CN111724199A - Intelligent community advertisement accurate delivery method and device based on pedestrian active perception - Google Patents
Intelligent community advertisement accurate delivery method and device based on pedestrian active perception Download PDFInfo
- Publication number
- CN111724199A CN111724199A CN202010480394.0A CN202010480394A CN111724199A CN 111724199 A CN111724199 A CN 111724199A CN 202010480394 A CN202010480394 A CN 202010480394A CN 111724199 A CN111724199 A CN 111724199A
- Authority
- CN
- China
- Prior art keywords
- pedestrian
- advertisement
- data
- community
- face
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 230000008447 perception Effects 0.000 title claims abstract description 21
- 238000002716 delivery method Methods 0.000 title claims abstract description 14
- 238000000034 method Methods 0.000 claims abstract description 26
- 238000002372 labelling Methods 0.000 claims abstract description 22
- 239000011159 matrix material Substances 0.000 claims description 45
- 238000001514 detection method Methods 0.000 claims description 20
- 238000012549 training Methods 0.000 claims description 19
- 238000013527 convolutional neural network Methods 0.000 claims description 8
- 230000008569 process Effects 0.000 claims description 7
- 238000004590 computer program Methods 0.000 claims description 6
- 239000013598 vector Substances 0.000 claims description 6
- 238000000354 decomposition reaction Methods 0.000 claims description 5
- 210000004205 output neuron Anatomy 0.000 claims description 5
- PXFBZOLANLWPMH-UHFFFAOYSA-N 16-Epiaffinine Natural products C1C(C2=CC=CC=C2N2)=C2C(=O)CC2C(=CC)CN(C)C1C2CO PXFBZOLANLWPMH-UHFFFAOYSA-N 0.000 claims description 3
- 230000009466 transformation Effects 0.000 claims description 3
- 210000002569 neuron Anatomy 0.000 claims description 2
- 230000020509 sex determination Effects 0.000 claims description 2
- 238000011156 evaluation Methods 0.000 description 6
- 230000000694 effects Effects 0.000 description 5
- 230000001815 facial effect Effects 0.000 description 4
- 210000004027 cell Anatomy 0.000 description 3
- 238000010586 diagram Methods 0.000 description 3
- 238000011835 investigation Methods 0.000 description 3
- 238000004458 analytical method Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 2
- 238000004891 communication Methods 0.000 description 2
- 238000013135 deep learning Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000003860 storage Methods 0.000 description 2
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000003796 beauty Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000012217 deletion Methods 0.000 description 1
- 230000037430 deletion Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 238000009826 distribution Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 238000011478 gradient descent method Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 238000005096 rolling process Methods 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/02—Marketing; Price estimation or determination; Fundraising
- G06Q30/0241—Advertisements
- G06Q30/0251—Targeted advertisements
- G06Q30/0269—Targeted advertisements based on user profile or attribute
- G06Q30/0271—Personalized advertisement
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
- G06V40/16—Human faces, e.g. facial parts, sketches or expressions
- G06V40/161—Detection; Localisation; Normalisation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
- G06V40/16—Human faces, e.g. facial parts, sketches or expressions
- G06V40/161—Detection; Localisation; Normalisation
- G06V40/166—Detection; Localisation; Normalisation using acquisition arrangements
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
- G06V40/16—Human faces, e.g. facial parts, sketches or expressions
- G06V40/172—Classification, e.g. identification
Abstract
The invention relates to a pedestrian active perception-based intelligent community advertisement accurate delivery method and device, wherein the method comprises the following steps: 1) acquiring images of pedestrians through a camera, wherein the camera is deployed on public facilities of a community; 2) carrying out face recognition on the pedestrian image, and judging whether the pedestrian is a person in the community or an external person; 3) for personnel in the community, calling corresponding pedestrian labeling data from a community database; if the pedestrian identification data is for the external person, reconstructing pedestrian identification data; 4) loading pedestrian labeling data into an advertisement recommendation system, and acquiring and displaying advertisement data; 5) and identifying the pose of the face of the pedestrian and the camera according to the image of the pedestrian, judging whether the pedestrian notices the delivered advertisement or not, and feeding back. Compared with the prior art, the method and the system can carry out more accurate and efficient advertisement putting according to the face information of the pedestrian user, and have the advantages of high advertisement putting efficiency, high accuracy, high user friendliness and the like.
Description
Technical Field
The invention relates to the field of advertisement putting, in particular to a smart community advertisement accurate putting method and device based on pedestrian active perception.
