CN110390226A - Crowd's event recognition method, device, electronic equipment and system - Google Patents
Crowd's event recognition method, device, electronic equipment and system Download PDFInfo
- Publication number
- CN110390226A CN110390226A CN201810340168.5A CN201810340168A CN110390226A CN 110390226 A CN110390226 A CN 110390226A CN 201810340168 A CN201810340168 A CN 201810340168A CN 110390226 A CN110390226 A CN 110390226A
- Authority
- CN
- China
- Prior art keywords
- crowd
- event
- image group
- pedestrian
- monitoring
- 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
- G06V20/40—Scenes; Scene-specific elements in video content
-
- 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
Abstract
The invention discloses a kind of crowd's event recognition method, device, electronic equipment and systems, belong to crowd's event recognition field.The described method includes: obtaining the monitor video of target area;At least one target monitoring image group is obtained from monitor video, each target monitoring image group includes continuous multiple monitoring video frames;Crowd's event recognition is carried out to each target monitoring image group by preset crowd's event model, crowd's event model is obtained according to the training of multiple crowd's event image groups, and each crowd's occurrence diagram picture group includes several crowd's event images of reflection crowd's event generating process;Determine whether target area crowd's event occurs according to recognition result.The present invention can be improved the accuracy of crowd's event recognition.
Description
Technical field
The present invention relates to crowd's event recognition field, in particular to a kind of crowd's event recognition method, device, electronic equipment
And system.
Background technique
With being continuously increased for economic fast development and people's social activities, transport hub, megastore and large size
It is higher and higher that a possibility that crowd's event, occurs for the public places such as site of activity.So-called crowd's event refers to that crowd is a large amount of suddenly
The event that aggregation or crowd largely disperse suddenly.Caused by crowd's event is likely due to burst occurred events of public safety, and very
It is easy to cause and the safety accidents such as tramples.For example, when the occurred events of public safety such as the attack of terrorism occur, it is likely that it is prominent to will appear crowd
The crowd's event so largely dispersed, and when crowd's a large amount of assembled crowd's events suddenly occur, crowd's event is likely to joint performance
The safety accidents such as become to trample.Therefore, crowd's event is accurately identified in quick disposition burst occurred events of public safety and avoids stepping on
It has great significance in terms of the safety accidents such as stepping on.
In the related technology, when identifying crowd's event, a monitoring can be obtained from monitor video at regular intervals
Video frame, and obtain the corresponding crowd density figure of the monitoring video frame, wherein crowd density figure can reflect the density of pedestrian,
When the density of the pedestrian of crowd's density map reflection is higher, i.e., it is believed that crowd's event has occurred.
In the implementation of the present invention, the inventor finds that the existing technology has at least the following problems:
The information of the corresponding crowd density figure reflection of one monitoring video frame is more single, is known using crowd's density map
Others is easy to cause the recognition accuracy of crowd's event lower at group's event.
Summary of the invention
The embodiment of the invention provides a kind of crowd's event recognition method, device, electronic equipment and systems, can be improved people
The accuracy rate of group's event recognition.The technical solution is as follows:
In a first aspect, providing a kind of crowd's event recognition method, which comprises
Obtain the monitor video of target area;
At least one target monitoring image group is obtained from the monitor video, each target monitoring image group includes
Continuous multiple monitoring video frames;
Crowd's event recognition, the people are carried out to each target monitoring image group by preset crowd's event model
Group's event model is obtained according to the training of multiple crowd's event image groups, and each crowd's event image group includes reflection crowd's thing
Several crowd's event images of part generating process;
Determine whether the target area crowd's event occurs according to recognition result.
Optionally, the monitor video includes multiple monitoring image groups, and each monitoring image group includes continuous more
A monitoring video frame, each monitoring video frame includes the image of pedestrian, described that at least one is obtained from the monitor video
A target monitoring image group, comprising:
For each monitoring image group, the figure of the pedestrian in each monitoring video frame of the monitoring image group is obtained
Picture;
According to the image of the pedestrian in each monitoring video frame, it is pre- to judge whether the change in location of the pedestrian meets
If condition, the preset condition is the change in location condition for reflecting pedestrian and assembling or dispersing;
When the change in location meets the preset condition, the monitoring image group is retrieved as the target monitoring figure
As group.
Optionally, the image according to the pedestrian in each monitoring video frame judges that the position of the pedestrian becomes
Change and whether meet preset condition, comprising:
Using identical generating mode, pedestrian's histogram is generated respectively for multiple monitoring video frames of the monitoring image group
Figure, wherein each monitoring video frame is equidistantly divided into multiple regions, each pedestrian's histogram in a first direction
Including multiple rectangular strips, position of the multiple rectangular strip in first axle and the multiple regions of corresponding monitoring video frame are one by one
Corresponding, each rectangular strip indicates the sum of the pixel value of pedestrian image in corresponding region in the length on the second axis, described
First axle and second axis are mutually perpendicular to;
Center rectangle item and edge rectangular strip are determined from each pedestrian's histogram, obtain multiple center rectangle items and
Multiple edge rectangular strips;
The pedestrian is judged according to the length situation of change of the multiple center rectangle item and the multiple edge rectangular strip
Change in location whether meet the preset condition.
Optionally, the distributing order of the multiple center rectangle item and the multiple edge rectangular strip and the monitoring image
The acquisition sequence consensus of multiple monitoring video frames of group, it is described according to the multiple center rectangle item and the multiple edge rectangle
The length situation of change of item judges whether the change in location of the pedestrian meets the preset condition, comprising:
When the length of the multiple center rectangle item gradually increases and the length of the multiple edge rectangular strip is gradually reduced
When, determine that the change in location of the pedestrian meets the preset condition.
Optionally, the distributing order of the multiple center rectangle item and the multiple edge rectangular strip and the monitoring image
The acquisition sequence consensus of multiple monitoring video frames of group, it is described according to the multiple center rectangle item and the multiple edge rectangle
The length situation of change of item judges whether the change in location of the multiple movement pedestrian meets the preset condition, comprising:
When the length of the multiple edge rectangular strip gradually increases and the length of the multiple center rectangle item is gradually reduced
When, determine that the change in location of the pedestrian meets the preset condition.
Optionally, each monitoring video frame is equidistantly divided into using target length as dividing unit in a first direction
Multiple regions, the target length are the length of pixel in said first direction, and the first direction is line direction or column side
To.
Optionally, described that crowd's event is carried out to each target monitoring image group by preset crowd's event model
Identification, comprising:
For each target monitoring image group, it is corresponding to obtain each monitoring video frame in the target monitoring image group
First crowd's density map, obtain several the first crowd density maps;
Obtain corresponding first light stream of image of pedestrian in multiple monitoring video frames that the target monitoring image group includes
Figure;
By crowd's event model, based on several described the first crowd density maps and the first light stream figure to described
Target monitoring image group carries out crowd's event recognition.
Optionally, described that crowd's event is carried out to each target monitoring image group by preset crowd's event model
Before identification, the method also includes:
For each crowd's occurrence diagram picture group in the multiple crowd's event image group, crowd's event image group is obtained
In the second crowd density figure corresponding to every width crowd event image, to obtain several the second crowd density figures;
It obtains in several crowd's event images that crowd's event image group includes second corresponding to the image of pedestrian
Light stream figure;
Model training is carried out based on several described the second crowd density figures and the second light stream figure, to obtain the crowd
Event model.
