CN106845325B - A kind of information detecting method and device - Google Patents
A kind of information detecting method and device Download PDFInfo
- Publication number
- CN106845325B CN106845325B CN201510885680.4A CN201510885680A CN106845325B CN 106845325 B CN106845325 B CN 106845325B CN 201510885680 A CN201510885680 A CN 201510885680A CN 106845325 B CN106845325 B CN 106845325B
- Authority
- CN
- China
- Prior art keywords
- region
- personage
- video frame
- target video
- suspicious
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
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
- 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
-
- 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
- G06V20/46—Extracting features or characteristics from the video content, e.g. video fingerprints, representative shots or key frames
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 discloses a kind of information detecting method and device, is related to technical field of image processing, wherein the above method includes: the personage detected in target video frame;Obtain character positions of the detected personage in reference frame, wherein the reference frame are as follows: the acquisition moment is located at the preset quantity video frame before the acquisition moment of the target video frame;According to preset Region detection algorithms, the suspicious region in the target video frame is detected;According to the character positions in detected personage and the reference frame, determine whether the suspicious region of the target video frame is region there are abnormal conditions.Infomation detection is carried out using scheme provided by the embodiments of the present application, the region in video frame there are abnormal conditions can be effectively detected, alleviate the operating pressure of staff.
Description
Technical field
This application involves technical field of image processing, in particular to a kind of information detecting method and device.
Background technique
With social progress and economic development, the business in each city is more and more flourishing, shop along street cumulative year after year.Through
Battalion person is often managed stand and is moved out of retail shop to outside retail shop to improve the turnover.Although in this way for operator with
Carried out benefit, but be easy to cause traffic jam, influenced city appearance, for this purpose, supervision of law enforcement personnel need to ceaselessly go on patrol with
Reduce the generation of such case.
Although operator can be reduced in such a way that law enfrocement official goes on patrol to move stand to outside retail shop out of retail shop
Phenomenon, but with the development of economy, retail shop is more and more, and the operating pressure of law enfrocement official is also increasing.
Summary of the invention
The embodiment of the present application discloses a kind of information detecting method and device, and to detect, there are abnormal conditions in video frame
Region mitigates the operating pressure of staff.
In order to achieve the above objectives, the embodiment of the present application discloses a kind of information detecting method, which comprises
Detect the personage in target video frame;
Obtain character positions of the detected personage in reference frame, wherein the reference frame are as follows: the acquisition moment is located at
Preset quantity video frame before the acquisition moment of the target video frame;
According to preset Region detection algorithms, the suspicious region in the target video frame is detected;
According to the character positions in detected personage and the reference frame, the suspicious of the target video frame is determined
Whether region is region there are abnormal conditions.
In a kind of specific implementation of the application, in the personage according to detected by and the reference frame
Character positions determine whether the suspicious region of the target video frame is region there are abnormal conditions, comprising:
Obtain character positions of the detected personage in the target video frame;
According to the character positions in the character positions and the reference frame in the target video frame, determine detected
Suspicious figure in personage;
Determine the suspicious region in the reference frame;
Obtain region corresponding with the suspicious region of the target video frame in the suspicious region of the reference frame;
Calculated in the target video frame between the suspicious figure and the suspicious region of the target video frame away from
From, and the distance between the suspicious figure and region obtained are calculated in the reference frame;
According to calculated distance, determine whether the suspicious region of the target video frame is area there are abnormal conditions
Domain.
In a kind of specific implementation of the application, character positions according in the target video frame and described
Character positions in reference frame determine the suspicious figure in detected personage, comprising:
Judge whether to meet expression formula one,
Wherein, the expression formula one are as follows:
meand> TmAnd vard< Tv1,
meandIndicate the mean value of the mean deviation distance of detected each personage, vardIt is each detected by indicating
The variance of the mean deviation distance of personage, TmIndicate preset mean value threshold value, Tv1It indicates preset first variance threshold value, is detected
The mean deviation distance of any personage i out indicates that the personage i is acquired in the target video frame and the reference frame
The average value of moving distance between moment adjacent two frames;
If not satisfied, judge whether to meet expression formula two,
Wherein, the expression formula two are as follows:
varp< Tv2And vard< Tv3,
varpIndicate the variance of the character positions in the target video frame, Tv2Indicate preset second variance threshold value, Tv3
Indicate preset third variance threshold values;
If not satisfied, mobile slow personage is determined from detected personage according to expression formula three, and according to expression formula
Four determine suspicious figure from the slow personage of the movement,
Wherein, the expression formula three are as follows:
frmd> Tcnt1And distbias< Tdist1,
frmdAny personage j in personage detected by indicating goes out in the target video frame and the reference frame
Existing number, distbiasIndicate the mean deviation distance of the personage j, Tdist1Indicate preset first offset distance threshold value,
Tcnt1Indicate preset first amount threshold;
The expression formula four are as follows:
distrela< Tdist2And frms> Tcnt2,
distrelaIndicate position and the pos of the slow personage k of any movementaverThe distance between, posaverIndicate each shifting
The history deflection for moving slow personage most locates position pos slowlyparLocation mean value, posparIndicate the slow personage k of any movement
The position in video frame f is acquired, the personage k is moved to the moving distance of the video frame f from the former frame of the video frame f
Are as follows: the minimum value of moving distance of the personage k between acquisition moment adjacent two video frames acquired, frmsIndicate institute
State the number that personage k is confirmed as slowly mobile personage in the video frame acquired, Tdist2Indicate preset second offset distance
From threshold value, Tcnt2Indicate preset second amount threshold.
It is described according to preset Region detection algorithms in a kind of specific implementation of the application, detect the target
Suspicious region in video frame, comprising:
According to preset regional model, the suspicious region in the target video frame is detected;Or
According to the region in the reference frame there are abnormal conditions, the suspicious region in the target video frame is determined;Or
According to preset regional model, the alternative suspicious region in the target video frame is detected;And according to the reference
Frame and the alternative suspicious region detected, determine the suspicious region in the target video frame.
In a kind of specific implementation of the application, the region according in the reference frame there are abnormal conditions,
Determine the suspicious region in the target video frame, comprising:
According to the acquisition moment, the target video frame F is obtainedCPrevious video frame FFIt is middle that there are the regions of abnormal conditions;
Determine the video frame FFIt is middle that there are the regions of abnormal conditions in the target video frame FCIn corresponding region;
Belong to the number of the pixel in the sport foreground region of the target video frame in each region determined by counting;
Judge whether the number for each pixel that statistics obtains meets following formula:
NumFAm/NumTAm< ThP,
Wherein, NumFamBelong to the picture in the sport foreground region of the target video frame in any region m determined by indicating
The number of vegetarian refreshments, NumTamIndicate the total number of pixel in the region m, ThPIndicate preset first pixel number ratio
Threshold value;
If satisfied, determining that the corresponding region of pixel number is the suspicious region in the target video frame.
It is described according to preset regional model in a kind of specific implementation of the application, detect the target video
Alternative suspicious region in frame, comprising:
Determine the foreground area in the target video frame;
According to preset regional model, alternative suspicious region is detected in the foreground area.
It is described according to the reference frame and the alternative suspicious area detected in a kind of specific implementation of the application
Domain determines the suspicious region in the target video frame, comprising:
Determine the target video frame FCPrevious video frame FFIn area identical with the alternative suspicious region position detected
Domain;
Belong to the video frame F in the pixel of each region determined by obtainingFIt is middle there are the region of abnormal conditions
The number of pixel;
Determine that meeting the corresponding alternative suspicious region in region of following formula in above-mentioned identified region is the mesh
The suspicious region in video frame is marked,
Na/Nt> TN,
Wherein, NaBelong to the video frame F in any region S determined by indicatingFIt is middle that there are the pictures in the region of abnormal conditions
The number of vegetarian refreshments, NtIndicate the number of pixel in the region S, TNIndicate preset second pixel number proportion threshold value.
Personage in a kind of specific implementation of the application, in the detection target video frame, comprising:
According to current time and/or present intensity, person detecting model is selected from preset person detecting model library;
According to the personage in selected person detecting model inspection target video frame.
In a kind of specific implementation of the application, personage position of the personage detected by the acquisition in reference frame
It sets, comprising:
Obtain personage associated with detected personage in reference frame;
According to the similarity degree and movement velocity between detected personage personage associated with it, calculating is detected
The confidence level of personage out, wherein the confidence level, for indicating that a personage personage associated there is same personage's
Credibility;
It obtains confidence level in detected personage and is greater than the personage of preset confidence threshold value in the reference frame
Character positions;
The personage according to detected by and the character positions in the reference frame, determine the target video frame
Whether suspicious region is region there are abnormal conditions, comprising:
It is greater than in personage and the reference frame of preset confidence threshold value according to confidence level in detected personage
Character positions, determine whether the suspicious region of the target video frame is region there are abnormal conditions.
