Advertisement attention detection method and device, equipment, medium and program product

Abstract

The invention discloses an advertisement attention detection method and device, equipment, a medium and a program product in the technical field of computers. According to the method, for the pedestrian video of the area where the advertisement space is located, the pedestrian position information and the pedestrian feature information in each frame of image of the pedestrian video are detected by using the pedestrian detection model, and then the orientation information of each pedestrian in each frame of image is identified by using the orientation identification model; processing the pedestrian position information, the pedestrian feature information and the orientation information by using a target tracking model so as to determine trajectory information of each pedestrian; and finally, based on the timestamp, the track information and the orientation information of each frame of image, obtaining a time point when each pedestrian enters and exits the area where the advertisement space is located and a duration facing the advertisement space, and obtaining the advertisement attention according to the time point and the duration. Therefore, the people flow near the advertisement space can be reflected, the attention degree of people near the advertisement space to the advertisement can be further determined, and advertisement adjustment can be achieved according to the attention degree.

Description

Technical Field

[0001] The present application relates to the field of computer technology, and in particular to a method, device, equipment, medium and program product for detecting advertising attention. Background Art

[0002] Currently, infrared sensors and pressure sensors are often installed near advertising spots. Signals collected by these sensors automatically count the number of pedestrians passing through the area, and use this information to estimate ad exposure. However, this solution only reflects the flow of people near the ad spot, but cannot determine whether people near the spot are actually paying attention to the ad.

[0003] Therefore, how to detect people's attention to advertisements is a problem that those skilled in the art need to solve. Summary of the Invention

[0004] In view of this, the purpose of this application is to provide a method, device, equipment, medium and program product for detecting advertising attention, so as to detect people's attention to advertisements.

[0005] In a first aspect, the present application provides a method for detecting advertising attention, comprising:

[0006] Obtain pedestrian videos in the area where the ad slot is located;

[0007] Detecting pedestrian position information and pedestrian feature information in each frame image of the pedestrian video using a pedestrian detection model;

[0008] Use the orientation recognition model to identify the orientation information of each pedestrian in each frame image;

[0009] Processing the pedestrian position information, the pedestrian feature information, and the orientation information using a target tracking model to determine trajectory information of each pedestrian;

[0010] Based on the timestamp of each frame image, the trajectory information and the orientation information, the time point when each pedestrian enters and exits the area where the advertising space is located and the duration of facing the advertising space are obtained, and the advertising attention is obtained based on the time point and the duration.

[0011] Optionally, detecting pedestrian position information and pedestrian feature information in each frame of the pedestrian video using a pedestrian detection model includes:

[0012] The pedestrian detection model is used to read each frame image from the pedestrian video, and the pedestrian coverage frame and pedestrian appearance features in the read image are detected, the coordinate information of the pedestrian coverage frame is used as the pedestrian position information, and the pedestrian appearance features are used as the pedestrian feature information.

[0013] Optionally, the orientation information of each pedestrian in each frame image is identified using an orientation recognition model, including:

[0014] Segmenting each pedestrian in each frame image from the corresponding image according to the pedestrian position information to obtain a pedestrian image area;

[0015] The orientation recognition model is used to extract orientation features of the pedestrian image area, and orientation information corresponding to the orientation features is obtained by recognition.

[0016] Optionally, the target tracking model includes: a classification module, a matching module and a trajectory prediction module;

[0017] Accordingly, the target tracking model is used to process the pedestrian position information, the pedestrian feature information, and the orientation information to determine the trajectory information of each pedestrian, including:

[0018] Processing the pedestrian position information, the pedestrian feature information, and the orientation information using the classification module to determine motion characteristics of each pedestrian;

[0019] Processing the motion features using the matching module to determine matching relationships between the same pedestrian in adjacent images;

[0020] The matching relationship is processed by the trajectory prediction module to predict the tracking frame position of each pedestrian, and the tracking frame position is used as the trajectory information.

