CN109697435A - Stream of people's quantity monitoring method, device, storage medium and equipment - Google Patents
Stream of people's quantity monitoring method, device, storage medium and equipment Download PDFInfo
- Publication number
- CN109697435A CN109697435A CN201910012764.5A CN201910012764A CN109697435A CN 109697435 A CN109697435 A CN 109697435A CN 201910012764 A CN201910012764 A CN 201910012764A CN 109697435 A CN109697435 A CN 109697435A
- Authority
- CN
- China
- Prior art keywords
- crowd
- people
- region
- model
- image
- 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.)
- Granted
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
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Artificial Intelligence (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Computational Biology (AREA)
- General Engineering & Computer Science (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Multimedia (AREA)
- Image Analysis (AREA)
Abstract
The present invention provides a kind of stream of people's quantity monitoring method, device, storage medium and equipment, is suitable for technical field of image processing.This method comprises: obtaining the target image of pedestrian to be monitored in video image；Utilize the model extraction graphic feature based on depth residual error network training；Using crowd monitoring category of model and orient the no man's land in the target image, single region and crowd region；The number of people distribution density figure that crowd monitoring is classified as crowd region is obtained using density regression model, and number in crowd region is calculated according to the number of people distribution density figure；Count the crowd monitoring be classified as in single region with crowd region in number, calculate total number of persons in the target image.The present invention counts the number in video image based on depth residual error network integration crowd monitoring and density analysis when flow of the people is monitored, can accurately and rapidly count in number in image, there is preferable robustness.
Description
Technical field
The present invention relates to information technology fields, more particularly to a kind of stream of people's quantity monitoring method, device, storage medium and set
It is standby.
Background technique
Recent years, crowd's counting technology are always the research hotspot that industry is concerned, it is also gradually applied to major quotient
Field chain store, supermarket, hotel, airport, subway, scenic spot etc., the flow of the people data generated under these scenes can be many fields
Of great value information is provided.For each emporium chain store, supermarket, the e-commerce system on line burning hot at present
It unites, such as Jingdone district, Taobao, day cat, Amazon, it is clearly to mention that the selling market under line, which is constantly subjected to crowded, scientific management,
The effective means of itself high competitiveness.Different periods in market, different zones stream of people's data in the section for improving its business decision
The property learned, comfort of the reasonability of scheduling of resource, consumer environment etc. play an important role, and business stream of people's data are to quotient
The performance appraisal of industry, commodity conversion rate, shop addressing, commodity display, advertisement value have critically important meaning.In addition, for
In the public places such as exhibition center, gymnasium, subway station, bus station, airport, the region of real-time and precise can be presented in stream of people's data
Number and crowd density, manager analyze dynamic adjustment staff by data and configure plan, and control area crowd's quantity makes
Resource more reasonable employment, while can also reinforce safety precaution.
However, traditional people flow rate statistical monitors precision according to can all influence it in dense population, a variety of different shielding lights,
For under sparse scene, it is worse that flow of the people precision is counted to the dense population for possessing complex background illumination.
Summary of the invention
In view of the foregoing deficiencies of prior art, the purpose of the present invention is to provide a kind of stream of people's quantity monitoring methods, dress
It sets, storage medium and equipment, for solving in the prior art in crowd monitoring and density analysis flow of the people, in complex background light
According to the problem for being directed to dense population statistics flow of the people inaccuracy under shade.
In order to achieve the above objects and other related objects, the application's in a first aspect, present invention provides a kind of flow of the people prison
Survey method, comprising:
Obtain the target image of pedestrian to be monitored in video image；
Target image characteristics described in model extraction based on depth residual error network training；
Using crowd monitoring category of model and orient the no man's land in the target image, single region and crowd area
Domain；
Number of people distribution density figure in crowd monitoring module in crowd region is obtained using density regression model, according to described
Number of people distribution density figure calculates number in crowd region；
It counts in the single region with number in crowd region, calculates total number of persons in the target image.
The second aspect of the application provides a kind of visitors flowrate monitor, comprising:
Image collection module, for obtaining the target image of pedestrian to be monitored in video image；
Characteristic extracting module extracts the target image characteristics using depth residual error network；
Crowd monitoring module using crowd monitoring category of model and orients the no man's land in the target image, list
People region and crowd region；
It is close to obtain the number of people distribution in crowd monitoring module in crowd region using density regression model for density regression block
Degree figure calculates number in crowd region according to the number of people distribution density figure；
Demographics module calculates the target image for counting in the single region with number in crowd region
Interior total number of persons.
The third aspect of the application provides a kind of storage medium, is stored with computer-readable instruction, can make at least one
Processor executes the process described above.
The fourth aspect of the application provides a kind of flow of the people monitoring device, comprising:
One or more processors；
Memory；And
One or more programs, wherein one or more of programs be stored in the memory and be configured as by
One or more of processors execute instruction, and execute instruction described in one or more of processors execution so that the electronics
Equipment executes above-described stream of people's quantity monitoring method.
