CN113992984B - Elevator advertisement monitoring and broadcasting method - Google Patents
Elevator advertisement monitoring and broadcasting method Download PDFInfo
- Publication number
- CN113992984B CN113992984B CN202111231251.7A CN202111231251A CN113992984B CN 113992984 B CN113992984 B CN 113992984B CN 202111231251 A CN202111231251 A CN 202111231251A CN 113992984 B CN113992984 B CN 113992984B
- Authority
- CN
- China
- Prior art keywords
- advertisement
- image
- screen
- elevator
- car
- 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.)
- Active
Links
- 238000012544 monitoring process Methods 0.000 title claims abstract description 55
- 238000000034 method Methods 0.000 title claims abstract description 32
- 238000012545 processing Methods 0.000 claims abstract description 20
- 238000013145 classification model Methods 0.000 claims abstract description 13
- 238000012360 testing method Methods 0.000 claims description 21
- 238000012549 training Methods 0.000 claims description 11
- 238000001514 detection method Methods 0.000 claims description 9
- 230000011218 segmentation Effects 0.000 claims description 9
- 230000006870 function Effects 0.000 claims description 6
- 230000008569 process Effects 0.000 claims description 6
- 230000003068 static effect Effects 0.000 claims description 5
- 238000013527 convolutional neural network Methods 0.000 claims description 3
- 238000013135 deep learning Methods 0.000 claims description 3
- 238000012216 screening Methods 0.000 claims description 3
- 238000000586 desensitisation Methods 0.000 abstract description 2
- 230000001133 acceleration Effects 0.000 description 4
- 238000000605 extraction Methods 0.000 description 4
- 230000009471 action Effects 0.000 description 2
- 230000008901 benefit Effects 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000012806 monitoring device Methods 0.000 description 2
- 230000002159 abnormal effect Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 239000012141 concentrate Substances 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000003203 everyday effect Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000009434 installation Methods 0.000 description 1
- 239000003550 marker Substances 0.000 description 1
- 230000000873 masking effect Effects 0.000 description 1
- 230000000284 resting effect Effects 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N21/00—Selective content distribution, e.g. interactive television or video on demand [VOD]
- H04N21/40—Client devices specifically adapted for the reception of or interaction with content, e.g. set-top-box [STB]; Operations thereof
- H04N21/43—Processing of content or additional data, e.g. demultiplexing additional data from a digital video stream; Elementary client operations, e.g. monitoring of home network or synchronising decoder's clock; Client middleware
- H04N21/442—Monitoring of processes or resources, e.g. detecting the failure of a recording device, monitoring the downstream bandwidth, the number of times a movie has been viewed, the storage space available from the internal hard disk
- H04N21/44204—Monitoring of content usage, e.g. the number of times a movie has been viewed, copied or the amount which has been watched
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
-
- 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
- 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
- G06N3/048—Activation functions
-
- 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
- G06N3/08—Learning methods
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N21/00—Selective content distribution, e.g. interactive television or video on demand [VOD]
- H04N21/20—Servers specifically adapted for the distribution of content, e.g. VOD servers; Operations thereof
- H04N21/23—Processing of content or additional data; Elementary server operations; Server middleware
- H04N21/234—Processing of video elementary streams, e.g. splicing of video streams or manipulating encoded video stream scene graphs
- H04N21/23418—Processing of video elementary streams, e.g. splicing of video streams or manipulating encoded video stream scene graphs involving operations for analysing video streams, e.g. detecting features or characteristics
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N21/00—Selective content distribution, e.g. interactive television or video on demand [VOD]
- H04N21/80—Generation or processing of content or additional data by content creator independently of the distribution process; Content per se
- H04N21/81—Monomedia components thereof
- H04N21/812—Monomedia components thereof involving advertisement data
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N7/00—Television systems
- H04N7/18—Closed-circuit television [CCTV] systems, i.e. systems in which the video signal is not broadcast
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Signal Processing (AREA)
- Multimedia (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Software Systems (AREA)
- Mathematical Physics (AREA)
- Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Databases & Information Systems (AREA)
- Evolutionary Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Business, Economics & Management (AREA)
- Marketing (AREA)
- Computer Networks & Wireless Communication (AREA)
- Indicating And Signalling Devices For Elevators (AREA)
Abstract
The invention relates to an elevator advertisement monitoring and broadcasting method, which comprises the following steps that S1, after an advertisement starts to be broadcast in an elevator car, an image in the car is shot by using camera equipment in the elevator car at a time point of delaying 1/4 to 3/4 of the advertisement duration, and the image is stored; s2, in the captured car image, according to the position of the advertisement screen in the car, which is determined in advance, capturing an advertisement screen scratch of the position of the advertisement screen in the car image; s3, classifying the advertisement screen scratch by using an image classification model: ending the monitoring program when the advertisement screen is a black screen image; or when the advertisement screen is a bright screen image, executing the following steps; s4, blurring processing S5 is carried out on the bright-screen advertisement screen image, and a monitoring and broadcasting result is generated and output. The invention can improve the accuracy of effective flow statistics to advertisers, and the desensitization treatment ensures that the advertisers can safely and effectively verify and compare data.