[0021] Optionally, based on the timestamp of each frame image, the trajectory information, and the orientation information, obtaining the time point at which each pedestrian enters and exits the area where the advertising space is located and the duration of time facing the advertising space includes:

[0022] For each frame of the pedestrian video, the marking information, the trajectory information, the direction information and the time points of entering and leaving the area where the advertising space is located of each pedestrian in each frame are recorded in sequence according to the timestamp order, and the duration is obtained by statistics.

[0023] Optionally, obtaining the advertisement attention according to the time point and the duration includes:

[0024] The advertising attention corresponding to each pedestrian is calculated based on the time point and the duration; and / or the comprehensive advertising attention is calculated based on the time point and the duration.

[0025] Optionally, it also includes:

[0026] The marking information, the trajectory information, the direction information and the time points of entering and leaving the area where the advertisement space is located of each pedestrian in each frame of the image are stored.

[0027] In a second aspect, the present application provides an advertising attention detection device, comprising:

[0028] An acquisition module is used to obtain pedestrian videos in the area where the ad slot is located;

[0029] A detection module, configured to detect pedestrian position information and pedestrian feature information in each frame of the pedestrian video using a pedestrian detection model;

[0030] A recognition module is used to identify the orientation information of each pedestrian in each frame image using an orientation recognition model;

[0031] a tracking module, configured to process the pedestrian position information, the pedestrian feature information, and the orientation information using a target tracking model to determine trajectory information of each pedestrian;

[0032] The calculation module is used to obtain the time point when each pedestrian enters and exits the area where the advertising space is located and the duration of facing the advertising space based on the timestamp of each frame image, the trajectory information and the orientation information, and obtain the advertising attention based on the time point and the duration.

[0033] Optionally, the detection module is specifically configured to:

[0034] The pedestrian detection model is used to read each frame image from the pedestrian video, and the pedestrian coverage frame and pedestrian appearance features in the read image are detected, the coordinate information of the pedestrian coverage frame is used as the pedestrian position information, and the pedestrian appearance features are used as the pedestrian feature information.

[0035] Optionally, the identification module is specifically configured to:

[0036] Segmenting each pedestrian in each frame image from the corresponding image according to the pedestrian position information to obtain a pedestrian image area;

[0037] The orientation recognition model is used to extract orientation features of the pedestrian image area, and orientation information corresponding to the orientation features is obtained by recognition.

[0038] Optionally, the target tracking model includes: a classification module, a matching module and a trajectory prediction module; accordingly, the tracking module is specifically used to:

[0039] Processing the pedestrian position information, the pedestrian feature information, and the orientation information using the classification module to determine motion characteristics of each pedestrian;

[0040] Processing the motion features using the matching module to determine matching relationships between the same pedestrian in adjacent images;

[0041] The matching relationship is processed by the trajectory prediction module to predict the tracking frame position of each pedestrian, and the tracking frame position is used as the trajectory information.

[0042] Optionally, the calculation module is specifically configured to:

[0043] For each frame of the pedestrian video, the marking information, the trajectory information, the direction information and the time points of entering and leaving the area where the advertising space is located of each pedestrian in each frame are recorded in sequence according to the timestamp order, and the duration is obtained by statistics.

[0044] Optionally, the calculation module is specifically configured to:

[0045] The advertising attention corresponding to each pedestrian is calculated based on the time point and the duration; and / or the comprehensive advertising attention is calculated based on the time point and the duration.

[0046] Optionally, it also includes:

[0047] The storage module is used to store the marking information of each pedestrian in each frame of the image, the trajectory information, the direction information, and the time points of entering and leaving the area where the advertising space is located.

[0048] In a third aspect, the present application provides an electronic device, comprising:

[0049] Memory for storing computer programs;

[0050] A processor is used to execute the computer program to implement the aforementioned disclosed advertising attention detection method.

[0051] In a fourth aspect, the present application provides a readable storage medium for storing a computer program, wherein the computer program, when executed by a processor, implements the aforementioned disclosed advertising attention detection method.

[0052] In a fifth aspect, the present application provides a computer program product, including a computer program / instruction, which, when executed by a processor, implements the steps of the aforementioned disclosed advertising attention detection method.