As described above, stream of people's quantity monitoring method, device, storage medium and equipment of the invention, have the advantages that
The present invention is when flow of the people monitors, based on depth residual error network integration crowd monitoring and density analysis to video image
Interior number is counted, since the feature extraction based on depth residual error network is not influenced vulnerable to the external world, it is ensured that extract number of people shoulder
Feature is more accurate, meanwhile, to Classification and Identification is carried out in image in the way of crowd monitoring, for different densities region using not
With demographics mode, can accurately and rapidly count in number in image, there is preferable robustness.
Detailed description of the invention
Fig. 1 is shown as a kind of stream of people's quantity monitoring method flow chart provided by the invention；
Fig. 2 is shown as the flow chart of training crowd's detection model in a kind of stream of people's quantity monitoring method provided by the invention；
Fig. 3 is shown as a kind of visitors flowrate monitor structural block diagram provided by the invention；
Fig. 4 is shown as training crowd's detection module structural block diagram in a kind of visitors flowrate monitor provided by the invention；
Fig. 5 is shown as a kind of flow of the people monitoring device structural schematic diagram provided by the invention.
Specific embodiment
Following will be combined with the drawings in the embodiments of the present invention, and technical solution in the embodiment of the present invention carries out clear, complete
Site preparation description, it is clear that the described embodiment is only a part of the embodiment of the present invention, instead of all the embodiments.Based on this
Embodiment in invention, every other reality obtained by those of ordinary skill in the art without making creative efforts
Example is applied, shall fall within the protection scope of the present invention.
Before to the embodiment of the present invention carrying out that explanation is explained in detail, first to the present embodiments relate to some names arrived
Word is explained.
Deep learning: this concept is derived from the research to artificial neural network.For example the multilayer perceptron containing more hidden layers is
For a kind of deep learning structure.Wherein, deep learning forms more abstract high-level characteristic by combination low-level feature, to excavate
The distributed nature of data indicates.
A kind of expression way is changed, deep learning is a kind of based on the method for carrying out representative learning to data.For observation
Various ways can be used to indicate in (such as piece image).For example it can be carried out with the vector of pixel intensity value each in the image
It indicates, or the image can be more abstractively expressed as to a series of sides, region of specific shape etc..And use certain specific tables
Show that method can be easier to carry out tasking learning from example, for example carries out recognition of face or human facial expression recognition etc..Wherein, depth
The benefit of study is that the feature learning and layered characteristic using non-supervisory formula or supervised extract highly effective algorithm to substitute manual obtain
Take feature.
Depth residual error network (ResNet): the depth of neural network is extremely important to its performance, therefore in the ideal case,
As long as over-fitting, depth should not be more deeper better to network.But one for encountering in hands-on neural network is excellent
The problem of change, i.e., with the continuous intensification of the depth of neural network, gradient gets over evanescence (i.e. gradient disperse) in the backward, it is difficult to
Optimized model instead results in the accuracy rate decline of network.A kind of expression way is changed, in the depth for being continuously increased neural network,
The problem of will appear Degradation (re-forming), i.e. accuracy rate, can first rise and then reach saturation, then continue to increase depth
Degree then will lead to accuracy rate decline.
Based on foregoing description it is found that the performance of network will be saturated, then be increased after the number that the network number of plies reaches certain
The performance of plus depth network will start to degenerate, but it is this degeneration be not as caused by over-fitting because training precision and
Measuring accuracy is all declining, and after network reaches certain depth, neural network is just difficult to have trained this explanation.And ResNet
Appearance is to solve the problems, such as that network depth is deepening later performance degradation.Specifically, ResNet proposes a depth
Residual error learn (deep residual learning) frame come solve the problems, such as it is this because depth increase and lead to performance degradation.
Assuming that having the accuracy rate for having reached saturation than shallower network, then along with several behind this network
A congruence mapping layer (Identity mapping), playing code error will not increase, i.e., deeper network should not bring training set
The rising of upper error.And the thought mentioned herein that preceding layer output is directly passed to back layer using congruence mapping, it is
The Inspiration Sources of ResNet.
Implementation environment involved in face retrieval method provided in an embodiment of the present invention is introduced below.
Embodiment 1
Referring to Fig. 1, being shown as a kind of stream of people's quantity monitoring method flow chart provided by the invention, details are as follows:
Step S101 obtains the target image of pedestrian to be monitored in video image；
Wherein, the video image source of acquisition can be the camera being mounted on everywhere, for example be mounted on market, station etc.
The image information of the corresponding region of camera acquisition at public place, can also be certain video images.
Step S102 extracts the target image characteristics using depth residual error network；
Wherein, based on each residual block being sequentially connected in depth residual error network, feature is carried out to the target image and is mentioned
It takes, the density map information of single number of people shoulder, the box in crowd region, confidence level and target image is obtained according to network structure；Appoint
Anticipating in a residual block includes an identical mapping and at least two convolutional layers, and the identical mapping of any one residual block is by institute
The input terminal for stating any one residual block is directed toward the output end of any one residual block；
Specifically, the target image is inputted to first residual block of the depth residual error network；For any one
Residual block, receives the output of the upper residual block, and based on first convolutional layer, second convolutional layer and described
Third convolutional layer carries out feature extraction to the output of a upper residual block；The output for obtaining the third convolutional layer, by institute
The output of the output and a upper residual block of stating third convolutional layer is transmitted to next residual block；It is residual to obtain the depth
The output of the last one residual block of poor network obtains people's head and shoulder feature of the target image.
