Advertisement delivery control method, advertisement delivery system, storage medium and program product

Abstract

The invention discloses an advertisement putting control method, an advertisement putting system, a storage medium and a program product, and the advertisement putting method comprises the steps: receiving a business district image uploaded by a user based on a user terminal node, and then determining a group portrait corresponding to the business district according to the business district image; and then one or more advertisements most adaptive to the group portrait are selected from the advertisement library according to the group portrait as target advertisements, and target advertisement putting device nodes in the business district are controlled to play the target advertisements, so that the purpose of accurately putting the advertisements is realized, and a plurality of decentralization processes are realized. In this way, the advertisement putting income in unit time is effectively improved, and advertisement putting is more accurate. Moreover, by implementing the above scheme based on the block chain, the purpose of decentration is better achieved, and the problem of low advertisement putting efficiency caused by artificial interference is effectively avoided.

Description

Technical Field

[0001] The present application relates to the field of advertising data processing, and in particular to an advertising delivery control method, an advertising delivery system, a storage medium, and a program product. Background Art

[0002] In the digital marketing and advertising industries, targeted advertising has become a crucial tool for improving advertising effectiveness and maximizing return on investment. In traditional advertising models, due to a lack of detailed analysis of multiple scenarios and customers, billboard owners repeatedly play ads placed by advertisers across their multiple billboards. Furthermore, the fact that most market data is currently collected and controlled by centralized platforms and companies significantly impacts the data's credibility and accuracy. This business model and delivery method results in poor advertising precision, a lack of analysis and feedback mechanisms for advertising effectiveness, and significant uncertainty regarding the actual economic value and commercial returns that advertising can provide. Summary of the Invention

[0003] The main purpose of this application is to provide an advertisement delivery control method, an advertisement delivery system and a storage medium, aiming to solve the technical problem of poor advertisement delivery accuracy in related technologies.

[0004] To achieve the above objectives, an embodiment of the present application provides an advertisement delivery control method, the method comprising:

[0005] Upload business district images through user terminal nodes;

[0006] Obtaining location information corresponding to the business district image;

[0007] Determining a target advertising device based on the business district where the location information is located;

[0008] Determining a group portrait corresponding to the business district area according to the business district image, and selecting a target advertisement in an advertisement library based on the group portrait;

[0009] Control the target advertisement delivery device to play the target advertisement.

[0010] In one embodiment, the advertisement delivery control method further includes:

[0011] Determining a sensitive area corresponding to the business district image, and determining a hash value of the sensitive area;

[0012] Get the hash value of the sensitive area of the image saved in the database;

[0013] If there is a hash value of the sensitive area whose matching degree with the hash value is greater than a preset threshold, the business district image is discarded.

[0014] In one embodiment, after the step of obtaining the hash value of the sensitive area of the image stored in the database, the advertisement delivery control method further includes:

[0015] If there is no hash value of the sensitive area whose matching degree with the hash value is greater than the preset threshold, determining whether there is a human body in the business district image;

[0016] If not, discard the business district image.

[0017] In one embodiment, after the step of discarding the business district image, the advertisement delivery control method further includes:

[0018] The user terminal node is controlled to output a prompt message indicating that the business district image does not meet the requirements.

[0019] In one embodiment, the step of obtaining the location information corresponding to the user terminal node includes:

[0020] Determining the location information corresponding to the business district image based on the positioning information uploaded by the user terminal node; or

[0021] Building features are extracted based on the commercial district image, and the location information corresponding to the commercial district image is determined based on a matching result between the building features and a preset city model.

[0022] In one embodiment, the step of determining the user group corresponding to the business district area according to the business district image, and selecting a target advertisement in the advertisement library based on the user group portrait includes:

[0023] Determining, based on the business district image, a human portrait corresponding to a human body in the business district image;

[0024] constructing the group portrait based on the human portraits corresponding to the multiple different business district images;

[0025] Target advertisements are selected from the advertisement library based on the group portrait.

[0026] In one embodiment, the step of selecting a target advertisement from an advertisement library based on the group portrait includes:

[0027] Inputting the group portrait into a pre-trained classification model and obtaining output data of the classification model;

[0028] One or more advertisements are selected from the advertisement library as the target advertisement according to a matching degree between the output data and the advertisements in the advertisement library.

[0029] In one embodiment, the advertisement delivery control method further includes:

[0030] Construct a sample group portrait based on the images corresponding to the preset business district, and mark the advertising label corresponding to the sample group portrait;

[0031] The classification model is trained based on the data pairs consisting of the sample group portrait and the advertising label as training samples.

[0032] In one embodiment, the advertisement delivery control method further includes:

[0033] Perform vector processing on the business district image at the user terminal and store it at the terminal;

[0034] Upload vectorized data to the blockchain through the user terminal node.

[0035] An embodiment of the present application further provides an advertisement delivery system, which includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the computer program is configured to implement the steps of the advertisement delivery control method described above.

[0036] An embodiment of the present application further provides a storage medium, which is a computer-readable storage medium. A computer program is stored on the storage medium. When the computer program is executed by a processor, the steps of the above-mentioned advertising delivery control method are implemented.

[0037] An embodiment of the present application further provides a computer program product, characterized in that it includes a computer program carried on a non-transitory computer-readable medium, and the computer program includes program code for executing the steps of the advertising delivery control method as described above.

[0038] The present application discloses an advertising delivery control method, an advertising delivery system, and a storage medium. These methods, based on blockchain technology, receive user-uploaded business district images from user terminal nodes. Based on the business district images and the vectorized data derived therefrom, the user determines the corresponding group profile for the business district. Based on the group profile, one or more advertisements that best match the group profile are selected from an advertisement library as target advertisements. The system then controls the target advertising delivery device nodes within the business district to play the target advertisements, thereby achieving precise advertising delivery. This effectively increases advertising delivery revenue per unit time and makes advertising delivery more accurate. Furthermore, implementing the above-mentioned solution based on blockchain further achieves decentralization, effectively avoiding issues such as manual interference, data verification, and centralized data manipulation that can lead to poor advertising delivery accuracy and efficiency. The present application can implement multiple decentralized processes (DA-PIN, Decentralized Asset-Physical Infrastructure Network), including but not limited to: acquiring target images through user terminals, vectorizing and storing the images at the user terminals, and uploading the vectorized data to the blockchain via the user terminals to implement related applications. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 This is a flow chart of an embodiment of an advertisement delivery control method according to an embodiment of the present application;

[0040] Figure 2 This is a flow chart of another embodiment of the advertising delivery control method involved in the embodiment of the present application;

[0041] Figure 3 It is a flowchart of an optional implementation scheme involved in another embodiment of the advertising delivery control method involved in the embodiment scheme of the present application;

[0042] Figure 4 This is a structural diagram of the advertising delivery system for this application.