[0053] From the above scheme, it can be seen that the present application provides a method for detecting advertising attention, including: obtaining a pedestrian video in the area where the advertising space is located; using a pedestrian detection model to detect the pedestrian position information and pedestrian feature information in each frame image of the pedestrian video; using a direction recognition model to identify the direction information of each pedestrian in each frame image; using a target tracking model to process the pedestrian position information, the pedestrian feature information and the direction information to determine the trajectory information of each pedestrian; based on the timestamp of each frame image, the trajectory information and the direction information, the time point of each pedestrian entering and leaving the area where the advertising space is located and the duration facing the advertising space are obtained, and the advertising attention is obtained according to the time point and the duration.

[0054] As can be seen, this application uses a pedestrian detection model to detect the pedestrian position information and pedestrian feature information in each frame of the pedestrian video in the area where the advertising space is located. It then uses a direction recognition model to identify the direction information of each pedestrian in each frame, and uses a target tracking model to process the pedestrian position information, pedestrian feature information and direction information to determine the trajectory information of each pedestrian. Finally, based on the timestamp, trajectory information and direction information of each frame, the time points when each pedestrian enters and exits the area where the advertising space is located and the duration of facing the advertising space are obtained, and the advertising attention is obtained based on the time points and duration. This not only reflects the size of the flow of people near the advertising space, but also further determines the degree of attention of people near the advertising space to the advertising, which is conducive to implementing advertising adjustments accordingly.

[0055] Correspondingly, the advertising attention detection device, equipment, medium and program product provided by this application also have the above-mentioned technical effects. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are merely embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without any creative work.

[0057] Figure 1 This is a flow chart of the first advertising attention detection method disclosed in this application;

[0058] Figure 2 This is a flow chart of the second advertising attention detection method disclosed in this application;

[0059] Figure 3 This is a flow chart of the third advertising attention detection method disclosed in this application;

[0060] Figure 4 This is a schematic diagram of an advertising attention detection device disclosed in this application;

[0061] Figure 5 This is a schematic diagram of an electronic device disclosed in this application. DETAILED DESCRIPTION

[0062] The following will be combined with the drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0063] Currently, infrared sensors and pressure sensors are often installed near advertising spots. Signals collected by these sensors automatically count the number of pedestrians passing through the area, and this information is used to estimate ad exposure, for example. This solution can only reflect the flow of people near the ad spot, but cannot determine whether people near the ad spot are actually paying attention to the ad. Therefore, this application provides an ad attention detection solution that not only reflects the flow of people near the ad spot, but also further determines the degree of attention paid by people near the ad spot, facilitating advertising adjustments accordingly.

[0064] See also Figure 1 As shown, the embodiment of the present application discloses a method for detecting advertising attention, including:

[0065] S1. Obtain a video of pedestrians in the area where the ad slot is located.

[0066] In this embodiment, the video of pedestrians in the area where the advertising space is located can be obtained by using a roadside camera to shoot the area; or a camera dedicated to capturing pedestrians in the area can be deployed next to the advertising space.

[0067] S2. Detecting pedestrian position information and pedestrian feature information in each frame of the pedestrian video using a pedestrian detection model.

[0068] It should be noted that the pedestrian position information in a single-frame image refers to the position of each pedestrian in the image, which is generally represented by a rectangular frame. Therefore, for a frame of image, the pedestrian detection model can obtain multiple rectangular frames, one for each person. Correspondingly, pedestrian feature information can be pixel features within a rectangular frame. Therefore, in one embodiment, the pedestrian detection model is used to detect pedestrian position information and pedestrian feature information in each frame of a pedestrian video, including: using the pedestrian detection model to read each frame of the pedestrian video, and detecting the pedestrian cover frame (i.e., the rectangular frame) and pedestrian appearance features in the read image, using the coordinate information of the pedestrian cover frame as the pedestrian position information, and using the pedestrian appearance features as the pedestrian feature information. Pedestrian detection models can be: Faster R-CNN, SPPNet, YOLOv3 / v4 / v5, and RetinaNet, etc.

[0069] S3. Using the orientation recognition model, the orientation information of each pedestrian in each frame image is identified.