In a kind of specific implementation of the application, the information detecting method further include:
In the suspicious region for determining the target video frame, there are in the case where abnormal conditions, obtain the suspicious region
It is confirmed as the number in the region there are abnormal conditions in the video frame of acquisition;
In the case where number obtained meets preset monitoring condition, prompt messages are sent.
In order to achieve the above objectives, the embodiment of the present application discloses a kind of information detector, and described device includes:
Person detecting module, for detecting the personage in target video frame;
Position obtains module, for obtaining character positions of the detected personage in reference frame, wherein the reference
Frame are as follows: the acquisition moment is located at the preset quantity video frame before the acquisition moment of the target video frame;
Suspicious region detection module, for according to preset Region detection algorithms, detect in the target video frame can
Doubt region;
Abnormal area determining module, for according to the character positions in detected personage and the reference frame, really
Whether the suspicious region of the fixed target video frame is region there are abnormal conditions.
In a kind of specific implementation of the application, the abnormal area determining module, comprising:
Character positions obtain submodule, for obtaining personage position of the detected personage in the target video frame
It sets;
Suspicious figure determines submodule, for according in the character positions and the reference frame in the target video frame
Character positions determine the suspicious figure in detected personage;
Corresponding region obtains submodule, in the suspicious region for obtaining the reference frame with the target video frame can
Doubt the corresponding region in region;
Apart from computational submodule, for calculating the suspicious figure and the target video frame in the target video frame
The distance between suspicious region, and calculate in the reference frame between the suspicious figure and region obtained away from
From;
Abnormal area determines submodule, for determining whether the suspicious region of the target video frame is that there are abnormal conditions
Region.
In a kind of specific implementation of the application, the suspicious figure determines submodule, comprising:
First information judging unit meets expression formula one for judging whether,
Wherein, the expression formula one are as follows:
meand> TmAnd vard< Tv1,
meandIndicate the mean value of the mean deviation distance of detected each personage, vardIt is each detected by indicating
The variance of the mean deviation distance of personage, TmIndicate preset mean value threshold value, Tv1It indicates preset first variance threshold value, is detected
The mean deviation distance of any personage i out indicates that the personage i is acquired in the target video frame and the reference frame
The average value of moving distance between moment adjacent two frames;
Second information judging unit, for sentencing in the case where the judging result of the first information judging unit is no
It is disconnected whether to meet expression formula two,
Wherein, the expression formula two are as follows:
varp< Tv2And vard< Tv3,
varpIndicate the variance of the character positions in the target video frame, Tv2Indicate preset second variance threshold value, Tv3
Indicate preset third variance threshold values;
Suspicious figure's determination unit, for the judging result of second information judging unit be it is no in the case where, root
Determine mobile slow personage from detected personage according to expression formula three, and according to expression formula four from the slow personage of the movement
Middle determining suspicious figure,
Wherein, the expression formula three are as follows:
frmd> Tcnt1And distbias< Tdist1,
frmdAny personage j in personage detected by indicating goes out in the target video frame and the reference frame
Existing number, distbiasIndicate the mean deviation distance of the personage j, Tdist1Indicate preset first offset distance threshold value,
Tcnt1Indicate preset first amount threshold;
The expression formula four are as follows:
distrela< Tdist2And frms> Tcnt2,
distrelaIndicate position and the pos of the slow personage k of any movementaverThe distance between, posaverIndicate each shifting
The history deflection for moving slow personage most locates position pos slowlyparLocation mean value, posparIndicate the slow personage k of any movement
The position in video frame f is acquired, the personage k is moved to the moving distance of the video frame f from the former frame of the video frame f
Are as follows: the minimum value of moving distance of the personage k between acquisition moment adjacent two video frames acquired, frmsIndicate institute
State the number that personage k is confirmed as slowly mobile personage in the video frame acquired, Tdist2Indicate preset second offset distance
From threshold value, Tcnt2Indicate preset second amount threshold.
In a kind of specific implementation of the application, the suspicious region detection module is specifically used for according to preset
Regional model detects the suspicious region in the target video frame;Or
The suspicious region detection module, specifically for determining according to the region in the reference frame there are abnormal conditions
Suspicious region in the target video frame;Or
The suspicious region detection module, comprising:
Alternative suspicious region detection sub-module, for detecting in the target video frame according to preset regional model
Alternative suspicious region;
First suspicious region determines submodule, for determining according to the reference frame and the alternative suspicious region detected
Suspicious region in the target video frame.
In a kind of specific implementation of the application, the suspicious region detection module, comprising:
Abnormal area obtains submodule, for obtaining the target video frame F according to the acquisition momentCPrevious video frame FF
It is middle that there are the regions of abnormal conditions;
Corresponding region determines submodule, for determining the video frame FFIt is middle that there are the regions of abnormal conditions in the target
Corresponding region in video frame;
Pixel number statistic submodule, for counting the fortune for belonging to the target video frame in identified each region
The number of the pixel of dynamic foreground area;
Ratio judging submodule, for judging whether the number for counting obtained each pixel meets following formula:
NumFAm/NumTAm< ThP,
Wherein, NumFamBelong to the picture in the sport foreground region of the target video frame in any region m determined by indicating
The number of vegetarian refreshments, NumTamIndicate the total number of pixel in the region m, ThPIndicate preset first pixel number ratio
Threshold value;
Second suspicious region determines submodule, is the case where being for the judging result in the ratio judging submodule
Under, determine that the corresponding region of pixel number is the suspicious region in the target video frame.
In a kind of specific implementation of the application, the alternative suspicious region detection sub-module, comprising:
Foreground area determination unit, for determining the foreground area in the target video frame;
Alternative suspicious region detection unit, for being detected in the foreground area alternative according to preset regional model
Suspicious region.
In a kind of specific implementation of the application, first suspicious region determines submodule, comprising:
Position same area determination unit, for determining the target video frame FCPrevious video frame FFIn with detect
The identical region in alternative suspicious region position;
Pixel number obtaining unit belongs to the video frame F in the pixel for each region determined by obtainingF
It is middle that there are the numbers of the pixel in the region of abnormal conditions;
Suspicious region determination unit, for determining that the region for meeting following formula in above-mentioned identified region is corresponding
Alternative suspicious region is the suspicious region in the target video frame,
Na/Nt> TN,
Wherein, NaBelong to the video frame F in any region S determined by indicatingFIt is middle that there are the pictures in the region of abnormal conditions
The number of vegetarian refreshments, NtIndicate the number of pixel in the region S, TNIndicate preset second pixel number proportion threshold value.
In a kind of specific implementation of the application, the person detecting module, comprising:
Person detecting model selects submodule, is used for according to current time and/or present intensity, from preset person detecting
Person detecting model is selected in model library;
Person detecting submodule, for according to the personage in selected person detecting model inspection target video frame.
In a kind of specific implementation of the application, the position obtains module, comprising:
It is associated with personage and obtains submodule, for obtaining personage associated with detected personage in reference frame;
Confidence calculations submodule, for according to the similarity degree between detected personage personage associated with it with
And movement velocity, calculate the confidence level of detected personage, wherein the confidence level is associated therewith for indicating a personage
The personage of connection is the credibility of same personage;
Position obtains submodule, the people for being greater than preset confidence threshold value for obtaining confidence level in detected personage
The character positions of object in the reference frame;
The abnormal area determining module, specifically for being greater than preset confidence according to confidence level in detected personage
The personage for spending threshold value and the character positions in the reference frame, determine whether the suspicious region of the target video frame is presence
The region of abnormal conditions.
In a kind of specific implementation of the application, the information detector further include:
Number obtains module, for determine the suspicious region of the target video frame there are in the case where abnormal conditions,
Obtain the number in the region that the suspicious region is confirmed as there are abnormal conditions in the video frame acquired;
Prompt messages sending module, for sending out in the case where number obtained meets preset monitoring condition
Send prompt messages.
As seen from the above, in scheme provided by the embodiments of the present application, first detection target video frame in personage and this
A little character positions of the personage in reference frame, then the suspicious region in target video frame is detected, finally combine detected people
Character positions in object and reference frame determine whether the suspicious region of target video frame is region there are abnormal conditions.It can
See and scheme provided by the embodiments of the present application is applied to be capable of detecting when the region in video frame there are abnormal conditions, is not necessarily to staff
The region there are abnormal conditions is found by way of patrol, can reduce the operating pressure of staff.