Background
With the development of information technology, the advertisement industry gradually transits from the traditional advertisement propagation modes of original newspapers, magazines and the like to diversified advertisement platforms, and advertisements are closer to the aspects of mass life.
As an advertisement appearing in a physical manner, a billboard and a rolling advertisement banner form are common in mass life, and advertisement placement in communities and residential areas is also the field most closely related to users who place advertisements. The population of residents in the community is complex, and thus a wide variety of advertising topics have high market potential.
However, the traditional advertisement delivery method cannot achieve efficient and targeted delivery. Most of the existing communities directly place advertisements on an electronic screen and make advertisement sequences to scroll back and forth. This broadcasting method often cannot deliver advertisements according to the preference of the user, and therefore the delivery efficiency is low.
Disclosure of Invention
The invention aims to overcome the defects of the prior art and provide the intelligent community advertisement accurate delivery method and device based on pedestrian active perception, which can achieve accurate and efficient advertisement delivery effect in a community.
The purpose of the invention can be realized by the following technical scheme:
a smart community advertisement accurate delivery method based on pedestrian active perception comprises the following steps:
a step of acquiring a pedestrian image: acquiring images of pedestrians through a camera, wherein the camera is deployed on public facilities of a community;
a pedestrian identification step: carrying out face recognition on the pedestrian image through a target detection module, if the pedestrian is detected, carrying out face comparison on the pedestrian and data in a preset community database, and judging whether the pedestrian is a person in the community or an external person;
a pedestrian labeling data acquisition step: if the pedestrian is detected to be a person in the community, calling pedestrian tagged data corresponding to the pedestrian from a community database, wherein the pedestrian tagged data comprises representation feature data and potential feature data; if the detected pedestrian is an external person, acquiring the appearance characteristic data of the external person, and taking the potential characteristic data of the person in the community with the most frequent appearance of the corresponding appearance characteristic data in the community database as the potential characteristic data of the external person, thereby constructing pedestrian labeling data of the external person;
and (3) advertisement putting step: loading pedestrian labeling data into a pre-trained advertisement recommendation system, acquiring advertisement data to be launched, and transmitting the advertisement data to a display screen, wherein the display screen is deployed on public facilities of a community;
detecting a face attitude angle: and identifying the pose of the face of the pedestrian and the camera according to the image of the pedestrian, judging whether the pedestrian notices the delivered advertisement or not, forming feedback data, and feeding the feedback data back to the advertisement recommendation system.
Further, in the step of pedestrian recognition, the pedestrian image is subjected to face recognition through a target detection algorithm.
Further, in the pedestrian labeling data acquisition step, the appearance characteristic data of the foreign person is acquired from the pedestrian image through a pre-trained ResNet18 convolutional neural network.
Further, the ResNet18 convolutional neural network is trained through an IMDB-WIKI face data set, the ResNet18 convolutional neural network comprising seven output neurons: two sex determination output neurons and five age group determination data neurons.
Further, the appearance feature data includes face, gender and age data.
Further, in the advertisement putting step, the expression of the advertisement recommendation system is as follows:
Gn×m＝Un×k×Ak×m
in the formula, n is the number of users in the advertisement recommendation system, the users construct according to the pedestrian labeling data, m is the number of advertisements, Gn×mScoring a matrix of user ratings for advertisements, Un×kImplicit to the user a dimensional matrix, Ak×mAnd k is the number of hidden dimensions, the user hidden dimension matrix is established based on the pedestrian labeling data, and the advertisement hidden dimension matrix is established in advance.
Further, the training process of the advertisement recommendation system is specifically to collect the scoring information of the received advertisements by the people in the community as training data, load the training data into the advertisement recommendation system for training, and optimize the parameters of the user hidden dimensional matrix and the advertisement hidden dimensional matrix in the training process by adopting a matrix decomposition mode.
Further, the advertisement putting step further comprises the step of playing the exposure type advertisement if the pedestrian is not detected or the detection fails.
Further, in the step of detecting the face pose angle, the process of identifying the pose of the face of the pedestrian and the camera is specifically,
the method comprises the steps of carrying out face detection on a pedestrian image through an MTCNN (multiple-transmission-network) network, obtaining position coordinates of face feature points in the pedestrian image, determining face rotation vectors through affine transformation on the basis of a preset six-feature-point face three-dimensional model, and converting the face rotation vectors into Euler angles, so that the poses of the face of the pedestrian and a camera are determined.
