CN110414360A - A kind of detection method and detection device of abnormal behaviour - Google Patents
A kind of detection method and detection device of abnormal behaviour Download PDFInfo
- Publication number
- CN110414360A CN110414360A CN201910589840.9A CN201910589840A CN110414360A CN 110414360 A CN110414360 A CN 110414360A CN 201910589840 A CN201910589840 A CN 201910589840A CN 110414360 A CN110414360 A CN 110414360A
- Authority
- CN
- China
- Prior art keywords
- target object
- target
- abnormal behaviour
- feature
- behavior
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 206010000117 Abnormal behaviour Diseases 0.000 title claims abstract description 72
- 238000001514 detection method Methods 0.000 title claims abstract description 65
- 230000006399 behavior Effects 0.000 claims abstract description 71
- 238000000034 method Methods 0.000 claims abstract description 65
- 238000012545 processing Methods 0.000 claims abstract description 37
- 238000003860 storage Methods 0.000 claims description 12
- 238000004590 computer program Methods 0.000 claims description 5
- 241000406668 Loxodonta cyclotis Species 0.000 claims description 3
- 230000006378 damage Effects 0.000 abstract description 5
- 238000004422 calculation algorithm Methods 0.000 description 32
- 230000009194 climbing Effects 0.000 description 22
- 238000004891 communication Methods 0.000 description 19
- 238000012544 monitoring process Methods 0.000 description 16
- 238000010586 diagram Methods 0.000 description 13
- 230000033001 locomotion Effects 0.000 description 10
- 238000012549 training Methods 0.000 description 9
- 230000001133 acceleration Effects 0.000 description 8
- 238000000605 extraction Methods 0.000 description 8
- 230000008569 process Effects 0.000 description 7
- 239000000284 extract Substances 0.000 description 6
- 230000006870 function Effects 0.000 description 6
- 230000006872 improvement Effects 0.000 description 6
- 230000004044 response Effects 0.000 description 6
- 230000002159 abnormal effect Effects 0.000 description 5
- 230000003542 behavioural effect Effects 0.000 description 4
- 238000004364 calculation method Methods 0.000 description 3
- 230000008859 change Effects 0.000 description 3
- 238000010168 coupling process Methods 0.000 description 3
- 238000005859 coupling reaction Methods 0.000 description 3
- 230000036544 posture Effects 0.000 description 3
- 238000012360 testing method Methods 0.000 description 3
- 230000009466 transformation Effects 0.000 description 3
- 238000004458 analytical method Methods 0.000 description 2
- 230000008901 benefit Effects 0.000 description 2
- 230000008878 coupling Effects 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 230000004927 fusion Effects 0.000 description 2
- 238000007689 inspection Methods 0.000 description 2
- 230000007774 longterm Effects 0.000 description 2
- 239000000463 material Substances 0.000 description 2
- 239000002245 particle Substances 0.000 description 2
- 230000000717 retained effect Effects 0.000 description 2
- 239000007787 solid Substances 0.000 description 2
- 238000012706 support-vector machine Methods 0.000 description 2
- 206010038743 Restlessness Diseases 0.000 description 1
- 208000027418 Wounds and injury Diseases 0.000 description 1
- 230000009471 action Effects 0.000 description 1
- 230000006978 adaptation Effects 0.000 description 1
- 230000003044 adaptive effect Effects 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 238000005336 cracking Methods 0.000 description 1
- 238000013500 data storage Methods 0.000 description 1
- 238000013502 data validation Methods 0.000 description 1
- 230000007423 decrease Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 230000001965 increasing effect Effects 0.000 description 1
- 208000014674 injury Diseases 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 239000013307 optical fiber Substances 0.000 description 1
- 238000005192 partition Methods 0.000 description 1
- 238000012827 research and development Methods 0.000 description 1
- 238000012216 screening Methods 0.000 description 1
- 230000011218 segmentation Effects 0.000 description 1
- 238000000926 separation method Methods 0.000 description 1
- 230000006641 stabilisation Effects 0.000 description 1
- 238000011105 stabilization Methods 0.000 description 1
- 238000010408 sweeping Methods 0.000 description 1
- 230000001052 transient effect Effects 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/194—Segmentation; Edge detection involving foreground-background segmentation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/20—Analysis of motion
- G06T7/246—Analysis of motion using feature-based methods, e.g. the tracking of corners or segments
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/40—Scenes; Scene-specific elements in video content
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/40—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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10016—Video; Image sequence
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V2201/00—Indexing scheme relating to image or video recognition or understanding
- G06V2201/07—Target detection
Abstract
The embodiment of the invention discloses a kind of detection method of abnormal behaviour and detection device, falls down for detecting old man in family and the abnormal behaviours such as disorderly climb with child and reduce the damage of abnormal behaviour so as to handle the generation in family due to abnormal behaviour in time.The method comprise the steps that obtaining the video data about target object；According to the video data by the processing of the Gaussian Background modeling method updated, location information, size and the area information of target object are obtained；According to the location information of the target object, size and area information, the target object is tracked by the target tracking algorism based on KCF and TLD, obtains the feature of the target object；The first behavior of the target object is detected when according to the feature and preset SVM classifier of the target object, when determining the first behavior abnormal behaviour of the target object, issues prompt messages.
Description
Technical field
The present invention relates to field of image processings, and in particular to a kind of detection method and detection device of abnormal behaviour.
Background technique
Along with the development of information-intensive society, safety becomes the problem of people increasingly pay close attention to, more and more video monitorings
Product is widely used in security fields.But current most of monitoring system, it is unable to the content of intellectual analysis video, is needed
It goes to analyze by people, such monitoring system can only be used to store and record video, and correlation could be used after abnormal behaviour occurs
Video is investigated, it is not possible to be prejudged the generation of abnormal behaviour, and then is avoided the generation of abnormal behaviour.Simultaneously because scientific skill
Art is maked rapid progress, and the higher and higher Qinghua of video monitoring system and intelligence, intelligent monitor system refer in inartificial operation
Under, system can make analytical judgment to the abnormal behaviour in monitor video.