[0070] In this embodiment, pedestrian orientation information can be considered as pedestrian attribute information. Specifically, ResNet50 can be used as an orientation recognition model for attribute recognition. In one embodiment, using the orientation recognition model to identify the orientation information of each pedestrian in each frame of image includes: segmenting each pedestrian in each frame of image from the corresponding image based on the pedestrian's position information to obtain a pedestrian image region (i.e., the region containing the rectangular frame); extracting orientation features from the pedestrian image region using the orientation recognition model, and identifying the orientation information corresponding to the orientation features. Orientation information includes front, side, and back views.

[0071] S4. Use the target tracking model to process pedestrian position information, pedestrian feature information and orientation information to determine the trajectory information of each pedestrian.

[0072] In a specific implementation, the target tracking model may include various functions such as ResNet, VGG, CNN, Hungarian algorithm, and distance measurement method. In one embodiment, the target tracking model includes a classification module, a matching module, and a trajectory prediction module. Accordingly, the target tracking model is used to process pedestrian position information, pedestrian feature information, and orientation information to determine the trajectory information of each pedestrian, including: using the classification module to process pedestrian position information, pedestrian feature information, and orientation information to determine the motion characteristics of each pedestrian; using the matching module to process the motion characteristics to determine the matching relationship of the same pedestrian in adjacent images, which is used to represent the relative movement relationship of the same pedestrian in different frames; using the trajectory prediction module to process the matching relationship, predict the tracking frame position of each pedestrian, and use the tracking frame position as the trajectory information.

[0073] S5. Based on the timestamp, trajectory information, and orientation information of each frame image, the time point at which each pedestrian enters and exits the area where the advertisement spot is located and the duration of time facing the advertisement spot are obtained, and the advertisement attention is obtained based on the time point and duration.

[0074] Since video playback follows a time relationship, the action trajectory of each person during the video recording period and the direction during each time are recorded in sequence according to the playback order of each frame image, and finally the time point when each pedestrian enters and exits the area where the ad space is located and the duration facing the ad space can be obtained. The various information finally obtained can be stored for subsequent calls, such as: storing the marking information, trajectory information, direction information and the time point when entering and exiting the area where the ad space is located of each pedestrian in each frame image. In one embodiment, based on the timestamp, trajectory information and direction information of each frame image, the time point when each pedestrian enters and exits the area where the ad space is located and the duration facing the ad space are obtained, including: for each frame image of the pedestrian video, the marking information (used to distinguish each behavior, such as 1, 2, 3, etc.), trajectory information, direction information and the time point when entering and exiting the area where the ad space is located of each pedestrian in each frame image are recorded in sequence according to the timestamp order, and the duration facing the ad space is obtained by statistics.

[0075] It should be noted that for each frame, the information that can be recorded includes: each pedestrian's marker information, location frame information, trajectory information, direction information, time of entry and exit from the advertising area, etc. By recording the timestamps of each frame in sequence, the duration of each pedestrian facing the advertising spot can be directly output in the final output, without the need for secondary calculation of the information recorded for each frame.

[0076] In one example, the overwriting recording in the order of the timestamps of the frames can be implemented by referring to the following process. The following information is recorded for the first frame:

[0077] Pedestrian 1: Location frame information, trajectory information, orientation (front), and the time point of entering the ad space is X;

[0078] Pedestrian 2: Location frame information, trajectory information, direction - side, and the time point of entering the ad area is Y;

[0079] The position frame information, trajectory information, direction-side, and time point of pedestrian 3 entering the area where the ad slot is located are X.

[0080] Accordingly, the following information is recorded for the second frame image:

[0081] Pedestrian 1: Location frame information, trajectory information, orientation (front), time of entry into the ad slot area (X), duration facing the ad slot (X+1);

[0082] Pedestrian 2: Location frame information, trajectory information, direction - side, and the time point of entering the ad area is Y;

[0083] The position frame information, trajectory information, direction-side, and time point of pedestrian 3 entering the area where the ad slot is located are X.

[0084] It can be seen that the positions and other information of pedestrians 1, 2, and 3 remain unchanged in the first and second frames, and since pedestrian 1 is facing the ad space, the duration of his facing the ad space can be directly recorded in the corresponding recording information of the second frame. For more detailed recording process, please refer to the following Figure 3 Related introduction.