The invention also provides a smart community advertisement accurate delivery device based on pedestrian active perception, which comprises a memory and a processor, wherein the memory stores a computer program, and the processor calls the computer program to execute the steps of the method.
Compared with the prior art, the invention has the following advantages:
(1) the invention utilizes public facilities in the community to deploy the display screen and the camera, converts the advertisement putting form of centralized broadcasting into the dispersed putting form, enlarges the advertisement putting range and improves the advertisement putting strength and efficiency.
(2) The method effectively utilizes the data resources of the residents in the community, combines the static information of the users with the dynamic information acquired in real time, improves the information utilization efficiency, and establishes the infrastructure of the deep learning-based recommendation system.
(3) According to the invention, through face recognition, pedestrian labeling data of a target pedestrian is obtained and loaded into an advertisement recommendation system, potential feature data of the pedestrian labeling data is combined with potential features of advertisements to obtain estimated values of interest degrees of the pedestrian to the advertisements, the largest corresponding advertisement in the estimated values is selected and released on a screen of a public facility, and the effect of accurate advertisement release is achieved to a certain extent, so that personalized and self-adaptive advertisement recommendation becomes possible, the traditional mechanical mode of circularly playing advertisements is changed, more accurate and efficient advertisement release can be performed according to face information of a user of the pedestrian, and the user friendliness is improved.
(4) The invention changes the traditional broadcast advertisement putting method, can feed back the condition that the pedestrian watches the advertisement in real time for the facial gesture recognition of the pedestrian, provides the feedback effect from the user side to the advertisement putting side, is beneficial to judging the efficiency and the accuracy degree of advertisement putting, and is beneficial to further improving and improving the advertisement recommending system in real time.
(5) According to the invention, the gender and age information of the external person is obtained from the pedestrian image through the ResNet18 convolutional neural network, the ResNet18 parameter quantity is small, and the rapid detection and judgment rate can be achieved.
Drawings
FIG. 1 is a schematic flow chart of an intelligent community advertisement accurate delivery method based on pedestrian active perception according to the present invention;
FIG. 2 is a schematic diagram of an advertisement recommendation system of the present invention;
fig. 3 is a schematic diagram illustrating the effect of the face feature point estimation according to the embodiment of the invention.
Detailed Description
The invention is described in detail below with reference to the figures and specific embodiments. The present embodiment is implemented on the premise of the technical solution of the present invention, and a detailed implementation manner and a specific operation process are given, but the scope of the present invention is not limited to the following embodiments.
Example 1
The invention relates to an intelligent community advertisement accurate delivery method based on pedestrian active perception, which is characterized in that target detection is carried out on a camera and a monitoring video image through a deep learning algorithm; analyzing the gender and the age of the obtained pedestrian target, and making labels of the pedestrian and the advertisement by comparing the information of the community face library with the advertisement information; according to the established label, advertisement recommendation is carried out on the pedestrian, and the most appropriate advertisement information is matched for putting; the face posture of the pedestrian is detected, and the advertisement putting efficiency is judged.
As shown in fig. 1, the method for accurately delivering the intelligent community advertisement includes the following steps:
a step of acquiring a pedestrian image: the pedestrian images are collected through the camera, the camera is deployed on public facilities of a community, the public facilities can be lamp posts, garbage cans, telegraph poles, wall surfaces and other public facilities, and the camera is deployed on the lamp posts in the embodiment;
a pedestrian identification step: carrying out face recognition on the pedestrian image through a target detection module, if the pedestrian is detected, carrying out face comparison on the pedestrian and data in a preset community database, and judging whether the pedestrian is a person in the community or an external person;
a pedestrian labeling data acquisition step: if the pedestrian is detected to be a person in the community, calling pedestrian tagged data corresponding to the pedestrian from a community database, wherein the pedestrian tagged data comprises potential characteristic data and expression characteristic data: face, gender and age group; if the detected pedestrian is an external person, acquiring the face, gender and age information of the external person, and taking the potential feature data of the person in the community with the corresponding gender and age as the potential feature data of the external person, wherein the potential feature data has the most frequent occurrence frequency, so as to construct the pedestrian labeling data of the external person;
and (3) advertisement putting step: the pedestrian labeling data is loaded into a pre-trained advertisement recommendation system, advertisement data to be launched are obtained and transmitted to a display screen, the display screen is deployed on public facilities of a community, the public facilities can be lamp posts, garbage cans, telegraph poles, wall surfaces and other public facilities, and the display screen and a camera are deployed on the lamp posts together in the embodiment;
detecting a face attitude angle: according to the pedestrian image, the pose of the face of the pedestrian and the camera is identified, whether the pedestrian notices the delivered advertisement is judged, feedback data is formed and fed back to the advertisement recommendation system, and the shooting area of the camera comprises a display screen.
