CN108549865A - A kind of people streams in public places amount adjusting householder method and system based on deep learning - Google Patents
A kind of people streams in public places amount adjusting householder method and system based on deep learning Download PDFInfo
- Publication number
- CN108549865A CN108549865A CN201810325232.2A CN201810325232A CN108549865A CN 108549865 A CN108549865 A CN 108549865A CN 201810325232 A CN201810325232 A CN 201810325232A CN 108549865 A CN108549865 A CN 108549865A
- Authority
- CN
- China
- Prior art keywords
- people
- flow
- time
- point
- following several
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/50—Context or environment of the image
- G06V20/52—Surveillance or monitoring of activities, e.g. for recognising suspicious objects
- G06V20/53—Recognition of crowd images, e.g. recognition of crowd congestion
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V30/00—Character recognition; Recognising digital ink; Document-oriented image-based pattern recognition
- G06V30/10—Character recognition
- G06V30/19—Recognition using electronic means
- G06V30/192—Recognition using electronic means using simultaneous comparisons or correlations of the image signals with a plurality of references
- G06V30/194—References adjustable by an adaptive method, e.g. learning
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Theoretical Computer Science (AREA)
- Databases & Information Systems (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a kind of, and the people streams in public places amount based on deep learning adjusts householder method and system, belongs to artificial intelligence field.People streams in public places amount provided by the invention based on deep learning adjusts householder method：Obtain the flow of the people at the top n time point of current point in time；Wherein, N is the integer more than or equal to 2；According to the flow of the people at the top n time point of the current point in time got, the flow of the people of the following several time points is predicted；According to the flow of the people of the following several time points predicted, the practical flow of the people of the following several time points is adjusted by mode recommended to the user.The present invention provides effective data by flow of the people prediction and the function of recommending of going on a journey, to target group and refers to, and help target group is flexible, freely arranges the travel time, to achieve the effect that adjust flow of the people, improve public place resource utilization.
Description
Technical field
The present invention relates to deep learnings, belong to artificial intelligence field.
Background technology
Currently, in many public places, on the problem of target group uses public resource, due to flow of the people unevenness, lead to
Often it will appear problems with：In flow of the people peak period, target group needs to wait in line to use public resource, causes target group
Temporal waste；And in the flow of the people low ebb phase, a large amount of public resources are vacant, the problem for causing public resource utilization rate low.
In the prior art, to solve the problems, such as due to caused by flow of the people unevenness, the common flow of the people management in public place
Method mainly has：
(1) using swiping the card occupancy system, vacant number of resources in real-time display public place.
(2) online order system is used, the usage time interval determined for target group's schedule ahead.
Above-mentioned stream of people's quantity management method (1) carries out occupancy of swiping the card to public resource by target group, and then obtains the public affairs
Public resource service condition in place altogether, and show the vacant number of resources in the public place so that target group can root
Decide whether to be waited in line according to the vacant number of resources in the public place, be wasted on target group's time to solve
The problem of；And target group is evenly distributed in the different time by above-mentioned stream of people's quantity management method (2) by online order
It is interior use public resource, it is above-mentioned due to caused by flow of the people unevenness to solve the problems, such as.
In the implementation of the present invention, inventor has found that the prior art at least has the following disadvantages：
In above-mentioned stream of people's quantity management method (1), target group still needs to that public place is gone to just to would know that vacant number of resources
Measure situation；And in above-mentioned stream of people's quantity management method (2), the arrangement of time of target group lacks flexibility and freedom.
Invention content
The present invention is directed to deficiencies of the prior art, provides a kind of people streams in public places amount based on deep learning
Householder method and system are adjusted, the technical solution is as follows：
A kind of people streams in public places amount adjusting householder method based on deep learning is provided, the method includes：
Obtain the flow of the people at the top n time point of current point in time；Wherein, N is the integer more than or equal to 2；
According to the flow of the people at the top n time point of the current point in time got, the following several time points is predicted
Flow of the people；
According to the flow of the people of the following several time points predicted, by mode recommended to the user to the future
The practical flow of the people of several time points is adjusted.