[0085] When calculating attention, the ad attention corresponding to each pedestrian can be calculated from the perspective of each pedestrian, or the ad attention of all pedestrians can be combined to obtain a comprehensive ad attention. In one embodiment, the ad attention is obtained based on the time point and duration, including: calculating the ad attention corresponding to each pedestrian based on the time point and the duration of the ad position; and / or calculating the comprehensive ad attention based on the time point and duration.

[0086] It should be noted that the structure and related functional designs of each model in this embodiment can refer to relevant existing models.

[0087] As can be seen, this embodiment uses a pedestrian detection model to detect the position and feature information of pedestrians in each frame of the pedestrian video in the area where the ad spot is located. It then uses a direction recognition model to identify the direction information of each pedestrian in each frame. The target tracking model processes the pedestrian position information, pedestrian feature information, and direction information to determine the trajectory information of each pedestrian. Finally, based on the timestamp, trajectory information, and direction information of each frame, the time points when each pedestrian enters and exits the area where the ad spot is located and the duration of time facing the ad spot are determined. The ad attention is then calculated based on the time points and duration. This not only reflects the size of the pedestrian flow near the ad spot, but also further determines the degree of attention paid to the ad by people near the ad spot, which is conducive to implementing advertising adjustments accordingly.

[0088] See Figure 2 ,Another method for detecting ad attention includes:

[0089] Step S101: input the real-time video stream to be detected, and use the pedestrian detection model to detect the pedestrian feature information and position information of each frame in turn.

[0090] Specifically, pedestrian detection models can be flexibly selected according to specific usage requirements, such as Faster R-CNN, SPPNet, YOLOv3 / v4 / v5, or RetinaNet, all of which can be pre-trained models. After determining the model, the selected model is trained using a pre-labeled pedestrian annotation dataset. Each frame image in the dataset is annotated with: (label, x, y, w, h); label represents the category (e.g., all are represented by 0, which is not processed in this embodiment), (x, y) are the coordinates of the center point of the rectangular box where the pedestrian is located, and w and h are the width and height of the box; among them, x, y, w, and h are all relative values, which are obtained by dividing the corresponding actual values ​​by the width and height of the image.

[0091] In one example, the dataset can be selected from general datasets such as COCO, VOC, and MPII; in order to increase the diversity of samples, pedestrian images from field recorded videos can also be added.

[0092] In step S102 , each frame image in the video and the pedestrian detection frame information obtained by the above steps are used as input, and the pedestrian attribute recognition model is used to obtain the pedestrian attribute information in each detection frame.

[0093] First, the pedestrian rectangle is segmented from the original image. A ResNet50 network is then used as the pedestrian attribute recognition model. The convolutional neural network extracts features, and the residual network, fully connected layer, and softmax layer output attribute probability vectors. Ultimately, attribute information is obtained, namely, the person's orientation, including front, side, and back. The cross-entropy function can be used as the loss function when training the pedestrian attribute recognition model.

[0094] In step S103, the pedestrian position information, pedestrian feature information and attribute information obtained above are used as input and passed into a tracking model, such as using a Deepsort algorithm to perform target tracking.

[0095] Step S103 specifically implements: target feature extraction, target matching, and target trajectory management.

[0096] Target feature extraction: Determine the feature representation of pedestrians in each frame and obtain the feature vector of the target.

[0097] Object matching: The feature vector of an object in the next frame is matched with the feature vector of the corresponding object in the previous frame to determine the location of the target pedestrian in the current frame. Specifically, the Hungarian algorithm can be used to solve the matching problem between multiple objects. Various distance metrics, such as Euclidean distance and cosine similarity, are used to determine the corresponding location of the same person in the next frame.

[0098] Target trajectory management: Manage and update the target trajectory based on the target matching results. Specifically, a Kalman filter can be used to predict and update the target position. The final recorded information includes: each person's tracking frame, the confidence level of the tracking frame, etc.

[0099] Step S104: the post-processing module performs information statistics and recording.

