Advertisement intelligent putting effect dynamic promotion method and system based on multi-target optimization

Abstract

The application provides an advertisement intelligent putting effect dynamic promotion method and system based on multi-target optimization, relates to the technical field of advertisement intelligent putting, and solves the technical problems of advertisement putting mechanism solidification, insufficient precision and low resource utilization in the prior art. The method comprises the following steps: obtaining advertisement information to be put, a putting position set and advertisement playing time; identifying and classifying the advertisement information to be put to obtain an advertisement type; screening the putting position set according to the advertisement type to obtain candidate putting positions; dividing the advertisement playing time into time periods according to the candidate putting positions to obtain candidate putting time periods; setting a candidate putting scheme based on the candidate putting positions and the candidate putting time periods; and verifying and screening the candidate putting scheme to obtain an optimal putting scheme. The application is used in the process of advertisement intelligent putting.

Description

Technical Field

[0001] This application relates to the field of intelligent advertising delivery technology, and in particular to a method and system for dynamically improving the effectiveness of intelligent advertising delivery based on multi-objective optimization. Background Technology

[0002] Existing advertising systems generally adopt a static, timed rotation model, relying solely on pre-set playlists to display ads. This approach has three key drawbacks: First, the delivery mechanism is rigid, failing to dynamically adjust content based on real-time scenarios, leading to user visual fatigue. Second, it lacks multi-dimensional data analysis capabilities, failing to effectively combine location features and time-of-day characteristics for precise ad matching. Third, the system's intelligence is insufficient, lacking dynamic optimization capabilities based on real-time feedback, resulting in wasted advertising resources and poor delivery performance, severely limiting the commercial value of advertising and the improvement of user experience. Summary of the Invention

[0003] This application provides a method and system for dynamically improving the effectiveness of intelligent advertising delivery based on multi-objective optimization, which solves the technical problems of rigid advertising delivery mechanisms, insufficient accuracy, and low resource utilization in the prior art.

[0004] To achieve the above objectives, this application adopts the following technical solution:

[0005] Firstly, a method for dynamically improving the effectiveness of intelligent ad delivery based on multi-objective optimization is provided, including:

[0006] The system obtains information on advertisements to be placed, a set of placement locations, and advertisement playback time; wherein, the information on advertisements to be placed includes several product names, advertising copy, and promotional videos, the set of placement locations includes the locations of several advertising playback devices, and the advertisement playback time is all the time during which advertisements are scheduled to play at a given location;

[0007] The advertising information to be delivered is identified and categorized to obtain the advertising type;

[0008] The candidate placement locations are obtained by filtering the set of placement locations based on the ad type;

[0009] The ad playback time is divided into time periods based on the candidate placement locations to obtain the candidate placement time periods;

[0010] Set up candidate delivery plans based on candidate delivery locations and candidate delivery time periods;

[0011] The candidate deployment schemes are verified and screened to obtain the optimal deployment scheme.

[0012] Based on the technical scheme, in the method for dynamically improving the effect of intelligent advertisement distribution based on multi-objective optimization provided in the application, the intelligent dynamic optimization of advertisement distribution is realized through the multi-dimensional matching mechanism of the fusion of space-time features and advertisement content. The precise distribution scene model is constructed based on the personnel feature recognition and time period division of the distribution position, and the multi-label classification technology of the advertisement content is combined, so that the system can automatically calculate the distribution effect score of the advertisement and the current scene, and realize personalized distribution through real-time verification. This method effectively solves the problem of mismatch between advertisement content and audience demand caused by the traditional fixed carousel mode, significantly improves the precision and resource utilization efficiency of advertisement distribution, and avoids user visual fatigue, providing a more intelligent and efficient solution for advertisement distribution scenarios.

[0013] In combination with the first aspect, in a possible implementation manner, the identifying and classifying the advertisement information to be distributed includes:

[0014] The advertisement script is cleaned through a text cleaning tool to obtain an advertisement text;

[0015] The effect keywords of the advertisement text are extracted through natural language processing to obtain a plurality of product keywords;

[0016] A type-keyword library is constructed, the product keywords are matched with the type-keyword library, and a plurality of product classifications are obtained;

[0017] The product name is classified through a preset product type classification table to obtain a product type;

[0018] It is judged whether the plurality of product classifications and the product type are the same. If yes, the product type is marked as an advertisement type. If no, the product classification with the most product keywords in the plurality of product classifications is marked as the advertisement type.

[0019] In combination with the first aspect, in a possible implementation manner, the filtering the set of distribution positions according to the advertisement type includes:

[0020] A plurality of personnel feature sets K of the set of distribution positions are obtained based on a position-personnel feature matching library; wherein the position-personnel feature matching library is constructed based on the position of the advertisement playing device and the personnel features corresponding to the position;

[0021] The keyword set F corresponding to the plurality of advertisement information to be distributed is extracted;

[0022] The matching degree of the keywords in the plurality of personnel feature sets K and the keyword set F is calculated and marked as a position matching degree;

[0023] The positions with the top N position matching degrees are selected to obtain candidate distribution positions; wherein the N is a positive integer.

[0024] In a possible implementation of the first aspect, the matching degree of the words in the plurality of personnel feature sets K and the keyword set F is calculated, including:

[0025] The number of identical words in the plurality of personnel feature sets K and the keyword set F is counted and marked as the identical number D;

[0026] The semantic similarity of a plurality of words in the plurality of personnel feature sets K and the keyword set F is calculated by Word2Vec, the number of words with a semantic similarity greater than a preset threshold is counted, and marked as the similar number R;

[0027] The identical number D and the similar number R are weighted and summed by a position weighted sum formula to obtain the position matching degree; wherein the expression of the position weighted sum formula is:

[0028] ;

[0029] In the formula, λ is the identical number weight, μ is the similar number weight, is the maximum value of the total number of words in the plurality of personnel feature sets K and the keyword set F.

[0030] In a possible implementation of the first aspect, the time period of the advertisement playing time is divided according to the candidate delivery position, including:

[0031] The historical demand of the candidate delivery position is obtained; wherein the historical demand includes product sales, market heat and user attention;

[0032] The time period of playing the advertisement in the candidate delivery position is set;

[0033] The time period is divided to obtain a plurality of delivery time periods;

[0034] The demand in the plurality of delivery time periods is calculated by a time matching formula to obtain a time matching degree;

[0035] The delivery time periods ranked in the top N of the time matching degree are selected to obtain candidate delivery time periods.

[0036] In a possible implementation of the first aspect, the expression of the time matching formula is:

[0037] ;

[0038] In the formula, t is the delivery time period of the advertisement, is a product sales index function in the delivery time period t, is a market heat index function in the delivery time period t, is a user attention index function in a delivery time period t, and a, β and γ are weight coefficients of product sales, market heat and user attention respectively, and a+β+γ=1.