Optionally, the flow of the people at the top n time point for obtaining current point in time, including：
Obtain crowd's image at the top n time point of the current point in time；
According to crowd's image at the top n time point of the current point in time got, with based on YOLOv2
The flow of the people computational methods of algorithm calculate the flow of the people at the top n time point of the current point in time.
Optionally, described with N before the flow of the people computational methods calculating current point in time based on YOLOv2 algorithms
The flow of the people at a time point, including：
Determine crowd's image frame one skilled in the art full-size function；
Crowd's image is divided into multiple pedestrian targets according to pedestrian's full-size function；
The flow of the people at the top n time point of the current point in time is calculated according to the pedestrian target.
Optionally, described with N before the flow of the people computational methods calculating current point in time based on YOLOv2 algorithms
The flow of the people at a time point further includes：
Crowd's image is laterally divided into four regions；
Three lines of demarcation delimited respectively in four regions；
Obtain coordinate of the position of the pedestrian target relative to three lines of demarcation delimited respectively in four regions
Change information；
The flow of the people at the top n time point of the current point in time is determined according to the changes in coordinates information.
Optionally, the flow of the people of the following several time points predicted described in the basis, passes through side recommended to the user
The practical flow of the people of the following several time points is adjusted in formula, including：
The attribute of the following several time points is obtained, the attribute includes date type and/or weather pattern；The day
Phase type includes working day, weekend and festivals or holidays, and the weather pattern includes fine day, cloudy day and rainy day；
It is calculated according to the flow of the people of the attribute of the following several time points and the following several time points predicted
The use recommendation of the following several time points；
The use recommendation highest time point is selected to recommend user, so that user selects.
On the other hand, a kind of people streams in public places amount adjusting auxiliary system based on deep learning, the system packet are provided
It includes：
Acquisition module, the flow of the people at the top n time point for obtaining current point in time；Wherein, N is more than or equal to 2
Integer；
Prediction module, the flow of the people at the top n time point of the current point in time for being got according to the acquisition module,
The flow of the people of the following several time points of prediction；
Adjustment module, the flow of the people of the following several time points for being predicted according to the prediction module, by with
The practical flow of the people of the following several time points is adjusted in the mode that family is recommended.
Optionally, the acquisition module, including
Crowd's image acquisition unit, crowd's image at the top n time point for obtaining the current point in time；
Computing unit, when the top n of the current point in time for being got according to crowd's image acquisition unit
Between crowd's image for putting, when calculating the top n of the current point in time with the flow of the people computational methods based on YOLOv2 algorithms
Between the flow of the people put.
Optionally, the computing unit, including：
Function determination subelement, for determining crowd's image frame one skilled in the art full-size function；
Subelement is divided, pedestrian's full-size function for determining according to the function determination subelement is by the crowd
Image is divided into multiple pedestrian targets；
Computation subunit, the pedestrian target for being divided according to the division subelement calculate before the current point in time
The flow of the people at N number of time point.
Optionally, the computing unit further includes：
Equal molecular cell, for crowd's image to be laterally divided into four regions；
Subelement delimited, for delimiting three lines of demarcation respectively in four regions after the equal molecular cell is divided equally；
Subelement is obtained, for obtaining the position of the pedestrian target relative to three delimited respectively in four regions
The changes in coordinates information in line of demarcation；
Flow of the people determination subelement, for being worked as described in the changes in coordinates information determination that subelement is got according to described obtain
The flow of the people at the top n time point at preceding time point.
Optionally, the adjustment module, including：
Attribute acquiring unit, the attribute for obtaining the following several time points, the attribute include date type and/
Or weather pattern；The date type includes working day, weekend and festivals or holidays, and the weather pattern includes fine day, cloudy day and rain
It；
Recommendation computing unit, the category of the following several time points for being got according to the attribute acquiring unit
Property and the flow of the people of the following several time points predicted calculate the use recommendation of the following several time points；
Recommendation unit is recommended for selecting the recommendation computing unit calculated using recommendation highest time point
To user, so that user selects.