Detailed description of the invention
In order to illustrate the technical solutions in the embodiments of the present application or in the prior art more clearly, to embodiment or will show below
There is attached drawing needed in technical description to be briefly described, it should be apparent that, the accompanying drawings in the following description is only this
Some embodiments of application for those of ordinary skill in the art without creative efforts, can be with
It obtains other drawings based on these drawings.
Fig. 1 is the flow diagram of the first information detecting method provided by the embodiments of the present application;
Fig. 2 is the flow diagram of second of information detecting method provided by the embodiments of the present application;
Fig. 3 is the flow diagram of the third information detecting method provided by the embodiments of the present application;
Fig. 4 is the flow diagram of the 4th kind of information detecting method provided by the embodiments of the present application;
Fig. 5 is the structural schematic diagram of the first information detector provided by the embodiments of the present application;
Fig. 6 is the structural schematic diagram of second of information detector provided by the embodiments of the present application;
Fig. 7 is the structural schematic diagram of the third information detector provided by the embodiments of the present application;
Fig. 8 is the structural schematic diagram of the 4th kind of information detector 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, it is clear that 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, it is obtained by those of ordinary skill in the art without making creative efforts every other
Embodiment shall fall in the protection scope of this application.
Fig. 1 is the flow diagram of the first information detecting method provided by the embodiments of the present application, this method comprises:
S101: the personage in detection target video frame.
Above-mentioned target video frame can be the image collected in real time by image capture device, wherein Image Acquisition
Equipment can be the image capture devices such as common ball machine in video monitoring.Those skilled in the art it is understood that
Multiple presetting bits can be set for a ball machine in practical application, it is corresponding to each presetting bit respectively according to the preset time interval
Region carry out Image Acquisition.It should be noted that in the case where above-mentioned image capture device is ball machine, using the application reality
When the scheme progress infomation detection of example offer is provided, it can be understood as carry out information in the acquired image of a presetting bit for ball machine
The case where detection.
Certainly, above-mentioned image capture device is not limited in ball machine, and the application is defined not to this.
It should be understood that detection target video frame in personage when, can by the human detection method based on statistics into
Row detection needs to construct model by the human detection method based on statistics when being detected, and by constructing model when institute structure
The factors such as the accuracy of established model and the scene of selected sample, intensity of illumination, brightness are related, therefore, can construct multiple moulds
Then type selects different models to be detected at different conditions, to improve the accuracy rate of person detecting.
It, can be first according to current time and/or present intensity, from pre- specifically, when personage in detection target video frame
If person detecting model library in select person detecting model, then according to selected person detecting model inspection target video
Personage in frame.
For example, may include using daytime in above-mentioned preset person detecting model library as the pedestrian dummy of background and with night
For the pedestrian dummy etc. of background, it is, of course, also possible to include using fine day as the pedestrian dummy of background, using the cloudy day as the pedestrian of background
Model etc., the application are defined not to this.
S102: character positions of the detected personage in reference frame are obtained.
Wherein, above-mentioned reference frame it is to be understood that acquisition the moment be located at target video frame acquisition the moment before preset
The value of quantity video frame, above-mentioned preset quantity is generally higher than equal to 1.
Since image capture device is to carry out Image Acquisition according to fixed time interval, and between this set time
Every typically small, for example, 0.04 second etc., so, it is usually identical when image capture device carries out Image Acquisition to Same Scene
Personage appear in continuous multiple frames image, so, the personage in S101 in detected target video frame generally can be
Acquisition the moment be located at the acquisition moment of target video frame before several video frames in occur.
Above-mentioned preset quantity can be determined according to practical situations, for example, it may be 5,10,20 etc..
It, can be according to these location informations obtained after obtaining character positions of the detected personage in reference frame
The information such as the detected motion profile of each personage are analyzed, may include: the movement of personage in the motion track information
The information such as path, movement velocity, further can analyze out the personage according to the motion profile of personage is to fast forward through, slowly
Speed advances, hovers or static etc..
It should be noted that above-mentioned reference frame and target video frame are the video frames for same presetting bit, or it is referred to as
For the video frame for Same Scene.
S103: according to preset Region detection algorithms, the suspicious region in target video frame is detected.
It should be noted that Region detection algorithms belong to the technology of comparative maturity, those skilled in the art being capable of root
It is easy to know a variety of Region detection algorithms according to the prior art, no longer repeat one by one here, more typical Region detection algorithms have CNN
(Convolutional Neural Networks, convolutional neural networks) Region detection algorithms etc..
Above-mentioned suspicious region can be doubtful operator in the stand region of outdoor setting, the parking stall area in doubtful parking lot
Domain etc., the application are not defined the specific appearance form of suspicious region.
Specifically, according to preset Region detection algorithms, it, can be according to pre- when detecting the suspicious region in target video frame
If regional model, detect the suspicious region in the target video frame, in which, target video frame can be carry out image
The first frame of acquisition, naturally it is also possible to be the non-first frame for carrying out Image Acquisition, the application is defined not to this.
In addition, those skilled in the art are it is understood that have the time between acquisition moment adjacent video frame
Correlation and spatial coherence, that is to say, that can be with reference target video frame when detecting the suspicious region in target video frame
Reference frame or in which region there are abnormal conditions, wherein the frame number of referenced reference frame can be a frame, be also possible to more
Frame, the application not to specific reference to frame number be defined.
In a kind of specific implementation of the application, according to the region in reference frame there are abnormal conditions, target is determined
When suspicious region in video frame, target video frame F can be obtained first according to the acquisition momentCPrevious video frame FFIn there are different
The region of reason condition determines video frame FFIt is middle that there are the regions of abnormal conditions in target video frame FCIn corresponding region, statistics is true
Belong to the number of the pixel in the sport foreground region of target video frame in fixed each region, then judgement statistics obtains every
Whether the number of one pixel meets following formula: NumFAm/NumTAm< ThP, if satisfied, determining that the pixel number is corresponding
Region be target video frame in suspicious region.
For the case where above-mentioned abnormal conditions are to manage stand, since stand is usually static constant, so, when
When judging to count obtained any pixel point number to meet above-mentioned expression formula, illustrate the pixel number pair in target video frame
A possibility that region answered is static stand is higher, hence, it can be determined that the corresponding region of pixel number is target video
Suspicious region in frame.
It should be noted that the specific implementation is mainly used in the case where non-first frame that Image Acquisition obtains.
Those skilled in the art are for a long time under normal circumstances it is understood that for static stand
It remains static, so same static stand region generally will appear and be in multiple continuous video frames
Stationary state in a kind of relatively good implementation of the application, can also judge obtained each in light of this situation
In the number of the pixel in the corresponding sport foreground region for belonging to target video frame in region and the region between pixel sum
Ratio be less than preset first pixel number proportion threshold value after, judge the region acquisition the moment be located at target video frame
Whether the number for being confirmed as the region there are abnormal conditions in video frame before is greater than preset numerical value, if more than then really
The fixed region is the suspicious region in target video frame in the corresponding region in target video frame.
Since for Same Scene, background area is almost unchanged, so, the movement in above-mentioned target video frame
Foreground area can use to be obtained in such a way that the background model that adaptive mode constructs extracts sport foreground region, in addition,
For the accuracy for guaranteeing background model, above-mentioned back can also be constantly updated according to image capture device each frame image collected
Scape model.
In addition, referring to fig. 2, providing second of information detecting method in another specific implementation of the application
Flow diagram in the present embodiment, according to preset Region detection algorithms, detect target video compared with previous embodiment
Suspicious region (S103) in frame, comprising:
S103A: according to preset regional model, the alternative suspicious region in target video frame is detected.
It is similar with person detecting is carried out in S101 in target video frame, detect the alternative suspicious region in target video frame
When can also be detected by model matching method, above-mentioned preset regional model can be to be constructed in advance.
For example, multiple operators can be obtained ahead of time in the image information of the stand of outdoor setting, machine may then pass through
Device study the methods of building stand regional model, wherein in each image information being obtained ahead of time stand it is more representative more
Help to construct preferable stand model, in addition, the image information being obtained ahead of time is more, more facilitates to construct preferable stand mould
Type.
Under normal conditions, image capture device carries out Image Acquisition for Same Scene, image acquired in this way
Background is almost unchanged, for example, street architecture object, retail shop's shop front etc., and prospect be then it is continually changing, therefore, detect target
It can be detected based on the foreground area of target video when alternative suspicious region in video, specifically, according to preset area
Domain model can first determine the foreground area in target video frame, then when detecting the alternative suspicious region in target video frame
According to preset regional model, alternative suspicious region is detected in above-mentioned foreground area.