[0043] The realization of the objectives, functional features and advantages of this application will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0044] It should be understood that the specific embodiments described herein are only used to explain the present application and are not intended to limit the present application.

[0045] In the digital marketing and advertising industries, targeted advertising has become a crucial tool for improving advertising effectiveness and maximizing return on investment. In the current model, billboard owners repeatedly play ads placed by advertisers across multiple billboards they own. This approach results in poor advertising precision.

[0046] Typically, during the advertising process, advertisers determine the demographic structure of each business district based on manual research. Based on this research, advertisers then search for billboards in business districts that match their product characteristics and place their ads there. However, because the demographics of each business district vary at different times and stages, this approach to advertising can result in poor accuracy. To address this shortcoming, the present invention proposes a blockchain-based advertising control method to address this issue.

[0047] In an embodiment of the present application, the advertising delivery system receives user-uploaded business district images from user terminal nodes. Based on the business district images, it then determines the corresponding group profile for the business district. Based on the group profile, it selects one or more advertisements from an advertising library that best match the group profile as target ads. The system then controls the target advertising delivery device nodes within the business district to play the target ads, achieving precise advertising delivery. This effectively increases advertising delivery revenue per unit time and makes advertising delivery more precise. Furthermore, implementing this solution based on blockchain further achieves decentralization and effectively avoids the problem of inefficient advertising delivery caused by human interference.

[0048] For ease of understanding, the technical solutions in the embodiments of the present application will be described below in conjunction with the drawings in the embodiments of the present application.

[0049] Please refer to Figure 1 In an optional implementation scheme, the advertisement delivery control method includes the following steps S10 to S50:

[0050] The advertising delivery control method provided in this embodiment can be executed in an advertising delivery system. The advertising delivery system may include a user terminal node, which is used for users to upload business district images captured by a camera, and an advertising delivery device node, which is used to deliver targeted advertisements. For example, a billboard set up in a shopping mall can log in to the advertising delivery device node based on a user account through a DAPP (i.e., a decentralized application based on blockchain), an APP, or a webpage. Alternatively, the advertising delivery device node can be logged in based on its own device ID or IP address. The processing node can be any computing power node in the advertising delivery system, which is used to perform data processing actions based on the smart contract deployed on the blockchain to disperse the computing power requirements of the user terminal node and the advertising delivery device node.

[0051] S10: Upload the business district image through the user terminal node;

[0052] After a user logs in to a user terminal node via a webpage, DAPP, or APP, the advertising delivery system can push a business district image acquisition interface to the user terminal node. This acquisition interface provides an image upload interface. This allows users to take a picture of the corresponding business district and then upload the image captured or saved on the user terminal to the advertising delivery system using this upload interface.

[0053] Optionally, to ensure the timeliness of user-uploaded commercial district images, upon receiving a commercial district image uploaded by a user, the user terminal node can first determine whether the image was taken by the user in real time. If so, the user is allowed to upload the image. If not, the uploaded image's capture time is obtained, and a determination is made as to whether the interval between the capture time and the current time is greater than or equal to a preset time interval. If so, the image is deemed unqualified; otherwise, the image is deemed qualified and the user is allowed to upload the image.

[0054] Exemplarily, when receiving a business district image uploaded by a user, the data source of the business district image is first determined. If the data source is a camera, the business district image is determined to be a real-time picture. If the data source is a storage device, the shooting time point of the picture is further obtained. For example, the shooting time point can be determined based on the time watermark in the picture, or the shooting time point can be determined based on the storage time of the picture. If the interval between its shooting time point and the current time point is greater than or equal to the preset threshold, or the shooting time point is earlier than the preset time point (wherein the preset time point is determined based on the current time point and the preset time interval), it is judged to be an invalid picture. The picture can be discarded and a prompt message indicating that the picture has expired is output. Otherwise, the picture uplink action is executed.

[0055] In this embodiment, users actively upload images of business districts, thereby increasing user engagement with the commercial environment and enhancing their user experience. This differs from existing technologies, where advertising push relies on the passive collection of user data. This passive data collection model often causes user dissatisfaction, can lead to users closing information collection channels, and in serious cases, can even cause privacy issues. In contrast, the present invention provides users with rewards such as business district discount coupons and gifts within a pre-configured business model. Users can only receive discounts by uploading images of the business district with location features. This approach transforms the passive acquisition of feature data into an active, authorized user engagement, improving the user experience while increasing the scale and richness of the acquired data and providing a robust sample base for subsequent data analysis. In this way, user-uploaded business district images can also be circulated on social media, attracting more potential users and relevant data samples, while also boosting the popularity of the business district and increasing traffic and profits.

[0056] S20: Acquire location information corresponding to the business district image;

[0057] After receiving the commercial district image uploaded by the user terminal node, the advertising delivery system can further obtain the location information of the commercial district image. The acquisition of the location information of the commercial district image can be implemented in different ways.

[0058] Optionally, in one embodiment, location information corresponding to the shopping district image may be determined based on the image information of the shopping district image, wherein the location information describes the shooting location of the shopping district image.

[0059] For example, in the process of determining location information based on a commercial district image, after receiving the commercial district image, building features in the commercial district image can be pre-extracted. For example, the received commercial district image can be loaded first. The commercial district image is then grayscaled to reduce computational complexity. Image smoothing is then performed to reduce the impact of noise on edge detection. Finally, edge detection is performed.

[0060] During edge detection, an appropriate edge detection operator, such as the Sobel operator, Prewitt operator, or Canny operator, is first selected. The selected edge detection operator is then convolved with the preprocessed image to detect edges. The Sobel and Prewitt operators detect edges by calculating the horizontal and vertical gradients of the image, while the Canny operator combines gradient information with a non-maximum suppression algorithm to more accurately detect edges.