The beneficial effects of the invention are as follows：
By flow of the people prediction and the function of recommending of going on a journey, provides effective data to target group and refer to, help
Target group is flexible, freely arranges the travel time, to reach the effect for adjusting flow of the people, improving public place resource utilization
Fruit；By using YOLOv2 as detection algorithm frame, under the premise of keeping high measurement accuracy, there is good detection speed
Degree, meets the requirement of real-time；By determine crowd's image frame one skilled in the art's full-size function and by crowd image it is lateral
Four regions are divided into, and delimit three lines of demarcation respectively in four regions, according to the position of target group relative to each
The variation in three lines of demarcation in region finally determines flow of the people, obtains preferable flow of the people computational accuracy, aobvious through testing
Show, the method for the present invention has preferable effect under the intensive scene of flow of the people and shaded by items scene, and average flow of the people calculates accurate
Exactness reaches 90.28%.
Description of the drawings
Fig. 1 is a kind of implementation environment schematic diagram of the embodiment of the present invention；
Fig. 2 is that the people streams in public places amount based on deep learning that the embodiment of the present invention one provides adjusts aided process flow sheet
Figure；
Fig. 3 is that the people streams in public places amount provided by Embodiment 2 of the present invention based on deep learning adjusts aided process flow sheet
Figure；
Fig. 4 is the division methods schematic diagram that crowd's image provided by Embodiment 2 of the present invention is laterally divided into four regions；
Fig. 5 is the knot that the people streams in public places amount based on deep learning that the embodiment of the present invention three provides adjusts auxiliary system
Structure schematic diagram；
Fig. 6 is the knot that the people streams in public places amount based on deep learning that the embodiment of the present invention four provides adjusts auxiliary system
Structure schematic diagram.
Specific implementation mode
To make the object, technical solutions and advantages of the present invention clearer, below in conjunction with attached drawing to embodiment party of the present invention
Formula is described in further detail.
Fig. 1 is a kind of implementation environment schematic diagram of the embodiment of the present invention, is obtained as shown in Figure 1, the implementation environment includes image
Taking equipment 120, processor 140 and terminal 160.
Image acquisition equipment 120 can be monitoring camera, and processor 140 can be a processor, if can also be
The processor cluster of dry-cure device composition, terminal 160 can be the mobile terminal devices such as mobile phone, tablet computer, terminal 160
It is provided with display panel.Image acquisition equipment 120 and processor 140 pass through wired or wireless network connection, 140 He of processor
Terminal 160 passes through wired or wireless network connection.Image acquisition equipment 120 obtains crowd's image in public place, by people
Group's image is sent to processor 140, and processor 140 to the processing of a large amount of crowd's images by predicting the following several time points
Flow of the people, then provides a user valid data reference by terminal 160, and user can be checked by the display panel of terminal 160
The recommendation information that processor 140 is sent.
Embodiment one：
The people streams in public places amount based on deep learning that the present embodiment provides a kind of adjusting householder method, described referring to Fig. 2
Method includes：
201：Obtain the flow of the people at the top n time point of current point in time；Wherein, N is the integer more than or equal to 2；
202：According to the flow of the people at the top n time point of the current point in time got, the following several time points is predicted
Flow of the people；
203：According to the flow of the people of the following several time points predicted, by mode recommended to the user to following several
The practical flow of the people at time point is adjusted.
The flow of the people at the top n time point for the current point in time that the present embodiment is got by basis, prediction are following several
The flow of the people at time point, to according to the flow of the people of the following several time points predicted, pass through mode pair recommended to the user
The practical flow of the people of the future several time points is adjusted so that user need not go to public place to would know that sky
Remaining resource quantity situation, and user can according to the flow of the people of the following several time points predicted come flexible arrangement when
Between.
Embodiment two
The people streams in public places amount based on deep learning that the present embodiment provides a kind of adjusting householder method, described referring to Fig. 3
Method includes：
With image acquisition equipment 120 it is monitoring camera in the present embodiment, processor 140 is computer, terminal 160 is hand
Machine, implementation environment are illustrated by taking the bathhouse of campus as an example；
301：Obtain crowd's image at the top n time point of current point in time；
Specifically, monitoring camera is mounted on campus bathhouse doorway direction, shooting is overlooked at a certain angle
The doorway stream of people passes in and out, and by the crowd's Image Real-time Transmission taken to computer, computer is receiving current point in time
Crowd's image at the top n time point received before crowd's image.