The steps are described in detail below.
1. Step of pedestrian image acquisition
All be equipped with camera device on the lamp pole in the community to carry out pedestrian's image acquisition. In addition, the side of the lamp post needs to be provided with an advertisement screen for subsequent estimation of the face posture angle of the pedestrian and advertisement showing.
2. Step of pedestrian recognition
And (3) carrying out age group judgment and gender identification by using an IMDB-WIKI face data set as a pre-trained ResNet18 convolutional neural network, wherein the output of the neural network comprises two gender judgments and five age group judgments, and seven output neurons in total. The ResNet18 has smaller parameter quantity, and can achieve faster detection and judgment rate.
The information of sex, age, interests and the like of people in the community needs to be collected and logged into a community database for storage.
3. Pedestrian labeling data acquisition step
Registering information such as human faces, interests and hobbies of people in the community, namely, logging tagged data of residents in the community into a database in advance, wherein the interest and hobbies information is used for constructing potential characteristic data; for the external personnel, the age group and the gender information are extracted according to the pictures collected by the camera and used as the labeling data of the user. Potential characteristics such as the interests and hobbies of the foreign people are to screen out a hobby group with the highest frequency from the data of community residents according to the age group and the gender of the pedestrian as the potential characteristics of the interests and hobbies of the pedestrian.
The potential characteristic data is information such as work type and hobby type.
4. Advertisement putting step
The advertisement recommendation system needs model training in advance. When collecting information about residents in a community, information about scores of received advertisements by some residents can be collected and used as training data. When designing a recommendation system, optimizing potential characteristic parameter matrixes of users and advertisements by adopting a matrix decomposition mode according to existing training data, and taking the potential characteristic parameter matrixes as parameters of an advertisement recommendation system model, wherein an expression of the advertisement recommendation system is as follows:
Gn×m＝Un×k×Ak×m
in the formula, n is the number of users in the collected data, m is the number of advertisements, k is the number of hidden dimensions, G is an evaluation score matrix of the users for the advertisements, U is a user hidden dimension matrix, and A is a hidden dimension matrix of the advertisements. The implicit matrix represents the implicit attributes of the user and the advertisement, the product of the corresponding attributes is the evaluation matrix of the user to the advertisement, and the matching degree of the implicit attributes corresponding to the user and the advertisement is also the matching degree of the user and the advertisement. And training the parameter matrixes of the users and the advertisements according to the models and the labeled data. After the tagged data of the pedestrians are collected and analyzed, the tagged data are sent to an advertisement recommendation system, the interest value of the user to the advertisement is calculated, and the advertisement with the highest score is selected for delivery.
The hidden dimension matrix of the user is established according to pedestrian labeling data and comprises information such as gender, age group, work type, hobby type and the like, information of personnel in a cell can be obtained by community investigation, for personnel in a non-cell, the information of the gender and the age group of the personnel can be detected firstly, and the work type and hobby with the highest frequency under the conditions of the gender and the age group are matched to jointly form the hidden dimension matrix of the user; after the information of personnel in the cell is obtained through investigation and the interest degree of each advertisement is scored, the implicit dimension matrix of each advertisement can be trained by using an advertisement recommendation system expression.
And reading potential characteristic data of each advertisement according to the received advertisement sequence, transmitting the data and the labeled data of the pedestrians into a trained advertisement recommendation system, and matching and showing the advertisement with the highest user evaluation predicted value.
When the face characteristics and gender and age group analysis cannot be identified, the exposure-type advertisement, namely the advertisement pursuing the exposure rate but not aiming at a specific age group, is automatically played.
5. Face attitude angle detection step
After the MTCNN network is used for detecting the faces of pedestrians, the position coordinates of the facial feature points on the two-dimensional image can be analyzed; according to a preset six-feature-point face three-dimensional model (feature points are constructed to comprise a left eye corner, a right eye corner, a nose tip, a left mouth corner, a right mouth corner and a lower jaw), a face rotation vector can be determined according to affine transformation and converted into an Euler angle, namely an angle of rotation along each axis on a three-dimensional plane, so that the poses of the face of a pedestrian and a camera are determined, whether the pedestrian notices the delivered advertisement is further calculated, and feedback information of the user end after the advertisement is delivered is formed. Therefore, the face detection and the position estimation of the feature points are the key points for the face pose detection.