[0061] Optionally, the detected edges can be further refined to reduce their width and make them closer to the actual contours, and morphological operations (such as erosion, dilation, opening, closing, etc.) can be performed to optimize the shape and continuity of the edges.

[0062] Based on the results of edge detection, the outline features of the landmark building can be extracted from the edges. This involves using a contour tracking algorithm to extract the complete contour line and calculate the contour's geometric features (such as length, area, and curvature). Key points on the contour, such as corners and inflection points, are then extracted as the landmark building's salient features.

[0063] After extracting the outline features, the extracted landmark building features are matched with those in the preset city model. For example, feature point matching algorithms such as SIFT (Scale-Invariant Feature Transform), SURF (Speeded-Up Robust Features), and ORB (Oriented FAST and Rotated BRIEF) can be used to find the most similar landmark building. Based on the matching results, the specific location of the landmark building in the image is determined. Finally, the building's location is used as the location information corresponding to the shopping district image, namely the location where it was taken.

[0064] For example, in another process of determining location information based on a shopping district image, feature point detection algorithms such as Moravec corner detection, Harris corner detection, and Shi-Tomasi corner detection can also be used to detect corners or key points in the image. This is because these key points are usually the points with the most dramatic changes in the image, such as building corners and edge intersections.

[0065] Each detected keypoint is then characterized to generate a feature descriptor. A feature descriptor is a multidimensional vector that contains the keypoint's location, scale, orientation, and surrounding image gradient information. Optional feature description algorithms include SIFT, SURF, and ORB.

[0066] After obtaining the feature descriptors corresponding to the shopping district image, the extracted feature descriptors are matched with the feature descriptors of landmark buildings in the pre-set city model to find the most similar feature pairs. The identity and location of the landmark buildings in the image are then determined based on the number and similarity of the matched feature pairs. Finally, the location of the building is used as the location information corresponding to the shopping district image, namely the location where the image was taken.

[0067] For example, in another process of determining location information based on a commercial district image, the latitude and longitude geographic coordinate information associated with the photo may be directly extracted based on the attribute information of the commercial district image, and the location information may be determined based on the latitude and longitude geographic coordinate information.

[0068] Alternatively, in another embodiment, the aforementioned location information may be determined using the positioning information of the user terminal node. Specifically, when a user uploads a shopping district image via the user terminal node, the user terminal's current location is simultaneously acquired, and the location information is then determined based on the user terminal's current location. Specifically, the location information corresponding to the shopping district image is determined based on the positioning information uploaded by the user terminal node.

[0069] It should be noted that the technical solution provided in this embodiment involves obtaining the location information and images of the terminal. This may involve the user's privacy. Therefore, before obtaining, an authorization request interface can be set. Only when the user grants the corresponding data acquisition permission through the authorization request interface will the corresponding data acquisition operation be performed. Otherwise, the data acquisition operation will not be performed to ensure the security of the user's private data. However, in this case, if the location information of the uploaded image cannot be obtained due to user selection or other reasons, the system can prompt the user to re-upload the image or allow the location information to be obtained before the user can participate in the business district activities. It is also possible to directly allow users to participate in business district activities, but give a lower limit of rewards.

[0070] S30: Determine a target advertising device based on the business district where the location information is located;

[0071] In this embodiment, the business district area corresponding to each business district can be pre-defined based on a map. After determining the location information, the business district area where the location information is located is determined. The advertising device corresponding to the online advertising device node within the business district area is set as the target advertising device. One or more advertising devices can be set within a business district area.

[0072] Optionally, as a way to define business districts, an administrator node can be set up in the advertising delivery system. The administrator node can upload a map with business district area definitions to the chain through the administrator node. The advertising delivery node then determines the specific ranges corresponding to each business district based on the business district area definitions uploaded by one or more administrator nodes. Alternatively, a user terminal node can upload a map with business district area definitions, and the system can then calculate the corresponding ranges for each business district based on the maps uploaded by multiple user terminals.

[0073] S40: Determine a group portrait corresponding to the business district area according to the business district image, and select a target advertisement from an advertisement library based on the group portrait;

[0074] In an optional embodiment, the group profile corresponding to the business district area may include two sets of features. The first set of features is determined based on the user profile corresponding to the user terminal node, and the second set of features is determined based on the human portrait in the business district image. Of course, in some variant embodiments, the user group profile may also be determined based solely on the human portraits of the people in the business district image, or the user profiles corresponding to the user terminal node. This embodiment does not exclude the option of determining the group profile based on a single type of portrait.

[0075] Optionally, when determining the user profile corresponding to a user terminal node, consumption data associated with the terminal device corresponding to the user terminal node can be obtained, provided the user grants data access permission. This consumption data can include online shopping data and offline payment data. The user profile is constructed based on the user's age, gender, preferences, and other data authorized during the user registration process. After constructing the user profile, the final user profile description is isolated from the data source to further protect the security of the user data. This only utilizes the informational power of data that cannot be traced back to its source.

[0076] In the process of obtaining human portraits in a business district image, the human portraits corresponding to the human bodies in the business district image can be determined based on the business district image, and the group portrait can be constructed based on the human portraits corresponding to multiple different business district images.

[0077] After acquiring the commercial district image, the first step is image preprocessing. This involves using a linear dimensionality reduction algorithm such as principal component analysis to reduce the dimensionality of the input image. The image is then encoded using methods such as One-Hot or Word2Vec and resized to a standard size.

[0078] PCA transforms a given set of correlated variables into another set of uncorrelated variables through linear transformation, that is, it projects high-dimensional data into a low-dimensional space, that is, it uses fewer variables in the input data to explain more variables, and these new variables are arranged in descending order of variance. The specific steps are:

[0079] Given a data matrix PCA works by solving the eigendecomposition of the covariance matrix:

[0080]

[0081] Among them, U is the eigenvector matrix and Λ is the eigenvalue diagonal matrix.