Description
Technical Field
The invention relates to advertisement supervision, in particular to a method for monitoring advertisements played in an elevator.
Background
With the popularization of intelligent advertising screens, more and more advertisers favor the advertising mode because the installation sites of the intelligent advertising screens are close to the audience and the playing frequency is high. The advertisement audiences in the elevator space concentrate on actively or passively receiving advertisement information in the elevator riding time, and the advertiser is especially favored to roll and play own advertisements on the intelligent advertisement screen installed in the elevator.
The monitoring method of advertisement has two methods of manual shooting monitoring and program automatic monitoring, and the monitoring method of real-time image when the advertisement is played is replaced gradually due to low efficiency, low reliability and high labor cost. CN112291606a discloses an advertisement monitoring method, an advertisement monitoring device and a computer readable storage medium. Firstly, receiving a switching instruction sent by monitored equipment, wherein the switching instruction is used for representing that the monitored equipment switches from a currently played multimedia resource to a next multimedia resource; and according to the switching instruction, at least one image is captured, wherein the at least one image comprises the monitored equipment. And actively reporting a switching instruction through the monitored equipment, and monitoring and broadcasting the multimedia resource currently played by the monitored equipment by the monitoring and broadcasting device through real-time snapshot. Meanwhile, CN112291606A also discloses an advertisement monitoring device and a computer readable storage medium, which are used for realizing the advertisement monitoring method, saving cost, improving the efficiency of advertisement monitoring and ensuring the timeliness and effectiveness of information acquisition by a monitoring end. However, the advertisement monitoring method according to CN112291606a cannot exclude a black screen which cannot be normally broadcast due to the failure of an intelligent screen when the effective broadcast flow is counted, so that the benefits of advertisers are damaged, and audience information in images and video data generated by a monitoring system is not subjected to desensitization fuzzy processing, so that the privacy rights of the audience are violated, and the advertisers may face the dilemma of legal accountability due to the monitoring work.
Disclosure of Invention
In order to solve the problems in the prior art, the invention aims to provide an elevator advertisement monitoring method so as to provide accurate and effective flow statistics on the premise of protecting public privacy.
In order to achieve the above purpose, the invention adopts the following technical scheme: an elevator advertisement monitoring method comprises the following steps:
s1, after an advertisement starts to be played in an elevator car, taking an image in the car by using an in-car camera device and storing the image at a time point of delaying 1/4 to 3/4 of the advertisement time length;
s2, in the captured car image, according to the position of the advertisement screen in the car, which is determined in advance, capturing an advertisement screen scratch of the position of the advertisement screen in the car image;
s3, classifying the advertisement screen scratch by using an image classification model according to the obtained advertisement screen scratch:
ending the monitoring program when the advertisement screen is a black screen image; or (b)
When the advertisement screen is a bright screen image, executing the following steps;
s4, carrying out fuzzy processing on the bright screen advertisement screen image:
(a) Obtaining the position of the image characteristics of the head;
(b) Generating a region to be blurred of the head image characteristic position, the watermark position, the elevator advertisement screen position and the floor region position according to the predetermined advertisement screen position, the watermark position and the predetermined floor region position;
(c) Carrying out fuzzy processing on the determined area needing to be fuzzy by using a Gaussian fuzzy algorithm;
s5, generating and outputting a monitoring and broadcasting result.
In the step S1:
the elevator inner camera device and the car inner advertisement screen are connected with a cloud server, and the cloud server sends a capture instruction to the camera device after delaying a certain time point after sending an advertisement playing instruction to the advertisement screen; a car image is captured that is playing the corresponding advertisement. And classifying and storing the shot images in the car into a cloud server according to a preset advertisement play list. The delay time is 1/4 to 3/4 of the corresponding advertisement duration; the delay time can be dynamically configured and adjusted at the cloud server.
