Advertisement putting monitoring method and system based on image AI analysis

Abstract

The invention discloses an advertisement putting monitoring method and system based on image AI analysis. The method comprises the following steps: integrating shooting equipment on advertisement putting equipment to obtain video data; identifying figures in the video data by using the target detection model, judging whether the eyes of the figures watch the advertisement area, and counting the number of the figures watching the advertisement area; obtaining human face features and human body posture features through a feature extraction model; classifying the figures according to ages and genders; identifying and acquiring interaction behaviors when the user watches the advertisement, and recording the number of times of the interaction behaviors; and calculating advertisement scores of different ages and different genders according to the number of people watching the advertisement, the number of interactive behaviors and the advertisement watching duration data. The advertisement putting effect can be accurately monitored, detailed audience behavior data can be provided for advertisers, the advertisement content, putting channels and strategies can be optimized, the attraction and putting efficiency of advertisements are effectively improved, and the return on investment and the propagation value of the advertisements are improved.

Description

Technical Field

[0001] The present invention relates to the field of advertising delivery technology, and in particular to an advertising delivery monitoring method and system based on image AI analysis. Background Art

[0002] In the current field of outdoor advertising monitoring, with the development of technology, camera monitoring technology has been widely used in the monitoring of advertising effects. By installing cameras in the advertising area, the situation around the advertisement can be recorded in real time, which improves the monitoring efficiency and data acquisition capabilities to a certain extent.

[0003] However, the existing camera monitoring technology still has obvious limitations. On the one hand, although the camera can obtain the image information of the advertising area, it is not possible to accurately identify the number of people watching the advertisement. In actual scenarios, due to the influence of factors such as the density of the crowd, the dynamic changes of people, and the shooting angle, it is difficult for the camera to accurately distinguish each individual watching the advertisement, and it is easy to have repeated counting or omissions, resulting in a large deviation between the number of viewers counted and the actual number. On the other hand, the current monitoring technology lacks in-depth analysis of the audience. Relying solely on the images recorded by the camera, it is impossible to classify and study the audience according to different characteristics such as age and gender. This makes it impossible for advertisers to understand the reaction and acceptance of different audience groups to advertisements, and it is difficult to adjust the advertising content and delivery strategy in a targeted manner. It is impossible to achieve accurate advertising delivery, resulting in a waste of advertising resources and reducing the effectiveness of advertising delivery and the return on investment. Summary of the invention

[0004] The purpose of this section is to summarize some aspects of embodiments of the present invention and briefly introduce some preferred embodiments. Some simplifications or omissions may be made in this section and the specification abstract and the invention title of this application to avoid blurring the purpose of this section, the specification abstract and the invention title, and such simplifications or omissions cannot be used to limit the scope of the present invention.

[0005] In view of the above existing problems, the present invention is proposed. Therefore, the present invention provides an advertising placement monitoring method based on image AI analysis to solve the problem that the existing technology cannot accurately identify the number of people watching advertisements and does not analyze according to the characteristics of different audience groups, resulting in incomplete and inaccurate evaluation of advertising effects.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0007] In a first aspect, the present invention provides an advertising delivery monitoring method based on image AI analysis, comprising:

[0008] Integrate a shooting device on the advertising placement device. When a person enters the visible area of the advertisement, the shooting function is enabled to record video data and obtain the video data.

[0009] Use a target detection model to identify the people in the video data, determine whether the eyes of the people are looking at the advertisement area, and count the number of people looking at the advertisement area.

[0010] Construct a feature extraction model, and obtain the facial features and body posture features of the people through the feature extraction model. Based on the facial features, classify the people according to age and gender. Based on the body posture features of the people, construct a mapping relationship model between the posture features and interaction behaviors, identify and obtain the interaction behaviors of users when watching advertisements, and record the number of interaction behaviors.

[0011] Calculate the advertisement scores for different ages and genders according to the number of people watching the advertisement, the number of interaction behaviors, and the advertisement viewing duration data.

[0012] As a preferred solution of the advertisement placement monitoring method based on image AI analysis described in the present invention, wherein: using a target detection model to identify the people in the video data includes:

[0013] Preprocess the video data, and the preprocessing includes video frame extraction, resolution adjustment, image normalization, and data augmentation operations.

[0014] Collect a large amount of image data containing people to construct a target detection data set, and annotate the bounding boxes and class labels of the people. Divide the annotated data set into a training set, a validation set, and a test set according to an appropriate ratio.

[0015] After setting the training rounds and batch size of the target detection model, use the training set to train the target detection model, and calculate the loss function to update the parameters of the target detection model.