Optionally, crowd's event model is convolutional neural networks model.
Second aspect, provides a kind of crowd's event recognition device, and described device includes:
Video acquiring module, for obtaining the monitor video of target area;
Image group obtains module, for obtaining at least one target monitoring image group, Mei Gesuo from the monitor video
Stating target monitoring image group includes continuous multiple monitoring video frames;
Identification module, for carrying out crowd's thing to each target monitoring image group by preset crowd's event model
Part identification, crowd's event model are obtained according to the training of multiple crowd's event image groups, each crowd's event image group
Several crowd's event images including reflecting crowd's event generating process;
Determining module, for determining whether the target area crowd's event occurs according to recognition result.
Optionally, the monitor video includes multiple monitoring image groups, and each monitoring image group includes continuous more
A monitoring video frame, each monitoring video frame include the image of pedestrian, and described image group obtains module, is used for:
For each monitoring image group, the figure of the pedestrian in each monitoring video frame of the monitoring image group is obtained
Picture;
According to the image of the pedestrian in each monitoring video frame, it is pre- to judge whether the change in location of the pedestrian meets
If condition, the preset condition is the change in location condition for reflecting pedestrian and assembling or dispersing;
When the change in location meets the preset condition, the monitoring image group is retrieved as the target monitoring figure
As group.
Optionally, described image group obtains module, is used for:
Using identical generating mode, pedestrian's histogram is generated respectively for multiple monitoring video frames of the monitoring image group
Figure, wherein each monitoring video frame is equidistantly divided into multiple regions, each pedestrian's histogram in a first direction
Including multiple rectangular strips, position of the multiple rectangular strip in first axle and the multiple regions of corresponding monitoring video frame are one by one
Corresponding, each rectangular strip indicates the sum of the pixel value of pedestrian image in corresponding region in the length on the second axis, described
First axle and second axis are mutually perpendicular to;
Center rectangle item and edge rectangular strip are determined from each pedestrian's histogram, obtain multiple center rectangle items and
Multiple edge rectangular strips;
The pedestrian is judged according to the length situation of change of the multiple center rectangle item and the multiple edge rectangular strip
Change in location whether meet the preset condition.
Optionally, the distributing order of the multiple center rectangle item and the multiple edge rectangular strip and the monitoring image
The acquisition sequence consensus of multiple monitoring video frames of group, described image group obtain module, are used for:
When the length of the multiple center rectangle item gradually increases and the length of the multiple edge rectangular strip is gradually reduced
When, determine that the change in location of the pedestrian meets the preset condition.
Optionally, the distributing order of the multiple center rectangle item and the multiple edge rectangular strip and the monitoring image
The acquisition sequence consensus of multiple monitoring video frames of group, described image group obtain module, are used for:
When the length of the multiple edge rectangular strip gradually increases and the length of the multiple center rectangle item is gradually reduced
When, determine that the change in location of the pedestrian meets the preset condition.
Optionally, each monitoring video frame is equidistantly divided into using target length as dividing unit in a first direction
Multiple regions, the target length are the length of pixel in said first direction, and the first direction is line direction or column side
To.
Optionally, the identification module, is used for:
For each target monitoring image group, it is corresponding to obtain each monitoring video frame in the target monitoring image group
First crowd's density map, obtain several the first crowd density maps;
Obtain corresponding first light stream of image of pedestrian in multiple monitoring video frames that the target monitoring image group includes
Figure;
By crowd's event model, based on several described the first crowd density maps and the first light stream figure to described
Target monitoring image group carries out crowd's event recognition.
Optionally, described device further includes training module, and the training module is used for:
For each crowd's occurrence diagram picture group in the multiple crowd's event image group, crowd's event image group is obtained
In the second crowd density figure corresponding to every width crowd event image, to obtain several the second crowd density figures;
It obtains in several crowd's event images that crowd's event image group includes second corresponding to the image of pedestrian
Light stream figure;
Model training is carried out based on several described the second crowd density figures and the second light stream figure, to obtain the crowd
Event model.
Optionally, crowd's event model is convolutional neural networks model.
The third aspect, provides a kind of electronic equipment, and the electronic equipment includes processor and memory,
Wherein, the memory, for storing computer program;
The processor, for executing the program stored in the memory, to realize that above-mentioned first aspect is any described
Crowd's event recognition method.
Fourth aspect, provides a kind of crowd's event recognition system, and crowd's event recognition system includes such as above-mentioned the
Two aspects any crowd's event recognition device and monitoring device
Technical solution provided in an embodiment of the present invention has the benefit that
By obtaining at least one target monitoring image group from the monitor video of target area, and utilize preset crowd
Event model carries out crowd's event recognition to each target monitoring image group, whether to determine target area according to recognition result
Appearance crowd's event, wherein each target monitoring image group includes continuous multiple monitoring video frames, due to the hair of crowd's event
Life is related with the moving situation of crowd in a period of time, and continuous multiple monitoring video frames can reflect target in a period of time
The change in location of pedestrian in region, to reflect the moving situation of crowd in a period of time region of interest within, therefore, using continuous
Multiple monitoring video frames carry out crowd's event recognition the accuracy of crowd's event recognition can be improved.
Detailed description of the invention
To describe the technical solutions in the embodiments of the present invention more clearly, make required in being described below to embodiment
Attached drawing is briefly described, it should be apparent that, drawings in the following description are only some embodiments of the invention, for
For those of ordinary skill in the art, without creative efforts, it can also be obtained according to these attached drawings other
Attached drawing.
Fig. 1 is a kind of crowd density figure provided in an embodiment of the present invention.
Fig. 2A is a kind of schematic diagram of implementation environment provided in an embodiment of the present invention.
Fig. 2 B is a kind of schematic diagram of implementation environment provided in an embodiment of the present invention.
Fig. 3 is a kind of flow chart of crowd's event recognition method provided in an embodiment of the present invention.
Fig. 4 A is a kind of flow chart of crowd's event recognition method provided in an embodiment of the present invention.
Fig. 4 B is a kind of histogram provided in an embodiment of the present invention.
Fig. 5 is a kind of block diagram of crowd's event recognition device provided in an embodiment of the present invention.
Fig. 6 is a kind of block diagram of crowd's event recognition device provided in an embodiment of the present invention.
Fig. 7 is the block diagram of a kind of electronic equipment provided in an embodiment of the present invention.
Fig. 8 is a kind of block diagram of crowd's event recognition system provided in an embodiment of the present invention.
Specific embodiment
To make the object, technical solutions and advantages of the present invention clearer, below in conjunction with attached drawing to embodiment party of the present invention
Formula is described in further detail.
Crowd's event recognition is in terms of quick disposition burst occurred events of public safety and avoiding such as tramples at the safety accidents
It has great significance.
In the prior art, for a certain region carry out crowd's event recognition when, server can at regular intervals from
A monitoring video frame is obtained in the monitor video in the region.Then, the corresponding people of the available monitoring video frame of server
Group's density map, wherein crowd density figure can reflect the density of pedestrian in the region.Fig. 1 is an illustrative crowd density
Figure uses the gray value size of pixel as shown in Figure 1, crowd's density map is grayscale image to characterize the density size of crowd,
The density of the bigger crowd of gray value is bigger, and the density of the smaller crowd of gray value is smaller.Obtaining, the monitoring video frame is corresponding
After crowd density figure, server can determine crowd's density map reflection the region in pedestrian density whether be higher than preset it is close
Threshold value is spent, when the density of pedestrian in the region of crowd's density map reflection is higher than pre-set density threshold value, server can be true
Crowd's event has occurred in the fixed region.