Because intelligent monitor system embodies increasing effect, lot of domestic and international security protection and prison in security fields
Control company has begun the product of research and development related fields.Now, domestic and international market has had already appeared intelligent video monitoring product.Intelligence
Energy video monitoring mainly automatically analyzes video, and crucial information is extracted from video sequence, finds and identifies and is interested
Abnormal behaviour, so as to replace artificial monitoring or assist artificial monitoring, analysis and the identifying system of video can pass through
The video flowing in real time or recorded is analyzed, detects abnormal behaviour.The intelligence of video monitoring, which refers to, is not needing manual operation
In the case where, by being automatically analyzed to video sequence image, so that the variation in monitoring scene is positioned, is identified,
Tracking, and early warning or alarm are made in time.
At present many places monitoring system all or traditional equipment, the work only recorded and cannot be in scene
Behavior judge, it usually needs personnel on duty are continual to be guarded, and a large amount of manpower and material resources are expended, most important
It is sometimes because can not achieve inerrancy without failing to report, after abnormal behaviour occurs the problem of fatigue or carelessness, it is also necessary to
Evidence is searched in a large amount of video resource.
Summary of the invention
The embodiment of the invention provides a kind of detection method of abnormal behaviour and detection devices, for detecting old man in family
It falls down and the abnormal behaviours such as disorderly climbs with child, so as to handle the generation in family due to abnormal behaviour in time, reduce abnormal row
For damage.
In view of this, first aspect present invention provides a kind of detection method of abnormal behaviour, may include:
Obtain the video data about target object；
According to the video data by the processing of the Gaussian Background modeling method updated, the position letter of target object is obtained
Breath, size and area information；
According to the location information of the target object, size and area information, pass through the target following based on KCF and TLD
Algorithm tracks the target object, obtains the feature of the target object；
The first behavior of the target object is carried out when according to the feature and preset SVM classifier of the target object
Detection when determining the first behavior abnormal behaviour of the target object, issues prompt messages.
Optionally, in some embodiments of the invention, described that the Gaussian Background of update is passed through according to the video data
The processing of modeling method obtains location information, size and the area information of target object, may include:
According to the video data by the processing of the Gaussian Background modeling method updated, target object is divided from background
It cuts out, the target object split includes location information, size and the area information of target object.
Optionally, in some embodiments of the invention, described according to the location information of the target object, size and area
Domain information tracks the target object by the target tracking algorism based on KCF and TLD, obtains the target object
Feature, may include:
According to the location information of the target object, size and area information, pass through the target following based on KCF and TLD
Algorithm tracks the target object, obtains target signature of the target object within the object time.
Optionally, in some embodiments of the invention, the target for obtaining the target object within the object time
Feature may include:
By obtaining target signature of the target object within the object time based on sliding window method.
Optionally, in some embodiments of the invention, the first behavior abnormal behaviour of the determination target object
When, prompt messages are issued, may include:
Prompt messages are sent to terminal device, the prompt messages include the text prompt about abnormal behaviour
Information, or the speech prompt information about abnormal behaviour.
Second aspect of the present invention provides a kind of detection device, may include:
Module is obtained, for obtaining the video data about target object；
Processing module, for, by the processing of the Gaussian Background modeling method updated, obtaining mesh according to the video data
Mark location information, size and the area information of object；According to the location information of the target object, size and area information, lead to
It crosses the target tracking algorism based on KCF and TLD to track the target object, obtains the feature of the target object；When
The first behavior of the target object is detected according to the feature of the target object and preset SVM classifier, is determined
When the first behavior abnormal behaviour of the target object, prompt messages are issued.
Optionally, in some embodiments of the invention,
The processing module, specifically for passing through the place of the Gaussian Background modeling method updated according to the video data
Reason, target object is split from background, and the target object split includes the location information, big of target object
Small and area information.
Optionally, in some embodiments of the invention,
The processing module, specifically for passing through base according to the location information of the target object, size and area information
The target object is tracked in the target tracking algorism of KCF and TLD, obtains the target object within the object time
Target signature.
Optionally, in some embodiments of the invention,
The processing module is specifically used for by obtaining the target object within the object time based on sliding window method
Target signature.
Optionally, in some embodiments of the invention,
The processing module is specifically used for sending prompt messages to terminal device, and the prompt messages include
About the text prompt information of abnormal behaviour, or about the speech prompt information of abnormal behaviour.
Third aspect present invention provides a kind of detection device, may include:
Transceiver, processor, memory, wherein the transceiver, the processor and the memory are connected by bus
It connects；
The memory, for storing operational order；
The transceiver, for obtaining the video data about target object；
The processor, for calling the operational order, execute such as first aspect present invention and first aspect is any can
The step of selecting the detection method of abnormal behaviour described in implementation.
Fourth aspect present invention provides a kind of readable storage medium storing program for executing, is stored thereon with computer program, the computer journey
The abnormal row as described in first aspect present invention and any optional implementation of first aspect is realized when sequence is executed by processor
For detection method the step of.
As can be seen from the above technical solutions, the embodiment of the present invention has the advantage that
In embodiments of the present invention, the video data about target object is obtained；Pass through update according to the video data
Gaussian Background modeling method processing, obtain location information, size and the area information of target object；According to the target pair
Location information, size and the area information of elephant carry out the target object by the target tracking algorism based on KCF and TLD
Tracking, obtains the feature of the target object；When the feature and preset SVM classifier according to the target object are to the mesh
First behavior of mark object is detected, and when determining the first behavior abnormal behaviour of the target object, issues warning note letter
Breath.The human body of camera video is transported using improved Gaussian Background modeling method and the TLD target tracking algorism based on KCF
Moving-target splits and carries out stable tracking, is made that improvement, improved Gauss to Gaussian Background modeling method first
It still can highly effective and high-precision in the case where background modeling method background variation indoors and indoor lighting conditions variation
Extract humanbody moving object in ground.TLD target tracking algorism based on KCF is also to be made that improvement to KCF algorithm and TLD algorithm,
The processing frame per second of algorithm is improved, while keeping the on-line training of detection module, improves the re-detection ability of algorithm, thus real
The quick tracking to target is showed.Realize detection to abnormal behaviour using SVM classifier, detect old man in family fall down with it is small
Child, which such as disorderly climbs at the abnormal behaviours, reduces the damage of abnormal behaviour so as to handle the generation in family due to abnormal behaviour in time.
Detailed description of the invention
Technical solution in order to illustrate the embodiments of the present invention more clearly, below will be to institute in embodiment and description of the prior art
Attached drawing to be used is needed to be briefly described, it should be apparent that, the accompanying drawings in the following description is only some implementations of the invention
Example, can also be obtained according to these attached drawings other attached drawings.