The delay time is determined according to the length of the advertisement time played, and the image capturing device captures images at the time point after the advertisement is played by the delay time, so that the receiving and reacting time can be reserved for the image capturing device and the advertisement screen, and the image capturing device can capture effective instant images in one cycle of advertisement playing.
In the step S2 of the process described above,
the position of the advertisement screen in the predetermined car is determined by the advertisement screen and a watermark detection algorithm.
Specifically, the cloud server sends a test image to the advertisement screen during the elevator stationary period, wherein the stationary period refers to a period when no person takes the elevator, and generally in a period from early morning to 3 points of each day, the cloud server issues a test image playing command, the test image is used for distinguishing advertisement screens of other manufacturers (a plurality of advertisement screens are arranged in some cabins), and meanwhile, the test image is also helpful for accurately determining the advertisement screen positions by a detection algorithm. In addition, considering that the positions of the watermark information added to the photographed images by each image pickup device of each elevator are not uniform, the position of each elevator needs to be detected by a yolov3 algorithm model. The image capturing device takes a car where an advertisement screen displaying a test image is located as a target capture image, watermark information is displayed on the capture image, a plurality of capture images are input into a cloud server, yolov3 model training is adopted, the advertisement screen with the test image is taken as a target, a watermark information area is taken as a target, and a corresponding capture image data set is acquired and constructed for the yolov3 model training. And outputting the advertisement screen position and the watermark position by the trained yolov3 network model.
In the step S3:
the image classification model adopts a resnet convolutional neural network, and the model is trained by utilizing the captured advertisement screen data set through a sigmod function as a loss function; and distinguishing the advertisement screen black screen image and the bright screen image through the trained resnet image classification model.
The cloud server classifies advertisement screen matting by using an image classification model, the images are sent into a cut-out resnet network with the resolution of 224 x 224, and the model outputs two categories which are respectively a black screen image and a bright screen image. Blackout may occur due to the advertisement order failing to play for a number of reasons, and the blackout image should be removed from the monitored image and may be considered as ineffective advertisement traffic. The bright screen image indicates that the corresponding advertising screen capture is an effective capture.
In said step S4, first
In the step S4 (a), a target position of all head feature images in the image is acquired by using a target detection algorithm yolov 3; the target detection algorithm takes the human head characteristics as the target output, the data set is formed by collecting images of mass elevator points, marking and arranging, the yolov3 model is adjusted, the resolution and the model structure are changed, the calculated amount is reduced, and the performance is improved.
In said step S4 (b), the predetermined advertisement screen position, watermark position are determined by the advertisement screen and watermark detection algorithm. Specifically, the cloud server sends a test image to the advertisement screen during the elevator stationary period, wherein the stationary period refers to a period when no person takes the elevator, and generally in a period from early morning to 3 points of each day, the cloud server issues a test image playing command, the test image is used for distinguishing advertisement screens of other manufacturers (a plurality of advertisement screens are arranged in some cabins), and meanwhile, the test image is also helpful for accurately determining the advertisement screen positions by a detection algorithm. In addition, considering that the positions of the watermark information added to the photographed images by each image pickup device of each elevator are not uniform, the position of each elevator needs to be detected by a yolov3 algorithm model. The image capturing device takes a car where an advertisement screen displaying a test image is located as a target capture image, watermark information is displayed on the capture image, a plurality of capture images are input into a cloud server, yolov3 model training is adopted, the advertisement screen with the test image is taken as a target, a watermark information area is taken as a target, and a corresponding capture image data set is acquired and constructed for the yolov3 model training. And outputting the advertisement screen position and the watermark position by the trained yolov3 network model.
The predetermined floor area position in the step S4 (b) is obtained by capturing images of the elevators in a static state through a floor area segmentation algorithm, judging whether the obtained captured images are in a door opening/closing model or not, and finally screening out an image of the door opening/closing unmanned state by each elevator; the stationary state of the elevator refers to a state in which the elevator does not open or close a door for a certain time and acceleration change does not occur during operation. And the cloud server analyzes the grabs through a bisetnet deep learning segmentation model, performs semantic segmentation on the floor region, and obtains the floor position region.