Specifically, since ResNet network introduces residual error network structure, as carry out feature extraction algorithm, with compared with
The deep network number of plies learns people's head and shoulder feature, has obtained more accurate monitoring effect, meanwhile, it solves because of network number of plies mistake
Gradient disperse problem caused by depth, can with deeper network structure carry out target image feature learning, it is ensured that number
The accuracy of statistics.
Step S103 using crowd monitoring category of model and orients the no man's land in the target image, single area
Domain and crowd region；
Single head and shoulder region and intensive crowd region are obtained by crowd monitoring model, when inputting original image,
Using full convolutional network, classification and Detection goes out the region or nobody region and in the single head and shoulder region in original image, crowd.
Specifically, by being divided to the crowd is dense in target image degree, by the target image of input be converted into it is multiple not
The region of same type individually counts convenient for subsequent for each region.
Step S104 obtains the number of people distribution density figure in crowd region using density regression model, according to the number of people
Distribution density figure calculates number in crowd region；
The Gaussian Profile for calculating the number of people shoulder in target image crowd region obtains being distributed based on the two-dimensional number of people close
Degree figure；Calculate the number that number of people sum in the number of people distribution map obtains crowd region；Wherein, the number of people distribution density figure is adopted
With impulse function convolution Gauss kernel representation.
Step S105 counts in the single region with number in crowd region, calculates total number of persons in the target image.
Wherein, different number statisticals are respectively adopted from sparse region for close quarters, unite respectively by region division
The number of each corresponding region is counted out, total number of persons in target image directly finally is can be obtained into number addition in each region.
In the present embodiment, by counting people respectively to each region delimited in full figure (target image or original image)
Number, wherein density map by the location information (position coordinates) in detection original image come, that is, using rectangle frame, pass through meter
The density in the rectangle frame is calculated, to obtain the number in rectangle frame region；Make for the dense population for having illumination shade
Can accurate statistics go out its corresponding number, improve crowd's estimated capacity, more can really reflect current crowd's number
Amount, since different number statisticals are respectively adopted by crowd density difference in the picture, Monitoring Population people more rapidly and accurately
Flow.
Embodiment two
Referring to Fig. 2, being a kind of flow chart of stream of people's quantity monitoring method training of human group's detection model provided by the invention, in detail
It states as follows:
Step S201 is labeled the crowd region of multiple sample images, when being single image in the sample image
When mark number of people shoulder, when in the sample image being crowd's image, frame is marked by group, and according to the mark of each sample image
Region constructs crowd monitoring model；
Step S202 is trained the crowd monitoring model by multiple training samples, and generation can be according to crowd
Feature realizes the crowd monitoring model of territorial classification positioning in the target image.
Specifically, in the training model, need to prepare data to be detected mark, that is, sample data (comprising it is a variety of not
With the sample image of density), for example, by the sample image of input by clearly number of people shoulder area marking (in the form of confining mark
It is fixed), label classification is 1；Crowd's mark that crowd density in sample image is concentrated, and marking classification is 2；And it is left in sample
Other regions be the zone marker classification of unmanned head and shoulder be e.g. 0, and do not mark.
In training density regression model, the end-to-end training of full figure can be used, seek Gauss for the number of people region of full figure
Distribution obtains a two-dimensional number of people distribution density figure, and number in full figure can be obtained in the addition of all values of the density map.
In addition, density map mark needs to be generated offline or online according to sampling multiple preset under network structure, training is learnt
An obtained number of people density profile, wherein the density map is defined by the way of impulse function convolution Gaussian kernel:
D indicates that final densities figure, N indicate number of people number in formula, and x indicates picture point, and xi indicates that number of people location point, δ indicate
Impulse function, G indicate gaussian kernel function.
For above-mentioned training, caffe algorithm also can be used, crowd monitoring model and density regression model, example is respectively trained
Such as, first train crowd's detection model, fix its corresponding network part, retraining density regression model (can be used Loss come into
Row supervised training, and softmax can be used in the Loss that classifies).
In addition, may be used also for the people's head and shoulder feature extracted in the target image are as follows:
It establishes training sample set: obtaining video monitoring frame image, a variety of pretreatments are carried out to acquired image, are adopted simultaneously
Manually mode determines crowd's quantity in image range；
According to Resnet Block come the network structure built, Block is stacked up Resnet network one by one
Residual error network structure, and network structure uses full convolutional network network, does not have to full articulamentum, and the presently disclosed net of Resnet
Network structure is least 18 layers, and there are many convolutional layer number of entire convolutional network structure, therefore, Resnet network at least one
A Block structure；
Training convolutional neural networks model: after initialization, using stochastic gradient descent method to the convolutional neural networks of building
Model is iterated training, the value of every one subgradient of iteration one-time detection and loss function, to obtain in network architecture
The optimal solution of each weighted value W and bias b obtain the optimal convolutional neural networks model of this training after iteration is multiple；
By the convolutional neural networks disaggregated model about far and near two subregions by obtaining, according to detection classification policy pair
The crowd density of overall region is estimated.