[0039] In combination with the first aspect, in a possible implementation manner, the verifying and screening of the candidate delivery scheme comprises:

[0040] acquiring video streams and code scanning data of the candidate delivery scheme; wherein the video streams are video streams acquired by delivery of the candidate delivery scheme in a preset time period, and the code scanning data is a number of times of scanning a two-dimensional code built in a promotional video in a preset time period;

[0041] identifying personnel watching the video streams through a target detection algorithm to obtain advertisement watching personnel;

[0042] counting a total number of the advertisement watching personnel to obtain an advertisement watching number;

[0043] tracking the advertisement watching personnel through a target tracking algorithm and counting a time length to obtain an advertisement gaze time length;

[0044] dividing the number of times of scanning the code by the advertisement watching number to obtain an advertisement conversion rate;

[0045] performing weighted summation on the advertisement watching number, the advertisement gaze time length and the advertisement conversion rate according to a preset effect weight coefficient to obtain a delivery effect score;

[0046] selecting a candidate delivery scheme corresponding to a maximum value of the delivery effect score and marking the candidate delivery scheme as an optimal delivery scheme.

[0047] In combination with the first aspect, in a possible implementation manner, the identifying personnel watching the video streams through the target detection algorithm comprises:

[0048] decoding and image preprocessing the video streams to obtain a plurality of standardized image frames;

[0049] performing personnel identification detection on the plurality of standardized image frames through a YOLO algorithm to obtain a plurality of target personnel detection boxes;

[0050] performing a de-redundancy operation on the plurality of target personnel detection boxes through a non-maximum suppression algorithm to obtain a plurality of effective detection boxes;

[0051] associating and matching the plurality of effective detection boxes through a DeepSORT algorithm, and assigning a unique ID to each target personnel to obtain a tracking trajectory with the ID; wherein the tracking trajectory comprises a plurality of effective detection boxes of the same ID;

[0052] The head region and the face key points are calculated by the head rotation angle calculation of the head region and the face key points through the HopeNet, to obtain the head posture Euler angle.

[0053] The head posture Euler angle is used to judge whether the target personnel is watching the advertisement, to obtain the advertisement watching personnel.

[0054] The head posture Euler angle is used to judge whether the target personnel is watching the advertisement, to obtain the advertisement watching personnel.

[0055] In combination with the above first aspect, in a possible implementation manner, the head posture Euler angle is used to judge whether the target personnel is watching the advertisement, including:

[0056] The head posture Euler angle is smoothed and denoised by a data smoothing algorithm, to obtain a smoothed Euler angle.

[0057] The smoothed Euler angle is compared with a preset angle threshold range.

[0058] When the smoothed Euler angle is within the angle threshold range, an effective detection frame corresponding to the smoothed Euler angle is marked as a watching advertisement personnel detection frame.

[0059] When the smoothed Euler angle is outside the angle threshold range, an effective detection frame corresponding to the smoothed Euler angle is marked as a non-watching advertisement personnel detection frame.

[0060] The target personnel corresponding to the watching advertisement personnel detection frame is marked as an advertisement watching personnel.

[0061] The second aspect provides an electronic device, including a communication unit and a processing unit; the communication unit is used to obtain to-be-launched advertisement information, a set of launching positions and advertisement playing time; the processing unit is used to identify and classify the to-be-launched advertisement information, to obtain an advertisement type; the set of launching positions is screened according to the advertisement type, to obtain a candidate launching position; the advertisement playing time is divided into time periods according to the candidate launching position, to obtain a candidate launching time period; a candidate launching scheme is set based on the candidate launching position and the candidate launching time period; and the candidate launching scheme is verified and screened, to obtain an optimal launching scheme.

[0062] The third aspect provides an electronic device, including a processor and a storage medium; the storage medium includes instructions, and the processor is used to run the instructions to implement the method described in the first aspect and any possible implementation manner of the first aspect. The electronic device can be an electronic device, or a chip in the electronic device.

[0063] In a fourth aspect, the present application provides a multi-objective optimization-based intelligent advertisement delivery effect dynamic improvement system, comprising: a data acquisition device, a cloud computing device, and a network device; wherein the data acquisition device is configured to obtain to-be-delivered advertisement information, a set of delivery locations, and advertisement play time; the cloud computing device is configured to identify and classify the to-be-delivered advertisement information to obtain an advertisement type; filter the set of delivery locations according to the advertisement type to obtain candidate delivery locations; divide the advertisement play time into time periods according to the candidate delivery locations to obtain candidate delivery time periods; set a candidate delivery scheme based on the candidate delivery locations and the candidate delivery time periods; verify and filter the candidate delivery scheme to obtain an optimal delivery scheme; and the network device is configured to transmit data obtained by the data acquisition device to the cloud computing device.

[0064] In a fifth aspect, the present application provides a computer-readable storage medium, which stores instructions, and when the instructions are executed on an electronic device, the electronic device performs the method described in the first aspect and any possible implementation manner of the first aspect.

[0065] In a sixth aspect, the present application provides a computer program product comprising instructions, and when the computer program product is executed on an electronic device, the electronic device performs the method described in the first aspect and any possible implementation manner of the first aspect.

[0066] The present application provides a multi-objective optimization-based intelligent advertisement delivery effect dynamic improvement method and system, which realizes intelligent dynamic optimization of advertisement delivery through a multi-dimensional matching mechanism that fuses spatiotemporal features and advertisement content. A precise delivery scenario model is constructed based on personnel feature recognition and time period division of delivery locations, and combined with multi-label classification technology of advertisement content, the system can automatically calculate the delivery effect score of advertisements and the current scenario, and realize personalized delivery through real-time verification. This method effectively solves the problem of mismatch between advertisement content and audience demand caused by traditional fixed rotation mode, significantly improves the precision and resource utilization efficiency of advertisement delivery, and avoids user visual fatigue, thereby providing a more intelligent and efficient solution for advertisement delivery scenarios.

[0067] It should be understood that the description of technical features, technical solutions, advantages or similar language in this application does not imply that all features and advantages can be achieved in any single embodiment. On the contrary, it can be understood that the description of a feature or advantage means that the specific technical feature, technical solution or advantage is included in at least one embodiment. Therefore, the description of technical features, technical solutions or advantages in this specification does not necessarily refer to the same embodiment. Further, the technical features, technical solutions and advantages described in this embodiment can be combined in any appropriate manner. Those skilled in the art will understand that the embodiments can be implemented without one or more specific technical features, technical solutions or advantages of the specific embodiments. In other embodiments, additional technical features and advantages can be identified in specific embodiments that do not embody all embodiments. BRIEF DESCRIPTION OF DRAWINGS

[0068] Figure 1 A system architecture diagram of an advertisement intelligent putting effect dynamic improvement system based on multi-objective optimization is provided for embodiments of the application.