And generating the areas needing fuzzy processing of the characteristic positions of the head images, the positions of the elevator advertisement screens, the positions of the watermarks and the positions of the floors according to the preset determined positions of the advertisement screens, the positions of the watermarks and the preset positions of the floors, namely, the characteristic positions of the head images in the original monitored captured images need to be subjected to fuzzy processing, and the advertisement screens, the watermarks and the areas of the floors do not need to be subjected to fuzzy processing. However, if the head features appear in the advertisement screen, the watermark and the floor area, the corresponding head feature image needs to be blurred to protect privacy of the passengers in the elevator so as to achieve the purpose of safely using the monitored picture without legal disputes, in this case, the image is still an effective bright screen image, and finally the image is counted into the effective flow statistical data of the corresponding advertisement. The advertisement screen area is used by an advertiser to check the consistency of the content played by the advertisement, and the watermark is used by the advertiser to check the effectiveness of the image; the advertising screen area and watermark must be clear; the floor area is selected not to be subjected to blurring processing in consideration of the fact that no sensitive information exists, blurring can influence the observation of the monitored image.
And (3) according to the determined area needing to be blurred, carrying out blurring processing on the original captured image of the effective bright screen image according to a Gaussian blurring algorithm, and generating a final required monitoring and broadcasting result. The monitoring result comprises effective monitoring pictures after fuzzy processing, elevator operation information at the moment of capturing pictures and corresponding advertisement information.
In step S5, the blurred monitored images are respectively stored according to the classification of the advertisement play list, and meanwhile, information of each image capturing object, such as time, elevator points, specific order of advertisements and the like, is stored in an electronic file form, so that an advertiser can conveniently and rapidly search and check; the grab pictures are corresponding to advertisement attribute extraction information, and can be visually marked on the images, so that owners can conveniently verify and count effective flow reports.
Compared with the prior art, the invention has the following advantages and beneficial effects:
the authenticity of the display media advertisement in the elevator car needs to be verified, so that the effectiveness of the advertisement owner in popularizing the product through the advertisement playing platform is ensured, and the abundantly used advertisement is prevented. The elevator advertisement monitoring and broadcasting method can provide guarantee for interests of all parties, not only can verify the broadcasting effect of advertisers, but also can effectively monitor an advertisement broadcasting platform.
The elevator advertisement pricing price is closely related to the corresponding advertisement monitoring and broadcasting, and the monitoring and broadcasting method can accurately verify the effective advertisement broadcasting times so as to influence the final pricing price.
The invention can automatically desensitize the monitoring and broadcasting images in batch in real time, save the monitoring and broadcasting images, greatly reduce the manpower demand and improve the operation efficiency of the elevator advertisement monitoring and broadcasting system.
The invention can grasp the images in the advertisement playing period of the advertisement screen, carry out fuzzy processing, filter out unqualified images, and filter out images as black screen images with abnormal advertisement screen, namely the advertisement screen images when the advertisement cannot be effectively played. And providing the monitoring image and the effective flow statistics report to the final advertiser, so that the monitoring image and the effective flow statistics report are convenient to check. The head characteristics of elevator passengers are removed from the monitored image after fuzzy processing, so that advertisers can safely and effectively utilize the monitored image to verify and compare effective flow statistics.
According to the method of the invention, a frame of screenshot is captured at a preset time delay point for the duration of each advertisement, and the advertisement monitoring and broadcasting result is obtained. Therefore, the data processing workload is reduced, and the monitoring efficiency of advertisement monitoring is improved. The model is trained repeatedly by utilizing the images intercepted by each frame, so that the model is accurate and close to the actual model, and the monitoring and broadcasting accuracy is improved. And invalid screenshot of the black screen is removed through classification and identification, so that the monitoring accuracy is improved. Meanwhile, the monitoring and broadcasting result obtained by the method is clean and safe through the protection treatment of the privacy information of the passengers.
Drawings
Fig. 1 is a flow diagram schematically illustrating an elevator advertisement monitoring method according to one embodiment of the present invention;
fig. 2 is a flow chart schematically illustrating the steps of a grip in an elevator advertising method according to an embodiment of the invention.
Detailed Description
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings that are required to be used in the embodiments will be briefly described below. It is apparent that the drawings in the following description are only some embodiments of the present invention, and that other drawings may be obtained from these drawings without inventive effort for a person of ordinary skill in the art.
The present invention will be described in detail below with reference to the drawings and the specific embodiments, which are not described in detail herein, but the embodiments of the present invention are not limited to the following embodiments.