However, the information of the corresponding crowd density figure reflection of a monitoring video frame is more single, therefore, using the crowd
Density map identifies that crowd's event is easy to cause the recognition accuracy of crowd's event lower.
The embodiment of the invention provides a kind of crowd's event recognition methods, can solve the above problem, and Fig. 2A is the crowd
A kind of schematic diagram of implementation environment involved by event recognition method, as shown in Figure 2 A, the implementation environment may include monitoring
Equipment 101 and server 102, monitoring device 101 can be communicated by wired or wireless mode with server 102.Prison
Controlling equipment 101 can be monitor camera, and server 102 can be a server, or be made of multiple servers
Server cluster, the embodiment of the present invention is not specifically limited in this embodiment.
Fig. 2 B is the schematic diagram of another implementation environment involved by crowd's event recognition method, as shown in Figure 2 B,
The implementation environment may include monitoring device 103, which can be monitor camera.
Fig. 3 is a kind of flow chart of crowd's event recognition method provided in an embodiment of the present invention, the crowd event recognition side
Method can be applied in implementation environment shown in Fig. 2A or Fig. 2 B, which can be by the server in Fig. 2A
102 or monitoring device 101 execute, can also be executed by the monitoring device 103 in Fig. 2 B, wherein the embodiment of the present invention only with by
Server is executed to be illustrated, and therewith similarly, the embodiment of the present invention exists the technical process executed by monitoring device
This is repeated no more.As shown in figure 3, crowd's event recognition method may comprise steps of:
Step 301, server obtain the monitor video of target area.
Step 302, server obtain at least one target monitoring image group, each target monitoring figure from the monitor video
As group includes continuous multiple monitoring video frames.
Step 303, server carry out crowd's event to each target monitoring image group by preset crowd's event model
Identification, wherein crowd's event model is obtained according to the training of multiple crowd's event image groups, and each crowd's occurrence diagram picture group includes
Several crowd's event images of reflection crowd's event generating process.
Step 304, server determine whether target area crowd's event occurs according to recognition result.
In conclusion crowd's event recognition method provided in an embodiment of the present invention, passes through the monitor video from target area
Middle at least one target monitoring image group of acquisition, and using preset crowd's event model to each target monitoring image group into
Pedestrian's group's event recognition, to determine whether target area crowd's event occurs according to recognition result, wherein each target monitoring figure
It include continuous multiple monitoring video frames as organizing, due to the generation of crowd's event and in a period of time, the moving situation of crowd has
It closes, and continuous multiple monitoring video frames can reflect the change in location of pedestrian in a period of time region of interest within, to reflect
Therefore the moving situation of crowd in a period of time region of interest within using continuous multiple monitoring video frames, that is to say and utilize mesh
Mark monitoring image group, which carries out crowd's event recognition, can be improved the accuracy of crowd's event recognition.
Fig. 4 A is the flow chart of another crowd's event recognition method provided in an embodiment of the present invention, crowd's event recognition
Method can be applied in implementation environment shown in Fig. 2A or Fig. 2 B, which can be by the service in Fig. 2A
Device 102 or monitoring device 101 execute, and can also be executed by the monitoring device 103 in Fig. 2 B, wherein the embodiment of the present invention only with
Executed by server to be illustrated, the technical process executed by monitoring device therewith similarly, the embodiment of the present invention
Details are not described herein, and as shown in Figure 4 A, which may comprise steps of:
Step 401, server obtain the monitor video of target area.
In embodiments of the present invention, server can in real time or periodically with the monitoring device that is set up in target area
It is communicated, to obtain monitor video captured by the monitoring device, to be carried out in the next steps according to the monitor video
Crowd's event recognition, to determine whether the target area has occurred crowd's event.
It should be pointed out that above-mentioned target area can be the pre-set region of technical staff, which can be with
For the biggish public place of flows of the people such as transport hub, megastore or large-scale activity scene.
Step 402, server obtain at least one target monitoring image group from the monitor video of target area.
The monitor video of target area may include multiple monitoring image groups, wherein each monitoring image group may include
Continuous multiple monitoring video frames, and, the image that each monitoring video frame may each comprise pedestrian (is referred to as foreground image
Or movement destination image).
In step 402, all monitoring image groups which can be included by server are retrieved as target prison
Image group is controlled, to get multiple target monitoring image groups.
In one embodiment of the invention, server can also be from multiple monitoring image groups that the monitor video includes
It determines at least one monitoring image group, and at least one determining monitoring image group is retrieved as at least one target monitoring image
Group, in this case, multiple continuous monitoring video frames that target monitoring image group includes can be doubtful crowd's event pair
The monitoring video frame answered, server can carry out the knowledge of crowd's event according at least one target monitoring image group in subsequent step
Not, in this way, server can supervise in the next steps according only to the target of the corresponding monitoring video frame composition of doubtful crowd's event
It controls image group and carries out crowd's event recognition, all monitoring image groups without including according to monitor video carry out crowd's event
Identification, so as to mitigate the computational burden of server.
In the following, the embodiment of the present invention will obtain at least server from multiple monitoring image groups that the monitor video includes
The technical process of one target monitoring image group is illustrated:
For each monitoring image group that monitor video includes, server can use frame difference method or background modeling method
Etc. modes obtain the image of the pedestrian in each monitoring video frame of the monitoring image group, then, server can be according to acquisition
To each monitoring video frame in pedestrian image, judge whether the change in location of pedestrian meets preset condition, when pedestrian's
When change in location meets preset condition, which can be retrieved as target monitoring image group by server.It is above-mentioned default
Condition is the change in location condition for reflecting pedestrian and assembling or dispersing.
Wherein, for each monitoring image group that monitor video includes, server is every according to the monitoring image group
The image of pedestrian in a monitoring video frame, judges whether the change in location of pedestrian meets the technical process of preset condition and can wrap
Include following sub-step:
Each monitoring video frame of the monitoring image group is divided into multiple regions by A1, server.
Optionally, each monitoring video frame can be equally spacedly divided into multiple regions by server in a first direction,
Wherein, the spacing of division can be target length, which can be for a pixel or multiple pixels in a first direction
Length, the first direction can be monitoring video frame line direction or column direction.In other words, server can be by each prison
Control video frame be divided into multiple strip regions, wherein each strip region may include at least one pixel column or at least one
Pixel column.
B1, for each monitoring video frame, in each region that the available monitoring video frame of server includes
The sum of pixel value of pedestrian image.
For example, some region in a certain monitoring video frame can be the pixel column a of a certain monitoring video frame, the pixel
Row a includes n (n is greater than 1 positive integer) a pixel, m (m is more than or equal to 0 natural number) a pixel in the n pixel
For the pixel in pedestrian image, rest of pixels is the pixel in background image, then in step bl is determined, server can be by the m
The pixel value of pixel is added, to obtain the sum of the pixel value of pedestrian image in some region.
C1, server use identical generating mode to generate pedestrian's histogram for each monitoring video frame is corresponding.