Embodiment three
Referring to Fig. 3, being a kind of visitors flowrate monitor structural block diagram provided by the invention, details are as follows:
Image collection module 31, for obtaining the target image of pedestrian to be monitored in video image；
Characteristic extracting module 32 extracts feature in the target image using depth residual error network；
Wherein, based on each residual block being sequentially connected in depth residual error network, feature is carried out to the target image and is mentioned
It takes, includes an identical mapping and at least two convolutional layers in any one residual block, the identical of any one residual block reflects
Penetrate the output end that any one residual block is directed toward by the input terminal of any one residual block.
Crowd monitoring module 33, using crowd monitoring category of model and orient the no man's land in the target image,
Single region and crowd region；
Wherein, the sample areas of multiple sample images is labeled, and according to the sample of each sample image
Region constructs crowd monitoring model；
The crowd monitoring model is trained by multiple training samples, generation can be according to people's head and shoulder feature in mesh
The crowd monitoring model of territorial classification is realized in logo image.
Density regression block 34 obtains the number of people distribution density figure in crowd region using density regression model, according to institute
It states number of people distribution density figure and calculates number in crowd region；
Wherein, the Gaussian Profile for calculating the number of people shoulder in target image crowd region is obtained based on the two-dimensional number of people
Distribution density figure；Calculate the number that number of people sum in the number of people distribution map obtains crowd region；Wherein, the number of people distribution is close
Degree figure uses impulse function convolution Gauss kernel representation.
Demographics module 35 calculates the target figure for counting in the single region with number in crowd region
As interior total number of persons.
Referring to Fig. 4, being crowd monitoring modular structure block diagram in a kind of visitors flowrate monitor provided by the invention, it is described in detail
It is as follows:
Model foundation unit 331 is labeled, when in the sample image for the crowd region to multiple sample images
Number of people shoulder is infused for single image markers, frame is marked by group when in the sample image being crowd's image, and according to each sample
The tab area of image constructs crowd monitoring model；
Model training unit 332 generates energy for being trained by multiple training samples to the crowd monitoring model
Enough crowd monitoring models for realizing territorial classification positioning in the target image according to crowd characteristic.
In embodiments of the present invention, each unit of number monitoring device can be realized by corresponding hardware or software unit, respectively
Unit can be independent soft and hardware unit, also can integrate as a soft and hardware unit, herein not to limit the present invention.
The specific embodiment of each unit can refer to the description of embodiment one, and details are not described herein.
Example IV
The structure that Fig. 5 shows stream of people's monitoring device of the offer of the embodiment of the present invention four illustrates only for ease of description
Part related to the embodiment of the present invention.
Stream of people's monitoring device 5 of the embodiment of the present invention includes processor 50, memory 51 and is stored in memory 51
And the computer program 52 that can be run on processor 50.The processor 50 realizes above-mentioned flow of the people when executing computer program 52
Step in monitoring method embodiment, such as step S101 to S105 shown in FIG. 1.Alternatively, processor 50 executes computer journey
The function of each unit in above-mentioned each Installation practice, such as the function of unit 31 to 35 shown in Fig. 3 are realized when sequence 52.
In embodiments of the present invention, by distinguishing statistical number of person to each region delimited in full figure (target image), i.e.,
Make for have the dense population of illumination shade also can accurate statistics go out its corresponding number, improve crowd's estimated capacity,
Current crowd's quantity more can be really reflected, is united since different numbers are respectively adopted by crowd density difference in the picture
Meter mode, Monitoring Population flow of the people more rapidly and accurately.
The calculating equipment of the embodiment of the present invention can be personal computer, smart phone, plate.In the calculating equipment 5
The step of reason device 50 is realized when realizing the fuzzy processing method of motion blur image when executing computer program 52 can refer to aforementioned
The description of embodiment of the method, details are not described herein.
Embodiment five
In embodiments of the present invention, a kind of computer readable storage medium is provided, which deposits
Computer program is contained, which realizes the step in the above-mentioned stream of people's quantity monitoring method embodiment when being executed by processor
Suddenly, for example, step S101 to S105 shown in FIG. 1.Alternatively, the computer program realizes above-mentioned each device when being executed by processor
The function of each unit in embodiment, such as the function of unit 31 to 35 shown in Fig. 3.
In embodiments of the present invention, by distinguishing statistical number of person to each region delimited in full figure (target image), i.e.,
Make for have the dense population of illumination shade also can accurate statistics go out its corresponding number, improve crowd's estimated capacity,
Current crowd's quantity more can be really reflected, is united since different numbers are respectively adopted by crowd density difference in the picture
Meter mode, Monitoring Population flow of the people more rapidly and accurately.
The computer readable storage medium of the embodiment of the present invention may include can carry computer program code any
Entity or device, recording medium, for example, the memories such as ROM/RAM, disk, CD, flash memory.
In addition, each embodiment of the application can pass through the data processing journey by data processing equipment such as computer execution
Sequence is realized.Obviously, data processor constitutes the application.In addition, being commonly stored at data in one storage medium
Reason program is by directly reading out storage medium for program or by installing or copying to depositing for data processing equipment for program
It stores up in equipment (such as hard disk and/or memory) and executes.Therefore, such storage medium also constitutes the application, present invention also provides
A kind of non-volatile memory medium, wherein being stored with data processor, this data processor can be used for executing the application
Any one of above method embodiment embodiment.