Fig. 1 is one embodiment schematic diagram of unusual checking system in the embodiment of the present invention；
Fig. 2 is schematic illustration applied by unusual checking system of the embodiment of the present invention；
Fig. 3 is one embodiment schematic diagram that the method for abnormal behaviour is detected in the embodiment of the present invention；
Fig. 4 is the frame diagram of KCF-TLD algorithm used in the embodiment of the present invention；
Fig. 5 is the TLD target tracking algorism flow chart based on KCF in the embodiment of the present invention；
Fig. 6 is the idiographic flow schematic diagram of classification and Detection in the embodiment of the present invention；
Fig. 7 is one embodiment schematic diagram of detection device in the embodiment of the present invention；
Fig. 8 is another embodiment schematic diagram of detection device in the embodiment of the present invention.
Specific embodiment
The embodiment of the invention provides a kind of detection method of abnormal behaviour and detection devices, for detecting old man in family
It falls down and the abnormal behaviours such as disorderly climbs with child, so as to handle the generation in family due to abnormal behaviour in time, reduce abnormal row
For damage.
In order to enable those skilled in the art to better understand the solution of the present invention, below in conjunction in the embodiment of the present invention
Attached drawing, technical solution in the embodiment of the present invention are described, it is clear that described embodiment is only present invention a part
Embodiment, instead of all the embodiments.Based on the embodiments of the present invention, it should fall within the scope of the present invention.
As shown in Figure 1, for one embodiment schematic diagram of unusual checking system in the embodiment of the present invention, abnormal behaviour
Detection system may include: camera module, holder, Wireless Fidelity (Wireless Fidelity, WIFI) module, communication mould
Block, embedded system, voice alarm module, hot electrostatic infrared module, application program of mobile phone (Application, APP), exception
Behavior algoritic module.Wherein, it should be noted that Ardroid embedded system passes through WIFI communication module or communication module
Connection network is protected as far as possible if there is one of module network when the signal is poor, the switching of network mode can be attached
Demonstrate,prove the stabilization of transmitted data on network.Communication module can be 4G communication module, be also possible to 5G communication module, be also possible to higher
The communication module of rank, specifically without limitation.Embedded system may include Android (Android) embedded system or other
The embedded system of type, it is not limited in the embodiment of the present invention.In the following embodiments, communication module can be communicated with 4G
It is illustrated for module, embedded system can be illustrated by taking Android embedded system as an example.
APP in Ardroid embedded system receives the data that camera module transmits, and passes through unusual checking algorithm
Module detects that (type of alarm can be by mobile phone terminal to personal mobile phone terminal alarms by communication module after abnormal behaviour
APP shows information, alternatively, directly being sent short messages to mobile phone terminal to realize, alternatively, carrying out alarm sound prompt on mobile phone terminal
Etc. modes), personal mobile phone terminal by APP realize by embedded system call camera shoot data validation abnormal behaviour
Information it is whether true.Unusual checking system can also by voice alarm module play warning information, allow it is indoor its
Other people know that something unexpected happened at the first time.
Therefore, intelligentized unusual checking system can not only prejudge abnormal behaviour, improve security capabilities, and
And do not need to expend too many human and material resources, cost is saved, contains huge business opportunity and economic benefit, so having
Very big realistic meaning and application value.
Most of the cracking a crib just for thief of household monitoring system of today, and old man and child can also be in generation
It is unexpected, and the unexpected probability of generation is not low, and common monitoring system precision is not high enough, leads to various erroneous judgements.Abnormal behaviour inspection
Examining system is for the various unexpected messages in family, and it is different quickly and accurately to detect that old man falls down and child disorderly climbs etc.
Chang Hangwei, when detect old man fall down the abnormal behaviours such as disorderly climb with child when, can be by communication module to personal mobile phone terminal report
It is alert, or warning information is played by voice alarm module, to have centainly to the safety of old man and child at home
It ensures.
Wherein, the flow diagram of unusual checking algorithm is as follows: as shown in Fig. 2, being the embodiment of the present invention in abnormal row
For schematic illustration applied by detection system.As shown in Figure 2, firstly, passing through after camera module reads video data
Improved Gaussian Background modeling method and be based on KCF (High-Speed Tracking with Kernelized
Correlation Filters) TLD (Tracking-Learning-Detection) target tracking algorism realize to human body
The segmentation and tracking of moving target, and the location information of humanbody moving object, human motion mesh can be accurately obtained very much
Target size and area information.Then the identification-state of humanbody moving object is established, then human body fortune is extracted by sliding window
The feature of moving-target, then with trained support vector machines (Support Vector Machine, SVM) classifier to human body
The behavior of moving target is classified, and when SVM classifier is correct to behavior classification and does not judge by accident, will issue abnormal letter
Breath.And current wrong data will be dropped when judging by accident and stored into embedded system, next time encounters such again
Judging by accident would not occur again in behavior.In this way by long-term use, the precision of the unusual checking system will be increasingly
Height reaches relatively high accuracy.Such unusual checking system allows injure caused by the abnormal behaviour in family and drop to
It is minimum.
In embodiments of the present invention, a kind of unusual checking system is provided, it, can by the long-term scanning of camera
Realization for example falls down the old man in family, child disorderly climbs such unusual checking, can be to family when being abnormal behavior
In other people carry out audio alert or make to send alarm short message and APP alarm sounds etc. to outgoing human hair, to allow indoor
Other people or outgoing people know whether old man and child are abnormal behavior in family at the first time.In such abnormal behaviour
Under detection system system, it can be reduced being injured as caused by abnormal behaviour in family, to the safety of old man in family and child
There is certain guarantee.
Technical solution of the present invention is described further by way of examples below, as shown in figure 3, real for the present invention
One embodiment schematic diagram for detecting the method for abnormal behaviour in example is applied, may include:
301, the video data about target object is obtained.
In embodiments of the present invention, Android embedded system connects net by WIFI communication module or 4G communication module
Network.Android embedded system realizes function of surfing the Net by calling WIFI communication module to connect accessing home gateway into network.4G
The interface connection that communication module function of surfing the Net is provided by point-to-point protocol (Point to Point Protocol, PPP) dialing
Network (executes PPP Dial Up script) after dial-up success, and kernel can generate the PPP network equipment, can be visited by creating socket
Ask network.4G communication module needs additional connection antenna and subscriber identification card (Subscriber Identification
Module, SIM) card, guarantee successfully network to enable 4G communication module.Embedded system can realize short message hair by order
It send, telephone receiving and dial, corresponding embedded device can by sending corresponding command calls 4G communication module to serial ports
To realize that short message is sent and phone answering and dials.