[0069] Figure 2 A flowchart of an advertisement intelligent putting effect dynamic improvement method based on multi-objective optimization is provided for embodiments of the application.

[0070] Figure 3 A flowchart of another advertisement intelligent putting effect dynamic improvement method based on multi-objective optimization is provided for embodiments of the application.

[0071] Figure 4 A flowchart of another advertisement intelligent putting effect dynamic improvement method based on multi-objective optimization is provided for embodiments of the application.

[0072] Figure 5 A flowchart of another advertisement intelligent putting effect dynamic improvement method based on multi-objective optimization is provided for embodiments of the application.

[0073] Figure 6 A flowchart of another advertisement intelligent putting effect dynamic improvement method based on multi-objective optimization is provided for embodiments of the application.

[0074] Figure 7 A structural schematic diagram of an electronic device is provided for embodiments of the application.

[0075] Figure 8 A hardware structure schematic diagram of an electronic device is provided for embodiments of the application. DETAILED DESCRIPTION

[0076] In the description of the present application, unless otherwise specified, " / " means "or", for example, A / B can mean A or B. "And / or" herein is only a description of the relationship between the associated objects, which means that there can be three relationships, for example, A and / or B, which can mean: A exists alone, A and B exist together, and B exists alone. In addition, "at least one" means one or more, and "multiple" means two or more. "First", "second", and the like do not limit the quantity and execution order, and "first", "second", and the like do not necessarily mean different.

[0077] It should be noted that in this application, "exemplary" or "for example" means to serve as an example, illustration or description. Any embodiment or design scheme described as "exemplary" or "for example" in this application should not be interpreted as more preferred or more advantageous than other embodiments or design schemes. Rather, the use of "exemplary" or "for example" is intended to present the relevant concept in a specific manner.

[0078] The method for dynamically improving the effect of intelligent advertisement placement based on multi-objective optimization provided by the embodiments of the present application can be applied to the system for dynamically improving the effect of intelligent advertisement placement based on multi-objective optimization 100 as shown in the figure, and the communication system includes a data acquisition device 10, a cloud computing device 20 and a network device 30 as shown in the figure. Figure 1 Figure 1 The communication system includes a data acquisition device 10, a cloud computing device 20 and a network device 30.

[0079] The data acquisition device 10 is configured to acquire the to-be-placed advertisement information, the set of placement positions and the advertisement playing time.

[0080] The cloud computing device 20 is configured to identify and classify the to-be-placed advertisement information to obtain the advertisement type, filter the set of placement positions according to the advertisement type to obtain the candidate placement positions, divide the advertisement playing time into time periods according to the candidate placement positions to obtain the candidate placement time periods, set the candidate placement scheme based on the candidate placement positions and the candidate placement time periods, and verify and filter the candidate placement scheme to obtain the best placement scheme.

[0081] The network device 30 is configured to transmit the data acquired by the data acquisition device 10 to the cloud computing device 20.

[0082] To solve the technical problems of the existing advertisement placement mechanism, such as rigidity, insufficient precision and low resource utilization, the embodiments of the present application provide a method for dynamically improving the effect of intelligent advertisement placement based on multi-objective optimization, which includes the following steps:

[0083] ​obtaining to-be-launched advertising information, a set of launching positions, and advertising playing time; wherein the to-be-launched advertising information includes a plurality of product names, advertising scripts, and promotional videos, the set of launching positions includes locations of a plurality of advertising playing devices, and the advertising playing time is all time arranged for playing advertising in the locations;

[0084] identifying and classifying the to-be-launched advertising information to obtain an advertising type;

[0085] screening the set of launching positions according to the advertising type to obtain candidate launching positions;

[0086] dividing the advertising playing time into time periods according to the candidate launching positions to obtain candidate launching time periods;

[0087] setting a candidate launching scheme based on the candidate launching positions and the candidate launching time periods;

[0088] verifying and screening the candidate launching scheme to obtain an optimal launching scheme.

[0089] Therefore, the technical problems of rigid advertising launching mechanism, insufficient precision, and low resource utilization rate in the prior art are solved.

[0090] As shown in Figure 2 the method for dynamically improving advertising intelligent launching effect based on multi-target optimization provided by the embodiments of the present application includes the following steps.

[0091] S201, obtaining to-be-launched advertising information, a set of launching positions, and advertising playing time.

[0092] The to-be-launched advertising information includes a plurality of product names, advertising scripts, and promotional videos, the set of launching positions includes locations of a plurality of advertising playing devices, and the advertising playing time is all time arranged for playing advertising in the locations.

[0093] For example, a food company launches a functional biscuit named “Sleeping Vitality”, and the to-be-launched advertising information provided by the company includes the product name “Sleeping Vitality γ-aminobutyric acid biscuit”, the advertising script “one piece of good sleep, double vitality: improve sleep and enhance immunity while enjoying delicious food”, and a promotional video combining leisure snack scenes and healthy life concepts; the set of launching positions includes locations of 600 intelligent elevator screens in the city science and technology park, high-end communities, and around comprehensive hospitals; and the advertising playing time is the bookable playing time periods of these locations.

[0094] S202, identifying and classifying the to-be-launched advertising information to obtain an advertising type.

[0095] In some implementations, the identifying and classifying the to-be-launched advertising information, as shown in Figure 3 includes the following steps.

[0096] The advertising text is obtained by cleaning the advertising copy using a text cleaning tool;

[0097] By using natural language processing, effective keywords are extracted from the advertising text to obtain several product keywords;

[0098] Build a type-keyword thesaurus, match product keywords with the type-keyword thesaurus to obtain several product categories;

[0099] The product names are categorized using a pre-defined product type classification table to obtain the product types;

[0100] Determine whether several product categories and product types are the same; if yes, mark the product type as the advertising type; otherwise, mark the product category with the most product keywords among the several product categories as the advertising type.

[0101] It should be noted that the product type classification table is a classification table formulated with reference to the "Commodity Classification Table", and the type-keyword thesaurus is a matching thesaurus built by those skilled in the art based on effect words and product types.

[0102] For example, the type-keyword thesaurus is shown in Table 1. Effective keywords are extracted from the advertising copy "One good night's sleep, double the energy: Improve sleep and boost immunity while enjoying delicious flavors," resulting in the product keywords "good sleep, improve sleep, boost immunity." These are then matched with the type-keyword thesaurus, resulting in the product category being "health supplements." The product name "Sleep Vitality Gamma-Aminobutyric Acid Cookies" is categorized as "food." Further refinement is then performed; since the product category and product type differ, the category with the most corresponding product keywords, "health supplements," is marked as the advertising type.