In conclusion the present invention is based on depth residual error network integration crowd monitoring and density analysis when flow of the people monitors
Number in video image is counted, since the feature extraction based on depth residual error network is not influenced vulnerable to the external world, it is ensured that
It is more accurate to extract people's head and shoulder feature, meanwhile, to Classification and Identification is carried out in image in the way of crowd monitoring, for different densities
Region uses different number statisticals, can accurately and rapidly count in number in image, there is preferable robustness.Institute
With the present invention effectively overcomes various shortcoming in the prior art and has high industrial utilization value.
The above-described embodiments merely illustrate the principles and effects of the present invention, and is not intended to limit the present invention.It is any ripe
The personage for knowing this technology all without departing from the spirit and scope of the present invention, carries out modifications and changes to above-described embodiment.Cause
This, institute is complete without departing from the spirit and technical ideas disclosed in the present invention by those of ordinary skill in the art such as
At all equivalent modifications or change, should be covered by the claims of the present invention.
Claims (10)
1. a kind of stream of people's quantity monitoring method, which is characterized in that the described method comprises the following steps:
Obtain the target image of pedestrian to be monitored in video image；
Target image characteristics described in model extraction based on depth residual error network training；
Using crowd monitoring category of model and orient the no man's land in the target image, single region and crowd region；
Number of people distribution density figure in crowd monitoring module in crowd region is obtained using density regression model, according to the number of people
Distribution density figure calculates number in crowd region；
It counts in the single region with number in crowd region, calculates total number of persons in the target image.
2. stream of people's quantity monitoring method according to claim 1, which is characterized in that described to be instructed using based on depth residual error network
Target image characteristics step described in experienced model extraction, comprising:
Based on each residual block being sequentially connected in depth residual error network, feature extraction is carried out to the target image, it is any one
It include an identical mapping and at least two convolutional layers in a residual block, the identical mapping of any one residual block is by described
The input terminal an of residual block of anticipating is directed toward the output end of any one residual block.
3. stream of people's quantity monitoring method according to claim 1, which is characterized in that the training package of the crowd monitoring model
It includes:
The crowd region of multiple sample images is labeled, when being single image markers note number of people shoulder in the sample image,
When in the sample image being crowd's image, frame is marked by group, and constructs crowd's inspection according to the tab area of each sample image
Survey model；
The crowd monitoring model is trained by multiple training samples, generation can be according to crowd characteristic in target image
The middle crowd monitoring model for realizing territorial classification positioning.
4. stream of people's quantity monitoring method according to claim 1, which is characterized in that described to obtain people using density regression model
Group region in number of people distribution density figure, according to the number of people distribution density figure calculate crowd region in number the step of, comprising:
Demarcate the number of people shoulder position in target image crowd region, be calculated based on two-dimensional number of people distribution density figure into
Row model training；Based on the crowd region of the crowd monitoring model inspection, the number of people is obtained according to trained density regression model
Distribution density figure；Calculate the number in its region number of people distribution map；Wherein, the number of people distribution density figure is rolled up using impulse function
Product Gauss kernel representation, the statistics of number is by the way of to the summation of density graph region.
5. a kind of visitors flowrate monitor, which is characterized in that described device includes:
Image collection module, for obtaining the target image of pedestrian to be monitored in video image；
Characteristic extracting module extracts the target image characteristics using depth residual error network；
Crowd monitoring module using crowd monitoring category of model and orients the no man's land in the target image, single area
Domain and crowd region；
Density regression block obtains the number of people distribution density in crowd monitoring module in crowd region using density regression model
Figure calculates number in crowd region according to the number of people distribution density figure；
Demographics module calculates total in the target image for counting in the single region with number in crowd region
Number.
6. visitors flowrate monitor according to claim 5, which is characterized in that the characteristic extracting module is further wrapped
It includes:
Based on each residual block being sequentially connected in depth residual error network, feature extraction is carried out to the target image, it is any one
It include an identical mapping and at least two convolutional layers in a residual block, the identical mapping of any one residual block is by described
The input terminal an of residual block of anticipating is directed toward the output end of any one residual block.
7. visitors flowrate monitor according to claim 5, which is characterized in that the training package of the crowd monitoring model
It includes:
Model foundation unit is labeled for the crowd region to multiple sample images, when being one in the sample image
Number of people shoulder is marked when image, frame is marked by group when in the sample image being crowd's image, and according to each sample image
Tab area constructs crowd monitoring model；
Model training unit, for being trained by multiple training samples to the crowd monitoring model, generation being capable of basis
Crowd characteristic realizes the crowd monitoring model of territorial classification positioning in the target image.
8. visitors flowrate monitor according to claim 5, which is characterized in that the density regression block is further wrapped
It includes:
Demarcate the number of people shoulder position in target image crowd region, be calculated based on two-dimensional number of people distribution density figure into
Row model training；Based on the crowd region of the crowd monitoring model inspection, the number of people is obtained according to trained density regression model
Distribution density figure；Calculate the number in its region number of people distribution map；Wherein, the number of people distribution density figure is rolled up using impulse function
Product Gauss kernel representation, the statistics of number is by the way of to the summation of density graph region.