[0082] In some embodiments, one-hot encoding is used to convert the extracted categorical features into numerical values. For a discrete feature with K categories, one-hot encoding converts it into a K-dimensional vector, where the i-th category is represented as a vector with 1 only in the i-th dimension and 0 in the rest:

[0083] For Category C i , coded as

[0084] In some embodiments, features extracted using the Word2Vec model are converted into dense vector representations. Taking the Skip-Gram model as an example, the goal is to maximize the probability of background words given a central word:

[0085]

[0086] in,

[0087] Afterwards, the obtained dense feature vector representation is subjected to nonlinear dimensionality reduction, and the sequence obtained after dimensionality reduction is learned using an autoencoder to obtain a potential representation of the data that is more representative.

[0088] In some embodiments, the t-SNE algorithm is applied to perform nonlinear dimensionality reduction. t-SNE minimizes the Kullback-Leibler divergence of the probability distribution of point pairs in high-dimensional space and low-dimensional space:

[0089]

[0090] Among them, the similarity p in high-dimensional space ij Defined as:

[0091]

[0092] Similarity q in low-dimensional space ij Defined as:

[0093]

[0094] Image features are extracted by constructing an autoencoder. This mechanism allows for automatic learning of the data's latent representation and high-order statistical properties from large amounts of unlabeled data, reducing the cost of manual annotation while also avoiding the subjectivity and uncertainty associated with manually designed features. The goal of the autoencoder is to minimize the difference between the input data and the reconstructed data. It accepts a sequence of image features transformed by the embedding layer and uses a convolutional neural network to move across the time series using a sliding window mechanism, capturing local features and contextual information to generate a more representative latent representation. This latent representation is then input into a decoder, which uses a deconvolutional neural network to reconstruct the original feature sequence by converting the latent representation back into a sequence of the same latitude as the input.

[0095] Train the feature extraction network and minimize the reconstruction error during training:

[0096]

[0097] Among them, f θ is the encoder function, g φ is the decoder function.

[0098] The autoencoder learns a sequence of image features that is efficiently and densely compressed and reconstructed.

[0099] For example, image recognition and basic feature extraction can be performed on the shopping district image to segment the human body in the shopping district image. For example, to achieve human body detection and segmentation, deep learning models such as YOLO and SSD can be used for human body detection to identify the location of the human body in the shopping district image. Then, image segmentation techniques (such as Mask R-CNN) can be applied to more accurately segment the human body area to reduce background interference.

[0100] Then, a pre-trained deep learning model is used to extract features from the segmented human body regions. The extracted features are input into a specially trained age and gender classifier to predict the age and gender of each person.

[0101] Specifically, a deep convolutional neural network can be used as a feature extractor, and the feature extraction process can be similar to the aforementioned embodiment.

[0102] For clothing and accessories analysis, images of human bodies can be analyzed to identify clothing type, color, brand, and accessory descriptions. Based on this information, individual consumer preferences and spending habits can be inferred. Finally, feature integration is performed. This integrates basic features such as age, gender, clothing and accessory information, and consumer preferences obtained from image recognition. Based on these integrated features, a user profile is constructed.

[0103] It's important to note that during model training, a large amount of labeled data can be used to train age and gender recognition, as well as clothing and accessories analysis models. New data is continuously collected and the models are iteratively optimized to improve recognition accuracy and personalized recommendations. Throughout this process, strict compliance with privacy regulations ensures the anonymity and security of user data.

[0104] Furthermore, after determining the user portrait and body portrait, a group portrait corresponding to the business district area can be generated based on the user portrait set and / or body portrait set. Then, based on the group portrait, target advertisements are selected from the advertisement library.

[0105] In an optional embodiment, a group portrait can be obtained by clustering and analyzing the set of user categories using the K-means method. Specifically, after obtaining a user feature vector based on user data, the user feature vector index is marked, and then multiple feature centroids are selected from the feature vector. The sum of the squares of the distances between the feature sample points and the centroid points within the group is determined using a distance function, and the points are classified into the class with the smallest distance to the centroid:

[0106]

[0107] Among them, μ k is the centroid of the k-th cluster.

[0108] Then recalculate the centroid of each class that has been obtained, and repeat it iteratively until the new centroid is less than or equal to the original centroid, so as to determine the K value, that is, the number of K classes, and then cluster and divide user portraits with different category characteristics through the K-means method. In some other methods, decision tree analysis, Bayesian classification, multivariate logistic regression analysis or other neural network methods can also be used to establish user portraits. In the process of extracting user preference rules using the decision tree model, the optimal feature division is selected based on information gain or Gini coefficient, and the user categories are divided in turn. Information gain is defined as:

[0109]

[0110] Where Ent(D) is the information entropy of data set D.

[0111] In an optional implementation scheme, the group portrait is input into a pre-trained classification model, and the output data of the classification model is obtained. Then, based on the matching degree between the output data and the advertisements in the advertisement library, one or more advertisements are selected from the advertisement library as the target advertisements.

[0112] Exemplarily, a sample group portrait can be constructed based on the image corresponding to the preset business district, and then the sample group portrait can be annotated with weights. In order to add corresponding labels to the sample group portrait. In this embodiment, during the training phase, the user terminal node can make annotations of the business district group portrait while uploading the business district picture. In this way, more labels can be obtained through the user terminal, as well as annotations associated with the business district image. After collecting enough labeled business district images, a group portrait can be constructed based on the business district image, and the group portrait can be annotated with weights based on the labels associated with the business district image. In the annotation process, the weight value of the group portrait under the label is determined based on the number of times the same label appears. Optionally, in some embodiments, in order to reduce data processing overhead, the labels can be standardized based on the semantic information of the labels provided by the user to reduce the number of labels. In this way, after multiple rounds of iterative training, the model can automatically predict the label corresponding to the group portrait based on the input group portrait determined based on the business district image.

[0113] Furthermore, similar to the annotation method used for group portraits, ads in the ad library can also be sent to user terminals, where they can be annotated. A classifier is then trained based on the annotated ads, allowing the classifier to predict the label for each ad after multiple rounds of iterative training.

[0114] Before predicting the label corresponding to an advertisement, the advertisement features must first be extracted. In some embodiments, a convolutional neural network (CNN) model is used to extract the advertisement visual features. The forward propagation of the CNN can be expressed as:

[0115] f=σ(W*x+b)

[0116] Among them, * represents the convolution operation and σ is the activation coefficient.

[0117] In some embodiments, the BERT model is used to extract advertisement semantic features. This model is based on the Transformer architecture and uses the self-attention mechanism. Its core calculation is:

[0118]

[0119] Among them, Q, K, and V are query, key, and value matrices respectively.