Camera module connects holder and reconnects hot electrostatic infrared module, is referred to as pyroelectric infrared-sensing module,
Hot electrostatic infrared module is connected on system board by serial ports interruption, be connected to hot electrostatic camera nobody when can auto sleep
It stops working, holder can control camera to rotate freely, and cover broader visual angle.Camera starts after system connects power supply
Work obtains video data.
302, the position of target object is obtained by the processing of the Gaussian Background modeling method updated according to the video data
Confidence breath, size and area information.
The processing for passing through the Gaussian Background modeling method updated according to the video data, obtains the position of target object
Confidence breath, size and area information may include: to pass through the place of the Gaussian Background modeling method updated according to the video data
Reason, target object is split from background, and the target object split includes the location information, big of target object
Small and area information.
Illustratively, camera is scanned indoor people, when camera scanning is to people, is carried on the back by improved Gauss
Scape modeling method splits moving target (such as people), which is that the value of each pixel in video image is recognized
Gaussian Profile for the stochastic variable for being Gaussian distributed, and each pixel be it is independent, by judging pixel
Color value whether the Gaussian Profile of this pixel to determine whether belonging to background pixel, Gauss formula is as follows:
Wherein, It(x, y) is the video image of t moment, μ (x, y) and σ2(x, y) is the mean value and variance of pixel value.Formula
(1-1) indicates every pixel color value all Gaussian distributed functions from the time.
It should be noted that detailed process is as follows for improved Gaussian Background modeling method:
Step 1: the video frame of input being divided into M*N region, each region includes 10*10 pixel, counts its gray scale picture
Plain histogram Ht(i, j), and preceding video frame is calculated with Pasteur's distance in region, result is saved in two-dimensional array DtIn (i, j):
Step 2: initialization first gives each position to establish a Gaussian Profile after reading in first frame image, mean value the
The pixel value of one frame corresponding position, when handling a new frame image, if the pixel value newly read in can be with original Gaussian Profile
Match, then only update that original Gaussian Profile, otherwise establishes one newly with the mean value and variance of the pixel value newly read in
Gaussian Profile, so go on, until the number of Gaussian Profile reaches K.
Step 3: every in t moment image being made the following judgment, output is exportedt(x, y) image, it is available
Foreground pixel, wherein k is fixed coefficient, generally takes 2.
Step 4: judged using formula (1-1), if it is determined that 0, then it needs to execute more background using following formula
New operation:
μt(x, y)=(1- α) μt-1(x,y)+αXt (1-5)
Wherein, XtFor the pixel value of the t moment point, σ value range is 0 < α < 1, its effect is control context update
Speed, the renewal speed of background is fast when σ value is arranged larger, and on the contrary then update is slowly.
Step 5: to each, new image program re-executes above-mentioned steps 2,3.
Foreground detection is calculated according to formula (1-4), and context update region is then sentenced according to formula (1-3)
It is disconnected, it is D for regiontGaussian Background update is carried out when (i, j)=1, for DtIt is updated when (i, j)=0 without Gaussian Background,
Wherein T=0.25.
Human body target is extracted by above-mentioned improved Gaussian Background modeling method.Illustratively, old man falls down
Disorderly climbing with child is a continuous dynamic process, i.e., human body target can occur in one section of continuous video, then next
Human body target can be tracked using a kind of TLD target tracking algorism based on KCF, enable the system to continuously obtain human body
The motion feature of target improves old man and falls down the accuracy rate for disorderly climbing detection with child.
303, according to the location information of the target object, size and area information, pass through the target based on KCF and TLD
Track algorithm tracks the target object, obtains the feature of the target object.
It is described according to the location information of the target object, size and area information, pass through the target based on KCF and TLD
Track algorithm tracks the target object, obtains the feature of the target object, may include: according to the target pair
Location information, size and the area information of elephant carry out the target object by the target tracking algorism based on KCF and TLD
Tracking, obtains target signature of the target object within the object time.
Believed above by position, size and region that improved Gaussian Background modeling method obtains human body target
Breath.It since the posture of human body target is changeable, and is easy to appear human body target and is blocked, it is existing that dimensional variation etc. occurs in human body target
As so having used a kind of TLD target tracking algorism based on KCF, adaptation of this method for the dimensional variation of human body target
Property it is strong, and the algorithm has re-detection ability, can detect line trace of going forward side by side rapidly after human body target appearance, and can be very
Handle human body target occlusion issue well, after target transient loss can recapture target rapidly, continue to track target.
As shown in figure 4, for the frame diagram of KCF-TLD algorithm used in the embodiment of the present invention.KCF- shown in Fig. 4
Comprising three tracking, detection, study modules in TLD algorithm, respective function is as follows:
Detection module: using the cascade classifier in TLD algorithm as detector, traversal obtains target given zone in the picture
Domain judges that it, what is obtained after screening is exactly most by variance classifier, Ensemble classifier and nearest neighbor classifier respectively
Whole testing result.
Tracking module: with KCF after improving (multiple dimensioned improvement carried out to KCF, it is additional choose the region more smaller than target with
And the region more bigger than target, it is obtained with the region of search of full size, smaller scale and large scale in this way, respectively at this
Three scales track target, then select optimal target response, the positions and dimensions of available target) algorithm be used as with
Track device.It is that, because tracking module time-consuming is shorter, algorithm can be improved based on tracking module in the TLD algorithm based on KCF
Treatment effeciency.After video input, tracking module is initially entered, track algorithm calculates target response as a result, root according to its model
According to the size of response, it can be determined that the reliability of tracking, if response is big, it is believed that tracking is accurate, using tracking result as calculation
The final result of method；If response is smaller.Think tracking failure, and detection module is activated to carry out re-detection, if detection module
Target position is found, then tracking module is trained using target.
Study module: study module is responsible for the training of detector and the update of tracker parameters.After tracking successfully, learn mould
The goal end position that root tuber is obtained according to algorithm extracts positive and negative samples and is trained to the classifier in detection module, positive sample
It is exactly the sample extracted from target area.Negative sample is exactly the sample extracted in background area, while goal-orientation is extracted
The final of algorithm can also drop in region of search, the correlation filtering parameter and target appearance model, study module updated in tracking module
As a result tracking module is passed to, for tracking next time.