[0100] See Figure 3 , create an empty list list to store information; traverse the real-time video stream; get the current timestamp Curtime; determine whether list.size() is greater than 1, if not, initialize the current information and store it; if so, traverse the current frame tracking information.

[0101] Specifically, traverse the current frame tracking information, search the historical records in the list and update the orientation, timestamp and other information of each ID; determine whether the object (i.e. a pedestrian) has disappeared for more than 20 frames. If so, generate the statistical information of the object; if not, update the orientation, timestamp and other information of the object to the list; determine whether list.size() is greater than 2. If so, remove the earliest record, that is: the list accumulates information as the video timestamp increases. After accumulation, the relevant information record of the previous frame is useless and can be deleted; if list.size() is not greater than 2, determine whether the current frame is the last frame. If so, end information recording; if not, continue to read the n+1th frame of the real-time video stream and repeat the above process.

[0102] The information of each frame includes:

[0103] tObjTrack: Contains the tracking ID (person marker information) and the tracking frame position.

[0104] state: The orientation information of each ID. state=0 means the front.

[0105] enterTime: The time when each ID enters the area. The first ID that appears uses the current time to indicate the time it enters the area.

[0106] leaveTime: The time when each ID leaves the area, initially set to 0.

[0107] durationState0: The time when each ID faces the ad slot. This value is obtained by subtracting the timestamps of each frame and is initialized to 0.

[0108] s32DisCount: The number of times each ID is not matched. If it reaches 20, it is considered that the corresponding pedestrian has left the current ad area. It is initialized to 0.

[0109] Based on this, the list can eventually accumulate information such as the time each pedestrian stayed in the advertising area and the time they faced the advertisement. Further analysis and statistics of this information can be used to obtain the degree of attention to the advertisement. In this embodiment, if a pedestrian faces the advertisement, it is considered that the pedestrian paid attention to and saw the advertisement.

[0110] As can be seen, this embodiment uses a detection model to detect pedestrians, a classification model to identify their orientation, a tracking model to track them, and a post-processing module to compare and analyze previous and subsequent frames. This allows for the completion of complex data analysis tasks in a much shorter timeframe, ultimately yielding relevant information for calculating ad placement attention. This solution is precise, objective, efficient, and timely.

[0111] An advertisement attention detection device provided in an embodiment of the present application is introduced below. The advertisement attention detection device described below and the advertisement attention detection method described above can be referenced to each other.

[0112] See also Figure 4 As shown, the embodiment of the present application discloses an advertisement attention detection device, comprising:

[0113] An acquisition module is used to obtain pedestrian videos in the area where the ad slot is located;

[0114] A detection module is used to detect pedestrian position information and pedestrian feature information in each frame image of the pedestrian video using a pedestrian detection model;

[0115] A recognition module is used to identify the orientation information of each pedestrian in each frame image using an orientation recognition model;

[0116] The tracking module is used to process pedestrian position information, pedestrian feature information and orientation information using the target tracking model to determine the trajectory information of each pedestrian;

[0117] The calculation module is used to obtain the time points when each pedestrian enters and exits the area where the advertising space is located and the duration of time facing the advertising space based on the timestamp, trajectory information and direction information of each frame image, and to obtain the advertising attention based on the time points and duration.

[0118] In one embodiment, the detection module is specifically configured to:

[0119] The pedestrian detection model is used to read each frame image from the pedestrian video, and the pedestrian coverage frame and pedestrian appearance features in the read image are detected. The coordinate information of the pedestrian coverage frame is used as the pedestrian position information, and the pedestrian appearance features are used as the pedestrian feature information.

[0120] In one embodiment, the identification module is specifically configured to:

[0121] Segment each pedestrian in each frame image from the corresponding image according to the pedestrian position information to obtain a pedestrian image area;

[0122] The orientation recognition model is used to extract the orientation features of the pedestrian image area, and the orientation information corresponding to the orientation features is identified.