[0120] In some embodiments, the implicit features of the advertisement may also be learned through a factorization machine (FFM) model, and the prediction method is as follows:

[0121]

[0122] in, Indicates that feature i is in field f j The hidden vector of .

[0123] Then, corresponding advertising features are obtained by obtaining the visual features, semantic features and implicit features of the advertisement.

[0124] After obtaining the tags corresponding to the ad and the group profile, the matching degree between the group profile and the ads in the ad library can be determined based on the matching degree between the tags. The ads in the ad library are then sorted based on the matching degree. Based on the sorting results, the ad with the highest matching degree, or the top n matching ads, is selected as the target ad.

[0125] It should be noted that if the top n matching ads are selected as target ads, the play duration of the top n ads in a carousel cycle can also be determined based on the matching degree. The matching degree can be set to be positively correlated with the play duration.

[0126] In some embodiments, the degree of matching is specifically manifested as an estimated click-through rate of the advertisement. The acquired advertisement features and user features are used to train a click-through rate prediction model. The click-through rate prediction model may be a gradient boosted decision tree (GBDT) model, which is trained using a training sample set formed by advertisement feature vectors and user feature vectors, and is validated using a test sample set to adjust the parameters of the click-through rate prediction model.

[0127] The gradient boosted decision tree (GBDT) model form is:

[0128]

[0129] Among them, h m (x) is the mth decision tree, γ m The corresponding weight.

[0130] Further, refer to Figure 3 , based on the trained click-through rate prediction model, the click-through rate in the ad library is estimated, and the ads in the ad library are arranged in descending order of click-through rate.

[0131] S50: Control the target advertisement delivery device to play the target advertisement.

[0132] After the target advertisement corresponding to the target advertisement delivery device is determined by the advertisement delivery system, the rendering data of the target advertisement can be read from the storage node of the target advertisement, and the rendering data can be sent to the node of the target advertisement delivery device, so that when the target advertisement delivery device receives the rendering data, it controls the target advertisement delivery device to play the target advertisement.

[0133] In the technical solution provided in this embodiment, a business district image is first uploaded through a user terminal node. The location information corresponding to the business district image is then obtained. Based on the business district area where the location information is located, a target advertising delivery device is determined. Furthermore, based on the business district image, a group profile corresponding to the business district area is determined. This allows the target advertisement to be selected from the advertisement library based on the group profile, and the target advertising delivery device is controlled to play the target advertisement. Because the group profile corresponding to the business district at the current moment can be determined based on the real-time business district image uploaded by the user terminal node, and advertisements matching the group profile can be selected for playback, this effectively improves the accuracy of advertisement delivery and enhances the user experience. Therefore, the advertising delivery revenue per unit time can be effectively increased. Furthermore, implementing the above solution based on blockchain better achieves the goal of decentralization and effectively avoids the problem of low advertising delivery efficiency caused by human interference.

[0134] In one embodiment, the above steps further include:

[0135] S51: Based on the target advertising effect indicators, an optimization algorithm is used to adjust the advertising delivery strategy.

[0136] The advertising effect indicator may include one or more of the following indicators:

[0137] Return on investment (ROI): The ratio between the net income generated by an investment activity and the investment cost;

[0138] Cost per thousand impressions (CPM): The cost per thousand ad impressions during the advertising process;

[0139] Click-through rate (CTR): The ratio between the number of times an ad or link is clicked and the number of times it is displayed.

[0140] Engagement rate: The ratio of the number of times users interact with an ad, content, or social media post (such as likes, comments, shares, etc.) to the number of times it was displayed.

[0141] Advertising effectiveness indicators may also include but are not limited to the following indicators: user stay time, brand reputation, retention rate, customer acquisition cost, and lifetime value.

[0142] Since the effect of advertising may take a long time to obtain, in order to further improve the efficiency of advertising, the deep Q network (DQN) is used to predict the advertising effect indicators. Its loss function is:

[0143]

[0144] Among them, θ and θ - are the parameters of the current and target networks, respectively.

[0145] The advertising delivery strategy includes the matching degree between the group portraits mentioned above and the advertisements in the advertisement library, and also includes the matching degree between the advertising delivery device and the advertisements. The matching degree between the above-mentioned advertising delivery device and the advertisement can refer to the matching degree between the advertising delivery devices and advertisements at different coordinates, that is, the spatial selection of advertisement delivery; it can also refer to the time matching degree between a specific advertising delivery device and the advertisement, that is, the time selection of advertisement delivery, and can also refer to the delivery frequency of a specific advertisement. Those skilled in the art can also know that different advertisements can be combined and delivered to the advertising delivery devices at the same location in the same time period. The combined advertisement delivery method on a specific advertising delivery device is not limited by the present application, that is, the spatial selection subdivision of the advertising delivery strategy can also refer to different delivery positions on a specific advertising delivery device.

[0146] In some embodiments, the optimization algorithm can be specifically a multi-armed bandit algorithm MBA, which is derived from a coin-operated slot machine with multiple lever arms in a casino. The customer's goal is to maximize the profit by making selection decisions under the condition of a limited number of pulling the lever arms. Assume that there are a total of K slot machines, and each slot machine has a certain probability of generating a reward after being pulled. The slot machines represent different selection strategies, and the customer needs to decide which slot machine's lever arm to pull each time, that is, to execute different selection strategies. Furthermore, the confidence upper bound algorithm UBC is used to maintain a confidence upper bound value for each slot machine, and each time the slot machine with the largest confidence upper bound value is selected to perform the operation. The higher the confidence upper bound value of the average reward of a slot machine, the more worthy it is to be selected. The calculation formula is as follows:

[0147]

[0148] Parameter explanation: Bi(t) is the UCB value; Xi(t) is the average reward obtained by slot machine i, i.e., the i-th advertising strategy, in the first t rounds; c is a constant greater than zero; t represents the current round number; n i (t) is the number of times slot machine i, i.e., the i-th advertising strategy, has been pulled and selected in the first t rounds. i (t) reflects the estimation of the reward level of the current slot machine, i.e., the advertising delivery strategy, and is the utilization part of the multi-armed slot machine algorithm; It increases slowly with the increase of the number of rounds t, and increases with n i (t) decreases as the number of times it is selected increases, thus encouraging the exploration of slot machine strategies that are selected less frequently, which constitutes the development part of the multi-armed slot machine algorithm. This method balances the use of the current optimal strategy with the development of new optimal strategies.