302：According to crowd's image at the top n time point of the current point in time got, with based on YOLOv2 algorithms
Flow of the people computational methods calculate current point in time top n time point flow of the people；
Specifically, computer is after crowd's image at the top n time point for receiving current point in time, with based on YOLOv2
The flow of the people computational methods of algorithm, YOLOv2 network structures include 19 convolutional layers and 5 maximum pond layers, are respectively intended to train
Sorter network and detection network, obtain the YOLO neural networks for detecting pedestrian target.Utilize trained YOLO nerve nets
Network detects pedestrian target, and rectangle frame is used in combination to mark；
Optionally, when calculating the top n of the current point in time with the flow of the people computational methods based on YOLOv2 algorithms
Between the flow of the people put include the following steps：
Step 1：Determine crowd's image frame one skilled in the art's full-size function；
Specifically, the present embodiment is analyzed by mass data, pedestrian's full-size function in crowd's image frame is determined；
Step 2：Crowd's image is divided into multiple pedestrian targets according to pedestrian's full-size function；
Specifically, with pedestrian's full-size function for the rectangle frame more than full-size range in crowd's image frame
It is cut, is divided into multiple pedestrian targets；In practical applications, crowd's image is intensive there may be flow of the people and generate pedestrian
Between partial occlusion the case where, multiple pedestrian targets can be caused to concentrate in a rectangle frame, and through the above steps one institute really
Rectangle frame more than full-size range can be divided into multiple pedestrian targets by fixed pedestrian's full-size function；
Step 3：The flow of the people at the top n time point of current point in time is calculated according to pedestrian target；
Specifically, the flow of the people in crowd's image can be calculated according to the pedestrian target number in crowd's image；
In practical applications, because there may be a pedestrian targets by another pedestrian in the crowd's image taken
Target is blocked the case where to by flase drop at a pedestrian target completely, and it is pre- to use kalman filter method in the present embodiment
The next frame moving region for surveying pedestrian target, matches pedestrian target in crowd's image using Hungary Algorithm, determines different frame
Between same pedestrian target information, ensure pedestrian target be overlapped several frames of missing inspection after, examined respectively after pedestrian target is divided into out
When survey, still and before correct matching, this just prevent in a certain range because missing inspection and the case where with losing；
In practical applications, the pedestrian target number in crowd's image not only included enter the people of public place, but also including from
The people that public place leaves, for the flow of the people for entering public place is determined more accurately, optionally, with based on YOLOv2
The flow of the people computational methods of algorithm calculate the flow of the people at the top n time point of the current point in time, further comprising the steps of：
Step 4：Crowd's image is laterally divided into four regions；
Specifically, the division methods that crowd's image is laterally divided into four regions are as shown in Figure 4；
Step 5：Delimit three lines of demarcation respectively in four regions；
Specifically, the method for delimiting three lines of demarcation in four regions respectively is as shown in Figure 4；
Step 6：The position for obtaining pedestrian target becomes relative to the coordinate in three lines of demarcation delimited respectively in four regions
Change information；
Specifically, judging whether the coordinate of pedestrian target has the behavior for crossing line of demarcation；
Step 7：The flow of the people at the top n time point of current point in time is determined according to changes in coordinates information.
Specifically, when pedestrian target forward direction crosses line of demarcation, judge that entering flow of the people increases, when pedestrian target is reversely got over
When crossing line of demarcation, flow of the people increase is walked out in judgement, and final flow of the people result takes the sum of maximum judgement result in each region；
303：According to the flow of the people at the top n time point of the current point in time got, the following several time points is predicted
Flow of the people；
Specifically, after getting the flow of the people at top n time point of current point in time, current time can also be obtained
The attribute at the top n time point of point, attribute include date type and/or weather pattern；Date type include working day, weekend and
Festivals or holidays, weather pattern include fine day, cloudy day and rainy day；
The attribute of the following several time points is obtained, attribute includes date type and/or weather pattern；Date type includes work
Make day, weekend and festivals or holidays, weather pattern includes fine day, cloudy day and rainy day；
In specific implementation process, after the flow of the people for calculating each time point, all to the attribute at the time point into rower
Note, and the attribute at the time point to be predicted also is marked, to be carried out according to the time point of same attribute during prediction
Prediction；Only simply the common attribute at time point is illustrated in the present embodiment, in practical application, date type can be thin
It is divided into working day, Sunday, winter vacation, summer vacation and a variety of different festivals or holidays, does not just illustrate one by one here, and weather class
Type can be specifically subdivided into shown in following table：Fine day, sweltering heat, cloudy, cloudy, light rain, moderate rain, heavy rain, slight snow, moderate snow, heavy snow,
Exceedingly odious weather (heavy rain, severe snow, hail etc.)；
In specific implementation process, the flow of the people after being predicted one time 30 minutes every 10 minutes
It is calculated according to the attribute of the following several time points and the flow of the people of the following several time points predicted following several
The use recommendation at time point；
Specifically, the attribute of flow of the people prediction data and time point is combined, the recommendation of each time point is calculated；
Selection recommends user using recommendation highest time point, so that user selects.