Wherein, each pedestrian's histogram may each comprise multiple rectangular strips, position of multiple rectangular strip in first axle
Corresponded with the multiple regions of corresponding monitoring video frame, and, multiple rectangular strip in first axle put in order with it is right
Identical, length of each rectangular strip on the second axis that puts in order of the multiple regions for the monitoring video frame answered in a first direction
Indicate the sum of the pixel value of pedestrian image in corresponding region.Wherein, first axle and the second axis are mutually perpendicular to, and above-mentioned first axle can
Think x-axis, the second axis can be y-axis;Alternatively, above-mentioned first axle can be y-axis, the second axis can be x-axis.
Fig. 4 B show illustrative pedestrian's histogram, and pedestrian's histogram is corresponding with monitoring video frame z, wherein
Monitoring video frame z may include p region, and each region is a pixel column of monitoring video frame z.As shown in Figure 4 B, should
Pedestrian's histogram includes p rectangular strip, the p region one which includes with monitoring video frame z
One is corresponding, put in order with the p region of monitoring video frame z putting in order in the row direction of the p rectangular strip in x-axis
It is identical, in the p rectangular strip length of each rectangular strip on the y axis indicate pedestrian image in corresponding region pixel value it
With.For example, k-th of rectangular strip in the p rectangular strip is corresponding with k-th of region in the p region that monitoring video frame z includes
(k-th of pixel column that k-th of region is monitoring video frame z), the length of k-th of rectangular strip on the y axis indicate monitoring
The sum of the pixel value of pedestrian image in k-th of pixel column of video frame z.
D1, server multiple pedestrian's histograms according to corresponding to multiple monitoring video frames that the monitoring image group includes are sentenced
Whether the change in location of line-break people meets preset condition.
Server can determine center rectangle item and side from each pedestrian's histogram in above-mentioned multiple pedestrian's histograms
Edge rectangular strip, to obtain multiple center rectangle items and multiple edge rectangular strips, wherein center rectangle item is referred in first axle
On position be in the rectangular strip of middle section, edge rectangular strip refers to that the position in first axle is in the square of fringe region
Shape item.It is pointed out that position phase of the server center rectangle item determining from each pedestrian's histogram in first axle
Together, and, position of the edge rectangular strip that server is determined from each pedestrian's histogram in first axle is also identical.It is obtaining
After multiple center rectangle items and multiple edge rectangular strips, server can be according to multiple center rectangle items and multiple edge rectangles
The length situation of change of item judges whether the change in location of pedestrian meets preset condition.
Optionally, server judges pedestrian according to the length situation of change of multiple center rectangle items and multiple edge rectangular strips
Change in location whether to meet the technical process of preset condition may include following sub-step:
A2, server arrange multiple center rectangle according to putting in order for multiple monitoring video frames in the monitoring image group
Item and multiple edge rectangular strip.
After arrangement, sequence of each center rectangle item in multiple center rectangle item is corresponding with the center rectangle item
Sequence of the monitoring video frame corresponding to pedestrian's histogram in multiple monitoring video frames that the monitoring image group includes is identical,
And sequence of each edge rectangular strip in multiple edge rectangular strip pedestrian's histogram institute corresponding with the edge rectangular strip is right
Sequence of the monitoring video frame answered in multiple monitoring video frames that the monitoring image group includes is identical.
B2, server determine that the length of multiple center rectangle item changes feelings according to the sequence of multiple center rectangle item
Condition.
C2, server determine that the length of multiple edge rectangular strip changes feelings according to the sequence of multiple edge rectangular strip
Condition.
D2, server judge pedestrian according to the length situation of change of multiple center rectangle item and multiple edge rectangular strip
Change in location whether meet preset condition.
Optionally, when the length of multiple center rectangle item gradually increases and the length of multiple edge rectangular strip gradually subtracts
Hour, illustrate that pedestrian gradually converges, server can determine that the change in location of pedestrian meets the default of reflection pedestrian's convergence at this time
Condition.Alternatively, when the length of multiple edge rectangular strip gradually increases and the length of multiple center rectangle item is gradually reduced,
Illustrate that pedestrian gradually disperses, server can determine that the change in location of pedestrian meets the preset condition of reflection pedestrian's dispersion at this time.
Step 403, server carry out crowd's event to each target monitoring image group by preset crowd's event model
Identification.
It should be pointed out that above-mentioned crowd's event model can be convolutional neural networks model, for example, crowd's event mould
Type can be Google's convolutional neural networks or residual error convolutional neural networks etc..
In the following, the embodiment of the present invention by with server to the first object monitoring figure at least one target monitoring image group
As being illustrated for group carry out crowd's event recognition to the technical process of step 403, which may include following son
Step:
A3, server obtain the corresponding first crowd density map of each monitoring video frame in first object monitoring image group,
Obtain several the first crowd density maps.
Server obtain the corresponding crowd density figure of monitoring video frame mode can there are many kinds, such as: one kind can
In the mode of energy, monitoring video frame can be divided into different regions by server, and calculate the pedestrian that each region is included
The number of image, then server can be determined based on the number of each region one skilled in the art image the corresponding gray value in the region or
Color value, then server can generate crowd density image according to the corresponding gray value in each region or color value;Another
In the possible mode of kind, server can determine each pedestrian image that monitoring video frame includes, for each pedestrian image
Speech, server can be a little that the center of circle determines a temperature region by radius of preset value with the pedestrian image, to obtain
Initial gray value or original pixel values can be arranged for each temperature region in multiple temperature regions, server, and then, server can
To determine intersection region, the intersection region is corresponding at least two temperature regions, which is corresponding at least two heat
The region of area coincidence is spent, server can change the intersection region according to the number in the corresponding temperature region in the intersection region
Pixel value or color value, to finally obtain crowd density figure.
The image that B3, server obtain pedestrian in multiple monitoring video frames that first object monitoring image group includes is corresponding
First light stream figure.
Wherein, the first light stream figure can be the motion vector figure of each pixel in the image of pedestrian.
Optionally, in t-th of monitoring video frame of the available multiple monitoring video frame of server pedestrian image packet
The each pixel included, for each pixel, server can traverse the figure of the pedestrian of the t+1 monitoring video frame
Picture, and obtained in the image of the pedestrian from the t+1 monitoring video frame and be greater than preset threshold with the similarity of the pixel
Target pixel points, then, server can be existed based on coordinate value of the pixel in image coordinate system and the target pixel points
Coordinate value in image coordinate system determines fortune of the pixel between t-th of monitoring video frame and the t+1 monitoring video frame
Dynamic vector, server can generate the first light stream figure according to the motion vector of each of the image of pedestrian pixel.
C3, server carry out crowd based on several the first crowd density maps and the first light stream figure by crowd's event model
Event recognition.
The first light stream figure and several the first crowd density maps can be input in crowd's event model by server, the people
Group's event model can carry out corresponding identification operation according to the first light stream figure and several the first crowd density maps, and export fortune
It calculates as a result, multiple monitoring video frames that server can determine that the first object monitoring image group includes according to the operation result are
No is the corresponding monitoring video frame of crowd's event.