As shown in figure 5, for the TLD target tracking algorism flow chart based on KCF in the embodiment of the present invention.It is as follows:
Specific step is as follows for TLD target tracking algorism based on KCF:
Firstly, being the initial phase of algorithm, video sequence is read, it then will using improved Gaussian Background modeling method
Humanbody moving object is split from background, then using the humanbody moving object split as tracking initial information,
Algorithm trains the correlation filter in tracking module according to the information of acquisition, meanwhile, positive negative sample is extracted in the video sequence,
After obtaining positive negative sample, in detection module Ensemble classifier and nearest classifier carry out initialization training.Tracking module with
After detection module initialization is completed, initial phase terminates.
In tracking phase, following steps are passed through for the processing of each frame of video:
(a) judging in previous frame, does algorithm successfully track target if tracking successfully, go to step (b), otherwise,
Go to step (c).
(b) dbjective state obtained according to previous frame is selected region of search, is tracked using KCF algorithm after improvement, if
The corresponding response of target is greater than preset threshold value, then it is assumed that tracks successfully, goes to step (e), otherwise go to step (c).
(c) start detection module, the overall situation is carried out to image using different scales and searches element, by cascade classifier to obtaining
Object candidate area is screened, and is retained by the conduct testing result of classifier, is gone to step (d).
(d) testing result is merged, the high multiple target frames of degree of overlapping is permeated a, to the detection after cluster
As a result it is analyzed.If number of results is 0, go to step (f), if number of results is 1, what is detected is target, is jumped
To step (e)；If the result detected is multiple, then it is assumed that detection failure gos to step (f).
(e) according to obtained target position, joined using the adaptive learning that the relative similarity of target calculates tracking module
Number extracts region of search feature, updates the parameter and display model of correlation filter；Extract positive negative sample, training detection module
In classifier；Label present frame successfully tracks target, goes to step (g).
(f) label present frame tracking failure, goes to step (g).
(g) judge whether the last frame for reaching sequence, if it is, algorithm terminates, exit tracking；Otherwise, step is gone to
Suddenly (a).
From algorithm above workflow it is recognised that the start-up trace module first of the TLD algorithm based on KCF, if tracking
It is successful then continue with next frame.Only tracking module fail when, just will start detection module, i.e., only target lose or
When being blocked, re-detection just will do it.So the operational efficiency of the algorithm is very high, and when humanbody moving object is blocked
Afterwards, can be by re-detection mechanism recapture target, the effect of target following when realizing long.
It may be implemented to carry out the human body target split accurately next to need with stable tracking by above method
Feature extraction carried out to human body target, do so the dimension that can reduce initial data, while the crucial letter of human body can be retained
Breath, to guarantee that the real-time and accuracy of the identification of these abnormal behaviours are disorderly climbed in old man's wrestling and child.Feature selecting and extraction
Main purpose be effectively to distinguish different behavior classifications by choosing least feature.In intelligent video monitoring system
In, the feature vector of selection is more, and characteristic vector space dimension is higher, and classifier design is more complicated, and the real-time of system is got over
Difference, but if feature is fewer, classification accuracy is lower, and discrimination is also lower, it is therefore desirable to suitable feature quantity is selected, from
And guarantee the accuracy and performance of system.
Before carrying out feature extraction, it is to be understood that old man and child common behavior indoors can just be selected effective
Feature creeps normal behaviour to distinguish with wrestling behavior and disorderly, and indoor human body movement normal behaviour has walking, jogs, is curved
Waist squats down, gets up, lies down, jumps, sits down and stands up.
Interior, which is fallen down, is divided into two types, is that original place is fallen down and fallen down in walking respectively, and the first kind is fallen down mobile in the direction x
It is unobvious, and the second class is fallen down movement in the x direction.Behavior is fallen down for these two types, the main distinction is people in level
Whether movement is had, and it is essentially the same for variation characteristic in resemblance and the direction y, therefore select feature i.e. selection and x coordinate
Change related feature, and child disorderly climbs and changes in the x direction unobvious, has movement in y-direction, therefore feature is selected to select
Select and change related feature with y-coordinate, but need to identify the behavior of falling down from normal behaviour and disorderly creep for, it is necessary to extract row
For multiple features, could effectively identify that old man falls down and the abnormal behaviours such as disorderly climb with child.
So the diagonal line intersection position for the minimum rectangular area that centroid position is human body can be defined, it is assumed that rectangle frame
Upper left side and lower right be (xl,yl), (xr,yr), then mass center of human body (xc,yc) can be obtained by following formula:
Obviously it can be seen that when wrestling behavior occurs in old man, mass center is obvious in the decline of y-coordinate direction: and when child occurs
Disorderly creep for when, mass center rises obvious in y-coordinate direction.
Simultaneously in order to avoid the height difference of different people causes error, needs that mass center is normalized in y-direction, see
Following formula:
ync=(yc-yb)/H (1-11)
Wherein, H is the maximum height of human body, can be obtained by above-mentioned target following, ybParticle when for human motion
Minimum point.This feature effectively will can walk and fall down behavior and disorderly creep for classification, but need by other behaviors with
It falls down behavior and disorderly creeps to distinguish, it is also necessary to which human body the ratio of width to height, horizontal vertical projection histogram and human motion are special
The behavioural characteristics such as sign.
Human body the ratio of width to height is defined as the ratio of human body width and height, for judging that human body is to stand or lie down.Preceding
Scape detects after obtaining target bianry image, calculates its horizontal and vertical projection histogram to indicate the contour feature of human body, horizontal
It is exactly the horizontal pixel quantity for calculating target and longitudinal pixel quantity with vertical projective histogram:
HZ(y)=| (xp,yp)∈F,yp=y | (1-12)
Vt(x)=| (xp,yp)∈F,xp=x | (1-13)
Since projection histogram can be different and different with the position of target, so after obtaining human body target, first to must
To profile be normalized, most common method is that the size of human body target is readjusted regular length M, still
Normalized parameter M needs different experiments to obtain different scenes, because it is high to human body attitude dependency degree, and very quick
Sense, in order to avoid these problems, can be used Fourier transformation to be normalized, specific practice is as follows:
Fourier coefficient is weak with the increase of k and v, and the difference of different postures is in initial part, the transformation of setting
Siding-to-siding block length is 30, is normalized after obtaining the Fourier transformation data of histogram by following formula:
It chooses(normalized horizontal histogram) is chosen as the feature for falling down detection(normalization
Vertical histogram) as the such spy of feature of climbing detection can reduce sign data redundancy, and extracting this feature can be effective
Distinguish wrestling and other behaviors and climbing and other behaviors.