When the present embodiment is by using the flow of the people computational methods based on YOLOv2 algorithms to calculate the top n of current point in time
Between the flow of the people put possess good detection speed under the premise of keeping high measurement accuracy, meet the requirement of real-time；It is logical
It crosses and determines crowd's image frame one skilled in the art's full-size function and crowd's image is laterally divided into four regions, and at four
Three lines of demarcation delimited in region respectively, the variation according to the position of target group relative to three lines of demarcation in each region
It finally determines flow of the people, obtains preferable flow of the people computational accuracy；By obtaining the attribute at time point, including date type
And/or weather pattern, the flow of the people of integration time point predict the flow of the people at the following time point with same alike result to calculate,
So that the flow of the people in the flow of the people data predicted more closing to reality situation, the attribute of integration time point calculate
The use recommendation of the following several time points recommends more rational usage time according to recommendation is used to user.
Embodiment three
The people streams in public places amount based on deep learning that the present embodiment provides a kind of adjusting auxiliary system, described referring to Fig. 5
System includes：
Acquisition module 510, the flow of the people at the top n time point for obtaining current point in time；Wherein, N is more than or equal to 2
Integer；
Prediction module 520, the top n time point of the current point in time for being got according to the acquisition module 510
Flow of the people predicts the flow of the people of the following several time points；
Adjustment module 530, the flow of the people of the following several time points for being predicted according to the prediction module 520 lead to
Mode recommended to the user is crossed the practical flow of the people of the following several time points is adjusted.
The flow of the people at the top n time point for the current point in time that the present embodiment is got by basis, prediction are following several
The flow of the people at time point, to according to the flow of the people of the following several time points predicted, pass through mode pair recommended to the user
The practical flow of the people of the future several time points is adjusted so that user need not go to public place to would know that sky
Remaining resource quantity situation, and user can according to the flow of the people of the following several time points predicted come flexible arrangement when
Between.
Example IV
The people streams in public places amount based on deep learning that the present embodiment provides a kind of adjusting auxiliary system, described referring to Fig. 6
System includes：
Acquisition module 600, the flow of the people at the top n time point for obtaining current point in time；Wherein, N is more than or equal to 2
Integer；
The acquisition module 600, including
Crowd's image acquisition unit 610, crowd's image at the top n time point for obtaining the current point in time；
Computing unit 620, before the current point in time for being got according to crowd's image acquisition unit 610
Crowd's image at N number of time point, with N before the flow of the people computational methods calculating current point in time based on YOLOv2 algorithms
The flow of the people at a time point.
The computing unit 620, including：
Function determination subelement 621, for determining crowd's image frame one skilled in the art full-size function；
Subelement 622 is divided, pedestrian's full-size function for determining according to the function determination subelement 621 is by institute
The crowd image of stating is divided into multiple pedestrian targets；
Computation subunit 623, the pedestrian target for being divided according to the division subelement 622 calculate the current time
The flow of the people at the top n time point of point.
The computing unit 620 further includes：
Equal molecular cell 624, for crowd's image to be laterally divided into four regions；
Subelement 625 delimited, for delimiting three points respectively in four regions after the equal molecular cell 624 is divided equally
Boundary line；
Subelement 626 is obtained, for obtaining the position of the pedestrian target relative to being delimited respectively in four regions
Three lines of demarcation changes in coordinates information；
Flow of the people determination subelement 627, the changes in coordinates information for being got according to the acquisition subelement 626 determine
The flow of the people at the top n time point of the current point in time.