It should be noted that above-mentioned crowd's event model can be obtained according to the training of multiple crowd's event image groups, wherein
Each crowd's occurrence diagram picture group includes several crowd's event images of reflection crowd's event generating process.The embodiment of the present invention below
The training process of crowd's event model will be briefly described:
For each of above-mentioned multiple crowd's event image groups group's event image group, the available crowd of server
Second crowd density figure corresponding to every width crowd event image in event image group, to obtain several the second crowd density figures,
Wherein, the technical process of several the second crowd density figures and the technology of several the first crowd density maps of acquisition described above are obtained
Similarly, the embodiment of the present invention repeats no more process.Server can also obtain several crowds that crowd's occurrence diagram picture group includes
Second light stream figure corresponding to the image of pedestrian in event image, wherein the technical process of the second light stream figure of acquisition and institute above
Similarly, the embodiment of the present invention repeats no more the technical process for acquisition the first light stream figure stated.Server is based on several second people
Group's density map and the second light stream figure carry out model training, to obtain crowd's event model.
Step 404, server determine whether target area crowd's event occurs according to recognition result.
When multiple monitoring view that any one target monitoring image group at least one above-mentioned target monitoring image group includes
When frequency frame is crowd's event corresponding monitoring video frame, server can determine that crowd's event occurs in target area.
Step 405, when determining that crowd's event occurs in target area, server is responded.
In one embodiment of the invention, after server determines that crowd's event occurs in target area, server can
To send warning message to designated terminal, so that the law enfrocement official for possessing the designated terminal can quickly ring crowd's event
It answers.
It should be pointed out that when executing above-mentioned crowd's event recognition method by monitoring device, in step 405, the prison
Response instruction can be sent to server, server is receiving when determining that crowd's event occurs in target area by controlling equipment
It after response instruction, can be responded accordingly to, for example, server can send warning message to designated terminal, so as to gather around
There is the law enfrocement official of the designated terminal that can carry out quick response to crowd's event.
Alternatively, in step 405, which can carry out phase when determining that crowd's event occurs in target area
It responds with answering, for example, monitoring device can send warning message to designated terminal, so as to possess the law enfrocement official of the designated terminal
Quick response can be carried out to crowd's event.In this case, communication module can be installed, so that monitoring in monitoring device
Equipment can send warning message to designated terminal by the communication module.
In conclusion crowd's event recognition method provided in an embodiment of the present invention, passes through the monitor video from target area
Middle at least one target monitoring image group of acquisition, and using preset crowd's event model to each target monitoring image group into
Pedestrian's group's event recognition, to determine whether target area crowd's event occurs according to recognition result, wherein each target monitoring figure
It include continuous multiple monitoring video frames as organizing, due to the generation of crowd's event and in a period of time, the moving situation of crowd has
It closes, and continuous multiple monitoring video frames can reflect the change in location of pedestrian in a period of time region of interest within, to reflect
Therefore the moving situation of crowd in a period of time region of interest within using continuous multiple monitoring video frames, that is to say and utilize mesh
Mark monitoring image group, which carries out crowd's event recognition, can be improved the accuracy of crowd's event recognition.
Fig. 5 is a kind of block diagram of crowd's event recognition device 500 provided in an embodiment of the present invention, as shown in figure 5, the crowd
Event recognition device 500 may include: video acquiring module 501, image group acquisition module 502, identification module 503 and determine mould
Block 504.
Wherein, video acquiring module 501, for obtaining the monitor video of target area.
The image group obtains module 502, for obtaining at least one target monitoring image group from the monitor video, each
The target monitoring image group includes continuous multiple monitoring video frames.
The identification module 503, for carrying out people to each target monitoring image group by preset crowd's event model
Group's event recognition, crowd's event model are obtained according to the training of multiple crowd's event image groups, each crowd's occurrence diagram picture group
Several crowd's event images including reflecting crowd's event generating process.
The determining module 504, for determining whether the target area crowd's event occurs according to recognition result.
In one embodiment of the invention, which includes multiple monitoring image groups, each monitoring image group
Including continuous multiple monitoring video frames, each monitoring video frame includes the image of pedestrian, which obtains module 502,
For: for each monitoring image group, obtain the image of the pedestrian in each monitoring video frame of the monitoring image group;According to
The image of pedestrian in each monitoring video frame, judges whether the change in location of the pedestrian meets preset condition, the default item
Part is the change in location condition for reflecting pedestrian and assembling or dispersing;When the change in location meets the preset condition, by the monitoring figure
As group is retrieved as the target monitoring image group.
In one embodiment of the invention, which obtains module 502, for using identical generating mode, is
Multiple monitoring video frames of the monitoring image group generate pedestrian's histogram respectively, wherein each monitoring video frame is in first party
Multiple regions are equidistantly divided into upwards, each pedestrian's histogram includes multiple rectangular strips, and multiple rectangular strip is in first axle
On the multiple regions of position and corresponding monitoring video frame correspond, length of each rectangular strip on the second axis indicates
The sum of the pixel value of pedestrian image in corresponding region, the first axle and second axis are mutually perpendicular to;From each pedestrian's histogram
Center rectangle item and edge rectangular strip are determined in figure, obtain multiple center rectangle items and multiple edge rectangular strips;According to multiple
It is default that the length situation of change of center rectangle item and multiple edge rectangular strip judges whether the change in location of the pedestrian meets this
Condition.
In one embodiment of the invention, the distributing order of multiple center rectangle item and multiple edge rectangular strip with
The acquisition sequence consensus of multiple monitoring video frames of the monitoring image group, the image group obtain module 502, in multiple
When the length of centre rectangular strip gradually increases and the length of multiple edge rectangular strip is gradually reduced, the change in location of the pedestrian is determined
Meet the preset condition.
In one embodiment of the invention, which obtains module 502, for the length in multiple edge rectangular strip
When degree gradually increases and the length of multiple center rectangle item is gradually reduced, determine that the change in location of the pedestrian meets the default item
Part.
In one embodiment of the invention, each monitoring video frame is in a first direction using target length as dividing unit
Multiple regions equidistantly are divided into, which is the length of pixel in the first direction, which is line direction
Or column direction.
In one embodiment of the invention, the identification module 503, is used for: for each target monitoring image group, obtaining
The corresponding first crowd density map of each monitoring video frame in the target monitoring image group is taken, several the first crowd densities are obtained
Figure;Obtain the corresponding first light stream figure of image of pedestrian in multiple monitoring video frames that the target monitoring image group includes;Pass through
Crowd's event model carries out people to the target monitoring image group based on several the first crowd density maps and the first light stream figure
Group's event recognition.
In one embodiment of the invention, which is convolutional neural networks model.
The embodiment of the invention also provides another crowd's event recognition device 600, which is removed
It can also include training module 505 including crowd's event recognition device 500 outside the modules for including.
Wherein, the training module 505, is used for: for each crowd's occurrence diagram picture group in multiple crowd's event image group,
The second crowd density figure corresponding to every width crowd event image in crowd's occurrence diagram picture group is obtained, to obtain several the second people
Group's density map;Obtain the second light corresponding to the image of pedestrian in several crowd's event images that crowd's occurrence diagram picture group includes
Flow graph;Model training is carried out based on several the second crowd density figures and the second light stream figure, to obtain crowd's event model.