[0103] Table 1

[0104]

[0105] S203. Filter the set of placement locations according to the type of advertisement to obtain candidate placement locations.

[0106] In some implementations, the set of placement locations is filtered according to the ad type, such as... Figure 4 As shown, it includes:

[0107] Several sets of personnel features K are obtained from the set of placement locations based on the location-person feature matching library; wherein, the location-person feature matching library is constructed based on the location of the advertising playback device and the personnel features corresponding to the location;

[0108] Extract the set of keywords F corresponding to several advertising messages to be delivered;

[0109] Calculate the matching degree of the words in the personnel feature set K and the keyword set F, and mark it as the location matching degree;

[0110] Select the locations with the top N location matching degrees to obtain the candidate delivery locations; wherein, the N is a positive integer.

[0111] For example, the location-personnel feature matching library is shown in Table 2. First, the personnel feature set K corresponding to all selectable screens is retrieved based on the location-personnel feature matching library (for example, “nursing home” corresponds to “poor sleep quality, low immunity, poor memory, and joint pain”); then, the keyword set F of this advertisement type is extracted {“improve sleep”, “enhance immunity”, “low immunity”, and “poor sleep quality”}, and the semantic correlation of each population feature and these keywords is calculated; finally, the locations with the top 3 matching degrees (such as the advertisement screens of nursing homes, high-end communities, and science and technology parks) are selected as the candidate delivery locations.

[0112] Table 2

[0113]

[0114] S204, time period division is performed on the advertisement playing time according to the candidate delivery location to obtain the candidate delivery time period.

[0115] In some implementations, the time period division on the advertisement playing time according to the candidate delivery location includes: Figure 5 As shown in the figure, it includes:

[0116] Obtain the historical demand quantity of the candidate delivery location; wherein, the historical demand quantity includes product sales, market heat, and user attention;

[0117] Set the time period for the candidate delivery location to play the advertisement;

[0118] Divide the time period to obtain a plurality of delivery time periods;

[0119] Calculate the demand quantity in the plurality of delivery time periods through a time matching formula to obtain a time matching degree;

[0120] Select the delivery time period with the top N time matching degrees to obtain the candidate delivery time period.

[0121] It should be noted that the product sales of the candidate delivery location can be analyzed and obtained through the e-commerce platform background and the like, the market heat can be analyzed and obtained through the search engine index and the like, and the user attention can be analyzed and obtained through the advertisement platform background and the like.

[0122] Exemplarily, first, the historical data of the candidate delivery location "high-end community" is acquired, wherein the historical demand of the high-end community shows that the sales of health products in the area are high in the evening, the search engine market heat rises on weekends, and the user attention of the advertising platform is most concentrated during the commuting period (8-9 am in the morning and 7-8 pm in the evening); based on this, the system sets the time period as the next week and divides it into multiple delivery time periods in units of hours; then, the comprehensive score of each time period is calculated through the time matching formula; finally, the top 3 matching periods are selected: working day 7-8 pm, working day 8-9 pm, and weekend 3-4 pm as the final candidate delivery time periods.

[0123] S205, setting a candidate delivery scheme based on the candidate delivery location and the candidate delivery time period.

[0124] Exemplarily, based on the candidate delivery location "high-end community" and the corresponding candidate delivery time period "working day 7-8 pm, working day 8-9 pm, and weekend 3-4 pm", three candidate delivery schemes are set: "high-end community, working day 7-8 pm", "high-end community, working day 8-9 pm", and "high-end community, weekend 3-4 pm".

[0125] S206, verifying and screening the candidate delivery scheme to obtain the best delivery scheme.

[0126] In some implementations, the verification and screening of the candidate delivery scheme includes Figure 6 as shown, including:

[0127] Acquiring video streams and scan code data of the candidate delivery scheme; wherein the video stream is a video stream acquired by delivering the candidate delivery scheme in a preset time period, and the scan code data is the number of scans of the two-dimensional code built in the promotional video in the preset time period;

[0128] Identifying the personnel watching the video stream through a target detection algorithm to obtain the advertisement watching personnel;

[0129] Counting the total number of the advertisement watching personnel to obtain the advertisement watching number;

[0130] Tracking the advertisement watching personnel through a target tracking algorithm and counting the duration to obtain the advertisement gaze duration;

[0131] Dividing the number of scans by the advertisement watching number to obtain the advertisement conversion rate;

[0132] According to the preset effect weight coefficient, the advertisement watching number, the advertisement gaze duration, and the advertisement conversion rate are weighted and summed to obtain the delivery effect score;

[0133] Select the candidate delivery scheme corresponding to the maximum delivery effect score and mark it as the best delivery scheme.

[0134] For example, after a three-day actual delivery of the two candidate delivery schemes (scheme one: high-end community, 7-8 pm on weekdays; scheme two: high-end community, 8-9 pm on weekdays), the system obtains the video stream in front of the advertising screen and records the scanning data of the two-dimensional code in the promotional video (120 scans for scheme one and 85 scans for scheme two); then, the system identifies 1500 people as the advertising viewers for scheme one and 1300 people for scheme two through the target detection algorithm from the video stream, and further calculates the average advertising attention duration as 2.1 seconds for scheme one and 1.7 seconds for scheme two through the target tracking algorithm; next, the system calculates the advertising conversion rate as 8% (120 / 1500) for scheme one and 6.5% (85 / 1300) for scheme two; finally, the system performs weighted summation according to the preset effect weight (viewer weight 0.3, attention duration weight 0.3, and conversion rate weight 0.4) to obtain the delivery effect score of scheme one as 0.3 x (1500 / 1500) + 0.3 x (2.1 / 2.1) + 0.4 x (0.08 / 0.08) = 1.0 and the score of scheme two as 0.3 x (1300 / 1500) + 0.3 x (1.7 / 2.1) + 0.4 x (0.065 / 0.08) = 0.885, so scheme one with the higher score is selected as the best delivery scheme.

[0135] Based on the above technical scheme, the method for dynamically improving the advertising intelligent delivery effect based on multi-target optimization provided in the application realizes intelligent dynamic optimization of advertising delivery through the multi-dimensional matching mechanism that fuses the spatiotemporal features and the advertising content. The precise delivery scene model is constructed based on the personnel feature recognition and the time period division of the delivery location, and the multi-label classification technology of the advertising content is combined, so that the system can automatically calculate the delivery effect score of the advertising and the current scene, and realize personalized delivery through real-time verification. This method effectively solves the problem of mismatch between the advertising content and the audience demand caused by the traditional fixed carousel mode, significantly improves the precision and resource utilization efficiency of advertising delivery, and avoids user visual fatigue, thereby providing a more intelligent and efficient solution for the advertising delivery scene.