The pedestrian's score for the advertisement can be determined by the time the pedestrian notices the advertisement, thereby obtaining feedback information for subsequent retraining of the advertisement recommendation system.
6. Display screen
The playing format and the advertisement playing mode set by the display screen comprise:
the lamp post display screen supports various communication modes such as a network, a serial port and a USB, wherein the network communication mode can transmit a video signal to be played to a screen by using a network cable or a wireless network, and the position and the mode of video playing can be freely set. After video information of all advertisements is collected, an advertisement playing interface is required to be designed, wherein the advertisement playing interface comprises the position of an advertisement window on a screen, window text information, a playing mode, video playing tone and the like; and then according to the RESTful network protocol, according to an Api interface conforming to the protocol, completing the operations of information acquisition, message establishment, playing box setting modification, corresponding content deletion and the like by methods of Get, Post, Put, Delete and the like. Obtaining the current information of the playing box, the version of the equipment, the name of the currently played program, the current storage, the starting time, the sensor information including brightness, temperature and the like, a screen capture, network setting, the time zone of the equipment, time synchronization and other information; programs can be published on a local area network through a Put method, and programs such as multi-window multi-picture, video, text, weather, clock and the like can be published; the setting of the playing box comprises automatic brightness and manual equipment setting, equipment program switching, equipment cloud account setting, screen rotation, an equipment time server, equipment time, a time zone, language and the like; and calling a Delete method to Delete the unnecessary programs and video resources on the equipment.
The following is a specific implementation provided by this example.
The intelligent community advertisement accurate delivery method based on pedestrian active perception comprises a lamp post facility provided with a camera and a projection screen, and comprises several modules of target detection, gender and age identification, an advertisement recommendation system, face pose estimation and the like.
The community needs to collect information of the resident firstly, including information of the face, gender, age group, work type and hobby type of the resident. The age groups are 20 years old or less, 20-30 years old, 30-40 years old, 40-50 years old and more than 50 years old; the work types comprise production transportation, technical personnel, business and service industries, offices and free occupations; the hobby types comprise movies, arts, literature, sports, beauty, home and the like. Each category of information will be a dimension of the potential features of the user and the data will be matched and stored in the database. Wherein, 0 in the sex information represents male, 1 represents female; the information of the work type and the hobby type can be selected more, the selected dimension is 1, and otherwise, the selected dimension is 0. In addition, the community also needs to classify whether the received advertisements belong to exposure advertisements, and when collecting the information of the residents, the community tests the advertisement matching degree of the partial residents, wherein 0 is uninteresting and 100 is interesting, so as to collect subsequent training data.
In the target detection module, if the pedestrian is not detected or the detection fails, the exposure advertisement is played. If the pedestrian target is detected, comparing the face of the pedestrian target according to the facial features of the pedestrian target, and if the pedestrian target is matched with records in the community database, automatically calling out information corresponding to the resident as potential feature data; if no data is matched with the data, the pedestrian is judged as an alien person, the age group and the gender of the pedestrian are detected and judged, data which are in accordance with the conditions in the community database are screened out, and the hobby and work type data with the most frequent frequency are taken as the data of the pedestrian, so that the potential characteristic data of the pedestrian is obtained.
And performing model training on the advertisement recommendation system by using the advertisement interest degree score obtained by the previous investigation and the potential characteristic data of the user. The advertisement recommendation system adopts a matrix decomposition mode, supposing that the interest degree matrix of the user for the advertisement can be decomposed into multiplication of a user hidden feature matrix and an advertisement hidden feature matrix, taking the advertisement hidden feature matrix as a parameter according to the collected data, and hopefully, the multiplication output of the two matrixes is as close as possible to the collected data, so as to finally obtain the optimized advertisement hidden feature matrix parameter. Before training, normalization processing is carried out on data, an advertisement hiding feature matrix is initialized randomly, a loss function selects a square average error, the advertisement hiding feature matrix is optimized by using a batch gradient descent method, and a final result is used as a parameter of an advertisement recommendation system.