What front was analyzed is the resemblance of human body, and the behavior of falling down can be efficiently differentiated from human body attitude and non-falls down row
For, and climbing behavior and non-climbing behavior are effectively distinguished, but human body is non-to be fallen down there are many behavior and non-climbing behaviors, it is some
Behavior easily occurs in similar to the behavior of falling down or climbing behavior in shape posture, and features described above cannot completely accurate area
Point, it is difficult to distinguish by features described above for example, people lies on the floor to lie down automatically with people after falling down, it may be with when people is after spring
Climbing behavior is difficult to distinguish by features described above, so being added to human body motion feature to increase the accuracy of classifier.
Behaviors, the particle such as falling down, lie down, squat down, jump and climb can all have significantly in the velocity and acceleration on y
Difference, wherein the acceleration climbed is contrary with the acceleration fallen down, lie down, squatted down, it is easy to distinguish climbing behavior and its
His behavior, then falls down and squats down.Velocity and acceleration of the equal behaviors on y of lying down has apparent difference, falls down y in behavior and becomes
The speed of change is significantly greater than and squats down, and the acceleration jumped is significantly greater than the acceleration climbed.The mass center for defining adjacent two frame is
(xi,yi),(xi+1,yi+1), then define the speed v of the human body target movement of i+1 framei+1:
vi+1=(yi+1-yi)/t (1-18)
Wherein, t is the time interval of two picture frames, defines the acceleration a of i+1 moment human body target movementi+1:
ai+1=(vi+1-v)/t (1-19)
By human body target in the available every frame of above two step in the acceleration in y-axis direction, have accurately by this feature
It distinguishes falling over of human body, lie down, squat down, jump and climb behavior.
Wherein, the target signature for obtaining the target object within the object time may include: by based on sliding
Windowhood method obtains target signature of the target object within the object time.
Which validity feature of selective extraction is described in detail above, but falling down behavior and climbing behavior is actually one
Whether a continuous action, cannot be only by the feature of human body target in one frame of identification to determine whether falling down and climbing.Otherwise,
It being easy to cause erroneous detection or missing inspection, the classification problem that detection is a time series in fact is fallen down and climb, when by obtaining one section
Between human body target feature, come judge behavior fall down behavior or climbing behavior be also it is non-fall down climbing behavior, for falling down
Feature is extracted based on sliding window method using one kind with climbing detection time sequence problem.
Sliding window method i.e. system using the sliding window of a fixed size come storage time sequence data, so
Afterwards as time goes by, sliding window is moved to the left, and the sequence data newly entered is put into window end, and the sequence in left side is moved out of
Window, sliding window are moved to the left, and the sequence data newly entered is put into window end, and the sequence in left side is moved out of window, sliding
Window is actually identical with common queuing nature.It is exactly sliding window that behavioural characteristic and climbing behavioural characteristic model are fallen down in extraction
Continuous human body target Fusion Features in mouthful are configured to characteristic vector space, this characteristic vector space is then sent into SVM points
Class device detection fall down, climbing behavior.
304, the first behavior when the feature and preset SVM classifier according to the target object to the target object
It is detected, when determining the first behavior abnormal behaviour of the target object, issues prompt messages.
When the first behavior abnormal behaviour of the determination target object, issue prompt messages, may include: to
Terminal device sends prompt messages, and the prompt messages include the text prompt information about abnormal behaviour, or
Speech prompt information about abnormal behaviour.
It should be noted that the climbing detection method of falling down of svm classifier is divided into two parts；Fall down behavior sample and climbing
Behavior sample off-line training and behavior of falling down and climbing behavior on-line checking, as shown in fig. 6, being examined to classify in the embodiment of the present invention
The idiographic flow schematic diagram of survey.As shown in Figure 6, it is mainly included in line monitoring modular and off-line training module, is said respectively
It is bright, as follows:
Off-line training: the process is exactly to need to fall down behavior sample data and child with a large amount of old men disorderly to creep as sample number
Accordingly and it is non-fall down climbing behavior sample data be sent in SVM, by training these sample data sets obtain old man fall down, it is small
The classifier that child disorderly climbs.
On-line checking: the process is that human body target feature is put into sliding window in the successive video frames video extraction, so
Feature is subjected to fusion extraction afterwards, is then fed into falling down in climbing behavior classifier for trained completion, carries out decision.
Finally, will pass through after embedded system receives old man and falls down the abnormal behaviour information disorderly climbed with child
Modules send Serial Port Information, after voice module receives exception information, can play alarm voice indoors immediately, allow room
Interior people knows that something unexpected happened at the first time；4G communication module also can just receive exception information, then can be by sending out to mobile phone
Send warning information, can also the alarm on the monitoring APP of mobile phone, in this case can allow outgoing people that can check information, sentence
It is disconnected whether accident really to occur.In this case the safety that old man and child stay at home can have largely been ensured.
In embodiments of the present invention, it is calculated using improved Gaussian Background modeling method and the TLD target following based on KCF
Method splits the humanbody moving object of camera video and carries out stable tracking, does first to Gaussian Background modeling method
Improvement is gone out, in the case where the background variation indoors of improved Gaussian Background modeling method and indoor lighting conditions variation still
Can be highly effective and accurately extract humanbody moving object.TLD target tracking algorism based on KCF is also to KCF algorithm
Improvement is made that with TLD algorithm, improves the processing frame per second of algorithm, while keeping the on-line training of detection module, improves calculation
The re-detection ability of method, to realize the quick tracking to target.And when humanbody moving object is blocked or occurs ruler
When degree variation, which still can stablize accurate tracking humanbody moving object in real time.
Further, the embodiment of the present invention utilizes feature extraction and SVM classifier based on sliding window to realize to exception
The detection of behavior not only reduces data redundancy in this way, and can reasonably extract effective feature, is then combined into
Feature vector is sent to trained SVM classifier and classifies.It in this way can be fast and accurately to humanbody moving object
Behavior is classified, and generalization ability is also very good.Old man in family is also detected by above method to fall down and child's unrest
It the abnormal behaviours such as climbs, and has built complete family's unusual checking system, and the system sweeping for a long time by camera
It retouches, can constantly improve in this way and old man is fallen down so that data are more and more perfect by collecting data and storing data
The precision of equal behavioral values is disorderly climbed with child so that in family the injury as caused by abnormal behaviour fall below it is minimum.