In conclusion crowd's event recognition device provided in an embodiment of the present invention, passes through the monitor video from target area
Middle at least one target monitoring image group of acquisition, and using preset crowd's event model to each target monitoring image group into
Pedestrian's group's event recognition, to determine whether target area crowd's event occurs according to recognition result, wherein each target monitoring figure
It include continuous multiple monitoring video frames as organizing, due to the generation of crowd's event and in a period of time, the moving situation of crowd has
It closes, and continuous multiple monitoring video frames can reflect the change in location of pedestrian in a period of time region of interest within, to reflect
Therefore the moving situation of crowd in a period of time region of interest within using continuous multiple monitoring video frames, that is to say and utilize mesh
Mark monitoring image group, which carries out crowd's event recognition, can be improved the accuracy of crowd's event recognition.
It should be understood that crowd's event recognition device provided by the above embodiment is when carrying out crowd's event recognition, only
The example of the division of the above functional modules, in practical application, can according to need and by above-mentioned function distribution by
Different functional modules is completed, i.e., the internal structure of device is divided into different functional modules, described above complete to complete
Portion or partial function.In addition, crowd's event recognition device provided by the above embodiment and crowd's event recognition method embodiment
Belong to same design, specific implementation process is detailed in embodiment of the method, and which is not described herein again.
Fig. 7 is the structural schematic diagram of a kind of electronic equipment shown according to an exemplary embodiment, which can be with
For server.The electronic equipment is for executing the crowd's event recognition method provided in Fig. 3 or Fig. 4 A illustrated embodiment.The electricity
Sub- equipment 700 includes central processing unit (CPU) 701 including random access memory (RAM) 702 and read-only memory (ROM)
703 system storage 704, and the system bus 705 of connection system storage 704 and central processing unit 701.The electricity
Sub- equipment 700 further includes the basic input/output (I/O system) that information is transmitted between each device helped in computer
706, and for the mass-memory unit 707 of storage program area 713, application program 714 and other program modules 715.
The basic input/output 706 includes display 708 for showing information and inputs letter for user
The input equipment 709 of such as mouse, keyboard etc of breath.Wherein the display 708 and input equipment 709 are all by being connected to
The input and output controller 710 of system bus 705 is connected to central processing unit 701.The basic input/output 706
Can also include input and output controller 710 with for receive and handle from keyboard, mouse or electronic touch pen etc. it is multiple its
The input of his equipment.Similarly, input and output controller 710 also provides output to display screen, printer or other kinds of defeated
Equipment out.
The mass-memory unit 707 is by being connected to the bulk memory controller (not shown) of system bus 705
It is connected to central processing unit 701.The mass-memory unit 707 and its associated computer-readable medium set for electronics
Standby 700 provide non-volatile memories.That is, the mass-memory unit 707 may include such as hard disk or CD-
The computer-readable medium (not shown) of ROM drive etc.
Without loss of generality, the computer-readable medium may include computer storage media and communication media.Computer
Storage medium includes information such as computer readable instructions, data structure, program module or other data for storage
The volatile and non-volatile of any method or technique realization, removable and irremovable medium.Computer storage medium includes
RAM, ROM, EPROM, EEPROM, flash memory or other solid-state storages its technologies, CD-ROM, DVD or other optical storages, tape
Box, tape, disk storage or other magnetic storage devices.Certainly, skilled person will appreciate that the computer storage medium
It is not limited to above-mentioned several.Above-mentioned system storage 704 and mass-memory unit 707 may be collectively referred to as memory.
According to various embodiments of the present invention, the electronic equipment 700 can also be connected to the network by internet etc.
Remote computer operation on to network.Namely electronic equipment 700 can be by the network that is connected on the system bus 705
Interface unit 711 is connected to network 712, in other words, Network Interface Unit 711 can be used also to be connected to other kinds of net
Network or remote computer system (not shown).
The memory further includes that one or more than one program, the one or more programs are stored in
In memory, central processing unit 701 realizes crowd's thing shown in Fig. 3 or Fig. 4 A by executing one or more programs
Part recognition methods.
In the exemplary embodiment, a kind of non-transitorycomputer readable storage medium including instruction, example are additionally provided
It such as include the memory of instruction, above-metioned instruction can be executed as the processor of electronic equipment to complete shown in each embodiment of the present invention
Crowd's event recognition method.For example, the non-transitorycomputer readable storage medium can be ROM, random access memory
Device (RAM), CD-ROM, tape, floppy disk and optical data storage devices etc..
Fig. 8 is a kind of block diagram of crowd's event recognition system 800 provided in an embodiment of the present invention, as shown in figure 8, the crowd
Event recognition system 800 may include: server 801 and monitoring device 802.
Wherein, server 801 is for executing crowd's event recognition method provided by Fig. 4 A illustrated embodiment.
Monitoring device 802 is sent to the clothes for the monitor video in photographic subjects region, and by the monitor video taken
Business device 801.
Those of ordinary skill in the art will appreciate that realizing that all or part of the steps of above-described embodiment can pass through hardware
It completes, relevant hardware can also be instructed to complete by program, the program can store in a kind of computer-readable
In storage medium, storage medium mentioned above can be read-only memory, disk or CD etc..
The foregoing is merely presently preferred embodiments of the present invention, is not intended to limit the invention, it is all in spirit of the invention and
Within principle, any modification, equivalent replacement, improvement and so on be should all be included in the protection scope of the present invention.
Claims (20)
1. a kind of crowd's event recognition method, which is characterized in that the described method includes:
Obtain the monitor video of target area;
At least one target monitoring image group is obtained from the monitor video, each target monitoring image group includes continuous
Multiple monitoring video frames;
Crowd's event recognition, crowd's thing are carried out to each target monitoring image group by preset crowd's event model
Part model is obtained according to the training of multiple crowd's event image groups, and each crowd's event image group includes reflection crowd's event hair
Several crowd's event images of raw process;
Determine whether the target area crowd's event occurs according to recognition result.
2. the method according to claim 1, wherein the monitor video includes multiple monitoring image groups, each
The monitoring image group includes continuous multiple monitoring video frames, and each monitoring video frame includes the image of pedestrian, described
At least one target monitoring image group is obtained from the monitor video, comprising:
For each monitoring image group, the image of the pedestrian in each monitoring video frame of the monitoring image group is obtained;
According to the image of the pedestrian in each monitoring video frame, judge whether the change in location of the pedestrian meets default item
Part, the preset condition are the change in location condition for reflecting pedestrian and assembling or dispersing;
When the change in location meets the preset condition, the monitoring image group is retrieved as the target monitoring image
Group.
3. according to the method described in claim 2, it is characterized in that, the pedestrian according in each monitoring video frame
Image, judges whether the change in location of the pedestrian meets preset condition, comprising:
Using identical generating mode, pedestrian's histogram is generated respectively for multiple monitoring video frames of the monitoring image group,
In, each monitoring video frame is equidistantly divided into multiple regions in a first direction, and each pedestrian's histogram includes
Multiple rectangular strips, position of the multiple rectangular strip in first axle and the multiple regions one of corresponding monitoring video frame are a pair of
It answers, each rectangular strip indicates the sum of the pixel value of pedestrian image in corresponding region in the length on the second axis, described
One axis and second axis are mutually perpendicular to;
Center rectangle item and edge rectangular strip are determined from each pedestrian's histogram, obtain multiple center rectangle items and multiple
Edge rectangular strip;
The position of the pedestrian is judged according to the length situation of change of the multiple center rectangle item and the multiple edge rectangular strip
Set whether variation meets the preset condition.