[0149] More specifically, the algorithm includes the following steps: S511: Initializing the average reward of each slot machine, i.e., the advertising delivery strategy, to 1 at the beginning of execution.

[0150] S512: Each slot machine is selected and pulled once; for each round t, the UCB value corresponding to each slot machine is calculated;

[0151] In this step, each slot machine is selected and activated once, which means that multiple aggregated advertising delivery strategies to be optimized are sequentially executed or simulated, and indicator data corresponding to the execution results is obtained. The indicator data can also be simulated indicator data. The indicator data can be one or more of the advertising delivery strategy effectiveness data described above. The weighted average of the indicator data or simulated indicator data is used as the reward value for that round and input into the algorithm model to obtain the UCB value.

[0152] S513: Select the largest slot machine in UCB to perform a pull operation.

[0153] S514: The average reward and the number of times the slot machine is pulled are updated according to the reward value obtained by the slot machine with the largest UCB in this round. The number of times the slot machine is pulled and the average reward of other unselected slot machines remain unchanged.

[0154] S515: Repeat steps S512 to S514 until a preset termination condition is reached. The preset termination condition may be that the total number of calculation rounds reaches an upper limit.

[0155] Furthermore, the output result is the advertising delivery strategy with the largest UCB under the current conditions, that is, the local optimal solution of the advertising delivery strategy under the current round or other restrictive conditions in the algorithm, that is, the optimized advertising delivery strategy.

[0156] In some embodiments, a greedy algorithm can be further used to further optimize the selection of slot machines, that is, advertising delivery strategies. After obtaining the optimal solution for the current round, a strategy can be randomly selected from multiple advertising delivery strategies with a probability e to be input into the algorithm, and several strategies with larger UCB values can be selected from the explored advertising delivery strategies with a probability (1-e) to be input into the algorithm. In other words, if the greedy algorithm finds that a certain strategy has a good indicator effect in the most recent executions or simulated executions, the utilization of the strategy will be increased. One way is to reduce the value of the constant c in the formula so that the algorithm focuses more on utilizing the current strategy rather than developing and exploring other strategies.

[0157] In some embodiments, the preset termination condition in step S515 may also be that the cumulative regret value is as small as possible, which can be calculated by the following formula:

[0158]

[0159] Where: R(T) is the cumulative regret value, μ represents the average reward score of the best advertising strategy in each round of advertising (that is, the average benefit of the best strategy), T represents the total number of rounds of advertising, that is, μT represents the total reward score that can be obtained if the best strategy can always be selected in T rounds, which is an ideal situation; r i (a t ) represents the adoption of a specific strategy a in the tth round of advertising. t The actual reward points earned.

[0160] This formula measures the difference between the actual score achieved up to the current round T and the score that a fixed strategy could have achieved. The larger the cumulative regret value, the greater the performance gap between the current strategy and the optimal target ad delivery strategy.

[0161] In some embodiments, different individual advertising delivery strategies can be combined based on the advertising delivery effectiveness indicators obtained through execution or simulation prediction to obtain multiple aggregated advertising delivery strategies to be optimized. These aggregated advertising delivery strategies to be optimized can focus on the initial optimal solution for a single advertising delivery effectiveness indicator, or can also take into account the initial optimal solutions for multiple advertising delivery effectiveness indicators. These aggregated advertising delivery strategies to be optimized are input into the aforementioned multi-armed bandit algorithm for calculation to obtain the optimized aggregated advertising delivery strategy.

[0162] After selecting the optimized aggregated advertising delivery strategy, the process returns to step S50, controlling the target advertising delivery device to deliver the target advertisement according to the optimized aggregated advertising delivery strategy. Simultaneously, based on the balance mechanism between exploration and exploitation in the multi-armed bandit algorithm, the process continues with the steps of selecting or simulating multiple aggregated advertising delivery strategies to be optimized, obtaining indicator data corresponding to the execution results, and obtaining the corresponding optimized aggregated advertising delivery strategy.

[0163] In some embodiments, different individual advertising delivery strategies can be combined based on the predicted advertising delivery performance indicators to obtain multiple aggregated advertising delivery strategies to be optimized. These aggregated advertising delivery strategies to be optimized can focus on the initial optimal solution for a single advertising delivery performance indicator, or can consider the initial optimal solutions for multiple advertising delivery performance indicators.

[0164] One of the aggregated advertising delivery strategies to be optimized is processed and calculated as a single arm in the multi-armed bandit algorithm, and the processed advertising delivery effect indicators are analyzed. Then, based on the processed advertising delivery effect indicators, the best one or the top-ranked advertising aggregation delivery strategy is selected as the optimized advertising aggregation delivery strategy.

[0165] Specifically, multiple aggregated advertising delivery strategies to be optimized are sequentially executed or simulated, and indicator data corresponding to the execution results is obtained. The indicator data may also be simulated indicator data. The indicator data may be one or more of the aforementioned advertising delivery strategy performance data. Based on the indicator data and the reward mechanism in the multi-armed bandit algorithm, different aggregated advertising delivery strategies to be optimized are scored. The aggregated advertising delivery strategy to be optimized whose score meets the preset rules is determined as the optimized aggregated advertising delivery strategy.

[0166] As a specific example, the optimized advertising aggregation delivery strategy may be: during holidays from 13:00 to 18:00, on Lan Kwai Fong advertising delivery device H, deliver advertisement A of advertiser X four times per hour.

[0167] After selecting the optimized aggregated advertising delivery strategy, the process returns to step S50, controlling the target advertising delivery device to deliver the target advertisement according to the optimized aggregated advertising delivery strategy. Simultaneously, based on the balancing mechanism between exploration and development within the multi-armed bandit algorithm, the process continues with the steps of selecting, executing, or simulating the execution of multiple aggregated advertising delivery strategies to be optimized, obtaining indicator data corresponding to the execution results, and obtaining the corresponding optimized aggregated advertising delivery strategy. This approach balances the application of the current optimal strategy with the development of the new optimal strategy.