Since for Same Scene, background area is almost unchanged, so, it can be not present in practical application
Image is acquired when abnormal conditions in advance as background model, when carrying out infomation detection later, by institute's acquired image with
Above-mentioned background model carries out Difference Calculation, is partitioned into foreground area from institute's acquired image.
S103B: according to reference frame and the alternative suspicious region detected, the suspicious region in target video frame is determined.
It when determining the suspicious region in target video frame, can be determined according to predeterminated position method, specifically, if standby
Suspicious region is selected to match with preset location information, then it is assumed that the alternative suspicious region is the suspicious area in target video frame
Domain, above-mentioned preset location information can be according to the experience of operation and maintenance personnel setting etc..
In a kind of relatively good implementation of the application, according to reference frame and the alternative suspicious region detected, mesh is determined
When marking the suspicious region in video frame, target video frame F can be first determinedCPrevious video frame FFIn with detect it is alternative can
The identical region of regional location is doubted, obtains in the pixel of identified each region and belongs to video frame FFIn there are abnormal conditions
Region pixel number, then determine it is above-mentioned determined by meet the region of following formula in region and corresponding alternatively may be used
Doubting region is the suspicious region in target video frame,
Na/Nt> TN,
Wherein, NaBelong to video frame F in any region S determined by indicatingFIt is middle that there are the pixels in the region of abnormal conditions
Number, NtIndicate the number of pixel in the S of region, TNIndicate preset second pixel number proportion threshold value.
It is stationary, thus above-mentioned there are abnormal conditions under normal conditions specifically, for managing stand
Region can be understood as the region with this stationary characteristic, in this case Na/Nt> TNWhen, it is believed that suspicious area
Domain is fixed, if it is fixed for being detected a certain region on the basis of meeting above-mentioned relation formula in continuous several frames
Motionless, it is believed that the region is doubtful stand region.
S104: according to the character positions in detected personage and reference frame, the suspicious area of target video frame is determined
Whether domain is region there are abnormal conditions.
It should be understood that above-mentioned suspicious region might not really there are the regions of abnormal conditions, so also need into
One step judges whether the suspicious region of target video frame is region there are abnormal conditions.For example, above-mentioned abnormal conditions can be
Operator is the stand of outdoor setting the case where, the case where the vehicle in parking lot crosses parking stall etc., and the application is not to upper
The specific appearance form for stating abnormal conditions is defined.
Specifically, there are the characteristics that in practical application the regions of abnormal conditions in addition to its own mutually outside the Pass, it is also possible to meeting
It is related to surrounding personage, for example, operator is normally near stand, and hover near stand etc..
In view of the foregoing, in a kind of optional implementation of the application, according to detected personage and reference
Character positions in frame can be obtained first when whether the suspicious region for determining target video frame is the region there are abnormal conditions
Character positions of the detected personage in target video frame, according in the character positions and reference frame in target video frame
Character positions determine the suspicious figure in detected personage, then determine the suspicious region in reference frame, and obtain reference
Region corresponding with the suspicious region of target video frame in the suspicious region of frame calculates suspicious figure and mesh in target video frame
Mark the distance between the suspicious region of video frame, and calculated in reference frame between suspicious figure and region obtained away from
From determining whether the suspicious region of target video frame is region there are abnormal conditions finally according to calculated distance.
It should be noted that scheme provided in this embodiment carries out infomation detection for target video frame, detected
Cheng Zhong can detect the suspicious region in target video frame in S103.For reference frame, since reference frame is the acquisition moment
Video frame before the acquisition moment of target video frame, so when carrying out infomation detection to target video frame, to ginseng
It examines frame and carried out identical infomation detection, detection process is identical as to the target video frame progress process of infomation detection, therefore,
For the suspicious region that can obtain reference frame during the infomation detection of reference frame.For convenience of each frame conduct in video sequence
The reference frame of subsequent video frame, after carrying out infomation detection to each frame in video sequence, what can first be will test should
The information of the suspicious region of video frame is locally stored, when reference frame of the video frame as other video frames, Ke Yizhi
Connect the information that the suspicious region of the video frame is obtained from the information being locally stored.
Optionally, according to calculated distance, determine whether the suspicious region of target video frame is that there are abnormal conditions
Region when, can distance obtained by calculation mean value, determine whether the suspicious region of target video frame is that there are exceptions
The region of situation, specifically, can be in the case where the mean value of above-mentioned calculated distance is less than preset distance threshold,
The suspicious region for thinking target video frame is the region there are abnormal conditions, opposite, in the equal of above-mentioned calculated distance
Value is not less than in the case where preset distance threshold, it is believed that the suspicious region of target video frame is that there is no the areas of abnormal conditions
Domain.
In addition, when whether the suspicious region for determining target video frame is the region there are abnormal conditions, in addition to can root
Except being determined according to the mean value of above-mentioned calculated distance, it is also contemplated that the information such as variance of above-mentioned calculated distance,
The application is defined not to this.
In a kind of relatively good implementation of the application, according to the people in the character positions and reference frame in target video frame
Object location determines the suspicious figure in detected personage, can be determined by following steps:
Judge whether to meet expression formula one,
Wherein, above-mentioned expression formula one are as follows:
meand> TmAnd vard< Tv1,
meandIndicate the mean value of the mean deviation distance of detected each personage, vardIt is each detected by indicating
The variance of the mean deviation distance of personage, TmIndicate preset mean value threshold value, Tv1It indicates preset first variance threshold value, is detected
The mean deviation distance of any personage i out indicates that personage i acquires moment adjacent two in target video frame and reference frame
The average value of moving distance between frame;
Meeting above-mentioned expression formula for the moment, it can be understood as personage is mobile in collective;
And it is being unsatisfactory for above-mentioned expression formula for the moment, it continues to determine whether to meet expression formula two,
Wherein, above-mentioned expression formula two are as follows:
varp< Tv2And vard< Tv3,
varpIndicate the variance of the character positions in target video frame, Tv2Indicate preset second variance threshold value, Tv3It indicates
Preset third variance threshold values;
When meeting above-mentioned expression formula two, it can be understood as current scene is personage's focusing field scape;
When being unsatisfactory for expression formula two, mobile slow personage is determined from detected personage according to expression formula three, and
Suspicious figure is determined from the slow personage of above-mentioned movement according to expression formula four,
Wherein, above-mentioned expression formula three are as follows:
frmd> Tcnt1And distbias< Tdist1,
frmdTime that any personage j in personage detected by indicating occurs in target video frame and reference frame
Number, distbiasIndicate the mean deviation distance of personage j, Tdist1Indicate preset first offset distance threshold value, Tcnt1Indicate default
The first amount threshold;
Above-mentioned expression formula four are as follows:
distrela< Tdist2And frms> Tcnt2,
distrelaIndicate position and the pos of the slow personage k of any movementaverThe distance between, posaverIndicate each shifting
The history deflection for moving slow personage most locates position pos slowlyparLocation mean value, posparIndicate the slow personage k of any movement
The position in video frame f is acquired, personage k is moved to the moving distance of video frame f from the former frame of video frame f are as follows: personage k is
The minimum value of moving distance between acquisition moment adjacent two video frames of acquisition, frmsIndicate personage k in the view acquired
It is confirmed as the number of slowly mobile personage, T in frequency framedist2Indicate preset second offset distance threshold value, Tcnt2It indicates
Preset second amount threshold.
It is operator in the case where the stand of outdoor setting in above-mentioned abnormal conditions in practical application, above-mentioned suspicious people
Object can be understood as operator, stall owner, above-mentioned in the case where the vehicle that above-mentioned abnormal conditions are parking lot crosses parking stall
It can be understood as car owner with personage.
It should be noted that embodiment illustrated in fig. 2 is also mainly used in the case where non-first frame that Image Acquisition obtains.
As seen from the above, personage and these people in scheme provided in this embodiment, first in detection target video frame
Character positions of the object in reference frame, then detect the suspicious region in target video frame, finally combine detected personage with
And the character positions in reference frame, determine whether the suspicious region of target video frame is region there are abnormal conditions.It can be seen that answering
The region in video frame there are abnormal conditions is capable of detecting when with scheme provided in this embodiment, passes through patrol without staff
Mode find the region there are abnormal conditions, can reduce the operating pressure of staff.
In another specific implementation of the application, referring to Fig. 3, the process of the third information detecting method is provided
Schematic diagram in the present embodiment, obtains detected personage at the acquisition moment and is located at target video frame compared with previous embodiment
Acquisition the moment before preset quantity video frame in position (S102), comprising:
S102A: personage associated with detected personage in reference frame is obtained.