Prediction module 700, the top n time point of the current point in time for being got according to the acquisition module 600
Flow of the people predicts the flow of the people of the following several time points；
Adjustment module 800, the flow of the people of the following several time points for being predicted according to the prediction module 700 lead to
Mode recommended to the user is crossed the practical flow of the people of the following several time points is adjusted；
The adjustment module 800, including：
Attribute acquiring unit 810, the attribute for obtaining the following several time points, the attribute includes date type
And/or weather pattern；The date type includes working day, weekend and festivals or holidays, the weather pattern include fine day, the cloudy day and
Rainy day；
Recommendation computing unit 820, the following some time for being got according to the attribute acquiring unit 810
The attribute of point and the flow of the people of the following several time points predicted calculate the use recommendation of the following several time points；
Recommendation unit 830, for selecting the recommendation computing unit 820 is calculated to use the recommendation highest time
Point recommends user, so that user selects.
When the present embodiment is by using the flow of the people computational methods based on YOLOv2 algorithms to calculate the top n of current point in time
Between the flow of the people put possess good detection speed under the premise of keeping high measurement accuracy, meet the requirement of real-time；It is logical
It crosses and determines crowd's image frame one skilled in the art's full-size function and crowd's image is laterally divided into four regions, and at four
Three lines of demarcation delimited in region respectively, the variation according to the position of target group relative to three lines of demarcation in each region
It finally determines flow of the people, obtains preferable flow of the people computational accuracy；By obtaining the attribute at time point, including date type
And/or weather pattern, the flow of the people of integration time point predict the flow of the people at the following time point with same alike result to calculate,
So that the flow of the people in the flow of the people data predicted more closing to reality situation, the attribute of integration time point calculate
The use recommendation of the following several time points recommends more rational usage time according to recommendation is used to user.
Part steps in the embodiment of the present invention can utilize software realization, and corresponding software program can be stored in can
In the storage medium of reading, such as CD or hard disk.
The foregoing is merely presently preferred embodiments of the present invention, is not intended to limit the invention, it is all the present invention spirit and
Within principle, any modification, equivalent replacement, improvement and so on should all be included in the protection scope of the present invention.
Claims (10)
1. a kind of people streams in public places amount based on deep learning adjusts householder method, which is characterized in that the method includes：
Obtain the flow of the people at the top n time point of current point in time；Wherein, N is the integer more than or equal to 2；
According to the flow of the people at the top n time point of the current point in time got, the stream of people of the following several time points is predicted
Amount；
It is several to the future by mode recommended to the user according to the flow of the people of the following several time points predicted
The practical flow of the people at time point is adjusted.
2. according to the method described in claim 1, it is characterized in that, the people at the top n time point for obtaining current point in time
Flow, including：
Obtain crowd's image at the top n time point of the current point in time；
According to crowd's image at the top n time point of the current point in time got, with based on YOLOv2 algorithms
Flow of the people computational methods calculate the current point in time top n time point flow of the people.
3. according to the method described in claim 2, it is characterized in that, described use the flow of the people calculating side based on YOLOv2 algorithms
Method calculates the flow of the people at the top n time point of the current point in time, including：
Determine crowd's image frame one skilled in the art full-size function；
Crowd's image is divided into multiple pedestrian targets according to pedestrian's full-size function；
The flow of the people at the top n time point of the current point in time is calculated according to the pedestrian target.
4. according to the method described in claim 3, it is characterized in that, described use the flow of the people calculating side based on YOLOv2 algorithms
Method calculates the flow of the people at the top n time point of the current point in time, further includes：
Crowd's image is laterally divided into four regions；
Three lines of demarcation delimited respectively in four regions；
Obtain changes in coordinates of the position of the pedestrian target relative to three lines of demarcation delimited respectively in four regions
Information；
The flow of the people at the top n time point of the current point in time is determined according to the changes in coordinates information.