[0136] In a possible implementation manner of the embodiment of the application, S203 can be implemented through the following S301, which is specifically described as follows.

[0137] S301, calculate the matching degree of the words in the plurality of personnel feature sets K and the keyword set F, and mark it as the location matching degree.

[0138] In some implementation manners, the calculation of the matching degree of the words in the plurality of personnel feature sets K and the keyword set F includes:

[0139] counting the number of identical words in the plurality of personnel characteristic sets K and the keyword set F, and marking as identical number D;

[0140] calculating the semantic similarity of a plurality of words in the plurality of personnel characteristic sets K and the keyword set F by Word2Vec, counting the number of words with semantic similarity greater than a preset threshold, and marking as similar number R;

[0141] performing weighted summation on the identical number D and the similar number R by a position weighted summation formula to obtain the position matching degree; wherein the expression of the position weighted summation formula is:

[0142]

[0143] In the formula, λ is the identical number weight, μ is the similar number weight, is the maximum value of the total number of words in the plurality of personnel characteristic sets K and the keyword set F.

[0144] For example, taking the personnel characteristic set K = {poor sleep quality, low immunity, poor memory, joint pain} and the keyword set F = {enhance immunity, low immunity, improve sleep, poor sleep quality} as an example, first count the number of identical words D = 2; then calculate the semantic similarity by the Word2Vec model, find that the semantic similarity of "poor sleep quality" and "improve sleep" is 0.82 (greater than the preset threshold 0.8), and the semantic similarity of "low immunity" and "enhance immunity" is 0.85 (greater than 0.8), so the similar number R = 2; finally, calculate the position matching degree by the position weighted summation formula (set λ = 0.6, μ = 0.4, |K| = 4, |F| = 4, Max(K, F) = 4): P = (0.6 × 2 + 0.4 × 2) / 4 = 2 / 4 = 0.5.

[0145] Based on the above technical solution, by fusing accurate matching and semantic similarity analysis, the fit degree of the advertisement placement position and the target population characteristics is significantly improved. The scheme uses the Word2Vec word vector technology to break through the semantic limitations of traditional keyword matching, which can effectively identify concept associations such as "poor sleep quality" and "improve sleep" that are semantically related but expressed differently; by setting a similarity threshold and a weighted summation formula, the accuracy of completely matching words is preserved, and the potential value of semantically similar words is captured; and the adjustable weight coefficients (λ, μ) make the model able to flexibly balance the emphasis of accurate matching and semantic expansion according to different marketing strategies. This multi-level matching mechanism overcomes the mechanical defects of single keyword matching, so that the selection of advertisement placement position not only reaches the population with surface characteristics, but also accurately covers the target audience with potential needs, greatly improving the conversion efficiency of advertisement placement. ​

[0146] In a possible implementation of the embodiment of the present application, the S204 can be implemented by the following S401, which is described in detail as follows.

[0147] S401, calculate the demand quantity in the several delivery time periods by a time matching formula to obtain a time matching degree.

[0148] In some implementations, the expression of the time matching formula is:

[0149] ;

[0150] wherein t is the delivery time period of the advertisement, is a product sales index function in the delivery time period t, is a market heat index function in the delivery time period t, is a user attention index function in the delivery time period t, and α, β and γ are weight coefficients of the product sales, the market heat and the user attention respectively, and α+β+γ=1.

[0151] It should be noted that the power function is used to simulate the law of diminishing marginal utility and fit the actual situation; the multiplication form is because each index is not independent but interrelated and mutually promoted; and the size of the weight coefficient can intuitively quantify the relative weight and importance of each index.

[0152] For example, the time matching degree of the delivery time period t of 8-9 pm needs to be calculated, and the system obtains the historical data of the time period as follows: the product sales index =80 (after normalization), the market heat index =75 (after normalization), and the user attention index =90 (after normalization); according to the business target, the weight coefficients are set as α=0.5 (emphasizing actual conversion), β=0.2 and γ=0.3; then the time matching formula is used for calculation: =(80^0.5)×(75^0.2)×(90^0.3)=(8.94)×(2.37)×(3.48)≈73.7; the score 73.7 is the time matching degree of the time period t, which will participate in ranking together with the scores of other time periods to screen the optimal candidate delivery time period.

[0153] Based on the above technical scheme, by introducing multi-dimensional index fusion calculation and nonlinear modeling, the accuracy and strategy of the advertisement time period selection are significantly improved. The model uses a power function structure to accurately depict the marginal utility diminishing law of each index, overcoming the defect of overestimating the benefit of resource input of the linear model; through the multiplication operator, the synergistic effect of product sales, market heat and user attention is organically integrated, which truly reflects the action mechanism of multiple factors promoting each other in the actual marketing scene; and the configurable weight system gives the operation personnel flexible strategy adjustment ability, which can dynamically adjust the importance of the index according to the product characteristics, marketing goals and other factors. This scientific modeling method realizes the transformation from experience-based scheduling to data-driven decision-making, so that the advertisement delivery can not only fit the fluctuation law of market demand, but also accurately connect the active time period of the target group, and finally maximize the delivery benefit.

[0154] In a possible implementation manner of the embodiment of the application, S206 can be implemented through S701, which is specifically described as follows.

[0155] S701, identifying the person watching the video stream through a target detection algorithm to obtain an advertisement watching person.

[0156] In some implementation manners, the identifying the person watching the video stream through the target detection algorithm comprises:

[0157] decoding and image preprocessing the video stream to obtain a plurality of standardized image frames;

[0158] performing personnel identification detection on the plurality of standardized image frames through a YOLO algorithm to obtain a plurality of target personnel detection boxes;

[0159] performing a de-redundancy operation on the plurality of target personnel detection boxes through a non-maximum suppression algorithm to obtain a plurality of effective detection boxes;

[0160] associating and matching the plurality of effective detection boxes through a DeepSORT algorithm to assign a unique ID to each target person to obtain a tracking trajectory with an ID; wherein the tracking trajectory comprises a plurality of effective detection boxes with the same ID;

[0161] detecting key points of the tracking trajectory through an HRNet to obtain head region and face key points;

[0162] calculating a head rotation angle of the head region and the face key points through a HopeNet to obtain a head pose Euler angle;

[0163] judging whether the target person is watching the advertisement through the head pose Euler angle to obtain the advertisement watching person.