The schematic diagram of the advertisement recommendation system is shown in fig. 2, wherein the advertisement evaluation matrix is a collected advertisement interest degree score matrix, the number of the collected tester data persons is N, the number of advertisements is M, the number of hidden dimensions is K, rN is an element in the user hidden dimension matrix, tM is an element in the advertisement hidden dimension matrix, and rNtM is an element in the advertisement evaluation score matrix of the user, for example, information of gender, age group, work type, and preference type is collected in a survey, the specific types are the same, that is, the age group is divided into 5 sections, the work type is divided into 4 types, and the preference type is divided into 6 types, wherein each type is represented by a separate dimension, and the gender dimension is added, so that a 16-dimensional user hidden vector can be formed, for example, a 35-year-old male technician preference, literature, and sports, and his hidden dimension matrix can be represented as [0,0,1,0,0,0,1,0,0,0,0,1,1,0,0,0]. And then, collecting the interest degree scores of each user for each advertisement, so that a complete advertisement evaluation matrix and a user hidden dimension matrix can be obtained. And randomly initializing an advertisement hidden dimension matrix, and iteratively training the parameters of the matrix to minimize the error of a matrix decomposition equation and obtain a final matrix result.
By utilizing the pedestrian labeling data and the trained advertisement recommendation system, the advertisement interest value of the pedestrian can be estimated, namely the hidden feature matrix of the pedestrian is multiplied by the hidden feature matrix of the advertisement, so that the estimated value of the interest degree of the pedestrian to each advertisement can be obtained. The advertisement corresponding to the maximum value in the estimated value is selected and put on the screen of the lamp post, so that the effect of accurate advertisement putting is achieved to a certain extent.
As shown in fig. 3, while the advertisement is delivered, the camera still needs to continuously track and recognize the face information of the pedestrian, and sends the extracted face information into the deep convolutional network, and determines the pose of the face according to the position of the facial feature point. The convolutional network for pose determination also requires model training. The pose detection of the face is generally determined according to the relative positions of dozens of feature points of the face, and the distribution rule of the feature points and the face pose present a certain correlation. Therefore, the positions of the human face and the feature points with different angles can be used as training data to train the convolution network, adjust the network parameters and finally obtain the estimator of the human face pose. After the face pose of the pedestrian is obtained, the advertisement screen is installed at the fixed position of the lamp post, so that the relative position of the face pose of the pedestrian and the screen can be calculated according to the relative relation, and whether the pedestrian watches the advertisement or not can be further judged. If the advertisement is judged to be watched and the duration exceeds a certain threshold, the interest of the pedestrian in the advertisement can be judged and positive feedback is given, otherwise, the interest of the pedestrian in the advertisement is not shown.
The embodiment also provides a smart community advertisement accurate delivery device based on pedestrian active perception, which comprises a memory and a processor, wherein the memory stores a computer program, and the processor calls the computer program to execute the steps of the smart community advertisement accurate delivery method based on pedestrian active perception.
The foregoing detailed description of the preferred embodiments of the invention has been presented. It should be understood that numerous modifications and variations could be devised by those skilled in the art in light of the present teachings without departing from the inventive concepts. Therefore, the technical solutions available to those skilled in the art through logic analysis, reasoning and limited experiments based on the prior art according to the concept of the present invention should be within the scope of protection defined by the claims.
Claims (10)
1. A smart community advertisement accurate delivery method based on pedestrian active perception is characterized by comprising the following steps:
a step of acquiring a pedestrian image: acquiring images of pedestrians through a camera, wherein the camera is deployed on public facilities of a community;
a pedestrian identification step: carrying out face recognition on the pedestrian image through a target detection module, if the pedestrian is detected, carrying out face comparison on the pedestrian and data in a preset community database, and judging whether the pedestrian is a person in the community or an external person;
a pedestrian labeling data acquisition step: if the pedestrian is detected to be a person in the community, calling pedestrian tagged data corresponding to the pedestrian from a community database, wherein the pedestrian tagged data comprises representation feature data and potential feature data; if the detected pedestrian is an external person, acquiring the appearance characteristic data of the external person, and taking the potential characteristic data of the person in the community with the most frequent appearance of the corresponding appearance characteristic data in the community database as the potential characteristic data of the external person, thereby constructing pedestrian labeling data of the external person;
and (3) advertisement putting step: loading pedestrian labeling data into a pre-trained advertisement recommendation system, acquiring advertisement data to be launched, and transmitting the advertisement data to a display screen, wherein the display screen is deployed on public facilities of a community;
detecting a face attitude angle: and identifying the pose of the face of the pedestrian and the camera according to the image of the pedestrian, judging whether the pedestrian notices the delivered advertisement or not, forming feedback data, and feeding the feedback data back to the advertisement recommendation system.