9. a kind of flow of the people monitoring device, which is characterized in that the equipment includes:
One or more processors；
Memory；And
One or more programs, wherein one or more of programs are stored in the memory and are configured as by described
One or more processors execute instruction, and execute instruction described in one or more of processors execution so that the electronic equipment
Execute stream of people's quantity monitoring method as described in any one of claims 1 to 4.
10. a kind of storage medium, which is characterized in that be stored with machine readable instructions, at least one processor is made to execute such as right
It is required that stream of people's quantity monitoring method described in any one of 1-4.
Applications Claiming Priority (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2018115292443 | 2018-12-14 | ||
| CN201811529244 | 2018-12-14 |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109697435A true CN109697435A (en) | 2019-04-30 |
| CN109697435B CN109697435B (en) | 2020-10-23 |
Family
ID=66232642
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910012764.5A Active CN109697435B (en) | 2018-12-14 | 2019-01-07 | People flow monitoring method and device, storage medium and equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109697435B (en) |
Cited By (23)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110287929A (en) * | 2019-07-01 | 2019-09-27 | 腾讯科技（深圳）有限公司 | The quantity of target determines method, apparatus, equipment and storage medium in group region |
| CN110852208A (en) * | 2019-10-29 | 2020-02-28 | 贵州民族大学 | Crowd density estimation method and readable storage medium |
| CN110855937A (en) * | 2019-09-25 | 2020-02-28 | 上海鑫湘机电设备工程有限公司 | Error control method of subway station hall people counting system |
| CN110874573A (en) * | 2019-10-30 | 2020-03-10 | 汇纳科技股份有限公司 | Density graph generation method and device based on residual error operation, electronic terminal and medium |
| CN111178276A (en) * | 2019-12-30 | 2020-05-19 | 上海商汤智能科技有限公司 | Image processing method, image processing apparatus, and computer-readable storage medium |
| CN111178235A (en) * | 2019-12-27 | 2020-05-19 | 卓尔智联(武汉)研究院有限公司 | Target quantity determination method, device, equipment and storage medium |
| CN111405239A (en) * | 2020-02-17 | 2020-07-10 | 浙江大华技术股份有限公司 | Monitoring method, server, monitoring system, and computer-readable storage medium |
| CN111639585A (en) * | 2020-05-21 | 2020-09-08 | 中国科学院重庆绿色智能技术研究院 | Self-adaptive crowd counting system and self-adaptive crowd counting method |
| CN111652161A (en) * | 2020-06-08 | 2020-09-11 | 上海商汤智能科技有限公司 | Crowd excess density prediction method and device, electronic equipment and storage medium |
| CN111652168A (en) * | 2020-06-09 | 2020-09-11 | 腾讯科技（深圳）有限公司 | Group detection method, device and equipment based on artificial intelligence and storage medium |
| CN112001274A (en) * | 2020-08-06 | 2020-11-27 | 腾讯科技（深圳）有限公司 | Crowd density determination method, device, storage medium and processor |
| CN112115900A (en) * | 2020-09-24 | 2020-12-22 | 腾讯科技（深圳）有限公司 | Image processing method, device, equipment and storage medium |
| CN112232236A (en) * | 2020-10-20 | 2021-01-15 | 城云科技（中国）有限公司 | Pedestrian flow monitoring method and system, computer equipment and storage medium |
| CN112364788A (en) * | 2020-11-13 | 2021-02-12 | 润联软件系统(深圳)有限公司 | Monitoring video crowd quantity monitoring method based on deep learning and related components thereof |
| CN112446406A (en) * | 2019-09-05 | 2021-03-05 | 中移信息技术有限公司 | Emergency point determination method and device and storage medium |
| CN112990517A (en) * | 2019-12-12 | 2021-06-18 | 中移雄安信息通信科技有限公司 | Crowd distribution prediction method and system |
| CN113052039A (en) * | 2021-03-16 | 2021-06-29 | 北京邮电大学 | Method, system and server for detecting pedestrian density of traffic network |
| CN113255430A (en) * | 2021-03-31 | 2021-08-13 | 中交第二公路勘察设计研究院有限公司 | Method for detecting and counting crowd distribution in video based on deep learning |
| CN113869269A (en) * | 2021-10-13 | 2021-12-31 | 平安银行股份有限公司 | Activity site congestion degree detection method and device, electronic equipment and storage medium |
| CN113963318A (en) * | 2021-12-22 | 2022-01-21 | 北京的卢深视科技有限公司 | People flow statistical method and device, electronic equipment and storage medium |
| CN116012776A (en) * | 2022-12-09 | 2023-04-25 | 北京数原数字化城市研究中心 | Method and device for monitoring number of people, electronic equipment and readable storage medium |
| CN118097553A (en) * | 2024-03-01 | 2024-05-28 | 北京数原数字化城市研究中心 | Pedestrian number determining method and device and related equipment |
| CN113869269B (en) * | 2021-10-13 | 2024-09-27 | 平安银行股份有限公司 | Method and device for detecting crowdedness degree of activity site, electronic equipment and storage medium |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103049787A (en) * | 2011-10-11 | 2013-04-17 | 汉王科技股份有限公司 | People counting method and system based on head and shoulder features |