[0164] Exemplarily, in a three-day test period of the three-day scheme, the system decodes and performs gray normalization preprocessing on the video stream (a total of 10800 frames, 30fps) collected by the "XX Science and Technology Park Elevator Screen", and detects a total of 15642 target personnel detection boxes through the YOLOv5s algorithm, and obtains 2381 effective detection boxes after redundancy removal through non-maximum suppression (confidence threshold 0.5, IoU threshold 0.45); then the DeepSORT algorithm (using Re-ID feature cosine distance matching) successfully assigns a unique ID to 317 independent pedestrians, generating 409 tracking trajectories; the HRNet-32 key point detection model locates 291 effective head regions on these trajectories, and then calculates the corresponding head Euler angles by HopeNet; the system screens out 142 pedestrians (accounting for 44.8% of the total number of detected persons) in 6742 frames of images as advertisement viewers according to the judgment rule (|yaw angle|<15°, |pitch angle|<20°).

[0165] Based on the above technical scheme, by constructing an end-to-end video analysis pipeline, fine perception and quantitative evaluation of advertisement viewing behavior are realized. The system realizes multi-target real-time detection and stable tracking by using YOLO+DeepSORT, effectively solving the problems of occlusion and ID jumping in complex scenes; combined with the HRNet-HopeNet double model architecture, through head key point detection and Euler angle calculation, the dependence on front-facing faces in traditional face detection is broken, and the robustness of pose estimation is significantly improved; finally, through time series smoothing processing and dynamic threshold judgment, active viewing and unintentional glancing are accurately distinguished, and the advertisement effect evaluation is improved from "exposure" to "effective attention" dimension, providing high-confidence behavior data support for optimizing the placement strategy.

[0166] In a possible implementation manner of the embodiment of the application, S701 can be implemented through S801, which is specifically described as follows.

[0167] S801, judging whether the target person is watching the advertisement by the head pose Euler angle to obtain an advertisement viewer.

[0168] In some implementation manners, the judging whether the target person is watching the advertisement by the head pose Euler angle comprises:

[0169] Smoothing and denoising the head pose Euler angle by a data smoothing algorithm to obtain a smoothed Euler angle;

[0170] Comparing the smoothed Euler angle with a preset angle threshold range;

[0171] When the smoothed Euler angle is in the angle threshold range, marking the effective detection box corresponding to the smoothed Euler angle as an advertisement viewer detection box;

[0172] When the smooth Euler angle is out of the angle threshold range, the effective detection frame corresponding to the smooth Euler angle is marked as an unwatched advertisement personnel detection frame;

[0173] The target personnel corresponding to the watched advertisement personnel detection frame is marked as an advertisement watching personnel.

[0174] It should be noted that the head pose Euler angle is smoothed to eliminate angle jitter caused by detection errors and slight head shaking, making the judgment more stable.

[0175] For example, the original head pose Euler angle sequence calculated by HopeNet is smoothed and denoised by the Savitzky-Golay convolution smoothing algorithm (window length of 5, polynomial order of 2) to obtain smooth Euler angles (instantaneous jitter noise is effectively suppressed); then, the smoothed angle is compared with the preset angle threshold range (yaw angle Yaw ∈ [-15°, 15°], pitch angle Pitch ∈ [-10°, 20°]); when the smooth Euler angle of a certain person in a certain frame meets the yaw angle and pitch angle threshold conditions at the same time, the effective detection frame corresponding to it is marked as a watched advertisement personnel detection frame, otherwise it is marked as an unwatched advertisement personnel detection frame; finally, the system marks the target personnel (identified by its unique ID) corresponding to all detection frames marked as "watched" in the entire time period as an advertisement watching personnel.

[0176] Based on the above technical solution, the combination of computer vision and behavior analysis technology realizes accurate identification and effective quantification of advertisement watching behavior. The system uses a stable head pose detection model combined with data smoothing to accurately capture the gaze direction and fixation state of personnel, effectively distinguishing between active watching and unintentional glancing. By setting reasonable angle thresholds, the system can reliably identify the target audience who truly pay attention to the advertisement. This technology breaks through the limitations of traditional advertisement exposure statistics, improves the effect evaluation from the number of exposures to the level of real attention, provides more scientific and detailed data support for the optimization of advertisement delivery effect, and realizes the upgrade of advertisement effect from "being seen" to "being concerned".

[0177] The above describes the scheme of the embodiments of the present application mainly from the perspective of device implementation. It can be understood that each device, for example, an electronic device, comprises at least one of a hardware structure and a software module for performing each function. Those skilled in the art should easily realize that, in combination with the units and algorithm steps of each example described in the embodiments disclosed herein, the present application can be realized in the form of hardware or a combination of hardware and computer software. Whether a certain function is realized in the form of hardware or computer software driving hardware depends on the specific application and design constraints of the technical scheme. Those skilled in the art can use different methods to realize the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0178] The embodiments of the present application can divide the functional units of the electronic device according to the above method examples. For example, each functional unit can be divided according to each function, or two or more functions can be integrated in one processing unit. The integrated unit can be realized in the form of hardware or software functional unit. It should be noted that the division of units in the embodiments of the present application is illustrative and is only a logical functional division. Actual implementation can have another division manner.

[0179] In the case of using an integrated unit, Figure 7 A possible structure diagram of the electronic device (denoted as electronic device 50) involved in the above embodiments is shown, which comprises a processing unit 501 and a communication unit 502, and can further comprise a storage unit 503. Figure 7 The shown structure diagram can be used to illustrate the structure of the electronic device involved in the above embodiments.

[0180] When Figure 7 When the shown structure diagram is used to illustrate the structure of the electronic device involved in the above embodiments, the processing unit 501 is used to control and manage the actions of the electronic device, the communication unit 502 is used for communication between the electronic device and other devices, and the storage unit 503 is used to store the program code and data of the electronic device.

[0181] For example, the communication unit 502 is used to obtain the to-be-launched advertising information, the set of launch positions, and the advertising play time;

[0182] The processing unit 501 is used to identify and classify the to-be-launched advertising information to obtain the advertising type, filter the set of launch positions according to the advertising type to obtain the candidate launch positions, divide the advertising play time into time periods according to the candidate launch positions to obtain the candidate launch time periods, set the candidate launch scheme based on the candidate launch positions and the candidate launch time periods, and verify and filter the candidate launch scheme to obtain the best launch scheme.

[0183] The processing unit 501 can be a processor or a controller, and the communication unit 502 can be a communication interface, a transceiver, a transceiver, a transceiver circuit, a transceiver device, etc. The communication interface is collectively referred to, and can include one or more interfaces. The storage unit 503 can be a memory. When the electronic device 50 is a chip, the processing unit 501 can be a processor or a controller, and the communication unit 502 can be an input interface and / or an output interface, a pin or a circuit, etc. The storage unit 503 can be a storage unit (for example, a register, a cache, etc.) within the chip, or a storage unit (for example, a read-only memory (ROM), a random access memory (RAM), etc.) located outside the chip.