Fig. 1 is a flow diagram schematically illustrating an elevator advertisement monitoring method according to one embodiment of the present invention; referring to fig. 1, an embodiment of an elevator advertisement monitoring method according to the present invention includes the following steps:
according to the elevator advertisement playing sequence table, after a new advertisement starts to be played on an elevator car advertisement screen, capturing images in a frame of elevator car by using a camera in the elevator car according to the length of the advertisement being played. In the present embodiment, the time point at which capturing of an image starts is set in the range of 1/4 to 3/4 of the advertisement duration. The capture is preferably performed at a time point of 1/2. If the advertisement duration is 10 seconds, capturing images in the car at 5 seconds when the advertisement starts playing. After the captured image is obtained, the captured image is classified based on an elevator advertisement playlist, and the image is stored in correspondence with the played advertisement. After the subsequent steps, the monitored advertisement can be obtained directly.
As shown in the flow of fig. 1, next, in the captured image in the car, the advertisement screen cut of the position of the advertisement screen is cut according to the predetermined position of the advertisement screen in the car.
In this step, it is first necessary to determine in advance the exact position of the advertising screen within the car. In order to accurately distinguish and determine which advertisement screen to be used for playing advertisements in a predetermined advertisement play list is to be played, in consideration of the fact that a plurality of advertisement screens set by different manufacturers may be installed in one elevator, according to the concept of the present invention, a test image is first transmitted to a reserved advertisement screen. In this embodiment, the test image may be a test card reserved for one screen content. According to the rules of the peak period and the valley period of elevator operation, the situation of taking or using the elevator is usually carried out in the early morning time period. Therefore, according to the present embodiment, it is set that the cloud server transmits the test image to the elevator advertisement screen at 3 am every day. The test card or image tested is utilized to precisely determine the location of the advertising screen to be used for playing the advertisement. The position of the advertisement screen is accurately calibrated, the number of invalid grabbers can be reduced, the grabber accuracy is improved, and the effectiveness of advertisement monitoring is further improved. According to one embodiment of the present invention, for convenience and accuracy of subsequent processing steps of extraction, resolution, and the like, the image capturing apparatus adds watermark marks (osd marks) containing time, advertisement numbers, elevator numbers, and the like to the captured image. This marker may be placed near the lower or longitudinal edges of the captured image to minimize masking or impact on the graphical content. Thus, according to embodiments of the present invention, the location of the watermark is further determined simultaneously with the location of the advertising screen. The predetermined position of the advertising screen within the car is determined by the advertising screen and watermark detection algorithm. In this embodiment, yolov3 model training is adopted, and the advertisement screen displaying the test card or the test image and the watermark mark on the corresponding capture image are taken as targets, and the corresponding data set is collected and constructed for yolov3 model training. And the trained yolov3 network model can output the advertisement screen position and the watermark position.
Meanwhile, according to the present embodiment, in the present step, the position of the elevator floor area is also determined at the same time.
In the process of predetermining the position of the floor area, carrying out drawing on the elevator after resting for 5 minutes through a floor area segmentation algorithm, judging whether the obtained drawing is in a door opening and closing mode or not, and finally screening a door closing unmanned image by each elevator; analyzing the capture map through a bisetnet deep learning segmentation model, and carrying out semantic segmentation on the floor region to obtain the floor position region. Wherein the elevator acceleration is collected by an elevator gyro sensor. Whether the elevator is in a static state or not is judged by opening and closing a door of the elevator and the movement acceleration of the elevator. Within 5 minutes, the elevator is judged to be in a static state without door opening and closing actions and without changing the acceleration of the elevator. The opening and closing of the elevator can be realized by a cloud server detecting the door state in the image and the like.
As shown in the flow of fig. 1, in the present embodiment, the advertisement screen matting is classified by using an image classification model according to the obtained advertisement screen matting.
In this embodiment, the captured advertising screen image is classified using an image classification model. The image classification model adopts a resnet convolutional neural network, and the model is trained by utilizing the captured advertisement screen data set through a sigmod function as a loss function; and distinguishing the advertisement screen black screen image and the bright screen image through the trained resnet image classification model. In this embodiment, training of the graph classification model is performed in the cloud server. Therefore, on one hand, the local equipment is reduced, and the cost is reduced; on the other hand, the training device can acquire data information of more models, different service lives, different models, products produced by different manufacturers and the like and input the data information into the model for training. Training the server with a richer source of data will help to improve the accuracy of the final result.
In this embodiment, the cloud server classifies the advertisement screen by using an image classification model, the image is sent to the cut-out resnet network with a resolution of 224×224, and the model outputs two categories, namely a black screen image and a bright screen image.