It is and above-mentioned pre- under normal conditions since image capture device is progress Image Acquisition according to the preset time interval
If time interval it is typically small, so the probability that same personage appears in continuous multiple frames image is higher, but due to shooting
The influence of the factors such as light, shooting angle, the travel speed of personage may be difficult to accurate detection under normal conditions and go out same people
Position of the object in different images, in view of the foregoing, in a kind of situation, above-mentioned personage associated with detected personage
Can be understood as may be with detected personage same personage personage.
S102B: it according to the similarity degree and movement velocity between detected personage personage associated with it, calculates
The confidence level of detected personage.
Wherein, above-mentioned confidence level, for indicating that a personage personage associated there is the credibility of same personage.
S102C: it obtains confidence level in detected personage and is greater than the personage of preset confidence threshold value in reference frame
Character positions.
Specifically, can by confidence level in detected personage greater than preset confidence threshold value Th1 personage and
The segment that associated personage of the personage in reference frame is constituted is referred to as high confidence level segment, on the contrary, by detected people
Associated personage structure of personage and the personage of the confidence level less than another preset confidence threshold value Th2 in reference frame in object
At segment be referred to as low confidence segment.Further, subsequent can be according to above-mentioned high confidence level segment and low confidence
Segment determines whether the suspicious region of target video frame is region there are abnormal conditions.
It should be noted that above-mentioned two confidence threshold value can be equal, and it can also be unequal, specifically, Th2 can be small
In Th1.
According to the character positions in detected personage and reference frame, determine target video frame suspicious region whether
For there are the region of abnormal conditions (S104), comprising:
S104A: it is greater than personage and the reference frame of preset confidence threshold value according to confidence level in detected personage
In character positions, determine whether the suspicious region of target video frame is region there are abnormal conditions.
As seen from the above, in scheme provided in this embodiment, it is greater than according to confidence level in detected personage preset
The personage of confidence threshold value and position obtained determine whether the suspicious region of target video frame is that there are abnormal conditions
Region, rather than be determined according to detected all persons, therefore can subtract on the basis of guaranteeing that definitive result is accurate
Few calculation amount, improves calculating speed.
In another specific implementation of the application, referring to fig. 4, the process of the 4th kind of information detecting method is provided
Schematic diagram, compared with previous embodiment, in the present embodiment, above- mentioned information detection method further include:
S105: in the suspicious region for determining target video frame, there are in the case where abnormal conditions, obtain the suspicious region to exist
It is confirmed as the number in the region there are abnormal conditions in the video frame acquired.
S106: in the case where number obtained meets preset monitoring condition, prompt messages are sent.
Specifically, above-mentioned preset monitoring condition can be and judge whether number obtained is greater than preset number threshold
Value, if more than determining that number obtained meets preset monitoring condition, otherwise, it is determined that number obtained is unsatisfactory for presetting
Monitoring condition.Above-mentioned preset frequency threshold value can be 0,1,2 etc..
As seen from the above, in scheme provided in this embodiment, there are abnormal feelings in the suspicious region for determining target video frame
In the case where condition, prompt messages are sent, facilitate staff found the abnormal situation in time, in addition, suspicious region is true
Be set to the region there are abnormal conditions number meet certain monitoring condition in the case where just send prompt messages, Neng Gouyou
Effect reduces wrong report phenomenon.
Corresponding with above- mentioned information detection method, the embodiment of the present application provides a kind of information detector.
Fig. 5 is the structural schematic diagram of the first information detector provided by the embodiments of the present application, which includes:
Person detecting module 501, for detecting the personage in target video frame;
Position obtains module 502, for obtaining character positions of the detected personage in reference frame, wherein described
Reference frame are as follows: the acquisition moment is located at the preset quantity video frame before the acquisition moment of the target video frame;
Suspicious region detection module 503, for detecting in the target video frame according to preset Region detection algorithms
Suspicious region;
Abnormal area determining module 504, for according to the character positions in detected personage and the reference frame,
Whether the suspicious region for determining the target video frame is region there are abnormal conditions.
Specifically, the abnormal area determining module 504 may include:
Character positions obtain submodule, for obtaining personage position of the detected personage in the target video frame
It sets;
Suspicious figure determines submodule, for according in the character positions and the reference frame in the target video frame
Character positions determine the suspicious figure in detected personage;
Corresponding region obtains submodule, in the suspicious region for obtaining the reference frame with the target video frame can
Doubt the corresponding region in region;
Apart from computational submodule, for calculating the suspicious figure and the target video frame in the target video frame
The distance between suspicious region, and calculate in the reference frame between the suspicious figure and region obtained away from
From;
Abnormal area determines submodule, for determining whether the suspicious region of the target video frame is that there are abnormal conditions
Region.
Specifically, the suspicious figure determines that submodule may include:
First information judging unit meets expression formula one for judging whether,
Wherein, the expression formula one are as follows:
meand> TmAnd vard< Tv1,
meandIndicate the mean value of the mean deviation distance of detected each personage, vardIt is each detected by indicating
The variance of the mean deviation distance of personage, TmIndicate preset mean value threshold value, Tv1It indicates preset first variance threshold value, is detected
The mean deviation distance of any personage i out indicates that the personage i is acquired in the target video frame and the reference frame
The average value of moving distance between moment adjacent two frames;
Second information judging unit, for sentencing in the case where the judging result of the first information judging unit is no
It is disconnected whether to meet expression formula two,
Wherein, the expression formula two are as follows:
varp< Tv2And vard< Tv3,
varpIndicate the variance of the character positions in the target video frame, Tv2Indicate preset second variance threshold value, Tv3
Indicate preset third variance threshold values;
Suspicious figure's determination unit, for the judging result of second information judging unit be it is no in the case where, root
Determine mobile slow personage from detected personage according to expression formula three, and according to expression formula four from the slow personage of the movement
Middle determining suspicious figure,
Wherein, the expression formula three are as follows:
frmd> Tcnt1And distbias< Tdist1,
frmdAny personage j in personage detected by indicating goes out in the target video frame and the reference frame
Existing number, distbiasIndicate the mean deviation distance of the personage j, Tdist1Indicate preset first offset distance threshold value,
Tcnt1Indicate preset first amount threshold;
The expression formula four are as follows:
distrela< Tdist2And frms> Tcnt2,
distrelaIndicate position and the pos of the slow personage k of any movementaverThe distance between, posaverIndicate each shifting
The history deflection for moving slow personage most locates position pos slowlyparLocation mean value, posparIndicate the slow personage k of any movement
The position in video frame f is acquired, the personage k is moved to the moving distance of the video frame f from the former frame of the video frame f
Are as follows: the minimum value of moving distance of the personage k between acquisition moment adjacent two video frames acquired, frmsIndicate institute
State the number that personage k is confirmed as slowly mobile personage in the video frame acquired, Tdist2Indicate preset second offset distance
From threshold value, Tcnt2Indicate preset second amount threshold.
Specifically, the suspicious region detection module 503, can be specifically used for detecting institute according to preset regional model
State the suspicious region in target video frame;Or
The suspicious region detection module 503 can be specifically used for according to the area in the reference frame there are abnormal conditions
Domain determines the suspicious region in the target video frame;Or
The suspicious region detection module 503 can also include:
Alternative suspicious region detection sub-module 5031, for detecting the target video frame according to preset regional model
In alternative suspicious region;
First suspicious region determines submodule 5032, for according to the reference frame and the alternative suspicious region that detects,
Determine the suspicious region in the target video frame.
In a kind of optional implementation of the application, the suspicious region detection module 503 may include:
Abnormal area obtains submodule, for obtaining the target video frame F according to the acquisition momentCPrevious video frame FF
It is middle that there are the regions of abnormal conditions;
Corresponding region determines submodule, for determining the video frame FFIt is middle that there are the regions of abnormal conditions in the target
Corresponding region in video frame;
Pixel number statistic submodule belongs to the target video frame F for counting in identified each regionC's
The number of the pixel in sport foreground region;
Ratio judging submodule, for judging whether the number for counting obtained each pixel meets following formula:
NumFAm/NumTAm< ThP,
Wherein, NumFamBelong to the picture in the sport foreground region of the target video frame in any region m determined by indicating
The number of vegetarian refreshments, NumTamIndicate the total number of pixel in the region m, ThPIndicate preset first pixel number ratio
Threshold value;
Second suspicious region determines submodule, is the case where being for the judging result in the ratio judging submodule
Under, determine that the corresponding region of pixel number is the suspicious region in the target video frame.
In a kind of specific implementation of the application, referring to Fig. 6, the structure for providing second of information detector is shown
It is intended to, compared with previous embodiment, in the present embodiment, the suspicious region detection module 503, comprising:
Alternative suspicious region detection sub-module 5031, for detecting the target video frame according to preset regional model
In alternative suspicious region;
First suspicious region determines submodule 5032, for according to the reference frame and the alternative suspicious region that detects,
Determine the suspicious region in the target video frame.