[0184] The communication unit can also be referred to as a transceiver unit. The antenna and control circuit with transceiver function in the electronic device 50 can be regarded as the communication unit 502 of the electronic device 50, and the processor with processing function can be regarded as the processing unit 501 of the electronic device 50. Optionally, the device for realizing the receiving function in the communication unit 502 can be regarded as a communication unit, and the communication unit is used to execute the receiving steps in the embodiments of the application. The communication unit can be a receiver, a receiver, a receiving circuit, etc. The device for realizing the sending function in the communication unit 502 can be regarded as a sending unit, and the sending unit is used to execute the sending steps in the embodiments of the application. The sending unit can be a transmitter, a sender, a sending circuit, etc.

[0185] Figure 7 The integrated units in the above embodiments can be stored in a computer readable storage medium if they are realized in the form of software function modules and sold or used as independent products. Based on such understanding, the technical solutions of the embodiments of the application or the whole or part of the technical solutions that make essential contributions to the prior art can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes a plurality of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) or a processor to execute all or part of the steps of the method described in the embodiments of the application. The storage medium storing the computer software product includes a U disk, a mobile hard disk, a read-only memory, a random access memory, a magnetic disk or an optical disk, and various media that can store program codes.

[0186] Figure 8 The units in the above embodiments can also be referred to as modules, for example, the processing unit can be referred to as a processing module.

[0187] The embodiments of the application also provide a hardware structure diagram of an electronic device (denoted as electronic device 60), which is shown in Figure 8The electronic device 60 comprises a processor 601, and optionally further comprises a memory 602 connected with the processor 601.

[0188] In the first possible implementation, referring to Figure 8 The electronic device 60 further comprises a transceiver 603. The processor 601, the memory 602 and the transceiver 603 are connected through a bus. The transceiver 603 is used for communicating with other devices or communication networks. Optionally, the transceiver 603 can comprise a transmitter and a receiver. The device for realizing the receiving function in the transceiver 603 can be regarded as a receiver, and the receiver is used for executing the receiving steps in the embodiments of the present application. The device for realizing the sending function in the transceiver 603 can be regarded as a transmitter, and the transmitter is used for executing the sending steps in the embodiments of the present application.

[0189] Based on the first possible implementation, Figure 8 The structure diagram shown can be used for illustrating the structure of the electronic device involved in the above embodiments.

[0190] Among them, ​ The system chip in the electronic device can also be illustrated. In this case, the actions performed by the above electronic device can be realized by the system chip, and the specific actions performed can be referred to in the above, and will not be described here.

[0191] In the implementation process, each step in the method provided by the embodiment can be completed by the integrated logic circuit of the hardware in the processor or the instructions in the software form. The steps of the method disclosed in the embodiment of the present application can be directly embodied as the execution completed by the hardware processor, or executed by the combination of the hardware and the software module in the processor.

[0192] The processor in the present application can include, but is not limited to, at least one of the following: a central processing unit (CPU), a microprocessor, a digital signal processor (DSP), a microcontroller unit (MCU), or an artificial intelligence processor, and various computing devices running software, each of which can include one or more cores for executing software instructions to perform operations or processing. The processor can be a separate semiconductor chip, or can be integrated with other circuits as a semiconductor chip, for example, it can be integrated with other circuits (such as coding and decoding circuits, hardware acceleration circuits, or various bus and interface circuits) to form a SoC (system on chip), or it can be integrated as a built-in processor in an ASIC. The ASIC integrated with the processor can be packaged separately or packaged together with other circuits. In addition to including cores for executing software instructions to perform operations or processing, the processor can further include necessary hardware accelerators, such as field programmable gate arrays (FPGAs), PLDs (programmable logic devices), or logic circuits that implement special logic operations.

[0193] The memory in the embodiments of the present application can include at least one of the following types: read-only memory (ROM) or other types of static storage devices that can store static information and instructions, random access memory (RAM) or other types of dynamic storage devices that can store information and instructions, and electrically erasable programmable read-only memory (EEPROM). In some scenarios, the memory can also be a compact disc read-only memory (CD-ROM) or other optical disc storage, optical disc storage (including compact discs, laser discs, optical discs, digital versatile discs, Blu-ray discs, etc.), magnetic disk storage medium or other magnetic storage device, or any other medium capable of carrying or storing desired program code in the form of instructions or data structures and capable of being accessed by a computer, but not limited thereto.

[0194] The embodiments of the present application also provide a computer readable storage medium including instructions, which, when executed on a computer, cause the computer to perform any of the above methods.

[0195] The embodiments of the present application also provide a computer program product including instructions, which, when executed on a computer, cause the computer to perform any of the above methods.

[0196] The embodiment of the present application further provides a chip, comprising a processor and an interface circuit, the interface circuit being coupled with the processor, the processor being used to run computer programs or instructions to realize the method described above, and the interface circuit being used to communicate with other modules outside the chip.

[0197] In the above embodiments, the implementation can be achieved by software, hardware, firmware or any combination thereof, entirely or partially. When implemented by software, the implementation can be achieved in the form of a computer program product, entirely or partially. The computer program product comprises one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the flow or function described in the embodiments of the present application is generated, entirely or partially. The computer can be a general-purpose computer, a special-purpose computer, a computer network or other programmable devices. The computer instructions can be stored in a computer readable storage medium or transferred from one computer readable storage medium to another, for example, the computer instructions can be transferred from one website, computer, server or data center to another website, computer, server or data center through wired (such as coaxial cable, optical fiber, digital subscriber line (digital subscriber line, DSL)) or wireless (such as infrared, wireless, microwave, etc.) mode. The computer readable storage medium can be any available medium that can be accessed by a computer or data storage device including one or more servers, data centers, etc. integrated with the medium. The available medium can be a magnetic medium (such as a floppy disk, a hard disk, a magnetic tape), an optical medium (such as a DVD), or a semiconductor medium (such as a solid state disk (solid state disk, SSD)), etc.

[0198] Although the present application is described herein in conjunction with various embodiments, other variations of the disclosed embodiments can be understood and implemented by those skilled in the art through viewing the drawings, the disclosure, and the appended claims. In the claims, the word "comprising" does not exclude other components or steps, and "a" or "one" does not exclude a plurality. A single processor or other unit can implement several functions listed in the claims. Some measures are described in mutually different dependent claims, but this does not mean that these measures cannot be combined to produce good results.

[0199] Although the present application has been described in connection with certain specific features and embodiments thereof, it is to be understood that it is provided as an example to the best of the applicant's knowledge and that various modifications and combinations of the described features and embodiments are possible and are within the spirit and scope of the application. Accordingly, the specification and drawings are to be regarded in an illustrative rather than a restrictive sense, and all such modifications and variations are considered within the scope of the present application as defined by the following claims and their equivalents. Obviously, many modifications and variations of the present application are possible in light of the above teachings. It is, therefore, to be understood that within the scope of the claims and their equivalents, the present application can be practiced otherwise than as specifically described.