[0016] Input the preprocessed video frames into the trained target detection model for inference. The target detection model outputs the bounding boxes and confidence scores of the areas where people may exist. Use the non-maximum suppression algorithm to post-process the bounding boxes, remove redundant and incorrect detection results, filter out the bounding boxes with confidence higher than the set threshold, and determine the specific positions and ranges of the people in the video frames by analyzing their position and size information, thus completing the identification of the people in the video data.

[0017] As a preferred solution of the advertisement placement monitoring method based on image AI analysis described in the present invention, wherein: determining whether the eyes of the people are looking at the advertisement area and counting the number of people looking at the advertisement area includes:

[0018] Decompose the video data into multiple frames of images, obtain the image of the human eye region, perform grayscale processing on the extracted eye region image to calculate the brightness values of the image pixel points, and obtain the pupil position of the person according to the brightness values;

[0019] Calculate the offset vector between the pupil center and the human eye region. If the horizontal and vertical components of the offset vector are within the preset pixel range, it is preliminarily determined that the person is looking at the advertisement area; Analyze 5 consecutive frames of images. If, among the 5 consecutive frames, the offset vectors calculated for each frame are within the preset pixel range, it is finally determined that the person is looking at the advertisement area;

[0020] In each frame of the video, mark the person determined to be looking at the advertisement area, use the feature matching algorithm for person tracking, record the person's contour and clothing color features, compare the features of the newly appeared suspected looking person with the recorded features to avoid duplicate counting, and count the number of different people who are marked and accurately compared.

[0021] As a preferred solution of the advertisement placement monitoring method based on image AI analysis according to the present invention, wherein: construct a feature extraction model, and obtain the human face features through the feature extraction model, including:

[0022] Design the feature extraction model through a convolutional neural network and a human pose estimation algorithm;

[0023] Decompose the video data into multiple frames of images, extract the human face image as the input, and input it into the convolutional neural network; The face image undergoes sliding convolution through the convolutional layer to generate multiple feature maps, transmit the feature maps output by the convolutional layer to the activation layer, perform non-linear transformation using the ReLU function, and the activated preliminary feature maps enter the pooling layer for downsampling through the max pooling operation;

[0024] Repeat the convolution, activation, and pooling operations to form a stack of multiple convolution-activation-pooling layers, obtain high-level feature maps, flatten the high-level feature maps into one-dimensional vectors, and input them into the fully connected layer to generate a fixed-length feature vector;

[0025] Convert the feature vector into the probability distribution of each category through the Softmax function, and classify the person according to age and gender according to the probability distribution.

[0026] As a preferred solution of the advertisement placement monitoring method based on image AI analysis according to the present invention, wherein: construct a feature extraction model, and obtain the human body pose features through the feature extraction model, including:

[0027] Obtain the coordinates of the human body key points of the video image person, and calculate the joint angles, limb length ratios, and relative position relationships according to the key point coordinates to obtain the human body pose features.

[0028] As a preferred solution of the advertising placement monitoring method based on image AI analysis according to the present invention, wherein: based on the body posture characteristics of a person, a mapping relationship model between the posture characteristics and interaction behaviors is constructed to identify and obtain the interaction behaviors of a user when watching an advertisement, and the number of interaction behaviors is recorded, including:

[0029] Collect a large amount of video data containing different interaction behaviors, use the method for extracting body posture characteristics to obtain body posture characteristic data, and perform detailed annotation on the interaction behaviors in the video to clarify the specific interaction behavior types corresponding to each posture characteristic;

[0030] Design the mapping relationship model using a support vector machine. The support vector machine takes the collected and annotated posture characteristic data as input and the corresponding interaction behavior types as output, and trains the mapping relationship model to improve the classification accuracy of the model;

[0031] Input the body posture characteristics of the person to be identified into the trained mapping relationship model. The mapping relationship model analyzes and judges the input posture characteristics according to the learned mapping relationship, and outputs the most likely corresponding interaction behavior type;

[0032] Adopt a state tracking mechanism. For each identified interaction behavior, set a corresponding state identifier. When the mapping relationship model identifies the start of an interaction behavior, start the state identifier of this behavior and start timing; when this behavior ends, stop timing and record the number of interaction behaviors.