As shown in fig. 7, may include: for one embodiment schematic diagram of detection device in the embodiment of the present invention
Module 701 is obtained, for obtaining the video data about target object；
Processing module 702, for, by the processing of the Gaussian Background modeling method updated, being obtained according to the video data
Location information, size and the area information of target object；According to the location information of the target object, size and area information,
The target object is tracked by the target tracking algorism based on KCF and TLD, obtains the feature of the target object；
The first behavior of the target object is detected when according to the feature and preset SVM classifier of the target object, really
When the first behavior abnormal behaviour of the fixed target object, prompt messages are issued.
Optionally, in some embodiments of the invention,
Processing module 702, specifically for passing through the processing of the Gaussian Background modeling method updated according to the video data,
Target object is split from background, the target object split includes the location information of target object, size
And area information.
Optionally, in some embodiments of the invention,
Processing module 702, specifically for passing through base according to the location information of the target object, size and area information
The target object is tracked in the target tracking algorism of KCF and TLD, obtains the target object within the object time
Target signature.
Optionally, in some embodiments of the invention,
Processing module 702 is specifically used for by obtaining the target object within the object time based on sliding window method
Target signature.
Optionally, in some embodiments of the invention,
Processing module 702 is specifically used for sending prompt messages to terminal device, and the prompt messages include closing
In the text prompt information of abnormal behaviour, or about the speech prompt information of abnormal behaviour.
As shown in figure 8, may include: for another embodiment schematic diagram of detection device in the embodiment of the present invention
Transceiver 801, processor 802, memory 803, wherein transceiver 801, processor 802 and memory 803 pass through
Bus connection；
Memory 803, for storing operational order；
Transceiver 801, for obtaining the video data about target object；
Processor 802 is modeled according to the video data by the Gaussian Background updated for calling the operational order
The processing of method obtains location information, size and the area information of target object；According to the location information of the target object,
Size and area information track the target object by the target tracking algorism based on KCF and TLD, obtain described
The feature of target object；When the feature and preset SVM classifier according to the target object are to the first of the target object
Behavior is detected, and when determining the first behavior abnormal behaviour of the target object, issues prompt messages.
Optionally, in some embodiments of the invention,
Processor 802 will specifically for the processing according to the video data by the Gaussian Background modeling method of update
Target object is split from background, the target object split include the location information of target object, size and
Area information.
Optionally, in some embodiments of the invention,
Processor 802, specifically for according to the location information of the target object, size and area information, by being based on
The target tracking algorism of KCF and TLD tracks the target object, obtains mesh of the target object within the object time
Mark feature.
Optionally, in some embodiments of the invention,
Processor 802, specifically for by obtaining the target object within the object time based on sliding window method
Target signature.
Optionally, in some embodiments of the invention,
Processor 802, be specifically used for terminal device send prompt messages, the prompt messages include about
The text prompt information of abnormal behaviour, or the speech prompt information about abnormal behaviour.
In the above-described embodiments, can come wholly or partly by software, hardware, firmware or any combination thereof real
It is existing.When implemented in software, it can entirely or partly realize in the form of a computer program product.
The computer program product includes one or more computer instructions.Load and execute on computers the meter
When calculation machine program instruction, entirely or partly generate according to process or function described in the embodiment of the present invention.The computer can
To be general purpose computer, special purpose computer, computer network or other programmable devices.The computer instruction can be deposited
Storage in a computer-readable storage medium, or from a computer readable storage medium to another computer readable storage medium
Transmission, for example, the computer instruction can pass through wired (example from a web-site, computer, server or data center
Such as coaxial cable, optical fiber, Digital Subscriber Line (DSL)) or wireless (such as infrared, wireless, microwave) mode to another website
Website, computer, server or data center are transmitted.The computer readable storage medium can be computer and can deposit
Any usable medium of storage either includes that the data storages such as one or more usable mediums integrated server, data center are set
It is standby.The usable medium can be magnetic medium, (for example, floppy disk, hard disk, tape), optical medium (for example, DVD) or partly lead
Body medium (such as solid state hard disk Solid State Disk (SSD)) etc..
It is apparent to those skilled in the art that for convenience and simplicity of description, the system of foregoing description,
The specific work process of device and unit, can refer to corresponding processes in the foregoing method embodiment, and details are not described herein.
In several embodiments provided by the present invention, it should be understood that disclosed system, device and method can be with
It realizes by another way.For example, the apparatus embodiments described above are merely exemplary, for example, the unit
It divides, only a kind of logical function partition, there may be another division manner in actual implementation, such as multiple units or components
It can be combined or can be integrated into another system, or some features can be ignored or not executed.Another point, it is shown or
The mutual coupling, direct-coupling or communication connection discussed can be through some interfaces, the indirect coupling of device or unit
It closes or communicates to connect, can be electrical property, mechanical or other forms.
The unit as illustrated by the separation member may or may not be physically separated, aobvious as unit
The component shown may or may not be physical unit, it can and it is in one place, or may be distributed over multiple
In network unit.It can select some or all of unit therein according to the actual needs to realize the mesh of this embodiment scheme
's.
It, can also be in addition, the functional units in various embodiments of the present invention may be integrated into one processing unit
It is that each unit physically exists alone, can also be integrated in one unit with two or more units.Above-mentioned integrated list
Member both can take the form of hardware realization, can also realize in the form of software functional units.
If the integrated unit is realized in the form of SFU software functional unit and sells or use as independent product
When, it can store in a computer readable storage medium.Based on this understanding, technical solution of the present invention is substantially
The all or part of the part that contributes to existing technology or the technical solution can be in the form of software products in other words
It embodies, which is stored in a storage medium, including some instructions are used so that a computer
Equipment (can be personal computer, server or the network equipment etc.) executes the complete of each embodiment the method for the present invention
Portion or part steps.And storage medium above-mentioned includes: USB flash disk, mobile hard disk, read-only memory (ROM, Read-Only
Memory), random access memory (RAM, Random Access Memory), magnetic or disk etc. are various can store journey
The medium of sequence code.
The above, the above embodiments are merely illustrative of the technical solutions of the present invention, rather than its limitations；Although referring to before
Stating embodiment, invention is explained in detail, those skilled in the art should understand that: it still can be to preceding
Technical solution documented by each embodiment is stated to modify or equivalent replacement of some of the technical features；And these
It modifies or replaces, the spirit and scope for technical solution of various embodiments of the present invention that it does not separate the essence of the corresponding technical solution.