5. according to the method described in claim 1, it is characterized in that, the following several time points that is predicted described in the basis
Flow of the people is adjusted the practical flow of the people of the following several time points by mode recommended to the user, including：
The attribute of the following several time points is obtained, the attribute includes date type and/or weather pattern；The date class
Type includes working day, weekend and festivals or holidays, and the weather pattern includes fine day, cloudy day and rainy day；
Future is calculated according to the flow of the people of the attribute of the following several time points and the following several time points predicted
The use recommendation of several time points；
The use recommendation highest time point is selected to recommend user, so that user selects.
6. a kind of people streams in public places amount based on deep learning adjusts auxiliary system, which is characterized in that the system comprises：
Acquisition module, the flow of the people at the top n time point for obtaining current point in time；Wherein, N is the integer more than or equal to 2；
Prediction module, the flow of the people at the top n time point of the current point in time for being got according to the acquisition module, prediction
The flow of the people of the following several time points；
Adjustment module, the flow of the people of the following several time points for being predicted according to the prediction module, by being pushed away to user
The practical flow of the people of the following several time points is adjusted in the mode recommended.
7. system according to claim 6, which is characterized in that the acquisition module, including
Crowd's image acquisition unit, crowd's image at the top n time point for obtaining the current point in time；
Computing unit, the top n time point of the current point in time for being got according to crowd's image acquisition unit
Crowd's image, the top n time point of the current point in time is calculated with the flow of the people computational methods based on YOLOv2 algorithms
Flow of the people.
8. system according to claim 7, which is characterized in that the computing unit, including：
Function determination subelement, for determining crowd's image frame one skilled in the art full-size function；
Subelement is divided, pedestrian's full-size function for determining according to the function determination subelement is by crowd's image
It is divided into multiple pedestrian targets；
Computation subunit, the pedestrian target for being divided according to the division subelement calculate the top n of the current point in time
The flow of the people at time point.
9. system according to claim 8, which is characterized in that the computing unit further includes：
Equal molecular cell, for crowd's image to be laterally divided into four regions；
Subelement delimited, for delimiting three lines of demarcation respectively in four regions after the equal molecular cell is divided equally；
Subelement is obtained, for obtaining the position of the pedestrian target relative to three points delimited respectively in four regions
The changes in coordinates information in boundary line；
Flow of the people determination subelement, when the changes in coordinates information for being got according to the acquisition subelement determines described current
Between point top n time point flow of the people.
10. system according to claim 6, which is characterized in that the adjustment module, including：
Attribute acquiring unit, the attribute for obtaining the following several time points, the attribute includes date type and/or day
Gas type；The date type includes working day, weekend and festivals or holidays, and the weather pattern includes fine day, cloudy day and rainy day；
Recommendation computing unit, the attribute of the following several time points for being got according to the attribute acquiring unit and
The flow of the people of the following several time points predicted calculates the use recommendation of the following several time points；
Recommendation unit recommends use for selecting the recommendation computing unit calculated using recommendation highest time point
Family, so that user selects.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810325232.2A CN108549865A (en) | 2018-04-12 | 2018-04-12 | A kind of people streams in public places amount adjusting householder method and system based on deep learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810325232.2A CN108549865A (en) | 2018-04-12 | 2018-04-12 | A kind of people streams in public places amount adjusting householder method and system based on deep learning |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108549865A true CN108549865A (en) | 2018-09-18 |
Family
ID=63514705
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810325232.2A Pending CN108549865A (en) | 2018-04-12 | 2018-04-12 | A kind of people streams in public places amount adjusting householder method and system based on deep learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108549865A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111553737A (en) * | 2020-04-30 | 2020-08-18 | 北京金山安全软件有限公司 | Business time prediction method and device |