In the classifying process, when the advertisement screen is determined to be a black screen image, the screenshot is meaningless for judging whether the advertisement is played or not, and whether the advertisement is played according to the specified content on time cannot be judged based on the black screen image. Therefore, in the present embodiment, after the determination result of the black screen image is obtained, the monitoring program is ended. When the advertisement screen matting is a bright screen image, the subsequent steps according to the present embodiment are performed.
As shown in fig. 1, in this step, first, head recognition is performed to determine that a passenger is present in the elevator car at the moment of the grip. And then determining the position and the size of the area which should be subjected to fuzzy processing according to the position of the head image characteristic of each passenger, in particular to the position relation between the passenger and the advertisement screen, the watermark mark on the advertisement screen and the elevator car floor.
According to the conception of the invention, not only the advertisement playing monitoring result with high accuracy is provided, but also privacy infringement or other action results possibly formed in the process of executing the advertisement playing monitoring work are avoided as much as possible. For example, infringement or damage to the elevator occupant may be constituted when the advertisement screen shot contains a clear image of the elevator occupant's face. For this reason, according to the inventive concept, before submitting a monitored result containing advertisement screen capturing information, blurring processing is performed on the submitted result. According to the present embodiment, all the occupant head areas are blurred according to different requirements. Therefore, the accuracy of the advertisement monitoring result is improved, and meanwhile, a clean monitoring result is provided.
In this embodiment, the target position of all the head feature images in the image is acquired by using the target detection algorithm yolov 3. In this step, the human head is targeted for output. The data set is constructed by collecting images of mass elevator points, marking and arranging, and the yolov3 model is adjusted, resolution and model structure are changed, so that the calculated amount is reduced, and meanwhile, the performance is improved.
And generating fuzzy rules of the characteristic positions of the human head images, the positions of the elevator advertisement screens, the positions of the watermark marks and the floor position areas according to the positions of the advertisement screens, the positions of the watermark marks and the positions of the floor areas which are determined in advance, namely, the advertisement screen areas, the watermark marks and the areas without shielding the heads of passengers in the original grabbing images in the monitored mode are not required to be fuzzy, and other areas are required to be fuzzy. And generating a region needing blurring according to the rule. The advertisement screen area is an area for approving and monitoring the content played by the advertisement, and the watermark mark is used for marking the corresponding relation between the captured image and the advertisement being played when the image is captured; the advertising screen area and the watermarking area must be clear. The floor area is selected not to be subjected to blurring processing in consideration of the fact that no sensitive information exists, blurring can influence the observation of the monitored image. However, if the head features appear on the advertising screen, the watermarking area and the floor area, the corresponding head feature images also need to be blurred to protect the privacy of the passengers. In this case, the image is still an effective bright screen image, and is finally counted into the effective flow statistics of the corresponding advertisement.
In the present embodiment, the determined region to be blurred is blurred by a gaussian blur algorithm. And finally generating and outputting a monitoring and broadcasting result.
Respectively storing the blurred monitoring images according to the advertisement play list classification, and simultaneously storing monitoring information captured by each image in the form of an electronic file, such as time, elevator points, advertisement specific orders and the like, so as to facilitate quick search and check; the grab images corresponding to the advertisement attribute extraction information can be visually marked on the images, so that verification is convenient, and effective flow reports are counted.
Fig. 2 is a flow chart schematically illustrating a grip in an elevator advertising method according to an embodiment of the present invention. As shown in fig. 2, in the present embodiment, the in-car camera device and the advertisement screen are connected to the cloud server. The cloud server sends an advertisement playing instruction to the advertisement screen according to the advertisement playing list, namely, sends a command to switch advertisements, meanwhile, the cloud server determines a time point for executing the capture after determining delay time according to the duration of the played advertisements, then sends a capture instruction to the camera equipment, and the camera captures images of the car according to the instruction. Usually, only one frame of image is needed to be grabbed in the process of playing an advertisement, so that the advertisement can be confirmed to be played once within a specified time according to a reservation schedule. Thus, the idea underlying the invention is to grab a frame of an advertisement being played on the advertisement screen in the elevator car after the advertisement has started to play for a period of time. According to one embodiment of the invention, the capture time point is set within a period of 1/4-3/4 of the advertising market being played. Typically, images captured during this period of time may indicate that the advertisement has been completely played. For example, when the playing of one advertisement is 10 seconds, a capture instruction is sent to the image capturing device after the advertisement starts to play for 5 seconds, and the 5 seconds is the delay time. The delay time can be dynamically configured and adjusted at the cloud server. After capturing the images, according to the embodiment, the captured images in the car are classified and stored in the cloud server according to a preset advertisement play list. Therefore, each advertisement and the captured image are stored in a one-to-one correspondence manner, and subsequent extraction and processing are facilitated. As shown in fig. 2, the capture step ends.