Specifically, the alternative suspicious region detection sub-module 5031 may include:
Foreground area determination unit, for determining the foreground area in the target video frame;
Alternative suspicious region detection unit, for being detected in the foreground area alternative according to preset regional model
Suspicious region.
Specifically, first suspicious region determines that submodule 5032 may include:
Position same area determination unit, for determining the target video frame FCPrevious video frame FFIn with detect
The identical region in alternative suspicious region position;
Pixel number obtaining unit belongs to the video frame F in the pixel for each region determined by obtainingF
It is middle that there are the numbers of the pixel in the region of abnormal conditions;
Suspicious region determination unit, for determining that the region for meeting following formula in above-mentioned identified region is corresponding
Alternative suspicious region is the suspicious region in the target video frame,
Na/Nt> TN,
Wherein, NaBelong to the video frame F in any region S determined by indicatingFIt is middle that there are the pictures in the region of abnormal conditions
The number of vegetarian refreshments, NtIndicate the number of pixel in the region S, TNIndicate preset second pixel number proportion threshold value.
Specifically, the person detecting module 501 may include:
Person detecting model selects submodule, is used for according to current time and/or present intensity, from preset person detecting
Person detecting model is selected in model library;
Person detecting submodule, for according to the personage in selected person detecting model inspection target video frame.
As seen from the above, personage and these people in scheme provided in this embodiment, first in detection target video frame
Character positions of the object in reference frame, then detect the suspicious region in target video frame, finally combine detected personage with
And the character positions of reference frame, determine whether the suspicious region of target video frame is region there are abnormal conditions.It can be seen that application
Scheme provided in this embodiment is capable of detecting when the region in video frame there are abnormal conditions, passes through patrol without staff
Mode finds the region there are abnormal conditions, can reduce the operating pressure of staff.
In another specific implementation of the application, referring to Fig. 7, the structure of the third information detector is provided
Schematic diagram, compared with previous embodiment, in the present embodiment, the position obtains module 502, comprising:
It is associated with personage and obtains submodule 5021, for obtaining personage associated with detected personage in reference frame;
Confidence calculations submodule 5022, for according to the similar journey between detected personage personage associated with it
Degree and movement velocity calculate the confidence level of detected personage, wherein the confidence level is for indicating a personage and its
Associated personage is the credibility of same personage;
Position obtains submodule 5023, is greater than preset confidence threshold value for obtaining confidence level in detected personage
Personage's character positions in the reference frame;
The abnormal area determining module 504, it is preset specifically for being greater than according to confidence level in detected personage
Character positions in the personage of confidence threshold value and the reference frame, determine the target video frame suspicious region whether be
There are the regions of abnormal conditions.
As seen from the above, in scheme provided in this embodiment, it is greater than according to confidence level in detected personage preset
The personage of confidence threshold value and position obtained determine whether the suspicious region of target video is area there are abnormal conditions
Domain, rather than be determined according to detected all persons, therefore can be reduced on the basis of guaranteeing that definitive result is accurate
Calculation amount improves calculating speed.
In another specific implementation of the application, referring to Fig. 8, the structure of the 4th kind of information detector is provided
Schematic diagram, compared with previous embodiment, in the present embodiment, above- mentioned information detection device further include:
Number obtains module 505, for the case where there are abnormal conditions for the suspicious region for determining the target video frame
Under, obtain the number in the region that the suspicious region is confirmed as there are abnormal conditions in the video frame acquired;
Prompt messages sending module 506, in the case where number obtained meets preset monitoring condition,
Send prompt messages.
As seen from the above, in scheme provided in this embodiment, there are abnormal feelings in the suspicious region for determining target video frame
In the case where condition, prompt messages are sent, facilitate staff found the abnormal situation in time, in addition, suspicious region is true
Be set to the region there are abnormal conditions number meet certain monitoring condition in the case where just send prompt messages, Neng Gouyou
Effect reduces wrong report phenomenon.
For device embodiment, since it is substantially similar to the method embodiment, related so being described relatively simple
Place illustrates referring to the part of embodiment of the method.
It should be noted that, in this document, relational terms such as first and second and the like are used merely to a reality
Body or operation are distinguished with another entity or operation, are deposited without necessarily requiring or implying between these entities or operation
In any actual relationship or order or sequence.Moreover, the terms "include", "comprise" or its any other variant are intended to
Non-exclusive inclusion, so that the process, method, article or equipment including a series of elements is not only wanted including those
Element, but also including other elements that are not explicitly listed, or further include for this process, method, article or equipment
Intrinsic element.In the absence of more restrictions, the element limited by sentence "including a ...", it is not excluded that
There is also other identical elements in process, method, article or equipment including the element.
Those of ordinary skill in the art will appreciate that all or part of the steps in realization above method embodiment is can
It is completed with instructing relevant hardware by program, the program can store in computer-readable storage medium,
The storage medium designated herein obtained, such as: ROM/RAM, magnetic disk, CD.
The foregoing is merely the preferred embodiments of the application, are not intended to limit the protection scope of the application.It is all
Any modification, equivalent replacement, improvement and so within spirit herein and principle are all contained in the protection scope of the application
It is interior.
Claims (18)
1. a kind of information detecting method, which is characterized in that the described method includes:
Detect the personage in target video frame;
Obtain character positions of the detected personage in reference frame, wherein the reference frame are as follows: the acquisition moment is located at described
Preset quantity video frame before the acquisition moment of target video frame;
According to preset Region detection algorithms, the suspicious region in the target video frame is detected;
According to the character positions in detected personage and the reference frame, the suspicious region of the target video frame is determined
It whether is region there are abnormal conditions;
Wherein, the personage according to detected by and the character positions in the reference frame, determine the target video frame
Suspicious region whether be region there are abnormal conditions, comprising:
Obtain character positions of the detected personage in the target video frame;
According to the character positions in the character positions and the reference frame in the target video frame, detected personage is determined
In suspicious figure;
Determine the suspicious region in the reference frame;
Obtain region corresponding with the suspicious region of the target video frame in the suspicious region of the reference frame;
Calculated in the target video frame suspicious figure between the suspicious region of the target video frame at a distance from, and
The distance between the suspicious figure and region obtained are calculated in the reference frame;
According to calculated distance, determine whether the suspicious region of the target video frame is region there are abnormal conditions.
2. the method according to claim 1, wherein the character positions according in the target video frame and
Character positions in the reference frame determine the suspicious figure in detected personage, comprising:
Judge whether to meet expression formula one,
Wherein, the expression formula one are as follows:
meand> TmAnd vard< Tv1,
meandIndicate the mean value of the mean deviation distance of detected each personage, vardIndicate detected each personage
Mean deviation distance variance, TmIndicate preset mean value threshold value, Tv1Indicate preset first variance threshold value, it is detected
The mean deviation distance of any personage i indicates that the personage i acquires the moment in the target video frame and the reference frame
The average value of moving distance between two adjacent frames;
If not satisfied, judge whether to meet expression formula two,
Wherein, the expression formula two are as follows:
varp< Tv2And vard< Tv3,
varpIndicate the variance of the character positions in the target video frame, Tv2Indicate preset second variance threshold value, Tv3It indicates
Preset third variance threshold values;
If not satisfied, mobile slow personage is determined from detected personage according to expression formula three, and according to expression formula four from
Suspicious figure is determined in the slow personage of movement,
Wherein, the expression formula three are as follows:
frmd> Tcnt1And distbias< Tdist1,
frmdWhat any personage j in personage detected by indicating occurred in the target video frame and the reference frame
Number, distbiasIndicate the mean deviation distance of the personage j, Tdist1Indicate preset first offset distance threshold value, Tcnt1Table
Show preset first amount threshold;
The expression formula four are as follows:
distrela< Tdist2And frms> Tcnt2,
distrelaIndicate position and the pos of the slow personage k of any movementaverThe distance between, posaverIndicate that each movement is slow
The history deflection of slow personage most locates position pos slowlyparLocation mean value, posparIndicate that the slow personage k of any movement is being acquired
Position in video frame f, the personage k are moved to the moving distance of the video frame f from the former frame of the video frame f are as follows:
The minimum value of moving distance of the personage k between acquisition moment adjacent two video frames acquired, frmsDescribed in expression
Personage k is confirmed as the number of slowly mobile personage, T in the video frame acquireddist2Indicate preset second offset distance
Threshold value, Tcnt2Indicate preset second amount threshold.