2. The method for accurately delivering the smart community advertisement based on the active perception of the pedestrian according to claim 1, wherein in the step of pedestrian recognition, the image of the pedestrian is subjected to face recognition through a target detection algorithm.
3. The method for accurately delivering intelligent community advertisements based on pedestrian active perception according to claim 1, wherein in the step of acquiring pedestrian tagged data, the representation feature data of the external people is acquired from the pedestrian image through a pre-trained ResNet18 convolutional neural network.
4. The method of claim 3, wherein the ResNet18 convolutional neural network is trained by IMDB-WIKI face data set, the ResNet18 convolutional neural network comprises seven output neurons: two sex determination output neurons and five age group determination data neurons.
5. The method for smart community advertisement precise delivery based on pedestrian active perception according to claim 1, wherein the appearance feature data comprises face, gender and age data.
6. The intelligent community advertisement accurate delivery method based on pedestrian active perception according to claim 1, wherein in the advertisement delivery step, the expression of the advertisement recommendation system is as follows:
Gn×m＝Un×k×Ak×m
in the formula, n is the number of users in the advertisement recommendation system, the users construct according to the pedestrian labeling data, m is the number of advertisements, Gn×mScoring a matrix of user ratings for advertisements, Un×kImplicit to the user a dimensional matrix, Ak×mAnd k is the number of hidden dimensions, the user hidden dimension matrix is established based on the pedestrian labeling data, and the advertisement hidden dimension matrix is established in advance.
7. The intelligent community advertisement accurate delivery method based on pedestrian active perception according to claim 6, characterized in that the training process of the advertisement recommendation system is specifically to collect the scoring information of the received advertisements by the personnel in the community as training data to be loaded into the advertisement recommendation system for training, and the advertisement recommendation system adopts a matrix decomposition mode to optimize the parameters of the user hidden dimension matrix and the advertisement hidden dimension matrix in the training process.
8. The method for smart community advertisement accurate delivery based on pedestrian active perception according to claim 1, wherein the advertisement delivery step further comprises playing an exposure type advertisement if no pedestrian is detected or the detection fails.
9. The intelligent community advertisement accurate delivery method based on pedestrian active perception according to claim 1, wherein in the face pose angle detection step, the identification process of the pose of the pedestrian face and the camera is specifically,
the method comprises the steps of carrying out face detection on a pedestrian image through an MTCNN (multiple-transmission-network) network, obtaining position coordinates of face feature points in the pedestrian image, determining face rotation vectors through affine transformation on the basis of a preset six-feature-point face three-dimensional model, and converting the face rotation vectors into Euler angles, so that the poses of the face of the pedestrian and a camera are determined.
10. An intelligent community advertisement accurate delivery device based on pedestrian active perception, which is characterized by comprising a memory and a processor, wherein the memory stores a computer program, and the processor calls the computer program to execute the steps of the method according to any one of claims 1 to 9.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010480394.0A CN111724199A (en) | 2020-05-30 | 2020-05-30 | Intelligent community advertisement accurate delivery method and device based on pedestrian active perception |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010480394.0A CN111724199A (en) | 2020-05-30 | 2020-05-30 | Intelligent community advertisement accurate delivery method and device based on pedestrian active perception |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN111724199A true CN111724199A (en) | 2020-09-29 |
Family
ID=72565379
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010480394.0A Pending CN111724199A (en) | 2020-05-30 | 2020-05-30 | Intelligent community advertisement accurate delivery method and device based on pedestrian active perception |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111724199A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112764378A (en) * | 2021-01-04 | 2021-05-07 | 国泰百安信息技术有限公司 | Intelligent community management system based on Internet of things |
| CN113657933A (en) * | 2021-08-16 | 2021-11-16 | 浙江新再灵科技股份有限公司 | Preparation method of elevator advertisement recommendation data |