4. according to the method described in claim 3, it is characterized in that, the multiple center rectangle item and the multiple edge rectangle
The acquisition sequence consensus of multiple monitoring video frames of the distributing order of item and the monitoring image group, it is described according in the multiple
It is described that the length situation of change of centre rectangular strip and the multiple edge rectangular strip judges whether the change in location of the pedestrian meets
Preset condition, comprising:
When the length of the multiple center rectangle item gradually increases and the length of the multiple edge rectangular strip is gradually reduced, really
The change in location of the fixed pedestrian meets the preset condition.
5. according to the method described in claim 3, it is characterized in that, the multiple center rectangle item and the multiple edge rectangle
The acquisition sequence consensus of multiple monitoring video frames of the distributing order of item and the monitoring image group, it is described according in the multiple
Whether the length situation of change of centre rectangular strip and the multiple edge rectangular strip judges the multiple change in location for moving pedestrian
Meet the preset condition, comprising:
When the length of the multiple edge rectangular strip gradually increases and the length of the multiple center rectangle item is gradually reduced, really
The change in location of the fixed pedestrian meets the preset condition.
6. according to any method of claim 3 to 5, which is characterized in that
Each monitoring video frame is equidistantly divided into multiple regions by dividing unit of target length in a first direction, institute
Stating target length is the length of pixel in said first direction, and the first direction is line direction or column direction.
7. the method according to claim 1, wherein it is described by preset crowd's event model to each described
Target monitoring image group carries out crowd's event recognition, comprising:
For each target monitoring image group, each monitoring video frame in the target monitoring image group corresponding is obtained
One crowd density figure obtains several the first crowd density maps;
Obtain the corresponding first light stream figure of image of pedestrian in multiple monitoring video frames that the target monitoring image group includes;
By crowd's event model, based on several described the first crowd density maps and the first light stream figure to the target
Monitoring image group carries out crowd's event recognition.
8. the method according to claim 1, wherein it is described by preset crowd's event model to each described
Before target monitoring image group carries out crowd's event recognition, the method also includes:
For each crowd's occurrence diagram picture group in the multiple crowd's event image group, obtain every in crowd's event image group
Second crowd density figure corresponding to width crowd's event image, to obtain several the second crowd density figures;
Obtain the second light stream corresponding to the image of pedestrian in several crowd's event images that crowd's event image group includes
Figure;
Model training is carried out based on several described the second crowd density figures and the second light stream figure, to obtain crowd's event
Model.
9. the method according to claim 1, wherein crowd's event model is convolutional neural networks model.
10. a kind of crowd's event recognition device, which is characterized in that described device includes:
Video acquiring module, for obtaining the monitor video of target area;
Image group obtains module, for obtaining at least one target monitoring image group, each mesh from the monitor video
Marking monitoring image group includes continuous multiple monitoring video frames;
Identification module, for carrying out the knowledge of crowd's event to each target monitoring image group by preset crowd's event model
Not, crowd's event model is obtained according to the training of multiple crowd's event image groups, and each crowd's event image group includes
Several crowd's event images of reflection crowd's event generating process;
Determining module, for determining whether the target area crowd's event occurs according to recognition result.
11. device according to claim 10, which is characterized in that the monitor video includes multiple monitoring image groups, often
A monitoring image group includes continuous multiple monitoring video frames, and each monitoring video frame includes the image of pedestrian, institute
It states image group and obtains module, be used for:
For each monitoring image group, the image of the pedestrian in each monitoring video frame of the monitoring image group is obtained;
According to the image of the pedestrian in each monitoring video frame, judge whether the change in location of the pedestrian meets default item
Part, the preset condition are the change in location condition for reflecting pedestrian and assembling or dispersing;
When the change in location meets the preset condition, the monitoring image group is retrieved as the target monitoring image
Group.
12. device according to claim 11, which is characterized in that described image group obtains module, is used for:
Using identical generating mode, pedestrian's histogram is generated respectively for multiple monitoring video frames of the monitoring image group,
In, each monitoring video frame is equidistantly divided into multiple regions in a first direction, and each pedestrian's histogram includes
Multiple rectangular strips, position of the multiple rectangular strip in first axle and the multiple regions one of corresponding monitoring video frame are a pair of
It answers, each rectangular strip indicates the sum of the pixel value of pedestrian image in corresponding region in the length on the second axis, described
One axis and second axis are mutually perpendicular to;
Center rectangle item and edge rectangular strip are determined from each pedestrian's histogram, obtain multiple center rectangle items and multiple
Edge rectangular strip;
The position of the pedestrian is judged according to the length situation of change of the multiple center rectangle item and the multiple edge rectangular strip
Set whether variation meets the preset condition.
13. device according to claim 12, which is characterized in that the multiple center rectangle item and the multiple edge square
The acquisition sequence consensus of multiple monitoring video frames of the distributing order of shape item and the monitoring image group, described image group obtain mould
Block is used for:
When the length of the multiple center rectangle item gradually increases and the length of the multiple edge rectangular strip is gradually reduced, really
The change in location of the fixed pedestrian meets the preset condition.
14. device according to claim 12, which is characterized in that the multiple center rectangle item and the multiple edge square
The acquisition sequence consensus of multiple monitoring video frames of the distributing order of shape item and the monitoring image group, described image group obtain mould
Block is used for:
When the length of the multiple edge rectangular strip gradually increases and the length of the multiple center rectangle item is gradually reduced, really
The change in location of the fixed pedestrian meets the preset condition.
15. 2 to 14 any device according to claim 1, which is characterized in that each monitoring video frame is in first party
Multiple regions are equidistantly divided by dividing unit of target length upwards, the target length is pixel in the first direction
On length, the first direction be line direction or column direction.
16. device according to claim 10, which is characterized in that the identification module is used for:
For each target monitoring image group, each monitoring video frame in the target monitoring image group corresponding is obtained
One crowd density figure obtains several the first crowd density maps;
Obtain the corresponding first light stream figure of image of pedestrian in multiple monitoring video frames that the target monitoring image group includes;
By crowd's event model, based on several described the first crowd density maps and the first light stream figure to the target
Monitoring image group carries out crowd's event recognition.
17. device according to claim 10, which is characterized in that described device further includes training module, the trained mould
Block is used for:
For each crowd's occurrence diagram picture group in the multiple crowd's event image group, obtain every in crowd's event image group
Second crowd density figure corresponding to width crowd's event image, to obtain several the second crowd density figures;
Obtain the second light stream corresponding to the image of pedestrian in several crowd's event images that crowd's event image group includes
Figure;
Model training is carried out based on several described the second crowd density figures and the second light stream figure, to obtain crowd's event
Model.
18. device according to claim 10, which is characterized in that crowd's event model is convolutional neural networks mould
Type.
19. a kind of electronic equipment, which is characterized in that the electronic equipment includes processor and memory,
Wherein, the memory, for storing computer program;
The processor, for executing the program stored in the memory, to realize any people of claim 1 to 9
Group's event recognition method.
