between the preset category labels and the target subgroup; Σ represents the summation symbol; Based on the recommendation matching degree function, the recommendation matching degree between the target sub-group and the target advertisement is calculated, the target advertisement corresponding to the maximum recommendation matching degree is determined, and recommended to the target sub-group.

7. The recommendation method of the advertising recommendation system based on the clothing, appearance, and body shape information of a group of people according to any one of claims 1-6, characterized in that, The method includes: Identify the target population and acquire initial population images corresponding to the target population using acquisition devices; The initial crowd image is preprocessed to generate the target crowd image; Extract the body posture features corresponding to the target population from the target population image; Based on the crowd classification algorithm, preset classification labels are retrieved. According to the correspondence between the preset classification labels and body features, the target crowd is divided into multiple target sub-groups, and each target sub-group corresponds to one preset classification label. Based on the advertising recommendation algorithm, a preset advertising library is retrieved, and the target advertisement in the preset advertising library corresponding to the preset category tag is matched to determine the target sub-group corresponding to the preset category tag, and the target advertisement is recommended to the target sub-group. Collect recommendation feedback data from the target sub-groups, and optimize the audience classification algorithm and the ad recommendation algorithm based on the recommendation feedback data.

Images