[0168] Please refer to Figure 2 Based on the above embodiment, in another optional implementation scheme, the advertisement delivery control method further includes:

[0169] S60: Determine a sensitive area corresponding to the business district image, and determine a hash value of the sensitive area;

[0170] S70: Obtaining a hash value of a sensitive area of an image stored in a database;

[0171] S80: discarding the business district image.

[0172] In this embodiment, in order to prevent different or the same user terminal nodes from repeatedly uploading the same business district picture, when the user terminal node receives the business district picture uploaded by the user, the picture can be deduplicated based on the hash value of the sensitive area.

[0173] For example, after acquiring a commercial district image, local features can be extracted from the image. For example, key points, edges, textures, etc. in the image can be set as sensitive areas as required, and then local features of the sensitive areas can be extracted from the image using a feature extraction algorithm (such as SIFT, SURF, ORB, etc.). Optionally, before extracting local features, the image can be scaled to a fixed size during image preprocessing to eliminate the impact of different sizes on feature extraction.

[0174] After extracting the local features, the extracted features are converted into feature vectors. Each feature can be a multidimensional vector, which contains information such as position, scale, and direction, depending on the feature extraction algorithm used. After being converted into a feature vector, a local sensitive hash function is constructed. For example, a set of hash functions can be selected or designed that can map similar feature vectors to similar hash values. After applying the hash function to each feature vector and generating one or more hash values, a hash table generated based on historically uploaded business district images or some other data structure that supports fast retrieval is obtained. It is also queried whether there are entries similar to the business district image to be uploaded. The judgment threshold for whether it is similar can be customized according to the deduplication requirements.

[0175] If a hash value for the sensitive area exists that matches the hash value by a degree greater than a preset threshold, the corresponding data structure is searched for a similar entry, and the shopping district image is determined to be a duplicate and discarded. Otherwise, the shopping district image is uploaded via the user terminal node, and the hash value is written into the corresponding data structure.

[0176] Alternatively, see Figure 3 In some variant embodiments, in order to meet further deduplication requirements, the method further includes:

[0177] S90: Determine whether there is a human body in the business district image;

[0178] Furthermore, based on the above embodiment, if a shopping district image does not contain human body information, it cannot provide valid information in the subsequent processing flow. Therefore, a filtering condition based on the presence of human body can be set. If no human body is contained, the shopping district image is discarded; otherwise, the upload action is performed.

[0179] It is understandable that when performing human body recognition, it can be implemented based on any human body recognition technology, and this embodiment will not be described in detail here.

[0180] The present application provides an advertising delivery system, which includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute the advertising delivery control method in the above-mentioned embodiment 1.

[0181] Reference below Figure 4 , which shows a schematic diagram of the structure of an advertisement delivery system suitable for implementing the embodiment of the present application. The advertisement delivery system in the embodiment of the present application may include but is not limited to a device with data processing functions such as a server. Figure 4 The advertisement delivery system shown is merely an example and should not limit the functions and scope of use of the embodiments of the present application.

[0182] like Figure 4As shown, the advertising delivery system may include a processing device 1001 (e.g., a central processing unit, a graphics processing unit, etc.), which can perform various appropriate actions and processes based on programs stored in a read-only memory (ROM) 1002 or programs loaded from a storage device 1003 into a random access memory (RAM) 1004. RAM 1004 also stores various programs and data required for the operation of the advertising delivery system. Processing device 1001, ROM 1002, and RAM 1004 are interconnected via a bus 1005. An input / output (I / O) interface 1006 is also connected to the bus. Typically, the following systems can be connected to I / O interface 1006: an input device 1007 including, for example, a touch screen, a touchpad, etc.; an output device 1008 including, for example, a liquid crystal display (LCD), a speaker, a vibrator, etc.; a storage device 1003 including, for example, a hard disk; and a communication device 1009. The communication device 1009 can allow the advertising delivery system to communicate with other devices wirelessly or wired to exchange data. Although the figure shows an advertising delivery system with various systems, it should be understood that it is not required to implement or have all of the systems shown. More or fewer systems can be implemented or provided instead.

[0183] In particular, according to the embodiments disclosed in the present application, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, the embodiments disclosed in the present application include a computer program product comprising a computer program carried on a computer-readable medium, the computer program comprising program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from a network via a communication device, or installed from a storage device 1003, or installed from a ROM 1002. When the computer program is executed by the processing device 1001, the above-mentioned functions defined in the method of the embodiment disclosed in the present application are executed.

[0184] The advertising delivery system provided in this application utilizes the advertising delivery system method of the aforementioned embodiment to resolve the technical issue of inaccurate advertising delivery. The beneficial effects of the advertising delivery system provided in this application are the same as those of the advertising delivery control method provided in the aforementioned embodiment. Other technical features of the advertising delivery system are the same as those disclosed in the aforementioned embodiment and are not further elaborated here.

[0185] It should be understood that the various parts disclosed in this application can be implemented using hardware, software, firmware, or a combination thereof. In the description of the above embodiments, specific features, structures, materials, or characteristics can be combined in any one or more embodiments or examples in a suitable manner.

[0186] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of this application. Therefore, the scope of protection of this application should be based on the scope of protection of the claims.

[0187] The present application provides a computer-readable storage medium having computer-readable program instructions (ie, computer program) stored thereon, and the computer-readable program instructions are used to execute the advertisement delivery control method in the above-mentioned embodiment.

[0188] The computer-readable storage medium provided in this application may be, for example, a USB flash drive, but is not limited to electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, systems or devices, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof. In this embodiment, the computer-readable storage medium may be any tangible medium that contains or stores a program that can be used by or in conjunction with an instruction execution system, system or device. The program code contained on the computer-readable storage medium may be transmitted using any appropriate medium, including but not limited to: wires, optical cables, RF (Radio Frequency), etc., or any suitable combination thereof.

[0189] The computer-readable storage medium may be included in the advertisement delivery system; or it may exist independently without being assembled into the advertisement delivery system.

[0190] The above-mentioned computer-readable storage medium carries one or more programs. When the above-mentioned one or more programs are executed by the advertising delivery system, the advertising delivery system first determines the outdoor ambient temperature after starting the noise reduction function of the outdoor unit, and then determines the frequency threshold at the current moment based on the outdoor ambient temperature, and controls the advertising delivery system to operate below the corresponding frequency threshold, thereby improving the user experience of the advertising delivery system.