20. a kind of crowd's event recognition system, which is characterized in that crowd's event recognition system include as claim 10 to
18 any crowd's event recognition devices and monitoring device.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810340168.5A CN110390226B (en) | 2018-04-16 | 2018-04-16 | Crowd event identification method and device, electronic equipment and system |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810340168.5A CN110390226B (en) | 2018-04-16 | 2018-04-16 | Crowd event identification method and device, electronic equipment and system |
Publications (2)
Publication Number | Publication Date |
---|---|
CN110390226A true CN110390226A (en) | 2019-10-29 |
CN110390226B CN110390226B (en) | 2021-09-21 |
Family
ID=68283893
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810340168.5A Active CN110390226B (en) | 2018-04-16 | 2018-04-16 | Crowd event identification method and device, electronic equipment and system |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110390226B (en) |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111666821A (en) * | 2020-05-12 | 2020-09-15 | 深圳力维智联技术有限公司 | Personnel gathering detection method, device and equipment |
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 |
CN112464898A (en) * | 2020-12-15 | 2021-03-09 | 北京市商汤科技开发有限公司 | Event detection method and device, electronic equipment and storage medium |
Citations (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20080031491A1 (en) * | 2006-08-03 | 2008-02-07 | Honeywell International Inc. | Anomaly detection in a video system |
CN103854027A (en) * | 2013-10-23 | 2014-06-11 | 北京邮电大学 | Crowd behavior identification method |
CN104813339A (en) * | 2012-09-12 | 2015-07-29 | 威智伦富智堡公司 | Methods, devices and systems for detecting objects in a video |
CN104820824A (en) * | 2015-04-23 | 2015-08-05 | 南京邮电大学 | Local abnormal behavior detection method based on optical flow and space-time gradient |
CN105447458A (en) * | 2015-11-17 | 2016-03-30 | 深圳市商汤科技有限公司 | Large scale crowd video analysis system and method thereof |
CN106022244A (en) * | 2016-05-16 | 2016-10-12 | 广东工业大学 | Unsupervised crowd abnormity monitoring and positioning method based on recurrent neural network modeling |
US9489582B2 (en) * | 2014-01-27 | 2016-11-08 | Xerox Corporation | Video anomaly detection based upon a sparsity model |
CN107229894A (en) * | 2016-03-24 | 2017-10-03 | 上海宝信软件股份有限公司 | Intelligent video monitoring method and system based on computer vision analysis technology |
CN107480578A (en) * | 2016-06-08 | 2017-12-15 | 国家计算机网络与信息安全管理中心 | A kind of video detection system and method using crowd behaviour analysis |
CN107729799A (en) * | 2017-06-13 | 2018-02-23 | 银江股份有限公司 | Crowd's abnormal behaviour vision-based detection and analyzing and alarming system based on depth convolutional neural networks |
CN107911653A (en) * | 2017-11-16 | 2018-04-13 | 王磊 | The module of intelligent video monitoring in institute, system, method and storage medium |
-
2018
- 2018-04-16 CN CN201810340168.5A patent/CN110390226B/en active Active
Patent Citations (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20080031491A1 (en) * | 2006-08-03 | 2008-02-07 | Honeywell International Inc. | Anomaly detection in a video system |
CN104813339A (en) * | 2012-09-12 | 2015-07-29 | 威智伦富智堡公司 | Methods, devices and systems for detecting objects in a video |
CN103854027A (en) * | 2013-10-23 | 2014-06-11 | 北京邮电大学 | Crowd behavior identification method |
US9489582B2 (en) * | 2014-01-27 | 2016-11-08 | Xerox Corporation | Video anomaly detection based upon a sparsity model |
CN104820824A (en) * | 2015-04-23 | 2015-08-05 | 南京邮电大学 | Local abnormal behavior detection method based on optical flow and space-time gradient |
CN105447458A (en) * | 2015-11-17 | 2016-03-30 | 深圳市商汤科技有限公司 | Large scale crowd video analysis system and method thereof |
CN107229894A (en) * | 2016-03-24 | 2017-10-03 | 上海宝信软件股份有限公司 | Intelligent video monitoring method and system based on computer vision analysis technology |
CN106022244A (en) * | 2016-05-16 | 2016-10-12 | 广东工业大学 | Unsupervised crowd abnormity monitoring and positioning method based on recurrent neural network modeling |
CN107480578A (en) * | 2016-06-08 | 2017-12-15 | 国家计算机网络与信息安全管理中心 | A kind of video detection system and method using crowd behaviour analysis |
CN107729799A (en) * | 2017-06-13 | 2018-02-23 | 银江股份有限公司 | Crowd's abnormal behaviour vision-based detection and analyzing and alarming system based on depth convolutional neural networks |
CN107911653A (en) * | 2017-11-16 | 2018-04-13 | 王磊 | The module of intelligent video monitoring in institute, system, method and storage medium |
Non-Patent Citations (3)
Title |
---|
TIAN WANG 等: ""Abnormal event detection based on analysis of movement information of video sequence"", 《OPTIK》 * |
冯亚闯: ""视频中的异常事件检测算法研究"", 《中国博士学位论文全文数据库 信息科技辑》 * |
吴新宇 等: ""基于视频的人群异常事件检测综述"", 《电子测量与仪器学报》 * |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111666821A (en) * | 2020-05-12 | 2020-09-15 | 深圳力维智联技术有限公司 | Personnel gathering detection method, device and equipment |
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 |
CN112464898A (en) * | 2020-12-15 | 2021-03-09 | 北京市商汤科技开发有限公司 | Event detection method and device, electronic equipment and storage medium |
Also Published As
Publication number | Publication date |
---|---|
CN110390226B (en) | 2021-09-21 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Islam et al. | A blockchain-based fog computing framework for activity recognition as an application to e-Healthcare services | |
Qureshi et al. | Surveillance camera scheduling: A virtual vision approach | |
Qureshi et al. | Smart camera networks in virtual reality | |
Yang et al. | ImgSensingNet: UAV vision guided aerial-ground air quality sensing system | |
CN111312367A (en) | Campus personnel abnormal psychological prediction method based on self-adaptive cloud management platform | |
CN110390226A (en) | Crowd's event recognition method, device, electronic equipment and system | |
US10289884B2 (en) | Image analyzer, image analysis method, computer program product, and image analysis system | |
CN109189878B (en) | A kind of crowd's thermodynamic chart preparation method and device | |
CN112052837A (en) | Target detection method and device based on artificial intelligence | |
CN108022171B (en) | Data processing method and equipment | |
CN111444786B (en) | Crowd evacuation method, device and system based on unmanned aerial vehicle group and storage medium | |
CN113366542A (en) | Techniques for implementing augmented based normalized classified image analysis computing events | |
KR20220024986A (en) | Target tracking method and device, storage medium and computer program | |
CN111008631B (en) | Image association method and device, storage medium and electronic device | |
Giyenko et al. | Application of convolutional neural networks for visibility estimation of CCTV images | |
CA3193958A1 (en) | Processing images using self-attention based neural networks | |
US20220138347A1 (en) | Risk evaluation of transmission of pathogens in a space | |
CN102867214B (en) | Counting management method for people within area range | |
JPWO2020105146A1 (en) | Information processing equipment, control methods, and programs | |
CN111582654B (en) | Service quality evaluation method and device based on deep cycle neural network | |
CN106713864B (en) | A kind of method for processing video frequency and device | |
WO2024041108A1 (en) | Image correction model training method and apparatus, image correction method and apparatus, and computer device | |
Elouali et al. | Data transmission reduction formalization for cloud offloading-based IoT systems | |
CN104809438B (en) | A kind of method and apparatus for detecting electronic eyes | |
CN114820684A (en) | Image segmentation mask correction method, device, equipment and medium thereof |
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 |