3. the method according to claim 1, wherein described according to preset Region detection algorithms, described in detection
Suspicious region in target video frame, comprising:
According to preset regional model, the suspicious region in the target video frame is detected;Or
According to the region in the reference frame there are abnormal conditions, the suspicious region in the target video frame is determined;Or
According to preset regional model, the alternative suspicious region in the target video frame is detected;And according to the reference frame and
The alternative suspicious region detected determines the suspicious region in the target video frame.
4. according to the method described in claim 3, it is characterized in that, the area according in the reference frame there are abnormal conditions
Domain determines the suspicious region in the target video frame, comprising:
According to the acquisition moment, the target video frame F is obtainedCPrevious video frame FFIt is middle that there are the regions of abnormal conditions;
Determine the video frame FFIt is middle that there are the regions of abnormal conditions in the target video frame FCIn corresponding region;
Belong to the number of the pixel in the sport foreground region of the target video frame in each region determined by counting;
Judge whether the number for each pixel that statistics obtains meets following formula:
NumFAm/NumTAm< ThP,
Wherein, NumFAmBelong to the pixel in the sport foreground region of the target video frame in any region m determined by indicating
Number, NumTAmIndicate the total number of pixel in the region m, ThPIndicate preset first pixel number ratio threshold
Value;
If satisfied, determining that the corresponding region of pixel number is the suspicious region in the target video frame.
5. according to the method described in claim 3, detecting the target it is characterized in that, described according to preset regional model
Alternative suspicious region in video frame, comprising:
Determine the foreground area in the target video frame;
According to preset regional model, alternative suspicious region is detected in the foreground area.
6. the method according to claim 3 or 5, which is characterized in that it is described according to the reference frame and detect it is alternative
Suspicious region determines the suspicious region in the target video frame, comprising:
Determine the target video frame FCPrevious video frame FFIn region identical with the alternative suspicious region position detected;
Belong to the video frame F in the pixel of each region determined by obtainingFIt is middle that there are the pixels in the region of abnormal conditions
Number;
It determines and meets the corresponding alternative suspicious region in region of following formula in above-mentioned identified region for target view
Suspicious region in frequency frame,
Na/Nt> TN,
Wherein, NaBelong to the video frame F in any region S determined by indicatingFIt is middle that there are the pixels in the region of abnormal conditions
Number, NtIndicate the number of pixel in the region S, TNIndicate preset second pixel number proportion threshold value.
7. the method according to claim 1, wherein the personage in the detection target video frame, comprising:
According to current time and/or present intensity, person detecting model is selected from preset person detecting model library;
According to the personage in selected person detecting model inspection target video frame.
8. the method according to claim 1, wherein people of the personage detected by the acquisition in reference frame
Object location, comprising:
Obtain personage associated with detected personage in reference frame;
According to the similarity degree and movement velocity between detected personage personage associated with it, calculate detected
The confidence level of personage, wherein the confidence level, for indicating that a personage personage associated there is the credible of same personage
Degree;
It obtains confidence level in detected personage and is greater than the personage of the personage of preset confidence threshold value in the reference frame
Position;
The personage according to detected by and the character positions in the reference frame, determine the suspicious of the target video frame
Whether region is region there are abnormal conditions, comprising:
It is greater than the people in the personage and the reference frame of preset confidence threshold value according to confidence level in detected personage
Object location determines whether the suspicious region of the target video frame is region there are abnormal conditions.
9. the method according to claim 1, wherein the method also includes:
In the suspicious region for determining the target video frame, there are in the case where abnormal conditions, obtain the suspicious region acquiring
Video frame in be confirmed as the number in region there are abnormal conditions;
In the case where number obtained meets preset monitoring condition, prompt messages are sent.
10. a kind of information detector, which is characterized in that described device includes:
Person detecting module, for detecting the personage in target video frame;
Position obtains module, for obtaining character positions of the detected personage in reference frame, wherein the reference frame
Are as follows: the acquisition moment is located at the preset quantity video frame before the acquisition moment of the target video frame;
Suspicious region detection module, for detecting the suspicious area in the target video frame according to preset Region detection algorithms
Domain;
Abnormal area determining module, for determining institute according to the character positions in detected personage and the reference frame
Whether the suspicious region for stating target video frame is region there are abnormal conditions;
Wherein, the abnormal area determining module, comprising:
Character positions obtain submodule, for obtaining character positions of the detected personage in the target video frame;
Suspicious figure determines submodule, for according to the personage in the character positions and the reference frame in the target video frame
Position determines the suspicious figure in detected personage;
Corresponding region obtains submodule, the suspicious area in the suspicious region for obtaining the reference frame with the target video frame
The corresponding region in domain;
It, can for calculate the suspicious figure and the target video frame in the target video frame apart from computational submodule
The distance between region is doubted, and calculates the distance between the suspicious figure and region obtained in the reference frame;
Abnormal area determines submodule, for determining whether the suspicious region of the target video frame is area there are abnormal conditions
Domain.
11. device according to claim 10, which is characterized in that the suspicious figure determines submodule, comprising:
First information judging unit meets expression formula one for judging whether,
Wherein, the expression formula one are as follows:
meand> TmAnd vard< Tv1,
meandIndicate the mean value of the mean deviation distance of detected each personage, vardIndicate detected each personage
Mean deviation distance variance, TmIndicate preset mean value threshold value, Tv1Indicate preset first variance threshold value, it is detected
The mean deviation distance of any personage i indicates that the personage i acquires the moment in the target video frame and the reference frame
The average value of moving distance between two adjacent frames;
Second information judging unit, in the case where the judging result of the first information judging unit is no, judgement to be
It is no to meet expression formula two,
Wherein, the expression formula two are as follows:
varp< Tv2And vard< Tv3,
varpIndicate the variance of the character positions in the target video frame, Tv2Indicate preset second variance threshold value, Tv3It indicates
Preset third variance threshold values;
Suspicious figure's determination unit, for the judging result of second information judging unit be it is no in the case where, according to table
Mobile slow personage is determined from detected personage up to formula three, and true from the slow personage of the movement according to expression formula four
Determine suspicious figure,
Wherein, the expression formula three are as follows:
frmd> Tcnt1And distbias< Tdist1,
frmdWhat any personage j in personage detected by indicating occurred in the target video frame and the reference frame
Number, distbiasIndicate the mean deviation distance of the personage j, Tdist1Indicate preset first offset distance threshold value, Tcnt1Table
Show preset first amount threshold;
The expression formula four are as follows:
distrela< Tdist2And frms> Tcnt2,
distrelaIndicate position and the pos of the slow personage k of any movementaverThe distance between, posaverIndicate that each movement is slow
The history deflection of slow personage most locates position pos slowlyparLocation mean value, posparIndicate that the slow personage k of any movement is being acquired
Position in video frame f, the personage k are moved to the moving distance of the video frame f from the former frame of the video frame f are as follows:
The minimum value of moving distance of the personage k between acquisition moment adjacent two video frames acquired, frmsDescribed in expression
Personage k is confirmed as the number of slowly mobile personage, T in the video frame acquireddist2Indicate preset second offset distance
Threshold value, Tcnt2Indicate preset second amount threshold.
12. device according to claim 10, which is characterized in that
The suspicious region detection module, be specifically used for according to preset regional model, detect in the target video frame can
Doubt region;Or
The suspicious region detection module, described in determining according to the region in the reference frame there are abnormal conditions
Suspicious region in target video frame;Or
The suspicious region detection module, comprising:
Alternative suspicious region detection sub-module, for detecting alternative in the target video frame according to preset regional model
Suspicious region;
First suspicious region determines submodule, described in determining according to the reference frame and the alternative suspicious region detected
Suspicious region in target video frame.
13. device according to claim 12, which is characterized in that the suspicious region detection module, comprising:
Abnormal area obtains submodule, for obtaining the target video frame F according to the acquisition momentCPrevious video frame FFIn deposit
In the region of abnormal conditions;
Corresponding region determines submodule, for determining the video frame FFIt is middle that there are the regions of abnormal conditions in the target video
Corresponding region in frame;
Pixel number statistic submodule belongs to the target video frame F for counting in identified each regionCMovement
The number of the pixel of foreground area;
Ratio judging submodule, for judging whether the number for counting obtained each pixel meets following formula:
NumFAm/NumTAm< ThP,
Wherein, NumFAmBelong to the pixel in the sport foreground region of the target video frame in any region m determined by indicating
Number, NumTAmIndicate the total number of pixel in the region m, ThPIndicate preset first pixel number ratio threshold
Value;
Second suspicious region determines submodule, for the judging result of the ratio judging submodule be in the case where, really
The fixed corresponding region of pixel number is the suspicious region in the target video frame.