The foregoing is merely exemplary of embodiments of the invention and, as regards devices and arrangements not explicitly described in this disclosure, it should be understood that this can be done by general purpose devices and methods known in the art.
The above is only one embodiment of the present invention, and is not limited thereto, and various modifications and variations of the present invention will be apparent to those skilled in the art. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.
Claims (4)
1. An elevator advertisement monitoring method comprises the following steps:
s1, after an advertisement starts to be played in an elevator car, capturing an image in the car by using an in-car camera device and storing the image at a time point of delaying 1/4 to 3/4 of the advertisement time length;
s2, in the captured car image, according to the position of the advertisement screen in the car, which is determined in advance, capturing an advertisement screen scratch of the position of the advertisement screen in the car image;
s3, classifying the advertisement screen scratch by using an image classification model:
ending the monitoring program when the advertisement screen is a black screen image; or (b)
When the advertisement screen is a bright screen image, executing the following steps;
s4, carrying out fuzzy processing on the advertisement screen matting:
(a) Obtaining the position of the image characteristics of the head;
(b) Generating a region to be blurred of the head image characteristic position, the watermark position, the elevator advertisement screen position and the floor region position according to the predetermined advertisement screen position, the watermark position and the predetermined floor region position;
(c) Carrying out fuzzy processing on the determined area needing to be fuzzy by using a Gaussian fuzzy algorithm;
s5, generating and outputting a monitoring and broadcasting result.
2. The elevator advertisement monitoring method according to claim 1, wherein: in the step S1 of the above-mentioned process,
the camera equipment in the elevator car and the advertisement screen in the car are connected with a cloud server; and
and classifying and storing the shot images in the car into a cloud server according to a preset advertisement play list.
3. The elevator advertisement monitoring method according to claim 2, wherein:
in the step S4 (a), a target position of all head image features in the image is obtained by using a target detection algorithm yolov 3;
the predetermined advertisement screen position and watermark position in the steps S2 and S4 (b) are that test images are sent to the advertisement screen during the static period of the elevator, the camera equipment takes the elevator car where the advertisement screen displaying the test images is positioned as a target capture image, watermarks are displayed on the capture image, a plurality of capture images are input to a cloud server, a yolov3 training model is adopted, and finally the advertisement screen position and watermark position information are determined;
the predetermined floor area position in the step S4 (b) is obtained by capturing images of the elevators in a static state through a floor area segmentation algorithm, judging whether the obtained captured images are in a door opening/closing model or not, and finally screening out an image of the door opening/closing unmanned state by each elevator; analyzing the grabs through a bisetnet deep learning segmentation model, and carrying out semantic segmentation on the floor area to obtain the floor position area.
4. The elevator advertisement monitoring method according to claim 1, wherein: in the step S3:
the image classification model is a resnet convolutional neural network, and is trained by using the captured advertising screen data set through a sigmod function as a loss function.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202111231251.7A CN113992984B (en) | 2021-10-22 | 2021-10-22 | Elevator advertisement monitoring and broadcasting method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202111231251.7A CN113992984B (en) | 2021-10-22 | 2021-10-22 | Elevator advertisement monitoring and broadcasting method |
Publications (2)
Publication Number | Publication Date |
---|---|
CN113992984A CN113992984A (en) | 2022-01-28 |
CN113992984B true CN113992984B (en) | 2024-02-23 |
Family
ID=79740187
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202111231251.7A Active CN113992984B (en) | 2021-10-22 | 2021-10-22 | Elevator advertisement monitoring and broadcasting method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN113992984B (en) |