| CN106778502A (en) * | 2016-11-21 | 2017-05-31 | 华南理工大学 | A kind of people counting method based on depth residual error network |
| CN107506692A (en) * | 2017-07-21 | 2017-12-22 | 天津大学 | A kind of dense population based on deep learning counts and personnel's distribution estimation method |
| CN107909044A (en) * | 2017-11-22 | 2018-04-13 | 天津大学 | A kind of demographic method of combination convolutional neural networks and trajectory predictions |
| CN108717528A (en) * | 2018-05-15 | 2018-10-30 | 苏州平江历史街区保护整治有限责任公司 | A kind of global population analysis method of more strategies based on depth network |
-
2019
- 2019-01-07 CN CN201910012764.5A patent/CN109697435B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103049787A (en) * | 2011-10-11 | 2013-04-17 | 汉王科技股份有限公司 | People counting method and system based on head and shoulder features |
| CN106778502A (en) * | 2016-11-21 | 2017-05-31 | 华南理工大学 | A kind of people counting method based on depth residual error network |
| CN107506692A (en) * | 2017-07-21 | 2017-12-22 | 天津大学 | A kind of dense population based on deep learning counts and personnel's distribution estimation method |
| CN107909044A (en) * | 2017-11-22 | 2018-04-13 | 天津大学 | A kind of demographic method of combination convolutional neural networks and trajectory predictions |
| CN108717528A (en) * | 2018-05-15 | 2018-10-30 | 苏州平江历史街区保护整治有限责任公司 | A kind of global population analysis method of more strategies based on depth network |
Non-Patent Citations (3)
| Title |
|---|
| ZHANG E: "A fast and robust people counting method in video surveillance", 《COMPUTATIONAL INTELLIGENCE AND SECURITY, 2007 INTERNATIONAL CONFERENCE ON. IEEE》 * |
| 曾令科: "基于深度卷积神经网络的人群计数算法研究", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
| 韩亚伟: "动态场景监控系统中人数统计算法的研究", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
Cited By (34)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110287929B (en) * | 2019-07-01 | 2023-09-05 | 腾讯科技（深圳）有限公司 | Method, device, equipment and storage medium for determining number of targets in group area |
| CN110287929A (en) * | 2019-07-01 | 2019-09-27 | 腾讯科技（深圳）有限公司 | The quantity of target determines method, apparatus, equipment and storage medium in group region |
| CN112446406A (en) * | 2019-09-05 | 2021-03-05 | 中移信息技术有限公司 | Emergency point determination method and device and storage medium |
| CN110855937A (en) * | 2019-09-25 | 2020-02-28 | 上海鑫湘机电设备工程有限公司 | Error control method of subway station hall people counting system |
| CN110852208A (en) * | 2019-10-29 | 2020-02-28 | 贵州民族大学 | Crowd density estimation method and readable storage medium |
| CN110852208B (en) * | 2019-10-29 | 2023-06-02 | 贵州民族大学 | Crowd density estimation method and readable storage medium |
| CN110874573A (en) * | 2019-10-30 | 2020-03-10 | 汇纳科技股份有限公司 | Density graph generation method and device based on residual error operation, electronic terminal and medium |
| CN110874573B (en) * | 2019-10-30 | 2022-05-13 | 汇纳科技股份有限公司 | Density graph generation method and device based on residual error operation, electronic terminal and medium |
| CN112990517A (en) * | 2019-12-12 | 2021-06-18 | 中移雄安信息通信科技有限公司 | Crowd distribution prediction method and system |
| CN111178235A (en) * | 2019-12-27 | 2020-05-19 | 卓尔智联(武汉)研究院有限公司 | Target quantity determination method, device, equipment and storage medium |
| CN111178276B (en) * | 2019-12-30 | 2024-04-02 | 上海商汤智能科技有限公司 | Image processing method, image processing apparatus, and computer-readable storage medium |
| CN111178276A (en) * | 2019-12-30 | 2020-05-19 | 上海商汤智能科技有限公司 | Image processing method, image processing apparatus, and computer-readable storage medium |
| CN111405239A (en) * | 2020-02-17 | 2020-07-10 | 浙江大华技术股份有限公司 | Monitoring method, server, monitoring system, and computer-readable storage medium |
| CN111639585A (en) * | 2020-05-21 | 2020-09-08 | 中国科学院重庆绿色智能技术研究院 | Self-adaptive crowd counting system and self-adaptive crowd counting method |
| CN111652161A (en) * | 2020-06-08 | 2020-09-11 | 上海商汤智能科技有限公司 | Crowd excess density prediction method and device, electronic equipment and storage medium |
| CN111652168A (en) * | 2020-06-09 | 2020-09-11 | 腾讯科技（深圳）有限公司 | Group detection method, device and equipment based on artificial intelligence and storage medium |
| CN111652168B (en) * | 2020-06-09 | 2023-09-08 | 腾讯科技（深圳）有限公司 | Group detection method, device, equipment and storage medium based on artificial intelligence |
| CN112001274B (en) * | 2020-08-06 | 2023-11-17 | 腾讯科技（深圳）有限公司 | Crowd density determining method, device, storage medium and processor |
| CN112001274A (en) * | 2020-08-06 | 2020-11-27 | 腾讯科技（深圳）有限公司 | Crowd density determination method, device, storage medium and processor |
| CN112115900A (en) * | 2020-09-24 | 2020-12-22 | 腾讯科技（深圳）有限公司 | Image processing method, device, equipment and storage medium |
| CN112115900B (en) * | 2020-09-24 | 2024-04-30 | 腾讯科技（深圳）有限公司 | Image processing method, device, equipment and storage medium |
| CN112232236A (en) * | 2020-10-20 | 2021-01-15 | 城云科技（中国）有限公司 | Pedestrian flow monitoring method and system, computer equipment and storage medium |
| CN112232236B (en) * | 2020-10-20 | 2024-02-06 | 城云科技（中国）有限公司 | Pedestrian flow monitoring method, system, computer equipment and storage medium |
| CN112364788A (en) * | 2020-11-13 | 2021-02-12 | 润联软件系统(深圳)有限公司 | Monitoring video crowd quantity monitoring method based on deep learning and related components thereof |
| CN112364788B (en) * | 2020-11-13 | 2021-08-03 | 润联软件系统(深圳)有限公司 | Monitoring video crowd quantity monitoring method based on deep learning and related components thereof |
| CN113052039B (en) * | 2021-03-16 | 2022-12-02 | 北京邮电大学 | Method, system and server for detecting pedestrian density of traffic network |
| CN113052039A (en) * | 2021-03-16 | 2021-06-29 | 北京邮电大学 | Method, system and server for detecting pedestrian density of traffic network |
| CN113255430A (en) * | 2021-03-31 | 2021-08-13 | 中交第二公路勘察设计研究院有限公司 | Method for detecting and counting crowd distribution in video based on deep learning |
| CN113869269A (en) * | 2021-10-13 | 2021-12-31 | 平安银行股份有限公司 | Activity site congestion degree detection method and device, electronic equipment and storage medium |
| CN113869269B (en) * | 2021-10-13 | 2024-09-27 | 平安银行股份有限公司 | Method and device for detecting crowdedness degree of activity site, electronic equipment and storage medium |
| CN113963318A (en) * | 2021-12-22 | 2022-01-21 | 北京的卢深视科技有限公司 | People flow statistical method and device, electronic equipment and storage medium |
| CN116012776B (en) * | 2022-12-09 | 2024-02-23 | 北京数原数字化城市研究中心 | Method and device for monitoring number of people, electronic equipment and readable storage medium |
| CN116012776A (en) * | 2022-12-09 | 2023-04-25 | 北京数原数字化城市研究中心 | Method and device for monitoring number of people, electronic equipment and readable storage medium |
| CN118097553A (en) * | 2024-03-01 | 2024-05-28 | 北京数原数字化城市研究中心 | Pedestrian number determining method and device and related equipment |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109697435B (en) | 2020-10-23 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109697435A (en) | Stream of people's quantity monitoring method, device, storage medium and equipment | |
| Sun et al. | Using long short-term memory recurrent neural network in land cover classification on Landsat and Cropland data layer time series | |
| Li et al. | Deep learning-based classification methods for remote sensing images in urban built-up areas | |
| Zhao et al. | Classifying airborne LiDAR point clouds via deep features learned by a multi-scale convolutional neural network | |
| CN109359538A (en) | Training method, gesture identification method, device and the equipment of convolutional neural networks | |
| CN106845499A (en) | A kind of image object detection method semantic based on natural language | |
| Jia et al. | Fruit recognition based on pulse coupled neural network and genetic Elman algorithm application in apple harvesting robot | |
| Jiang et al. | GraphGST: Graph generative structure-aware transformer for hyperspectral image classification | |
| Kong et al. | Hierarchical attention and context modeling for group activity recognition | |
| CN109978074A (en) | Image aesthetic feeling and emotion joint classification method and system based on depth multi-task learning | |
| Xue et al. | Attention‐based spatial–temporal hierarchical ConvLSTM network for action recognition in videos | |
| Cheng et al. | Analysis and forecasting of the day-to-day travel demand variations for large-scale transportation networks: a deep learning approach | |
| Wang et al. | Quantifying legibility of indoor spaces using Deep Convolutional Neural Networks: Case studies in train stations | |
| Erhan et al. | Analyzing objective and subjective data in social sciences: Implications for smart cities | |
| Alashban et al. | Single convolutional neural network with three layers model for crowd density estimation | |
| Andersson et al. | Combining street-level and aerial images for dengue incidence rate estimation | |
| Zhang et al. | Object counting method based on dual attention network | |
| Gao et al. | Applying optimized YOLOv8 for heritage conservation: enhanced object detection in Jiangnan traditional private gardens | |
| Andersson et al. | Towards predicting dengue fever rates using convolutional neural networks and street-level images | |
| Wang et al. | Efficient convolutional neural architecture search for LiDAR DSM classification | |
| Xu et al. | Real-time target detection and recognition with deep convolutional networks for intelligent visual surveillance | |
| Zhang et al. | A Survey of Generative Techniques for Spatial-Temporal Data Mining | |
| Fu et al. | The comparison of different graph convolutional neural networks for image recognition | |
| Jin et al. | Learning context-aware region similarity with effective spatial normalization over Point-of-Interest data | |
| Zhong et al. | Multimodal cooperative self‐attention network for action recognition |
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 |