[0191] Computer program code for performing the operations of the present application may be written in one or more programming languages, or a combination thereof, including object-oriented programming languages such as Java, Smalltalk, C++, and conventional procedural programming languages such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a stand-alone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving a remote computer, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0192] The flow charts and block diagrams in the accompanying drawings illustrate the possible architecture, functions and operations of the systems, methods and computer program products according to various embodiments of the present application. In this regard, each box in the flow chart or block diagram can represent a module, program segment or a part of code, and the module, program segment or a part of code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order than that marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram and / or flow chart, and the combination of the boxes in the block diagram and / or flow chart can be implemented by a dedicated hardware-based system that performs the specified function or operation, or can be implemented by a combination of dedicated hardware and computer instructions.

[0193] The modules described in the embodiments of the present application may be implemented in software or hardware, wherein the name of a module does not necessarily limit the unit itself.

[0194] The computer-readable storage medium provided in this application stores computer-readable program instructions (i.e., a computer program) for executing the aforementioned advertising delivery control method, thereby resolving the technical issue of inaccurate advertising delivery. Compared to the prior art, the beneficial effects of the computer-readable storage medium provided in this application are the same as those of the advertising delivery control method provided in the aforementioned embodiment, and are not further elaborated here.

[0195] An embodiment of the present application provides a computer program product, including a computer program, which implements the steps of the above-mentioned advertising delivery control method when executed by a processor.

[0196] The computer program product provided in this application can solve the technical problem of inaccurate advertising delivery. Compared with the prior art, the beneficial effects of the computer program product provided in the embodiment of this application are the same as the beneficial effects of the advertising delivery control method provided in the above embodiment, and will not be repeated here.

[0197] The above are only preferred embodiments of the present application and do not limit the scope of the present invention. Any equivalent structure or equivalent process transformation made by using the contents of the present application specification and drawings, or directly or indirectly applied to other related technical fields, are also included in the scope of the present invention.

[0198] It should be noted that, in this article, the terms "comprise", "include" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or system comprising a series of elements includes not only those elements, but also includes other elements not explicitly listed, or also includes elements inherent to such process, method, article or system. In the absence of further restrictions, an element defined by the sentence "comprises a ..." does not exclude the presence of other identical elements in the process, method, article or system comprising the element. In the intervals given in this application, all include boundary values unless explicitly defined.

[0199] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above embodiment methods can be implemented by means of software plus the necessary general hardware platform, and of course can also be implemented by hardware, but in many cases the former is a better implementation method.

[0200] The above are only preferred embodiments of the present application and do not limit the patent scope of the present application. Any equivalent structure or equivalent process transformation made using the contents of the present application specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present application.

Claims

1. An advertisement delivery control method, characterized in that: The advertisement delivery control method includes: Upload business district images through user terminal nodes; Obtaining location information corresponding to the business district image; Determining a target advertising device based on the business district where the location information is located; Determining a group portrait corresponding to the business district area according to the business district image, and selecting a target advertisement in an advertisement library based on the group portrait; Control the target advertisement delivery device to play the target advertisement.

2. The advertising delivery control method according to claim 1, wherein: The advertisement delivery control method further includes: Determining a sensitive area corresponding to the business district image, and determining a hash value of the sensitive area; Get the hash value of the sensitive area of the image saved in the database; If there is a hash value of the sensitive area whose matching degree with the hash value is greater than a preset threshold, the business district image is discarded.

3. The advertising delivery control method according to claim 2, wherein: After the step of obtaining the hash value of the sensitive area of the image stored in the database, the advertisement delivery control method further includes: If there is no hash value of the sensitive area whose matching degree with the hash value is greater than the preset threshold, determining whether there is a human body in the business district image; If not, discard the business district image.

4. The advertisement delivery control method according to any one of claims 2 to 3, characterized in that: After the step of discarding the business district image, the advertisement delivery control method further includes: The user terminal node is controlled to output a prompt message indicating that the business district image does not meet the requirements.

5. The advertisement delivery control method according to claim 1, wherein: The step of obtaining the location information corresponding to the user terminal node includes: Determining the location information corresponding to the business district image based on the positioning information uploaded by the user terminal node; or Building features are extracted based on the commercial district image, and the location information corresponding to the commercial district image is determined based on a matching result between the building features and a preset city model.

6. The advertisement delivery control method according to claim 1, wherein: The step of determining the user group corresponding to the business district area according to the business district image, and selecting a target advertisement in the advertisement library based on the group portrait includes: Determining, based on the business district image, a human portrait corresponding to a human body in the business district image; constructing the group portrait based on the human portraits corresponding to the multiple different business district images; Target advertisements are selected from the advertisement library based on the group portrait.

7. The advertisement delivery control method according to claim 6, wherein: The step of selecting a target advertisement in the advertisement library based on the group portrait includes: Inputting the group portrait into a pre-trained classification model and obtaining output data of the classification model; One or more advertisements are selected from the advertisement library as the target advertisements according to a matching degree between the output data and the advertisements in the advertisement library.

8. The advertisement delivery control method according to claim 7, wherein: The advertisement delivery control method further includes: Construct a sample group portrait based on the images corresponding to the preset business district, and mark the advertising label corresponding to the sample group portrait; The classification model is trained based on the data pairs consisting of the sample group portrait and the advertising label as training samples.

9. The advertisement delivery control method according to claim 1, wherein: The advertisement delivery control method further includes: Performing vectorization processing on the commercial district image at the user terminal and storing the same at the user terminal; The vectorized data is uploaded to the blockchain through the user terminal node.

10. An advertisement delivery system, characterized in that: The advertisement delivery system includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the computer program is configured to implement the steps of the advertisement delivery control method according to any one of claims 1 to 9.

11. A storage medium, characterized in that: The storage medium is a computer-readable storage medium, and a computer program is stored on the storage medium. When the computer program is executed by a processor, the steps of the advertisement delivery control method according to any one of claims 1 to 9 are implemented.

12. A computer program product, characterized in that It comprises a computer program carried on a non-transitory computer-readable medium, the computer program containing program code for executing the method as claimed in any one of claims 1 to 9.

Images