[0033] As a preferred solution of the advertising placement monitoring method based on image AI analysis according to the present invention, wherein: according to the number of people watching the advertisement, the number of interaction behaviors, and the advertisement viewing duration data, calculate the advertisement scores for different ages and genders, including:

[0034] Respectively assign weights to the number of people watching the advertisement, the total viewing duration, and the number of interaction behaviors, and the sum of the weights is 1. Standardize the number of people watching the advertisement, the total viewing duration, and the number of interaction behaviors in each age interval group, and perform weighted calculation based on the weight coefficients, the number of people watching the advertisement, the total viewing duration, and the number of interaction behaviors to obtain the advertisement scores for different age interval groups;

[0035] Respectively assign weights to the number of people watching the advertisement, the total viewing duration, and the number of interaction behaviors, and the sum of the weights is 1. Standardize the number of people watching the advertisement, the total viewing duration, and the number of interaction behaviors in the gender group, and perform weighted calculation based on the weight coefficients, the number of people watching the advertisement, the total viewing duration, and the number of interaction behaviors to obtain the advertisement scores for different gender groups.

[0036] Second aspect, the present invention provides a system for monitoring advertisement placement based on image AI analysis, including:

[0037] A data acquisition module, which is used to integrate a shooting device on the advertisement placement device. When it detects that a person enters the advertisement visible area, it turns on the shooting function to record video data and obtain the video data.

[0038] A statistics module, which is used to identify the people in the video data by using a target detection model, judge whether the eyes of the people are looking at the advertisement area, and count the number of people looking at the advertisement area.

[0039] A feature extraction module, which is used to build a feature extraction model, and obtain the facial features and body posture features of the people through the feature extraction model; classify the people according to age and gender based on the facial features; based on the body posture features of the people, build a mapping relationship model between the posture features and the interaction behaviors, identify and obtain the interaction behaviors of the users when watching the advertisement, and record the number of interaction behaviors.

[0040] An evaluation and monitoring module, which is used to calculate the advertisement scores of different ages and genders according to the number of people watching the advertisement, the number of interaction behaviors, and the advertisement viewing duration data.

[0041] Third aspect, the present invention provides a computing device, including:

[0042] A memory and a processor;

[0043] The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions. When the computer-executable instructions are executed by the processor, the steps of the method for monitoring advertisement placement based on image AI analysis are implemented.

[0044] Fourth aspect, the present invention provides a computer-readable storage medium, which stores computer-executable instructions. When the computer-executable instructions are executed by a processor, the steps of the method for monitoring advertisement placement based on image AI analysis are implemented.

[0045] Compared with the prior art, the beneficial effects of the present invention are as follows: The advertising placement monitoring method based on image AI analysis of the present invention integrates a shooting device on the advertising placement device to obtain video data, uses a target detection model to accurately identify the people in the video, determines whether the people are looking at the advertising area based on the pupil position and counts the number of people. At the same time, a feature extraction model is constructed to obtain the facial and body posture features of the people, realizing classification by age and gender, recognition and counting of interactive behaviors. Finally, the advertising scores of different ages and genders are calculated by combining the number of people watching the advertisement, the number of interactive behaviors, and the advertisement viewing duration, which can accurately monitor the advertising placement effect, provide detailed audience behavior data for advertisers, so as to optimize the advertisement content, placement channels and strategies, effectively improve the attractiveness and placement efficiency of the advertisement, realize the reasonable allocation of advertising resources, and enhance the return on investment and communication value of the advertisement. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained according to these drawings. Among them:

[0047] Figure 1 It is a schematic diagram of the overall process of the advertising placement monitoring method based on image AI analysis according to an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0048] In order to make the above objects, features and advantages of the present invention more obvious and understandable, the following will make a detailed description of the specific embodiments of the present invention in conjunction with the drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0049] In the following description, many specific details are set forth in order to fully understand the present invention. However, the present invention can also be implemented in other ways different from those described herein. Those skilled in the art can make similar extensions without departing from the connotation of the present invention. Therefore, the present invention is not limited by the specific embodiments disclosed below.

[0050] Secondly, the so-called "one embodiment" or "embodiment" herein refers to a specific feature, structure or characteristic that can be included in at least one implementation manner of the present invention. The "in one embodiment" that appears in different places in this specification does not all refer to the same embodiment, nor is it a separate or alternative embodiment that mutually excludes other embodiments.

[0051] The present invention will be described in detail with reference to the schematic diagrams. When describing the embodiments of the present invention in detail, for the convenience of explanation, the cross-sectional views showing the device structure will be enlarged locally not in accordance with the general scale, and the schematic diagrams are only examples and should not limit the scope of protection of the present invention herein. In addition, in actual production, three-dimensional spatial dimensions including length, width, and depth should be included.