Claims

1. A method for dynamically improving the effect of intelligent delivery of advertisements based on multi-objective optimization, characterized in that, The application comprises the following steps: acquiring to-be-launched advertising information, a set of launching positions, and advertising playing time; wherein the to-be-launched advertising information comprises a plurality of product names, advertising scripts, and promotional videos, the set of launching positions comprises the positions of a plurality of advertising playing devices, and the advertising playing time is all the time when the advertising is arranged to be played in the positions; identifying and classifying the to-be-launched advertising information to obtain an advertising type; screening the set of launching positions according to the advertising type to obtain candidate launching positions; dividing the advertising playing time into time periods according to the candidate launching positions to obtain candidate launching time periods; setting a candidate launching scheme based on the candidate launching positions and the candidate launching time periods; verifying and screening the candidate launching scheme to obtain an optimal launching scheme; the verification and screening of the candidate launching scheme comprises the following steps: acquiring video streams and scanning code data of the candidate launching scheme; wherein the video streams are video streams acquired by launching a plurality of candidate launching schemes within a preset time period, and the scanning code data is the number of times of scanning a two-dimensional code built in the promotional video within a preset time period; identifying the personnel watching the video streams through a target detection algorithm to obtain advertising watching personnel; counting the total number of the advertising watching personnel to obtain an advertising watching number; tracking the advertising watching personnel through a target tracking algorithm and counting the time length to obtain an advertising gazing time length; dividing the number of times of scanning by the advertising watching number to obtain an advertising conversion rate; weighting and summing the advertising watching number, the advertising gazing time length, and the advertising conversion rate according to a preset effect weight coefficient to obtain a launching effect score; selecting the candidate launching scheme corresponding to the maximum value of the launching effect score and marking it as the optimal launching scheme; the identification of the personnel watching the video streams through the target detection algorithm comprises the following steps: decoding and image pre-processing the video streams to obtain a plurality of standardized image frames; identifying and detecting the personnel through a YOLO algorithm on the plurality of standardized image frames to obtain a plurality of target personnel detection boxes; performing a de-redundancy operation on the plurality of target personnel detection boxes through a non-maximum suppression algorithm to obtain a plurality of effective detection boxes; associating and matching the plurality of effective detection boxes through a DeepSORT algorithm, assigning a unique ID to each target personnel, and obtaining a tracking trajectory with the ID; wherein the tracking trajectory comprises a plurality of effective detection boxes with the same ID; detecting key points of the tracking trajectory through an HRNet to obtain head region and facial key points; calculating a head rotation angle of the head region and the facial key points through a HopeNet to obtain a head posture Euler angle; judging whether the target personnel is watching the advertising through the head posture Euler angle to obtain the advertising watching personnel.

2. The method of claim 1, wherein, the identification and classification of the to-be-launched advertising information comprises the following steps: cleaning the advertising scripts through a text cleaning tool to obtain advertising texts; extracting product keywords through natural language processing on the advertising texts to obtain a plurality of product keywords; constructing a type-keyword library, matching the product keywords with the type-keyword library, and obtaining a plurality of product classifications; classifying the product names through a preset product type classification table to obtain product types; determining whether the product categories are the same as the product type; if yes, marking the product type as the advertisement type; if no, marking the product category with the most product keywords in the product categories as the advertisement type.

3. The method of claim 1, wherein, The filtering of the set of delivery locations according to the advertisement type comprises: obtaining a set of personnel features K of the set of delivery locations based on a location-personnel feature matching library, wherein the location-personnel feature matching library is constructed based on the location of the advertisement playing device and the personnel features corresponding to the location; extracting a keyword set F corresponding to the plurality of advertisement information to be delivered; calculating the lexical matching degree of the set of personnel features K and the keyword set F and marking it as the location matching degree; selecting the locations with the top N location matching degrees to obtain candidate delivery locations, wherein N is a positive integer.

4. The method of claim 3, wherein, The calculation of the lexical matching degree of the set of personnel features K and the keyword set F comprises: counting the number of identical words in the set of personnel features K and the keyword set F and marking it as the identical number D; calculating the semantic similarity of a plurality of words in the set of personnel features K and the keyword set F by Word2Vec, counting the number of words with a semantic similarity greater than a preset threshold, and marking it as the similar number R; performing weighted summation on the identical number D and the similar number R by a location weighted summation formula to obtain the location matching degree, wherein the expression of the location weighted summation formula is: ； where λ is a same quantity weight, and μ is a similar quantity weight, is the maximum value of the total number of words in the set of personnel features K and the set of keywords F.

5. The method of claim 1, wherein, The time period division of the advertisement playing time according to the candidate delivery location comprises: obtaining the historical demand of the candidate delivery location, wherein the historical demand includes product sales, market heat, and user attention; setting a time period for playing the advertisement in the candidate delivery location; dividing the time period to obtain a plurality of delivery time periods; calculating the demand in the plurality of delivery time periods by a time matching formula to obtain a time matching degree; selecting the delivery time periods with the top N time matching degrees to obtain candidate delivery time periods.

6. The method of claim 5, wherein, The expression of the time matching formula is: ； wherein t is a time period of the advertisement, is a product sales index function in the time period t, is a market heat index function in the time period t, is a user attention index function in the time period t, and α, β and γ are weight coefficients of the product sales, the market heat and the user attention, respectively, and α+β+γ=1.

7. The method of claim 1, wherein, The judgment of whether the target personnel is watching the advertisement by the head pose Euler angle comprises: smoothing and denoising the head pose Euler angle by a data smoothing algorithm to obtain a smoothed Euler angle; comparing the smoothed Euler angle with a preset angle threshold range; when the smoothed Euler angle is within the angle threshold range, marking the valid bounding box corresponding to the smoothed Euler angle as a watching advertisement personnel bounding box; when the smoothed Euler angle is outside the angle threshold range, marking the valid bounding box corresponding to the smoothed Euler angle as a non-watching advertisement personnel bounding box; marking the target personnel corresponding to the watching advertisement personnel bounding box as an advertisement watching personnel.

8. An electronic device for use in the method of any one of claims 1-7, characterized by comprises: a communication unit and a processing unit; the communication unit is configured to obtain the advertisement information to be delivered, the set of delivery locations, and the advertisement playing time; The processing unit is configured to identify and classify the to-be-launched advertising information to obtain an advertising type, filter a set of launching positions according to the advertising type to obtain candidate launching positions, divide a playing time of the advertising into time periods according to the candidate launching positions to obtain candidate launching time periods, set a candidate launching scheme based on the candidate launching positions and the candidate launching time periods, and verify and filter the candidate launching scheme to obtain an optimal launching scheme.

Images