[0123] In one embodiment, the target tracking model includes: a classification module, a matching module, and a trajectory prediction module; accordingly, the tracking module is specifically used to:

[0124] The classification module processes pedestrian location information, pedestrian feature information, and orientation information to determine the motion characteristics of each pedestrian;

[0125] The matching module is used to process motion features to determine the matching relationship between the same pedestrian in adjacent images;

[0126] The trajectory prediction module is used to process the matching relationship, predict the tracking frame position of each pedestrian, and use the tracking frame position as the trajectory information.

[0127] In one embodiment, the calculation module is specifically configured to:

[0128] For each frame of the pedestrian video, the marking information, trajectory information, direction information and the time points of entering and leaving the area where the advertising space is located of each pedestrian in each frame are recorded in sequence according to the timestamp order, and the duration is calculated.

[0129] In one embodiment, the calculation module is specifically configured to:

[0130] The advertising attention corresponding to each pedestrian is calculated based on the time point and duration; and / or the comprehensive advertising attention is calculated based on the time point and duration.

[0131] In one embodiment, it further includes:

[0132] The storage module is used to store the marking information, trajectory information, direction information and the time points of entering and exiting the area where the advertising space is located of each pedestrian in each frame of the image.

[0133] Among them, for more specific working processes of each module and unit in this embodiment, reference can be made to the corresponding contents disclosed in the aforementioned embodiments, which will not be repeated here.

[0134] It can be seen that this embodiment provides an advertisement attention detection device that can not only reflect the size of the flow of people near the advertisement position, but also further determine the degree of attention paid by people near the advertisement position to the advertisement, which is conducive to implementing advertisement adjustments accordingly.

[0135] An electronic device provided in an embodiment of the present application is introduced below. The electronic device described below and the advertising attention detection method and device described above can be referenced to each other.

[0136] See also Figure 5 As shown, the embodiment of the present application discloses an electronic device, including:

[0137] Memory 501, used for storing computer programs;

[0138] The processor 502 is configured to execute the computer program to implement the method disclosed in any of the above embodiments.

[0139] A readable storage medium provided in an embodiment of the present application is introduced below. The readable storage medium described below and the advertising attention detection method, device, and apparatus described above can be referenced to each other.

[0140] A readable storage medium for storing a computer program, wherein the computer program, when executed by a processor, implements the advertising attention detection method disclosed in the aforementioned embodiment. The specific steps of the method can be referred to the corresponding contents disclosed in the aforementioned embodiment and will not be repeated here.

[0141] A computer program product provided in an embodiment of the present application is introduced below. The computer program product described below can be referenced with other embodiments described herein.

[0142] A computer program product includes a computer program / instruction, which implements the steps of the aforementioned advertising attention detection method when executed by a processor.

[0143] References to "first," "second," "third," "fourth," and the like (if any) herein are intended to distinguish similar objects and are not necessarily intended to describe a particular order or precedence. It should be understood that the terms used in this manner are interchangeable where appropriate so that the embodiments described herein can be implemented in an order other than that illustrated or described herein. In addition, the terms "including" and "having," as well as any variations thereof, are intended to cover non-exclusive inclusions, e.g., a process, method, or apparatus comprising a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units that are not explicitly listed or that are inherent to such processes, methods, or apparatus.

[0144] It should be noted that the descriptions of "first", "second", etc. in this application are for descriptive purposes only and should not be understood as indicating or implying their relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" or "second" may explicitly or implicitly include at least one of such features. In addition, the technical solutions between the various embodiments can be combined with each other, but this must be based on the fact that they can be implemented by ordinary technicians in this field. When the combination of technical solutions is contradictory or cannot be implemented, it should be deemed that such a combination of technical solutions does not exist and is not within the scope of protection required by this application.

[0145] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referenced to each other.

[0146] The steps of the methods or algorithms described in conjunction with the embodiments disclosed herein may be implemented directly using hardware, a software module executed by a processor, or a combination of the two. The software module may be placed in random access memory (RAM), internal memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, a hard disk, a removable disk, a CD-ROM, or any other form of readable storage medium known in the art.

[0147] This document uses specific examples to illustrate the principles and implementation methods of this application. The description of the above embodiments is only used to help understand the method and core ideas of this application. At the same time, for those skilled in the art, based on the ideas of this application, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting this application.