| CN117370668A (en) * | 2023-10-25 | 2024-01-09 | 广州极数科技有限公司 | Community data management analysis method and device based on digital intelligence and terminal equipment |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106296307A (en) * | 2016-08-24 | 2017-01-04 | 郑州天迈科技股份有限公司 | Electronic stop plate advertisement delivery effect based on recognition of face analyzes method |
| CN108846694A (en) * | 2018-06-06 | 2018-11-20 | 厦门集微科技有限公司 | A kind of elevator card put-on method and device, computer readable storage medium |
| CN109003134A (en) * | 2018-07-20 | 2018-12-14 | 云南航伴科技有限公司 | Intelligent advertisement delivery system and method based on recognition of face |
| WO2019149005A1 (en) * | 2018-01-30 | 2019-08-08 | 深圳冠思大数据服务有限公司 | Offline interactive advertisement system |
| CN111062752A (en) * | 2019-12-13 | 2020-04-24 | 浙江新再灵科技股份有限公司 | Elevator scene advertisement putting method and system based on audience group |
-
2020
- 2020-05-30 CN CN202010480394.0A patent/CN111724199A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106296307A (en) * | 2016-08-24 | 2017-01-04 | 郑州天迈科技股份有限公司 | Electronic stop plate advertisement delivery effect based on recognition of face analyzes method |
| WO2019149005A1 (en) * | 2018-01-30 | 2019-08-08 | 深圳冠思大数据服务有限公司 | Offline interactive advertisement system |
| CN108846694A (en) * | 2018-06-06 | 2018-11-20 | 厦门集微科技有限公司 | A kind of elevator card put-on method and device, computer readable storage medium |
| CN109003134A (en) * | 2018-07-20 | 2018-12-14 | 云南航伴科技有限公司 | Intelligent advertisement delivery system and method based on recognition of face |
| CN111062752A (en) * | 2019-12-13 | 2020-04-24 | 浙江新再灵科技股份有限公司 | Elevator scene advertisement putting method and system based on audience group |
Non-Patent Citations (1)
| Title |
|---|
| 王正友: "《大数据信息推荐理论与关键技术》", 科学技术文献出版社, pages: 297 - 298 * |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112764378A (en) * | 2021-01-04 | 2021-05-07 | 国泰百安信息技术有限公司 | Intelligent community management system based on Internet of things |
| CN112764378B (en) * | 2021-01-04 | 2021-08-10 | 国泰百安信息技术有限公司 | Intelligent community management system based on Internet of things |
| CN113657933A (en) * | 2021-08-16 | 2021-11-16 | 浙江新再灵科技股份有限公司 | Preparation method of elevator advertisement recommendation data |
| CN117370668A (en) * | 2023-10-25 | 2024-01-09 | 广州极数科技有限公司 | Community data management analysis method and device based on digital intelligence and terminal equipment |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108446390B (en) | Method and device for pushing information | |
| CN107197384B (en) | The multi-modal exchange method of virtual robot and system applied to net cast platform | |
| CN108269128B (en) | Advertisement putting method, device, equipment and storage medium | |
| CN108234591B (en) | Content data recommendation method and device based on identity authentication device and storage medium | |
| CN109635680B (en) | Multitask attribute identification method and device, electronic equipment and storage medium | |
| CN111724199A (en) | Intelligent community advertisement accurate delivery method and device based on pedestrian active perception | |
| CN108494836A (en) | Information-pushing method, device and equipment | |
| CN113395542B (en) | Video generation method and device based on artificial intelligence, computer equipment and medium | |
| CN110633669A (en) | Mobile terminal face attribute identification method based on deep learning in home environment | |
| CN101668176A (en) | Multimedia content-on-demand and sharing method based on social interaction graph | |
| CN111814655B (en) | Target re-identification method, network training method thereof and related device | |
| CN111444826A (en) | Video detection method and device, storage medium and computer equipment | |
| Yu et al. | AI-based targeted advertising system | |
| CN110796828A (en) | Method and system for searching lost population based on Internet of vehicles | |
| CA3050456C (en) | Facial modelling and matching systems and methods | |
| CN113850627A (en) | Elevator advertisement display method and device and electronic equipment | |
| CN112580599A (en) | Video identification method and device and computer readable storage medium | |
| CN107563362B (en) | Method, client and system for evaluation operation | |
| CN111768729A (en) | VR scene automatic explanation method, system and storage medium | |
| US20220319232A1 (en) | Apparatus and method for providing missing child search service based on face recognition using deep-learning | |
| CN105323143A (en) | Network information pushing method, apparatus and system based on instant message | |
| CN114283349A (en) | Data processing method and device, computer equipment and storage medium | |
| CN112101387A (en) | Salient element identification method and device | |
| CN111160946A (en) | Advertisement accurate delivery privacy protection method and system based on video technology | |
| CN116863860B (en) | Intelligent energy-saving advertising machine |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |