CN109446989A - Crowd massing detection method, device and storage medium - Google Patents
Crowd massing detection method, device and storage medium Download PDFInfo
- Publication number
- CN109446989A CN109446989A CN201811270525.1A CN201811270525A CN109446989A CN 109446989 A CN109446989 A CN 109446989A CN 201811270525 A CN201811270525 A CN 201811270525A CN 109446989 A CN109446989 A CN 109446989A
- Authority
- CN
- China
- Prior art keywords
- set image
- frame
- region
- crowd
- pedestrian area
- 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
- G06V20/00—Scenes; Scene-specific elements
- G06V20/40—Scenes; Scene-specific elements in video content
- G06V20/41—Higher-level, semantic clustering, classification or understanding of video scenes, e.g. detection, labelling or Markovian modelling of sport events or news items
- G06V20/42—Higher-level, semantic clustering, classification or understanding of video scenes, e.g. detection, labelling or Markovian modelling of sport events or news items of sport video content
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Theoretical Computer Science (AREA)
- Computational Linguistics (AREA)
- Software Systems (AREA)
- Image Analysis (AREA)
Abstract
The embodiment of the present application provides a kind of crowd massing detection method, device and storage medium, which comprises obtains the continuous multiframe pre-set image of target video；According to crowd density, pedestrian area is obtained from pre-set image described in every frame；Flowing pedestrian area is obtained from pre-set image described in every frame；According to the pedestrian area and the flowing pedestrian area, region to be confirmed is obtained；When the number that the goal-selling in the region to be confirmed occurs in the continuous multiple frames pre-set image reaches threshold value, there is default aggregation event in confirmation.It can accurately predict whether default aggregation event occur.
Description
Technical field
This application involves intelligent video-image monitoring field, in particular to a kind of crowd massing detection method, device and deposit
Storage media.
Background technique
The rapid development of computer technology provides bigger display stage for the application of video monitoring, and intelligence is in number
Also it is more and more applied in safety precaution field.It, can be to institute if the intellectual analysis of video is added in video monitoring
The video source of concern is analyzed in real time, effectively information is avoided to omit, and is greatly reduced labor management cost, is provided for city management
Effective guarantee.Currently, crowd massing event analysis at the biggish scene of flow of the people (such as subway, square), can all enter and leave normal
Crowd be mistaken for crowd massing event, to be unable to judge accurately crowd massing event.
Summary of the invention
The embodiment of the present application provides a kind of crowd massing detection method, device and storage medium, and crowd massing can be improved
The accuracy of event judgement.
The embodiment of the present application provides a kind of crowd massing detection method comprising:
Obtain the continuous multiframe pre-set image of target video；
According to crowd density, pedestrian area is obtained from pre-set image described in every frame；
Flowing pedestrian area is obtained from pre-set image described in every frame；
According to the pedestrian area and the flowing pedestrian area, region to be confirmed is obtained；
When the number that the goal-selling in the region to be confirmed occurs in the continuous multiple frames pre-set image reaches threshold
When value, there is default aggregation event in confirmation.
The embodiment of the present application provides a kind of crowd massing detection device comprising:
Pre-set image obtains module, for obtaining the continuous multiframe pre-set image of target video；
Pedestrian area obtains module, for obtaining pedestrian area from pre-set image described in every frame according to crowd density；
It flows pedestrian area and obtains module, for obtaining flowing pedestrian area from pre-set image described in every frame；
Region to be confirmed obtains module, for obtaining to be confirmed according to the pedestrian area and the flowing pedestrian area
Region；
Judgment module, time for occurring in the every frame image of the multiframe when the goal-selling in the region to be confirmed
When number reaches threshold value, there is default aggregation event in confirmation.
The embodiment of the present application also provides a kind of storage medium, computer program is stored in the storage medium, when described
When computer program is run on computers, so that the computer executes crowd massing detection method as described above.
In crowd massing detection method provided by the embodiments of the present application, device and storage medium, the default figure of every frame is first obtained
The pedestrian area of picture, then obtains the proper flow pedestrian area of every frame pre-set image, to obtain region to be confirmed, finally when
When the number that goal-selling in region to be confirmed occurs in continuous multiple frames image reaches threshold value, there is default aggregation thing in confirmation
Part.It can accurately predict whether default aggregation event occur.
Detailed description of the invention
In order to more clearly explain the technical solutions in the embodiments of the present application, make required in being described below to embodiment
Attached drawing is briefly described.It should be evident that the drawings in the following description are only some examples of the present application, for
For those skilled in the art, without creative efforts, it can also be obtained according to these attached drawings other attached
Figure.
Fig. 1 is the flow diagram of crowd massing detection method provided by the embodiments of the present application.
Fig. 2 is the flow diagram that pedestrian area provided by the embodiments of the present application obtains.
Fig. 3 is the flow diagram that flowing pedestrian area provided by the embodiments of the present application obtains.
Fig. 4 is the flow diagram that crowd massing region provided by the embodiments of the present application obtains.
Fig. 5 is schematic diagram before closed operation provided by the embodiments of the present application.
Fig. 6 is schematic diagram after closed operation provided by the embodiments of the present application.
Fig. 7 is multiple target tracking treatment process schematic diagram provided by the embodiments of the present application.
Fig. 8 is crowd massing detection device schematic diagram provided by the embodiments of the present application.
Fig. 9 is another schematic diagram of crowd massing detection device provided by the embodiments of the present application.
Specific embodiment
Below in conjunction with the attached drawing in the embodiment of the present application, technical solutions in the embodiments of the present application carries out clear, complete
Site preparation description.Obviously, described embodiments are only a part of embodiments of the present application, instead of all the embodiments.It is based on
Embodiment in the application, those skilled in the art's every other implementation obtained without creative efforts
Example, shall fall in the protection scope of this application.
Referring to Fig. 1, Fig. 1 is the flow diagram of crowd massing detection method provided by the embodiments of the present application, process can
To include:
101, obtain the continuous multiframe pre-set image of target video.(successive frame in video is handled frame by frame).
The embodiment of the present application is based on video frequency graphic monitoring, and video image includes continuous multiframe pre-set image, then from
The each frame pre-set image obtained frame by frame in the continuous multiple frames pre-set image of one video image.Wherein, video image can be
The image of the less monitoring scene of flow of the people (such as suburb) is also possible to the biggish scene of flow of the people (such as subway, square)
Image.
102, according to crowd density, pedestrian area is obtained from every frame pre-set image.
Continuous multiple frames pre-set image in video can be handled frame by frame.
In some embodiments, pedestrian area can be obtained from every frame pre-set image according to crowd density estimation algorithm.
In some embodiments, density of stream of people figure can be obtained from every frame pre-set image according to crowd density estimation algorithm,
Wherein, each pixel corresponds to pedestrian's probability of occurrence value in density of stream of people figure；Binary conversion treatment is carried out to each pixel and obtains pedestrian
Region.
Specifically, referring to Fig. 2, the step obtains density of stream of people figure and corresponding pedestrian area based on pre-set image.Core
Center algorithm is crowd density estimation algorithm, and density estimation algorithm is commonly used to carry out whole pre-set image people stream counting, input
Whole pre-set image can obtain the density of stream of people figure of whole pre-set image, density of stream of people figure by density of stream of people algorithm process
In each pixel correspond to pedestrian's probability of occurrence value, value range [0.0~1.0), sum to each pixel of density of stream of people figure
The number for obtaining whole pre-set image, carrying out binary conversion treatment to each pixel can get pedestrian area.
Wherein, crowd density estimation algorithm is made of training process, detection process, and training process is off-line training processing,
It is not repeated to train in actual operation, the detection model for crowd density estimation can be obtained by training, run
It then calls the detection model input picture to handle in the process, obtains density of stream of people figure, carrying out binary conversion treatment to density map can be with
Obtain pedestrian area.
Wherein training process is inputted as image pattern and corresponding calibration information, is exported as crowd's Density estimating model.It is first
First to collection in worksite video, processing obtains certain amount (such as 30,000) sample, formulates pedestrian's calibration rule and demarcates, then
Image calibration data are converted into data format required by trained frame；Then using training frame, sample data and place are read in
The label data obtained after reason, by loop iteration training, average every sample is iterated training 50 times, and final acquisition crowd is close
Degree estimation model.
Detection process, inputs as current frame image data, calls Density estimating model to image using density estimation algorithm
It is handled, is exported as the density of stream of people figure of whole figure.
Pedestrian area extraction process, the density of stream of people figure that detection process obtains, each pixel correspond to pedestrian's probability value, non-row
People region, probability value are usually 0, are threshold value with fixed value 0.0001, carry out binary conversion treatment, will be above the threshold pixels point value
255 are set to, the threshold pixels point value set 0 will be less than, the point for belonging to pedestrian area is combined as pedestrian area.
103, flowing pedestrian area is obtained from every frame pre-set image.
It in some embodiments, can be according to foreground extraction algorithm, from pre- described in pre-set image and every frame described in every frame
If obtaining background area in the pre-set image of the preceding first default frame number of image；According to the background area of pre-set image described in every frame
Domain obtains the flowing pedestrian area of every frame pre-set image.
In some embodiments, it is flat that Gauss can be carried out to pre-set image before the step of according to foreground extraction algorithm
The pretreatment of sliding and gray processing.
Specifically, referring to Fig. 3, the step is based on pre-set image, acquisition flowing pedestrian area.It needs pre- according to present frame
If image and the current time background image for combining past default frame number pre-set image study to update.Default frame number can be with
For 10 frames, 20 frames, or all frames of present frame.It is pre- Gaussian smoothing, gray processing can be carried out to every frame pre-set image first
Processing is then fed into foreground extraction algorithm process, obtains flowing pedestrian area.The core algorithm of the step uses mixed Gaussian mould
Type foreground extraction algorithm, detailed process include: to be established by counting each coordinate points pixel value variation for each point multiple random
Normal distribution, that is, background model judges that the point is background dot if the pixel value meets wherein any one normal distribution, and
And background model is updated, and otherwise it is considered flowing pedestrian area, traverses each point of image according to this logic, it is final all
The point for belonging to flowing pedestrian area combines as flowing pedestrian area.
104, according to pedestrian area and flowing pedestrian area, obtain region to be confirmed.
In some embodiments, it can specifically include:
By removing the flowing pedestrian area in the pedestrian area, the region corresponding second to be confirmed is obtained
Crowd density figure；
Morphological scale-space is carried out to the second crowd density figure, and extracts profile；
The boundary rectangle for extracting the profile obtains regional location to be confirmed；
The total number of persons in the region to be confirmed is obtained according to second density map；
When total number of persons is greater than the first preset threshold in the region to be confirmed, confirm that the region to be confirmed is people's clustering
Collect region.Specifically, referring to Fig. 4, this step is the pedestrian area obtained based on step 101 and the flowing that step 102 obtains
Pedestrian area obtains the boundary rectangle position of the pedestrian area of region to be confirmed and aggregation.
The pedestrian area that step 101 obtains includes all pedestrian areas of current scene, not only includes the pedestrian of aggregation,
It also include the pedestrian of proper flow, and the flowing pedestrian area that step 102 passes through foreground extraction only includes the pedestrian area of movement
(and moving region of other moving objects) traverses pedestrian area image each point, removes the motion parts (pixel in pedestrian area
0) value is set, remaining is crowd massing region；Crowd massing area coverage density map is further corresponded to, by density map corresponding points
Value, which is done to add up, can be obtained aggregation zone people's sum.
Then think the region as crowd massing region, Morphological scale-space when people's sum is higher than the first preset threshold T1 set
And profile is extracted, the boundary rectangle for further extracting profile is crowd massing zone position, otherwise then thinks that the region is positive
Normal region is without any processing.To all areas, successively processing terminate, will be determined as crowd massing zone position (boundary rectangle)
It is output in next step.Wherein Morphological scale-space operation is closed operation of progress.
The Morphological scale-space that wherein the present embodiment uses is closed operation, and closed operation can make the profile of object become smooth, energy
Narrow interruption and elongated wide gap are eliminated, while the small cavity that target internal can be filled and the fracture in image outline.It closes
Before operation as shown in figure 5, after closed operation shown in Fig. 6.
105, when the number that the goal-selling in region to be confirmed occurs in continuous multiple frames image reaches threshold value, confirmation
There is default aggregation event.
In some embodiments, which can specifically include: obtain the second default frame number before pre-set image described in every frame
At least one of pre-set image target to be detected；
When the goal-selling in the crowd massing regional location is identical as the target to be detected, and the goal-selling
When the number occurred in the described second default frame number pre-set image reaches threshold value, there is default aggregation event in confirmation.
Specifically, this step is the crowd massing regional location (boundary rectangle) and previous frame obtained based on step 104
The multiple target tracking of image obtains present frame multiple target tracking as a result, analyzing each target as a result, after multiple target matching treatment
The coherent condition of the continuous certain frame number in region makes final illegal aggregation event judgement.The module uses IOU matching strategy, due to
Substantially it does not move in crowd massing region, then it is assumed that the same crowd region appears in the essentially identical position of next frame, is based on
Target is divided into different target class to handle by class again, is illustrated by taking the processing of adjacent two frame as an example, other frames are successively updated by this rule.
Referring to Fig. 7, in some embodiments, passing through the target and previous frame multiple target tracking knot for detecting present frame
Fruit carries out the matching of IOU threshold value, the target and previous frame multiple target tracking result that present frame detects can be classified respectively.
The objective result that the present frame of acquisition detects can be divided into two classes: 1. target that class is detected for the first time；Class is 2. before
It is detected and has been tracked, this frame is detected again.Previous frame target following result can be divided into three kinds: class is 1. current
The target that frame is detected again；2. class is not detected in present frame, but continuously being not detected number does not reach drop threshold
T2；3. class is not detected in present frame, and continuously being not detected number reaches target drop threshold.
Handled by class: by class that present frame detects, 1. target is created as new tracking target and tracks, under
One frame can carry out first tracking prediction at the beginning and go out new target location；The class that present frame detects 2. with previous frame target following knot
1. class in fruit is same class target, then corrected using present frame testing result to tracking result, to update acquisition pair
It should previous frame tracking target information；Class in previous frame target following result is 2. due to being not up to drop threshold, then it is assumed that the mesh
Mark is still only not detected in monitoring area, and using monotrack, (this programme is thought since crowd massing region is constant
Previous frame position is current frame position) result be updated as present frame target information；Previous frame target following result
In class 3. reached drop threshold T2 since continuous multiple frames do not detect, therefore, it is considered that the target has disappeared and has been screened out not
It tracks again.By obtaining present frame multiple target tracking result after handling above.
Illegal aggregation event judgement is carried out, record has each target in the past period continuous one in multiple target tracking result
There is aggregation number in framing number T3, these information are stored by frame sequential, when in 3 frame of past continuous T appearance aggregation number
Then judge that crowd massing event occurs in the suspect region more than threshold value T4, then carries out illegal aggregation event and report.
It should be noted that the algorithm configuration in the embodiment of the present application includes many places threshold value, need to be carried out according to field demand
Setting, if default value can be enabled by being not provided with inside.Wherein, T1 be when people's sum be higher than the threshold value T1 that sets then think the region as
Crowd massing region；T2 be for deleting the aggregation crowd to have disappeared, when 2 frame of crowd's continuous T do not occur again then not with
Track；T3 is the frame number for retrospect needed for judging aggregation event；T4 is that aggregation event is gathered in continuous certain frame number moon T3
Collecting number is more than that threshold value T4 then just thinks illegal aggregation event occur.
The embodiment of the present application in machine vision technique density of stream of people estimation technique, foreground extraction technology, at morphology
Based on reason technology, multiple target matching correlation technology, make final judgement in conjunction with the state in continuous multiple frames crowd region in video flowing
To complete crowd massing event detection function.Specifically, the embodiment of the present application uses based on depth learning technology first
Crowd density estimation algorithm passes through training by a large amount of pedestrian samples by the calibration conversion of rule set by crowd density estimation algorithm
Study obtain crowd density estimation model, using call the Density estimating model detector to monitoring area image carry out crowd
Density estimation obtains crowd density estimation figure, further by obtaining pedestrian area to the density map binary conversion treatment；It is same with this
When, moving region in input figure (flowing pedestrian area) is extracted by foreground detection techniques, further to the motion diagram of acquisition two
Value processing obtains the pedestrian area of proper flow；Then, it is poor to be made by pedestrian area and moving region, eliminates in pedestrian area
Moving region (proper flow part in crowd), it is remaining be true assembled crowd region, only to the pedestrian area of aggregation
Domain is counted the number of people by scanning crowd density figure, judges whether current region occurs crowd massing according to the region number, and
Extracting the region boundary rectangle is crowd massing region；It finally designs and only matches correlation technology by the multiple target of foundation of position,
Each crowd massing region is carried out to be associated with tracking in timing, dynamic surveillance updates crowd region id information and records continuous state,
Judge to make final event according to the status information in the timing of crowd region.The embodiment of the present application is practical, adapts to scene,
And unusual precision is achieved in actual monitored video.
The embodiment of the present application extracts pedestrian area by the crowd density estimation algorithm of deep learning, rather than simple meter
Number.Single frames judgement different from the past, but illegal aggregation event judgement is carried out based on video continuous information, and be based on this Shen
Please embodiment multiple target match association algorithm, can to each crowd massing region carry out timing on be associated with match, dynamic surveillance.
Accuracy rate is high, and stability is good.
Referring to Fig. 8, the embodiment of the present application also provides a kind of crowd massing detection device, crowd massing detection device is main
Including 4 modules: crowd density estimation module, foreground extracting module, crowd massing area extraction module and illegal aggregation event are sentenced
Disconnected module.
Wherein, crowd density estimation module uses the crowd density estimation algorithm based on depth learning technology, will be big
Pedestrian sample is measured by the calibration conversion of rule set by crowd density estimation algorithm, learns to obtain crowd density estimation mould by training
Type is estimated using calling the detector of the Density estimating model to carry out crowd density estimation acquisition crowd density to monitoring area image
Meter figure, further by obtaining pedestrian area to the density map binary conversion treatment.Foreground extracting module passes through foreground detection techniques
Moving region in input figure is extracted, the pedestrian area that binary conversion treatment obtains proper flow further is carried out to the motion diagram of acquisition
Domain.Crowd massing area extraction module, it is poor to be made by pedestrian area and moving region, eliminates the moving region (people in pedestrian area
Proper flow part in group), remaining is true assembled crowd region, only passes through scanning crowd to the pedestrian area of aggregation
Density map counts the number of people, and judges whether current region occurs crowd massing according to the region number, and it is external to extract the region
Rectangle is crowd massing region.Illegal aggregation event judge module, design are right only using position as the Multitarget Tracking of foundation
Each crowd massing region carries out being associated with tracking in timing, dynamic surveillance, updates crowd region id information and records continuous each frame shape
State, to make final illegal aggregation event judgement according to the status information in the timing of crowd region.
Referring to Fig. 9, the embodiment of the present application also provides a kind of crowd massing detection device 300, crowd massing detection device
300 include that pre-set image obtains module 310, pedestrian area obtains module 320, flowing pedestrian area obtains module 330, to be confirmed
Region obtains module 340 and judgment module 350.
Wherein, pre-set image obtains module 310 for obtaining the continuous multiframe pre-set image of target video；.
Pedestrian area obtains module 320 and is used to obtain pedestrian area from every frame pre-set image according to crowd density.
It flows pedestrian area and obtains module 330 and be used to obtain from every frame pre-set image and flow pedestrian area.
Region to be confirmed obtains module 340 and is used to obtain region to be confirmed according to pedestrian area and flowing pedestrian area.
Judgment module 360 is used to reach when the number that the goal-selling in region to be confirmed occurs in multiframe pre-set image
When threshold value, there is default aggregation event in confirmation.
In some embodiments, pedestrian area obtains module 320 and is also used to according to crowd density estimation algorithm, pre- from every frame
If obtaining pedestrian area in image.
In some embodiments, pedestrian area obtains module 320 and is also used to according to crowd density estimation algorithm, pre- from every frame
If image obtains density of stream of people figure, wherein each pixel corresponds to pedestrian's probability of occurrence value in density of stream of people figure；Each pixel is clicked through
Row binary conversion treatment obtains pedestrian area.
In some embodiments, flowing pedestrian area obtains module 330 and is also used to according to foreground extraction algorithm, from every frame institute
It states in the pre-set image of the preceding first default frame number of pre-set image described in pre-set image and every frame and obtains background area；According to
The background area of pre-set image described in every frame obtains the flowing pedestrian area of pre-set image described in every frame.
In some embodiments, flowing pedestrian area obtain module 330 be also used to pre-set image carry out Gaussian smoothing, with
And gray processing pretreatment.
In some embodiments, region to be confirmed obtains module 340 and is also used to by removing the institute in the pedestrian area
Flowing pedestrian area is stated, the corresponding second crowd density figure in the region to be confirmed is obtained；To the second crowd density figure into
Row Morphological scale-space, and extract profile；The boundary rectangle for extracting the profile obtains regional location to be confirmed；According to described second
Density map obtains the total number of persons in the region to be confirmed；When total number of persons is greater than the first preset threshold in the region to be confirmed,
Confirm that the region to be confirmed is crowd massing region.
In some embodiments, judgment module 360 is also used to obtain the second default frame number before pre-set image described in every frame
At least one of pre-set image target to be detected；When in the crowd massing regional location goal-selling with it is described to be detected
Target is identical, and when the number that occurs in the described second default frame number pre-set image of the goal-selling reaches threshold value, confirmation
There is default aggregation event.
The embodiment of the present application also provides a kind of storage medium, computer program is stored in the storage medium, when described
When computer program is run on computers, the computer executes crowd massing detection side described in any of the above-described embodiment
Method, such as: obtain the continuous multiframe pre-set image of target video；According to crowd density, obtained from pre-set image described in every frame
Pedestrian area；Flowing pedestrian area is obtained from pre-set image described in every frame；According to the pedestrian area and the flowing pedestrian
Region obtains region to be confirmed；When time that the goal-selling in the region to be confirmed occurs in the multiframe pre-set image
When number reaches threshold value, there is default aggregation event in confirmation.
In the embodiment of the present application, storage medium can be magnetic disk, CD, read-only memory (Read OnlyMemory,
) or random access memory (RandomAccess Memory, RAM) etc. ROM.
In the above-described embodiments, it all emphasizes particularly on different fields to the description of each embodiment, there is no the portion being described in detail in some embodiment
Point, reference can be made to the related descriptions of other embodiments.
It should be noted that for the crowd massing detection method of the embodiment of the present application, this field common test personnel
It is understood that realize all or part of the process of the embodiment of the present application crowd massing detection method, is that can pass through computer program
It is completed to control relevant hardware, computer program can be stored in a computer-readable storage medium, be such as stored in electricity
It in the memory of sub- equipment, and is executed by least one processor in the electronic equipment, in the process of implementation may include as answered
With the process of the embodiment of program management-control method.Wherein, storage medium can be magnetic disk, CD, read-only memory, arbitrary access
Memory body etc..
For the crowd massing detection device of the embodiment of the present application, each functional module be can integrate in a processing core
In piece, it is also possible to modules and physically exists alone, can also be integrated in two or more modules in a module.On
It states integrated module both and can take the form of hardware realization, can also be realized in the form of software function module.Integrated
If module is realized and when sold or used as an independent product in the form of software function module, also can store at one
In computer-readable storage medium, storage medium is for example read-only memory, disk or CD etc..
Detailed Jie has been carried out to crowd massing detection method, device and storage medium provided by the embodiments of the present application above
It continues, specific examples are used herein to illustrate the principle and implementation manner of the present application, and the explanation of above embodiments is only
It is to be used to help understand the application.Meanwhile for those skilled in the art, according to the thought of the application, in specific embodiment party
There will be changes in formula and application range, in conclusion the contents of this specification should not be construed as limiting the present application.
Claims (10)
1. a kind of crowd massing detection method characterized by comprising
Obtain the continuous multiframe pre-set image of target video；
According to crowd density, pedestrian area is obtained from pre-set image described in every frame；
Flowing pedestrian area is obtained from pre-set image described in every frame；
According to the pedestrian area and the flowing pedestrian area, region to be confirmed is obtained；
When the number that the goal-selling in the region to be confirmed occurs in the multiframe pre-set image reaches threshold value, confirmation
There is default aggregation event.
2. crowd massing detection method according to claim 1, which is characterized in that it is described according to crowd density, from every frame
The step of pedestrian area is obtained in the pre-set image, comprising:
According to crowd density estimation algorithm, pedestrian area is obtained from pre-set image described in every frame.
3. crowd massing detection method according to claim 2, which is characterized in that described to be calculated according to crowd density estimation
Method, from pre-set image described in every frame the step of acquisition pedestrian area, comprising:
It is described according to crowd density estimation algorithm, obtain density of stream of people figure from pre-set image described in every frame, wherein the stream of people is close
Each pixel corresponds to pedestrian's probability of occurrence value in degree figure；
Binary conversion treatment is carried out to each pixel and obtains pedestrian area.
4. crowd massing detection method according to claim 1, which is characterized in that described from pre-set image described in every frame
The step of obtaining flowing pedestrian area, comprising:
According to foreground extraction algorithm, from the first default frame number before pre-set image described in pre-set image and every frame described in every frame
Pre-set image in obtain background area；
According to the background area of pre-set image described in every frame, the flowing pedestrian area of pre-set image described in every frame is obtained.
5. crowd massing detection method according to claim 4, which is characterized in that the step according to foreground extraction algorithm
Before rapid, further includes:
Gaussian smoothing and gray processing pretreatment are carried out to the pre-set image.
6. crowd massing detection method according to claim 1, which is characterized in that described according to the pedestrian area and institute
The step of stating flowing pedestrian area, obtaining region to be confirmed, comprising:
By removing the flowing pedestrian area in the pedestrian area, corresponding second crowd in the region to be confirmed is obtained
Density map；
Morphological scale-space is carried out to the second crowd density figure, and extracts profile；
The boundary rectangle for extracting the profile obtains regional location to be confirmed；
The total number of persons in the region to be confirmed is obtained according to second density map；
When total number of persons is greater than the first preset threshold in the region to be confirmed, confirm that the region to be confirmed is crowd massing area
Domain；
When the number that the goal-selling in the region to be confirmed occurs in the multiframe pre-set image reaches threshold value, confirmation
There is the step of default aggregation event, comprising:
When the number that the goal-selling in the crowd massing regional location occurs in the multiframe pre-set image reaches threshold value
When, there is default aggregation event in confirmation.
7. crowd massing detection method according to claim 6, which is characterized in that described when the crowd massing region position
When the number that goal-selling in setting occurs in the multiframe pre-set image reaches threshold value, there is default aggregation event in confirmation
Step, comprising:
Obtain at least one of pre-set image of the second default frame number target to be detected before pre-set image described in every frame；
When the goal-selling in the crowd massing regional location is identical as the target to be detected, and the goal-selling is in institute
When stating the number occurred in the second default frame number pre-set image and reaching threshold value, there is default aggregation event in confirmation.
8. a kind of crowd massing detection device characterized by comprising
Pre-set image obtains module, for obtaining the continuous multiframe pre-set image of target video；
Pedestrian area obtains module, for obtaining pedestrian area from pre-set image described in every frame according to crowd density；
It flows pedestrian area and obtains module, for obtaining flowing pedestrian area from pre-set image described in every frame；
Region to be confirmed obtains module, for obtaining region to be confirmed according to the pedestrian area and the flowing pedestrian area；
Judgment module, the number for occurring in the every frame image of the multiframe when the goal-selling in the region to be confirmed reach
When to threshold value, there is default aggregation event in confirmation.
9. crowd massing detection device according to claim 8, which is characterized in that
The region to be confirmed obtains module, is also used to obtain by removing the flowing pedestrian area in the pedestrian area
Take the corresponding second crowd density figure in the region to be confirmed；Morphological scale-space is carried out to the second crowd density figure, and is mentioned
Contouring；The boundary rectangle for extracting the profile obtains regional location to be confirmed；According to second density map obtain it is described to
Confirm the total number of persons in region；When total number of persons is greater than the first preset threshold in the region to be confirmed, the area to be confirmed is confirmed
Domain is crowd massing region；
The judgment module is also used to when the goal-selling in the crowd massing regional location is in the multiframe pre-set image
When the number of appearance reaches threshold value, there is default aggregation event in confirmation.
10. a kind of storage medium, which is characterized in that computer program is stored in the storage medium, when the computer journey
When sequence is run on computers, so that the computer executes crowd massing detection side as described in any one of claim 1 to 7
Method.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811270525.1A CN109446989A (en) | 2018-10-29 | 2018-10-29 | Crowd massing detection method, device and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811270525.1A CN109446989A (en) | 2018-10-29 | 2018-10-29 | Crowd massing detection method, device and storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109446989A true CN109446989A (en) | 2019-03-08 |
Family
ID=65548628
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811270525.1A Pending CN109446989A (en) | 2018-10-29 | 2018-10-29 | Crowd massing detection method, device and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109446989A (en) |
Cited By (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110032947A (en) * | 2019-03-22 | 2019-07-19 | 深兰科技（上海）有限公司 | A kind of method and device that monitor event occurs |
| CN111291690A (en) * | 2020-02-17 | 2020-06-16 | 深圳市联合视觉创新科技有限公司 | Route planning method, route planning device, robot, and medium |
| CN111507367A (en) * | 2019-11-15 | 2020-08-07 | 杭州商警云智能科技有限公司 | Aggregation point analysis method and system based on face files |
| CN111586369A (en) * | 2020-06-05 | 2020-08-25 | 上海商汤智能科技有限公司 | Aggregation detection method and device, electronic equipment and readable storage medium |
| CN111597461A (en) * | 2020-05-08 | 2020-08-28 | 北京百度网讯科技有限公司 | Target object aggregation prediction method and device and electronic equipment |
| CN111639597A (en) * | 2020-05-29 | 2020-09-08 | 上海闪马智能科技有限公司 | Detection method of flag-raising touring event |
| 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 |
| CN112883768A (en) * | 2019-11-29 | 2021-06-01 | 华为技术有限公司 | Object counting method and device, equipment and storage medium |
| CN113837034A (en) * | 2021-09-08 | 2021-12-24 | 云从科技集团股份有限公司 | Aggregated population monitoring method, device and computer storage medium |
| CN113963316A (en) * | 2021-11-25 | 2022-01-21 | 上海闪马智能科技有限公司 | Target event determination method and device, storage medium and electronic device |
| CN117156259A (en) * | 2023-10-30 | 2023-12-01 | 海信集团控股股份有限公司 | Video stream acquisition method and electronic equipment |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105930814A (en) * | 2016-04-28 | 2016-09-07 | 天津大学 | Method for detecting personnel abnormal gathering behavior on the basis of video monitoring platform |
| CN106296721A (en) * | 2015-05-14 | 2017-01-04 | 株式会社理光 | Object based on stereoscopic vision assembles detection method and device |
| CN107371158A (en) * | 2017-06-13 | 2017-11-21 | 华南理工大学 | The investigation of mobile communication carrier user accounting and crowd measure estimating and measuring method in region |
| CN107393313A (en) * | 2017-09-04 | 2017-11-24 | 郑州大学 | A kind of intelligent Control System for Traffic Lights based on Pedestrian flow detection |
-
2018
- 2018-10-29 CN CN201811270525.1A patent/CN109446989A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106296721A (en) * | 2015-05-14 | 2017-01-04 | 株式会社理光 | Object based on stereoscopic vision assembles detection method and device |
| CN105930814A (en) * | 2016-04-28 | 2016-09-07 | 天津大学 | Method for detecting personnel abnormal gathering behavior on the basis of video monitoring platform |
| CN107371158A (en) * | 2017-06-13 | 2017-11-21 | 华南理工大学 | The investigation of mobile communication carrier user accounting and crowd measure estimating and measuring method in region |
| CN107393313A (en) * | 2017-09-04 | 2017-11-24 | 郑州大学 | A kind of intelligent Control System for Traffic Lights based on Pedestrian flow detection |
Cited By (16)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110032947A (en) * | 2019-03-22 | 2019-07-19 | 深兰科技（上海）有限公司 | A kind of method and device that monitor event occurs |
| CN110032947B (en) * | 2019-03-22 | 2021-11-19 | 深兰科技（上海）有限公司 | Method and device for monitoring occurrence of event |
| CN111507367A (en) * | 2019-11-15 | 2020-08-07 | 杭州商警云智能科技有限公司 | Aggregation point analysis method and system based on face files |
| CN112883768A (en) * | 2019-11-29 | 2021-06-01 | 华为技术有限公司 | Object counting method and device, equipment and storage medium |
| CN112883768B (en) * | 2019-11-29 | 2024-02-09 | 华为云计算技术有限公司 | Object counting method and device, equipment and storage medium |
| CN111291690A (en) * | 2020-02-17 | 2020-06-16 | 深圳市联合视觉创新科技有限公司 | Route planning method, route planning device, robot, and medium |
| CN111291690B (en) * | 2020-02-17 | 2023-12-05 | 深圳市联合视觉创新科技有限公司 | Route planning method, route planning device, robot and medium |
| CN111597461A (en) * | 2020-05-08 | 2020-08-28 | 北京百度网讯科技有限公司 | Target object aggregation prediction method and device and electronic equipment |
| CN111597461B (en) * | 2020-05-08 | 2023-11-17 | 北京百度网讯科技有限公司 | Target object aggregation prediction method and device and electronic equipment |
| CN111639597A (en) * | 2020-05-29 | 2020-09-08 | 上海闪马智能科技有限公司 | Detection method of flag-raising touring event |
| CN111586369A (en) * | 2020-06-05 | 2020-08-25 | 上海商汤智能科技有限公司 | Aggregation detection method and device, electronic equipment and readable storage medium |
| 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 |
| CN113837034A (en) * | 2021-09-08 | 2021-12-24 | 云从科技集团股份有限公司 | Aggregated population monitoring method, device and computer storage medium |
| CN113963316A (en) * | 2021-11-25 | 2022-01-21 | 上海闪马智能科技有限公司 | Target event determination method and device, storage medium and electronic device |
| CN117156259A (en) * | 2023-10-30 | 2023-12-01 | 海信集团控股股份有限公司 | Video stream acquisition method and electronic equipment |
| CN117156259B (en) * | 2023-10-30 | 2024-03-22 | 海信集团控股股份有限公司 | Video stream acquisition method and electronic equipment |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109446989A (en) | Crowd massing detection method, device and storage medium | |
| CN102542289B (en) | Pedestrian volume statistical method based on plurality of Gaussian counting models | |
| CN105184818B (en) | A kind of video monitoring anomaly detection method and its detecting system | |
| CN106096577B (en) | A kind of target tracking method in camera distribution map | |
| CN102324016B (en) | Statistical method for high-density crowd flow | |
| CN104835179B (en) | Based on the adaptive improvement ViBe background modeling methods of dynamic background | |
| Kamthe et al. | Suspicious activity recognition in video surveillance system | |
| CN105335701B (en) | A kind of pedestrian detection method based on HOG Yu D-S evidence theory multi-information fusion | |
| CN105513053B (en) | One kind is used for background modeling method in video analysis | |
| CN104992453A (en) | Target tracking method under complicated background based on extreme learning machine | |
| Gong et al. | Local distinguishability aggrandizing network for human anomaly detection | |
| Malhi et al. | Vision based intelligent traffic management system | |
| Liang et al. | Methods of moving target detection and behavior recognition in intelligent vision monitoring. | |
| Tang et al. | Hybrid blob and particle filter tracking approach for robust object tracking | |
| CN103577804A (en) | Abnormal human behavior identification method based on SIFT flow and hidden conditional random fields | |
| Choubisa et al. | Object tracking in intelligent video surveillance system based on artificial system | |
| CN106384089A (en) | Human body reliable detection method based on lifelong learning | |
| Rao et al. | Abnormal activity detection in video sequences using learnt probability densities | |
| Chen et al. | Moving objects detection based on background subtraction combined with consecutive frames subtraction | |
| Leyva et al. | Video anomaly detection based on wake motion descriptors and perspective grids | |
| CN111008601A (en) | Fighting detection method based on video | |
| Yang et al. | Crowd gathering and commotion detection based on the stillness and motion model | |
| Deepak et al. | Design and utilization of bounding box in human detection and activity identification | |
| CN109389048A (en) | Pedestrian detection and tracking in a kind of monitor video | |
| Yan et al. | Crowd counting via scale-adaptive convolutional neural network in extremely dense crowd images |
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 | ||
| WD01 | Invention patent application deemed withdrawn after publication | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20190308 |