Claims

1. A method for detecting advertising attention, characterized in that: include: Obtain pedestrian videos in the area where the ad slot is located; Detecting pedestrian position information and pedestrian feature information in each frame image of the pedestrian video using a pedestrian detection model; Use the orientation recognition model to identify the orientation information of each pedestrian in each frame image; Processing the pedestrian position information, the pedestrian feature information, and the orientation information using a target tracking model to determine trajectory information of each pedestrian; Based on the timestamp of each frame image, the trajectory information and the orientation information, the time point when each pedestrian enters and exits the area where the advertising space is located and the duration of facing the advertising space are obtained, and the advertising attention is obtained based on the time point and the duration.

2. The method according to claim 1, characterized in that Detecting pedestrian position information and pedestrian feature information in each frame of the pedestrian video using a pedestrian detection model, including: The pedestrian detection model is used to read each frame image from the pedestrian video, and the pedestrian coverage frame and pedestrian appearance features in the read image are detected, the coordinate information of the pedestrian coverage frame is used as the pedestrian position information, and the pedestrian appearance features are used as the pedestrian feature information.

3. The method according to claim 1, characterized in that The orientation recognition model is used to identify the orientation information of each pedestrian in each frame image, including: Segmenting each pedestrian in each frame image from the corresponding image according to the pedestrian position information to obtain a pedestrian image area; The orientation recognition model is used to extract orientation features of the pedestrian image area, and orientation information corresponding to the orientation features is obtained by recognition.

4. The method according to claim 1, wherein The target tracking model includes: a classification module, a matching module and a trajectory prediction module; Accordingly, the target tracking model is used to process the pedestrian position information, the pedestrian feature information, and the orientation information to determine the trajectory information of each pedestrian, including: Processing the pedestrian position information, the pedestrian feature information, and the orientation information using the classification module to determine motion characteristics of each pedestrian; Processing the motion features using the matching module to determine matching relationships between the same pedestrian in adjacent images; The matching relationship is processed by the trajectory prediction module to predict the tracking frame position of each pedestrian, and the tracking frame position is used as the trajectory information.

5. The method according to claim 1, characterized in that Based on the timestamp of each frame image, the trajectory information, and the orientation information, the time point at which each pedestrian enters and exits the area where the advertising space is located and the duration of time facing the advertising space are obtained, including: For each frame of the pedestrian video, the marking information, the trajectory information, the direction information and the time points of entering and leaving the area where the advertising space is located of each pedestrian in each frame are recorded in sequence according to the timestamp order, and the duration is obtained by statistics.

6. The method according to any one of claims 1 to 5, characterized in that Obtaining advertisement attention according to the time point and the duration includes: The advertising attention corresponding to each pedestrian is calculated based on the time point and the duration; and / or the comprehensive advertising attention is calculated based on the time point and the duration.

7. The method according to any one of claims 1 to 5, characterized in that Also includes: The marking information, the trajectory information, the direction information and the time points of entering and leaving the area where the advertisement space is located of each pedestrian in each frame of the image are stored.

8. An advertising attention detection device, characterized in that: include: An acquisition module is used to obtain pedestrian videos in the area where the ad slot is located; A detection module, configured to detect pedestrian position information and pedestrian feature information in each frame of the pedestrian video using a pedestrian detection model; A recognition module is used to identify the orientation information of each pedestrian in each frame image using an orientation recognition model; a tracking module, configured to process the pedestrian position information, the pedestrian feature information, and the orientation information using a target tracking model to determine trajectory information of each pedestrian; The calculation module is used to obtain the time point when each pedestrian enters and exits the area where the advertising space is located and the duration of facing the advertising space based on the timestamp of each frame image, the trajectory information and the orientation information, and obtain the advertising attention based on the time point and the duration.

9. An electronic device, characterized in that: include: Memory for storing computer programs; A processor, configured to execute the computer program to implement the method according to any one of claims 1 to 7.

10. A readable storage medium, characterized in that: Used to store a computer program, wherein when the computer program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.

11. A computer program product comprising a computer program / instructions, characterized in that When the computer program / instructions are executed by a processor, the method according to any one of claims 1 to 7 is implemented.

Images