| CN113128831A (en) * | 2021-03-11 | 2021-07-16 | 特斯联科技集团有限公司 | People flow guiding method and device based on edge calculation, computer equipment and storage medium |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20160358074A1 (en) * | 2015-06-05 | 2016-12-08 | Cisco Technology, Inc. | Methods and Systems for Counting People |
| CN106503829A (en) * | 2016-09-26 | 2017-03-15 | 同济大学 | A kind of crowding Forecasting Methodology of the Urban Public Open Space based on multi-source data |
| CN106960250A (en) * | 2017-03-03 | 2017-07-18 | 陕西师范大学 | A kind of method of dynamic prediction tourist attractions passenger flow |
| CN106960195A (en) * | 2017-03-27 | 2017-07-18 | 深圳市丰巨泰科电子有限公司 | A kind of people counting method and device based on deep learning |
| CN107368790A (en) * | 2017-06-27 | 2017-11-21 | 上海汇纳信息科技股份有限公司 | Pedestrian detection method, system, computer-readable recording medium and electronic equipment |
-
2018
- 2018-04-12 CN CN201810325232.2A patent/CN108549865A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20160358074A1 (en) * | 2015-06-05 | 2016-12-08 | Cisco Technology, Inc. | Methods and Systems for Counting People |
| CN106503829A (en) * | 2016-09-26 | 2017-03-15 | 同济大学 | A kind of crowding Forecasting Methodology of the Urban Public Open Space based on multi-source data |
| CN106960250A (en) * | 2017-03-03 | 2017-07-18 | 陕西师范大学 | A kind of method of dynamic prediction tourist attractions passenger flow |
| CN106960195A (en) * | 2017-03-27 | 2017-07-18 | 深圳市丰巨泰科电子有限公司 | A kind of people counting method and device based on deep learning |
| CN107368790A (en) * | 2017-06-27 | 2017-11-21 | 上海汇纳信息科技股份有限公司 | Pedestrian detection method, system, computer-readable recording medium and electronic equipment |
Non-Patent Citations (2)
| Title |
|---|
| 刘建国 等: "基于 YOLOv2 的行人检测方法研究", 《数字制造科学》 * |
| 张卿: "基于视频的人流检测与分析研究", 《中国优秀硕士学位论文全文数据库信息科技辑》 * |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111553737A (en) * | 2020-04-30 | 2020-08-18 | 北京金山安全软件有限公司 | Business time prediction method and device |
| CN111553737B (en) * | 2020-04-30 | 2023-08-22 | 北京金山安全软件有限公司 | Business time prediction method and device |
| CN113128831A (en) * | 2021-03-11 | 2021-07-16 | 特斯联科技集团有限公司 | People flow guiding method and device based on edge calculation, computer equipment and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US10735694B2 (en) | System and method for activity monitoring using video data | |
| CN103985250B (en) | The holographic road traffic state vision inspection apparatus of lightweight | |
| CN104215249B (en) | Smoothening method of driving track | |
| CN108322891B (en) | Traffic area congestion identification method based on user mobile phone signaling | |
| US20100286899A1 (en) | Combining Road and Vehicle Traffic Information | |
| CN109446920B (en) | Urban track traffic passenger's degree of crowding detection method based on convolutional neural networks | |
| CN105628951A (en) | Method and device for measuring object speed | |
| CN104408746B (en) | A kind of passenger flow statistical system based on depth information | |
| CN110874583A (en) | Passenger flow statistics method and device, storage medium and electronic equipment | |
| CN110692083A (en) | Block-matched optical flow and stereo vision for dynamic vision sensors | |
| CN104517125B (en) | The image method for real time tracking and system of high-speed object | |
| CN110264495A (en) | A kind of method for tracking target and device | |
| CN103164711A (en) | Regional people stream density estimation method based on pixels and support vector machine (SVM) | |
| CN106251695A (en) | Destination's parking stall intelligent recommendation system and method based on parking space state monitoring | |
| CN109583505A (en) | A kind of object correlating method, device, equipment and the medium of multisensor | |
| CN108171993A (en) | A kind of vehicle on highway speed calculation method based on mobile phone signaling big data | |
| CN105956632A (en) | Target detection method and device | |
| CN106558051A (en) | A kind of improved method for detecting road from single image | |
| CN104794790B (en) | A kind of scenic spot visitor statistical method | |
| CN108471497A (en) | A kind of ship target real-time detection method based on monopod video camera | |
| Zhang et al. | Vehicle re-identification for lane-level travel time estimations on congested urban road networks using video images | |
| CN109389016A (en) | A kind of method and system that the number of people counts | |
| CN108549865A (en) | A kind of people streams in public places amount adjusting householder method and system based on deep learning | |
| CN107264570A (en) | steel rail light band distribution detecting device and method | |
| CN110473428A (en) | A kind of intelligent parking method, apparatus and system |
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 | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20180918 |