14. device according to claim 12, which is characterized in that the alternative suspicious region detection sub-module, comprising:
Foreground area determination unit, for determining the foreground area in the target video frame;
Alternative suspicious region detection unit is alternative suspicious for being detected in the foreground area according to preset regional model
Region.
15. device described in 2 or 14 according to claim 1, which is characterized in that first suspicious region determines submodule, packet
It includes:
Position same area determination unit, for determining the target video frame FCPrevious video frame FFIn with detect it is standby
Select the identical region in suspicious region position;
Pixel number obtaining unit belongs to the video frame F in the pixel for each region determined by obtainingFIn deposit
In the number of the pixel in the region of abnormal conditions;
Suspicious region determination unit, for determining that the region for meeting following formula in above-mentioned identified region is corresponding alternative
Suspicious region is the suspicious region in the target video frame,
Na/Nt> TN,
Wherein, NaBelong to the video frame F in any region S determined by indicatingFIt is middle that there are the pixels in the region of abnormal conditions
Number, NtIndicate the number of pixel in the region S, TNIndicate preset second pixel number proportion threshold value.
16. device according to claim 10, which is characterized in that the person detecting module, comprising:
Person detecting model selects submodule, is used for according to current time and/or present intensity, from preset person detecting model
Person detecting model is selected in library;
Person detecting submodule, for according to the personage in selected person detecting model inspection target video frame.
17. device according to claim 10, which is characterized in that the position obtains module, comprising:
It is associated with personage and obtains submodule, for obtaining personage associated with detected personage in reference frame;
Confidence calculations submodule, for according to the similarity degree and fortune between detected personage personage associated with it
Dynamic speed, calculates the confidence level of detected personage, wherein the confidence level is associated there for indicating a personage
Personage is the credibility of same personage;
Position obtains submodule, exists for obtaining the personage that confidence level is greater than preset confidence threshold value in detected personage
Character positions in the reference frame;
The abnormal area determining module, specifically for being greater than preset confidence level threshold according to confidence level in detected personage
Character positions in the personage of value and the reference frame, determine whether the suspicious region of the target video frame is that there are exceptions
The region of situation.
18. device according to claim 10, which is characterized in that described device further include:
Number obtains module, for, there are in the case where abnormal conditions, obtaining in the suspicious region for determining the target video frame
The suspicious region is confirmed as the number in the region there are abnormal conditions in the video frame acquired;
Prompt messages sending module, for sending report in the case where number obtained meets preset monitoring condition
Alert prompt information.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510885680.4A CN106845325B (en) | 2015-12-04 | 2015-12-04 | A kind of information detecting method and device |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510885680.4A CN106845325B (en) | 2015-12-04 | 2015-12-04 | A kind of information detecting method and device |
Publications (2)
Publication Number | Publication Date |
---|---|
CN106845325A CN106845325A (en) | 2017-06-13 |
CN106845325B true CN106845325B (en) | 2019-10-22 |
Family
ID=59150583
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201510885680.4A Active CN106845325B (en) | 2015-12-04 | 2015-12-04 | A kind of information detecting method and device |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN106845325B (en) |
Families Citing this family (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109871730A (en) * | 2017-12-05 | 2019-06-11 | 杭州海康威视数字技术股份有限公司 | A kind of target identification method, device and monitoring device |
CN110163029B (en) * | 2018-02-11 | 2021-03-30 | 中兴飞流信息科技有限公司 | Image recognition method, electronic equipment and computer readable storage medium |
WO2019196934A1 (en) | 2018-04-13 | 2019-10-17 | Shanghai Truthvision Information Technology Co., Ltd. | System and method for abnormal scene detection |
CN108921083B (en) * | 2018-06-28 | 2021-07-27 | 浙江工业大学 | Illegal mobile vendor identification method based on deep learning target detection |
CN111402192B (en) * | 2018-12-28 | 2023-10-27 | 杭州海康威视数字技术股份有限公司 | Inspection well cover detection method and inspection well cover detection device |
CN111126252B (en) * | 2019-12-20 | 2023-08-18 | 浙江大华技术股份有限公司 | Swing behavior detection method and related device |
CN113705274B (en) * | 2020-05-20 | 2023-09-05 | 杭州海康威视数字技术股份有限公司 | Climbing behavior detection method and device, electronic equipment and storage medium |
CN112700657B (en) * | 2020-12-21 | 2023-04-28 | 阿波罗智联(北京)科技有限公司 | Method and device for generating detection information, road side equipment and cloud control platform |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101344920A (en) * | 2008-07-21 | 2009-01-14 | 北大方正集团有限公司 | Method and device for detecting specific area in video data frame |
CN102063614A (en) * | 2010-12-28 | 2011-05-18 | 天津市亚安科技电子有限公司 | Method and device for detecting lost articles in security monitoring |
CN102348102A (en) * | 2010-07-30 | 2012-02-08 | 鸿富锦精密工业(深圳)有限公司 | Roof safety monitoring system and method thereof |
CN104284143A (en) * | 2013-07-03 | 2015-01-14 | 智原科技股份有限公司 | Image monitoring system and method thereof |
Family Cites Families (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP4568009B2 (en) * | 2003-04-22 | 2010-10-27 | パナソニック株式会社 | Monitoring device with camera cooperation |
-
2015
- 2015-12-04 CN CN201510885680.4A patent/CN106845325B/en active Active
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101344920A (en) * | 2008-07-21 | 2009-01-14 | 北大方正集团有限公司 | Method and device for detecting specific area in video data frame |
CN102348102A (en) * | 2010-07-30 | 2012-02-08 | 鸿富锦精密工业(深圳)有限公司 | Roof safety monitoring system and method thereof |
CN102063614A (en) * | 2010-12-28 | 2011-05-18 | 天津市亚安科技电子有限公司 | Method and device for detecting lost articles in security monitoring |
CN104284143A (en) * | 2013-07-03 | 2015-01-14 | 智原科技股份有限公司 | Image monitoring system and method thereof |
Non-Patent Citations (3)
Title |
---|
《Detecting suspicious background changes in video surveillance of busy scenes》;D. Gibbins等;《Proceedings Third IEEE Workshop on Applications of Computer Vision. WACV"96》;19961231;第22-26页 * |
《智能视频监控技术与应用》;郑世宝;《电视技术》;20090131;第33卷(第1期);第94-96页 * |
《用于监控智能报警***的图像识别技术》;周小四等;《上海交通大学学报》;20020430;第36卷(第4期);第498-501页 * |
Also Published As
Publication number | Publication date |
---|---|
CN106845325A (en) | 2017-06-13 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN106845325B (en) | A kind of information detecting method and device | |
CN108062349B (en) | Video monitoring method and system based on video structured data and deep learning | |
CN106980829B (en) | Abnormal behaviour automatic testing method of fighting based on video analysis | |
CN103164858B (en) | Adhesion crowd based on super-pixel and graph model is split and tracking | |
CN102542289B (en) | Pedestrian volume statistical method based on plurality of Gaussian counting models | |
CN104303193B (en) | Target classification based on cluster | |
CN106203513B (en) | A kind of statistical method based on pedestrian's head and shoulder multi-target detection and tracking | |
CN109344690B (en) | People counting method based on depth camera | |
CN102004920B (en) | Method for splitting and indexing surveillance videos | |
CN105844234A (en) | People counting method and device based on head shoulder detection | |
CN104978567B (en) | Vehicle checking method based on scene classification | |
CN103986910A (en) | Method and system for passenger flow statistics based on cameras with intelligent analysis function | |
CN107229894A (en) | Intelligent video monitoring method and system based on computer vision analysis technology | |
JP2019505866A (en) | Passerby head identification method and system | |
CN103425967A (en) | Pedestrian flow monitoring method based on pedestrian detection and tracking | |
CN107948465A (en) | A kind of method and apparatus for detecting camera and being disturbed | |
CN103617410A (en) | Highway tunnel parking detection method based on video detection technology | |
CN107483894B (en) | The high-speed rail station video monitoring system of realization passenger transportation management is judged based on scene | |
CN106210634A (en) | A kind of wisdom gold eyeball identification personnel fall down to the ground alarm method and device | |
CN102930248A (en) | Crowd abnormal behavior detection method based on machine learning | |
CN106570449B (en) | A kind of flow of the people defined based on region and popularity detection method and detection system | |
CN106548488A (en) | It is a kind of based on background model and the foreground detection method of inter-frame difference | |
CN106296677A (en) | A kind of remnant object detection method of double mask context updates based on double-background model | |
CN111881749A (en) | Bidirectional pedestrian flow statistical method based on RGB-D multi-modal data | |
CA2670021A1 (en) | System and method for estimating characteristics of persons or things |
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 |