Families Citing this family (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN116453098A (en) * | 2023-04-21 | 2023-07-18 | 杭州硕泰科技有限公司 | Advertisement monitoring analysis method and system based on block chain |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2006117355A (en) * | 2004-10-20 | 2006-05-11 | Hitachi Building Systems Co Ltd | Elevator advertisement watching and listening system |
WO2014104518A1 (en) * | 2012-12-27 | 2014-07-03 | 전자부품연구원 | System and method for providing target advertisement |
CN208071039U (en) * | 2018-03-15 | 2018-11-09 | 广东工匠特种设备有限公司 | A kind of elevator with advertisement dispensing and monitoring device |
KR101952623B1 (en) * | 2017-09-14 | 2019-02-27 | (주)글루버 | Method and apparatus for video service with privacy protection |
CN112291606A (en) * | 2020-10-30 | 2021-01-29 | 成都新潮传媒集团有限公司 | Advertisement monitoring method, advertisement monitoring device and computer readable storage medium |
JP2021070555A (en) * | 2019-10-30 | 2021-05-06 | 西日本エレベータ株式会社 | Remote monitoring display system and monitoring display method for elevator |
-
2021
- 2021-10-22 CN CN202111231251.7A patent/CN113992984B/en active Active
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2006117355A (en) * | 2004-10-20 | 2006-05-11 | Hitachi Building Systems Co Ltd | Elevator advertisement watching and listening system |
WO2014104518A1 (en) * | 2012-12-27 | 2014-07-03 | 전자부품연구원 | System and method for providing target advertisement |
KR101952623B1 (en) * | 2017-09-14 | 2019-02-27 | (주)글루버 | Method and apparatus for video service with privacy protection |
CN208071039U (en) * | 2018-03-15 | 2018-11-09 | 广东工匠特种设备有限公司 | A kind of elevator with advertisement dispensing and monitoring device |
JP2021070555A (en) * | 2019-10-30 | 2021-05-06 | 西日本エレベータ株式会社 | Remote monitoring display system and monitoring display method for elevator |
CN112291606A (en) * | 2020-10-30 | 2021-01-29 | 成都新潮传媒集团有限公司 | Advertisement monitoring method, advertisement monitoring device and computer readable storage medium |
Non-Patent Citations (3)
Title |
---|
Real-time monitoring of television advertisement using BoW;Da Pan et al,;《2014 7th International Congress on Image and Signal Processing》;109-113 * |
一种基于视频特征的智能电视广告监播***设计方案;李厦;《广播与电视技术》;第46卷(第6期);137-141 * |
基于图像识别的电视广告监播***;陈海东等;《电脑与电信》(第9期);25-28 * |
Also Published As
Publication number | Publication date |
---|---|
CN113992984A (en) | 2022-01-28 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
WO2019105342A1 (en) | Method and apparatus for detecting fake license plates of vehicles, readable storage medium, and electronic device | |
CN105448103B (en) | Vehicle fake-license detection method and system | |
CN107886722A (en) | Driving information handling method and system, terminal and computer-readable recording medium | |
CN111783744A (en) | Operation site safety protection detection method and device | |
CN113947731B (en) | Foreign matter identification method and system based on contact net safety inspection | |
CN111539313A (en) | Examination cheating behavior detection method and system | |
CN110738178A (en) | Garden construction safety detection method and device, computer equipment and storage medium | |
CN108230669B (en) | Road vehicle violation detection method and system based on big data and cloud analysis | |
CN113992984B (en) | Elevator advertisement monitoring and broadcasting method | |
CN110928620B (en) | Evaluation method and system for driving distraction caused by automobile HMI design | |
CN106571040B (en) | Suspicious vehicle confirmation method and equipment | |
CN112381014A (en) | Illegal parking vehicle detection and management method and system based on urban road | |
CN106530414A (en) | Person, certificate and ticket unified secondary detection system | |
CN113422975A (en) | Background live video auditing method | |
CN114926824A (en) | Method for judging bad driving behavior | |
CN113945990A (en) | Safety inspection method, device and system for passenger car | |
CN109344829A (en) | A kind of Train number recognition method and device of High Speed Railway Trains | |
CN116935659B (en) | High-speed service area bayonet vehicle auditing system and method thereof | |
CN112016509B (en) | Personnel station abnormality reminding method and device | |
CN112149586A (en) | Automatic video clip extraction system and method based on neural network | |
CN115204678B (en) | Tourist evaluation analysis system based on scenic spot tourist platform | |
CN114926776A (en) | Supervision method and system for muck vehicle loading and transporting in construction site | |
CN114998880B (en) | Method for detecting violent low-custom barrage based on LSTM technology | |
CN116453098A (en) | Advertisement monitoring analysis method and system based on block chain | |
CN113132796B (en) | AI edge terminal safe playing method based on PID algorithm and related equipment |
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 | ||
GR01 | Patent grant | ||
GR01 | Patent grant |