[0052] Meanwhile, in the description of the present invention, it should be noted that the orientation or positional relationship indicated by terms such as "upper, lower, inner, and outer" is based on the orientation or positional relationship shown in the drawings. It is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the system or component referred to must have a specific orientation, be constructed and operated in a specific orientation, and thus cannot be construed as a limitation to the present invention. In addition, the terms "first, second, or third" are only used for descriptive purposes and cannot be construed as indicating or implying relative importance.

[0053] Unless otherwise clearly defined and limited in the present invention, the terms "installed, connected, and coupled" should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or an integral connection; it can also be a mechanical connection, an electrical connection, or a direct connection, and can also be indirectly connected through an intermediate medium, or can be the communication inside two components. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood according to specific situations.

[0054] Embodiment 1

[0055] Referring to Figure 1 , an embodiment of the present invention provides an advertising placement monitoring method based on image AI analysis, including:

[0056] S1: Integrate a shooting device on the advertising placement device. When it is detected that a person enters the advertising visible area, turn on the shooting function to record video data and obtain the video data;

[0057] Preferably, preprocess the video data, and the preprocessing includes video frame extraction, resolution adjustment, image normalization, and data augmentation operations;

[0058] Specifically, video frame extraction is to divide continuous video data into frames of images at a certain time interval to provide basic data units for subsequent image analysis. Since different shooting devices or shooting conditions may result in inconsistent video resolutions, a resolution adjustment operation is performed to uniformly adjust the resolutions of all video frames to a size suitable for subsequent processing to ensure the accuracy and consistency of the analysis. Normalizing the pixel values of the image is to map the pixel values to a specific range to eliminate the influence brought by the difference in pixel value distribution between different images, making the subsequent processing of the image more stable and efficient.

[0059] Data augmentation includes rotation, flipping, cropping, adding noise, etc. By performing data augmentation operations on the original video frames, new image samples can be generated, increasing the quantity of training data, thereby improving the generalization ability of the subsequent model and enabling it to perform better in the face of various different actual scenarios.

[0060] S2: Use the object detection model to identify the people in the video data, determine whether the eyes of the people are looking at the advertisement area, and count the number of people looking at the advertisement area;

[0061] Preferably, using the object detection model to identify the people in the video data specifically includes the following steps:

[0062] Collect a large amount of image data containing people to construct an object detection dataset, and annotate the people with bounding boxes and class labels. Divide the annotated dataset into a training set, a validation set, and a test set in a suitable proportion;

[0063] Set the number of training epochs and batch size of the object detection model. The number of training epochs represents the number of times the entire training dataset is traversed by the model, and the batch size is the number of samples used for parameter update each time. Too few training epochs will cause the model to learn insufficiently, while too many may lead to overfitting; too large a batch size may cause the model to fall into a local optimum, and too small a batch size will reduce the training efficiency. During training, the object detection model processes the training data in batches and performs forward propagation to obtain the prediction results. To measure the difference between the prediction results and the true labels, the mean squared error is used as the loss function. The mean squared error reflects the degree of deviation of the prediction by calculating the average of the sum of the squares of the differences between the predicted values and the true values. Based on this loss function value, the model parameters are updated. After multiple rounds of training, the object detection model continuously learns the data features and gradually improves its ability to detect people in the video.

[0064] Evaluate the performance of the object detection model during training and adjust the hyperparameters through the validation set. During the training process of the object detection model, after every certain number of training epochs, input the validation set into the model in the current training state. The model will perform object detection on the image data in the validation set and output the corresponding bounding boxes and confidence scores; compare the detection results of the model with the true data already annotated in the validation set. Measure the detection effect of the model on the validation set by calculating metrics such as accuracy and recall. If it is found that the performance metrics of the model on the validation set do not meet the expectations, adjust the parameters of the model.

[0065] After the target detection model is adjusted and optimized multiple times through the validation set to make its performance reach a relatively satisfactory level, the trained target detection model is then used to perform inference with the test set input. The target detection model outputs the bounding boxes and confidence scores of the regions where there may be people. The non-maximum suppression algorithm is used to post-process the bounding boxes to remove redundant and incorrect detection results, and filter out the bounding boxes with confidence higher than the set threshold. By analyzing their position and size information, the specific position and range of the people in the video frame are determined, and the recognition of the people in the video data is completed.

[0066] Preferably, the number of people looking at the advertisement area is counted, which specifically includes the following steps:

[0067] The video data is decomposed into multiple frames of images, the eye region images of the people are obtained, the extracted eye region images are grayscale processed to calculate the brightness values of the image pixel points, and the pupil positions of the people are obtained according to the brightness values;

[0068] The offset vector between the pupil center and the eye region of the person is calculated. If the horizontal and vertical components of the offset vector are within the preset pixel range, it is preliminarily determined that the person is looking at the advertisement area; the continuous 5 frames of images are analyzed. If the offset vectors calculated for each of the 5 consecutive frames are within the preset pixel range, it is finally determined that the person is looking at the advertisement area;

[0069] In each frame of the video image, the people determined to be looking at the advertisement area are marked, the feature matching algorithm is used for person tracking, the person contour and clothing color features are recorded, and the features of the newly appeared suspected people looking are compared with the recorded features to avoid double counting, and the number of different people who are marked and accurately compared is counted.

[0070] It should be noted that the present invention uses a target detection model to identify people in video data, conducts scientific training based on a dataset constructed with a large amount of labeled data, optimizes through the validation set and performs inference with the test set, realizes the accurate positioning of people, effectively avoids problems such as insufficient learning and overfitting of the model, and can improve the detection ability and generalization. On the basis of the target detection model detecting people, the video data is decomposed into multiple frames of images, the eye region images of the people are grayscale processed to calculate the pixel point brightness values to obtain the pupil positions, and then combined with the calculation of the offset vector and the analysis of multiple frames of images, it is accurately determined whether the people are looking at the advertisement area. And when counting the number of people looking, the feature matching algorithm is used for person tracking, the person contour and clothing color features are recorded, and double counting is avoided, realizing the accurate counting of the number of people looking at the advertisement area.

[0071] S3 constructs a feature extraction model, and obtains the facial features and body posture features of the person through the feature extraction model; classifies the person according to age and gender based on the facial features; constructs a mapping relationship model between the posture features and interaction behaviors based on the body posture features of the person, identifies and obtains the interaction behaviors of the user when watching the advertisement, and records the number of interaction behaviors;

[0072] Preferably, the feature extraction model is designed through a convolutional neural network and a human pose estimation algorithm;

[0073] Preferably, obtaining the facial features of the person through the feature extraction model specifically includes the following steps:

[0074] Decompose the video data into multiple frames of images, extract the facial images of the person as the input, and input them into the convolutional neural network; the facial images are subjected to sliding convolution through the convolutional layer to generate multiple feature maps, and the feature maps output by the convolutional layer are transmitted to the activation layer, and the ReLU function is used for non-linear transformation. The activated preliminary feature maps enter the pooling layer and are downsampled through the max pooling operation;

[0075] Repeat the convolution, activation, and pooling operations to form a stack of multiple convolutional-activation-pooling layers, obtain the high-level feature maps, flatten the high-level feature maps into one-dimensional vectors, and input them into the fully connected layer to generate a fixed-length feature vector;

[0076] Convert the feature vector into the probability distribution of each category through the Softmax function, and classify the person according to age and gender based on the probability distribution;

[0077] Specifically, obtain the shape and position features of parts such as eyes, nose, and mouth through the convolutional layer of the convolutional neural network. After multiple layers of convolution, activation, and pooling operations, extract the facial bone structure and skin texture features. Combine and transform these features at different levels to generate a fixed-length feature vector.

[0078] Input the feature vector into the Softmax function, and the Softmax function will convert it into the probability distribution of each category. In terms of age classification, according to the principle of the largest probability, the person is divided into different age ranges such as children, teenagers, young people, middle-aged people, and elderly people. In terms of gender classification, it is judged whether the person is male or female according to the probability, and the category with the largest probability is the age or gender classification result to which the person belongs.

[0079] Preferably, obtaining the body posture features of the person through the feature extraction model specifically includes the following steps:

[0080] Obtain the coordinates of the human key points of the video image person, and calculate the joint angles, limb length ratios, and relative position relationships according to the key point coordinates to obtain the body posture features of the person;

[0081] The human body key points include key parts such as the midpoint of the neck, the left and right shoulders, the left and right elbows, the left and right wrists, the left and right hip joints, the left and right knee joints, and the left and right ankles;

[0082] Collect a large amount of video data containing different interaction behaviors, use the method of extracting human body posture features to obtain body posture feature data, and perform detailed annotation on the interaction behaviors in the video to clarify the specific interaction behavior types corresponding to each posture feature;

[0083] Adopt a support vector machine to design a mapping relationship model. The support vector machine takes the collected and labeled posture feature data as input and the corresponding interaction behavior types as output, and trains the mapping relationship model to improve the classification accuracy of the model;

[0084] Input the human body posture features of the person to be recognized into the trained mapping relationship model. The mapping relationship model analyzes and judges the input posture features according to the learned mapping relationship, and outputs the most likely corresponding interaction behavior type;

[0085] Adopt a state tracking mechanism. For each recognized interaction behavior, set a corresponding state identifier. When the mapping relationship model recognizes the start of an interaction behavior, start the state identifier of this behavior and start timing; when this behavior ends, stop timing and record the number of times of the interaction behavior.

[0086] Specifically, according to the calculation results of joint angles, limb length ratios, and relative positions, construct a mapping relationship model. The mapping relationship model is pre-trained through a large amount of human body posture data with annotations, and these annotations clarify the specific interaction behaviors corresponding to different posture feature combinations. When the body posture features calculated in real time are input, the mapping relationship model will make a judgment based on the learned patterns, recognize the specific interaction behaviors of the person when watching the advertisement, and record the number of times each interaction behavior appears.

[0087] It should be noted that through multi-layer operations of the convolutional neural network, the present invention can comprehensively and deeply capture facial features from basic parts to high-level abstractions. The generated feature vectors can be accurately classified by age and gender through the Softmax function, providing a data basis for the subsequent advertisement scoring of different audiences; in the extraction of human body posture features, the present invention obtains posture feature data by calculating the relevant relationships of key point coordinates, combines a large amount of annotated video data, and the mapping relationship model trained by the support vector machine can accurately identify interaction behaviors, and at the same time uses the state tracking mechanism to ensure accurate counting.

[0088] S4: Calculate the advertisement scores for different ages and genders according to the number of people watching the advertisement, the number of interaction behaviors, and the advertisement viewing duration data;

[0089] Preferably, weights are respectively assigned to the number of people watching the advertisement, the total viewing duration, and the number of interaction behaviors, and the sum of the weights is 1. The number of people watching the advertisement, the total viewing duration, and the number of interaction behaviors in each age interval group are standardized, and weighted calculations are performed based on the weight coefficients, the number of people watching the advertisement, the total viewing duration, and the number of interaction behaviors to obtain the advertisement scores for different age interval groups.

[0090] Preferably, weights are respectively assigned to the number of people watching the advertisement, the total viewing duration, and the number of interaction behaviors, and the sum of the weights is 1. The number of people watching the advertisement, the total viewing duration, and the number of interaction behaviors in the gender group are standardized, and weighted calculations are performed based on the weight coefficients, the number of people watching the advertisement, the total viewing duration, and the number of interaction behaviors to obtain the advertisement scores for different gender groups.

[0091] According to the calculation results, the advertising placement effect can be comprehensively and deeply monitored and analyzed. For the advertisement scores of different age interval groups, it is possible to clearly understand the acceptance degree and interest level of the audience in each age group towards the advertisement. For the advertisement scores of different gender groups, it is possible to judge the performance differences of the advertisement among male and female audiences. If the score of a certain gender is not ideal, relevant advertisement elements can be specifically monitored, such as whether the visual style and copywriting expression of the advertisement are unattractive to that gender. By continuously monitoring these calculation results, advertisers and placement parties can timely discover problems existing in the advertising placement process, dynamically adjust the advertising strategy, so as to improve the effect of the advertisement among different age and gender audience groups, realize the precision and efficiency of advertising placement, and further enhance the overall communication value and commercial benefits of the advertisement, ensuring that the advertising placement can reach the target audience to the greatest extent and generate positive feedback.

[0092] The above is a schematic solution of an advertising placement monitoring method based on image AI analysis in this embodiment. It should be noted that the technical solution of the advertising placement monitoring system based on image AI analysis belongs to the same concept as the technical solution of the above-mentioned advertising placement monitoring method based on image AI analysis. For the details not described in detail in the technical solution of the advertising placement monitoring system based on image AI analysis in this embodiment, reference can be made to the description of the technical solution of the above-mentioned advertising placement monitoring method based on image AI analysis.

[0093] The advertising placement monitoring system based on image AI analysis in this embodiment includes:

[0094] A data acquisition module, which is used to integrate a shooting device on the advertising placement device, and when it detects that a person enters the advertisement visible area, it turns on the shooting function to record video data and obtain the video data;

[0095] A statistics module, configured to use a target detection model to identify people in video data, determine whether the eyes of a person are looking at an advertisement area, and count the number of people looking at the advertisement area;

[0096] A feature extraction module, configured to build a feature extraction model, and obtain a person's facial features and body posture features through the feature extraction model; classify people according to age and gender based on the facial features; based on the person's body posture features, build a mapping relationship model between the posture features and interaction behaviors, identify and obtain the interaction behaviors of users when watching advertisements, and record the number of interaction behaviors;

[0097] An evaluation and monitoring module, configured to calculate advertisement scores for different ages and genders according to the number of people watching advertisements, the number of interaction behaviors, and advertisement viewing duration data.

[0098] This embodiment also provides a computing device, applicable to the situation of advertisement placement monitoring based on image AI analysis, including:

[0099] A memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the method for realizing advertisement placement monitoring based on image AI analysis as proposed in the above embodiment.

[0100] This embodiment also provides a storage medium, on which a computer program is stored, and when the program is executed by a processor, it implements the method for realizing advertisement placement monitoring based on image AI analysis as proposed in the above embodiment.

[0101] The storage medium proposed in this embodiment and the method for realizing advertisement placement monitoring based on image AI analysis proposed in the above embodiment belong to the same inventive concept. Technical details not described in detail in this embodiment can be referred to in the above embodiment, and this embodiment has the same beneficial effects as the above embodiment.

[0102] Through the above description of the implementation manners, those skilled in the art can clearly understand that the present invention can be implemented by means of software and necessary general-purpose hardware. Of course, it can also be implemented by hardware, but in many cases the former is a better implementation manner. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. The computer software product can be stored in a computer-readable storage medium, such as a floppy disk, a read-only memory (ROM), a random access memory (RAM), a flash memory (FLASH), a hard disk, or an optical disc of a computer, etc., including several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods of various embodiments of the present invention.

[0103] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical solutions of the present invention, and they should all be covered by the scope of the claims of the present invention.

Claims

1. An advertising placement monitoring method based on image AI analysis, characterized in that: include: Integrate a camera on the advertising device, and when a person is detected to enter the visible area of ​​the advertisement, start the camera function to record and obtain video data; Use the object detection model to identify people in the video data, determine whether the people's eyes are looking at the advertising area, and count the number of people looking at the advertising area; Construct a feature extraction model to obtain facial features and body posture features of characters through the feature extraction model; classify characters by age and gender based on facial features; construct a mapping relationship model between posture features and interactive behaviors based on body posture features of characters, identify and obtain interactive behaviors of users when watching advertisements, and record the number of interactive behaviors; Ad scores for different ages and genders are calculated based on the number of people who viewed the ad, the number of interactive behaviors, and the ad viewing time.

2. The advertising placement monitoring method based on image AI analysis according to claim 1, characterized in that: Use object detection models to identify people in video data, including: Preprocess the video data, including video frame extraction, resolution adjustment, image normalization and data enhancement operations; Collect a large amount of image data containing people to build an object detection dataset, annotate the people with bounding boxes and category labels, and divide the annotated dataset into training set, validation set, and test set in appropriate proportions; After setting the number of training rounds and batch size for the target detection model, use the training set to train the target detection model, calculate the loss function and update the target detection model parameters; The preprocessed video frames are input into the trained target detection model for inference. The target detection model outputs the bounding boxes and confidence scores of the areas where people may exist. The bounding boxes are post-processed using the non-maximum suppression algorithm to remove redundant and erroneous detection results, and the bounding boxes with confidence levels higher than the set threshold are screened out. The specific position and range of the person in the video frame are determined by analyzing their position and size information, thereby completing the recognition of the person in the video data.

3. The method for monitoring advertising placement based on image AI analysis according to claim 1 or 2, characterized in that: Determine whether the person's eyes are looking at the advertising area, and count the number of people looking at the advertising area, including: Decomposing the video data into multiple frames of images, obtaining an eye region image of a person, graying the extracted eye region image to calculate the brightness value of the image pixel, and obtaining the pupil position of the person according to the brightness value; Calculate the offset vector between the pupil center and the person's eye area. If the horizontal component and vertical component of the offset vector are within the preset pixel range, it is preliminarily determined that the person is looking at the advertising area. Analyze 5 consecutive frames of images. If the offset vector calculated in each of the 5 consecutive frames is within the preset pixel range, it is finally determined that the person is looking at the advertising area. In each frame of the video, the person who is determined to be looking at the advertising area is marked, and the feature matching algorithm is used to track the person. The person's outline and clothing color characteristics are recorded, and the features of the newly appeared suspected person are compared with the recorded features to avoid double counting. The number of different people that have been marked and accurately matched is counted.

4. The method for monitoring advertisement placement based on image AI analysis according to claim 3, characterized in that: Construct a feature extraction model to obtain facial features of the person, including: Design a feature extraction model using convolutional neural networks and human posture estimation algorithms; The video data is decomposed into multiple frames of images, and the facial image of the person is extracted as input and input into the convolutional neural network; the facial image is subjected to sliding convolution in the convolutional layer to generate multiple feature maps, and the feature map output by the convolutional layer is transmitted to the activation layer, and the ReLU function is used for nonlinear transformation. The activated preliminary feature map enters the pooling layer and is downsampled through the maximum pooling operation; Repeat convolution, activation, and pooling operations to form a stack of multiple convolution-activation-pooling layers to obtain high-level feature maps, flatten the high-level feature maps into one-dimensional vectors, and input them into the fully connected layer to generate a fixed-length feature vector; The feature vector is converted into a probability distribution for each category through a Softmax function, and the characters are classified according to age and gender based on the probability distribution.

5. The method for monitoring advertisement placement based on image AI analysis according to claim 4, characterized in that: Construct a feature extraction model to obtain the body posture features of the person, including: The coordinates of the key points of the human body of the person in the video image are obtained, and based on the key point coordinates, the joint angles, limb length ratios and relative position relationships are calculated to obtain the person's body posture characteristics.

6. The method for monitoring advertisement placement based on image AI analysis according to claim 5, characterized in that: Based on the body posture characteristics of the characters, a mapping relationship model between posture characteristics and interactive behaviors is constructed to identify and obtain the interactive behaviors of users when watching advertisements, and record the number of interactive behaviors, including: Collect a large amount of video data containing different interactive behaviors, use the character body posture feature extraction method to obtain body posture feature data, annotate the interactive behaviors in the video in detail, and clarify the specific interactive behavior type corresponding to each posture feature; A support vector machine is used to design a mapping relationship model. The support vector machine takes the collected and labeled posture feature data as input and the corresponding interactive behavior type as output. The mapping relationship model is trained to improve the classification accuracy of the model. The body posture features of the person to be identified are input into the trained mapping relationship model. The mapping relationship model analyzes and judges the input posture features based on the learned mapping relationship and outputs the most likely corresponding interactive behavior type. A state tracking mechanism is used to set a corresponding state identifier for each interactive behavior identified. When the mapping relationship model recognizes the start of an interactive behavior, the state identifier of the behavior is started and the timing begins; when the behavior ends, the timing is stopped and the number of interactive behaviors is recorded.

7. The method for monitoring advertisement placement based on image AI analysis according to claim 6, characterized in that: Based on the number of people who viewed the ad, the number of interactive behaviors, and the ad viewing time, the ad scores for different ages and genders are calculated, including: The number of people who watched the ad, the total viewing time, and the number of interactive behaviors were weighted respectively, and the sum of the weights was 1. The number of people who watched the ad, the total viewing time, and the number of interactive behaviors in each age group were standardized, and weighted calculations were performed based on the weight coefficient, the number of people who watched the ad, the total viewing time, and the number of interactive behaviors to obtain the ad scores for different age groups; Weights are assigned to the number of people who watch the ad, total viewing time, and number of interactive behaviors, and the sum of the weights is 1. The number of people who watch the ad, total viewing time, and number of interactive behaviors in the gender group are standardized, and a weighted calculation is performed based on the weight coefficient, the number of people who watch the ad, total viewing time, and number of interactive behaviors to obtain the ad scores for different gender groups.

8. A system for monitoring advertising based on image AI analysis, characterized in that: include, The data acquisition module is used to integrate a shooting device on the advertising device, and when a person is detected to enter the visible area of ​​the advertisement, the shooting function is turned on to record video data and obtain video data; A statistics module, used to identify people in the video data using the target detection model, determine whether the people's eyes are looking at the advertising area, and count the number of people looking at the advertising area; The feature extraction module is used to construct a feature extraction model, and obtain the facial features and body posture features of the characters through the feature extraction model; based on the facial features, the characters are classified according to age and gender; based on the body posture features of the characters, a mapping relationship model between posture features and interactive behaviors is constructed to identify and obtain the interactive behaviors of users when watching advertisements, and record the number of interactive behaviors; The evaluation and monitoring module is used to calculate the advertising scores of different ages and genders based on the number of people who watched the advertisement, the number of interactive behaviors and the advertisement viewing time data.

9. An electronic device, characterized in that: include: Memory and processor; The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions. When the computer-executable instructions are executed by the processor, the steps of the advertising delivery monitoring method based on image AI analysis as described in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium, characterized in that: It stores computer executable instructions, which, when executed by a processor, implement the steps of the advertising delivery monitoring method based on image AI analysis as described in any one of claims 1 to 7.

Images