Claims (10)
1. a kind of detection method of abnormal behaviour characterized by comprising
Obtain the video data about target object；
According to the video data by update Gaussian Background modeling method processing, obtain target object location information,
Size and area information；
According to the location information of the target object, size and area information, pass through the target tracking algorism based on KCF and TLD
The target object is tracked, the feature of the target object is obtained；
The first behavior of the target object is examined when according to the feature and preset SVM classifier of the target object
It surveys, when determining the first behavior abnormal behaviour of the target object, issues prompt messages.
2. the method according to claim 1, wherein described carried on the back according to the video data by the Gauss updated
The processing of scape modeling method obtains location information, size and the area information of target object, comprising:
According to the video data by the processing of the Gaussian Background modeling method updated, target object is partitioned into from background
Come, the target object split includes location information, size and the area information of target object.
3. method according to claim 1 or 2, which is characterized in that the location information according to the target object, big
Small and area information tracks the target object by the target tracking algorism based on KCF and TLD, obtains the mesh
Mark the feature of object, comprising:
According to the location information of the target object, size and area information, pass through the target tracking algorism based on KCF and TLD
The target object is tracked, target signature of the target object within the object time is obtained.
4. according to the method described in claim 3, it is characterized in that, the mesh for obtaining the target object within the object time
Mark feature, comprising:
By obtaining target signature of the target object within the object time based on sliding window method.
5. method according to claim 1 or 2, which is characterized in that the first behavior of the determination target object is different
When Chang Hangwei, prompt messages are issued, comprising:
Prompt messages are sent to terminal device, the prompt messages include the text prompt letter about abnormal behaviour
Breath, or about the speech prompt information of abnormal behaviour.
6. a kind of detection device characterized by comprising
Module is obtained, for obtaining the video data about target object；
Processing module, for, by the processing of the Gaussian Background modeling method updated, obtaining target pair according to the video data
Location information, size and the area information of elephant；According to the location information of the target object, size and area information, pass through base
The target object is tracked in the target tracking algorism of KCF and TLD, obtains the feature of the target object；Work as basis
The feature of the target object and preset SVM classifier detect the first behavior of the target object, described in determination
When the first behavior abnormal behaviour of target object, prompt messages are issued.
7. detection device according to claim 6, which is characterized in that
The processing module will specifically for the processing according to the video data by the Gaussian Background modeling method of update
Target object is split from background, the target object split include the location information of target object, size and
Area information.
8. detection device according to claim 6 or 7, which is characterized in that
The processing module, specifically for according to the location information of the target object, size and area information, by being based on
The target tracking algorism of KCF and TLD tracks the target object, obtains mesh of the target object within the object time
Mark feature；
The processing module, specifically for by obtaining mesh of the target object within the object time based on sliding window method
Mark feature；
The processing module, be specifically used for terminal device send prompt messages, the prompt messages include about
The text prompt information of abnormal behaviour, or the speech prompt information about abnormal behaviour.
9. a kind of detection device characterized by comprising
Transceiver, processor, memory, wherein the transceiver, the processor and the memory are connected by bus；
The memory, for storing operational order；
The transceiver, for obtaining the video data about target object；
The processor executes abnormal behaviour according to any one of claims 1 to 5 for calling the operational order
The step of detection method.
10. a kind of readable storage medium storing program for executing, is stored thereon with computer program, which is characterized in that the computer program is processed
The step of device realizes the detection method of abnormal behaviour according to any one of claims 1 to 5 when executing.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910589840.9A CN110414360A (en) | 2019-07-02 | 2019-07-02 | A kind of detection method and detection device of abnormal behaviour |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910589840.9A CN110414360A (en) | 2019-07-02 | 2019-07-02 | A kind of detection method and detection device of abnormal behaviour |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110414360A true CN110414360A (en) | 2019-11-05 |
Family
ID=68358672
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910589840.9A Pending CN110414360A (en) | 2019-07-02 | 2019-07-02 | A kind of detection method and detection device of abnormal behaviour |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110414360A (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113065500A (en) * | 2021-04-15 | 2021-07-02 | 电子科技大学 | Abnormal behavior control system for special actions |
| CN113158858A (en) * | 2021-04-09 | 2021-07-23 | 苏州爱可尔智能科技有限公司 | Behavior analysis method and system based on deep learning |
| CN113518201A (en) * | 2020-07-14 | 2021-10-19 | 阿里巴巴集团控股有限公司 | Video processing method, device and equipment |
| CN114187666A (en) * | 2021-12-23 | 2022-03-15 | 中海油信息科技有限公司 | Identification method and system for watching mobile phone while walking |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105718857A (en) * | 2016-01-13 | 2016-06-29 | 兴唐通信科技有限公司 | Human body abnormal behavior detection method and system |
| US20170206423A1 (en) * | 2014-07-22 | 2017-07-20 | S-1 Corporation | Device and method surveilling abnormal behavior using 3d image information |
| CN108509938A (en) * | 2018-04-16 | 2018-09-07 | 重庆邮电大学 | A kind of fall detection method based on video monitoring |
| CN109670396A (en) * | 2018-11-06 | 2019-04-23 | 华南理工大学 | A kind of interior Falls Among Old People detection method |
-
2019
- 2019-07-02 CN CN201910589840.9A patent/CN110414360A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20170206423A1 (en) * | 2014-07-22 | 2017-07-20 | S-1 Corporation | Device and method surveilling abnormal behavior using 3d image information |
| CN105718857A (en) * | 2016-01-13 | 2016-06-29 | 兴唐通信科技有限公司 | Human body abnormal behavior detection method and system |
| CN108509938A (en) * | 2018-04-16 | 2018-09-07 | 重庆邮电大学 | A kind of fall detection method based on video monitoring |
| CN109670396A (en) * | 2018-11-06 | 2019-04-23 | 华南理工大学 | A kind of interior Falls Among Old People detection method |
Non-Patent Citations (5)
| Title |
|---|
| JIA DONGYAO 等: "A method of detecting human body falling action in a complex background", 《2016 IEEE 20TH JUBILEE INTERNATIONAL CONFERENCE ON INTELLIGENT ENGINEERING SYSTEMS》 * |
| 禹明娟: "基于视频的室内老人摔倒检测研究", 《中国优秀博硕士学位论文全文数据库(硕士)信息科技辑》 * |
| 艾达等: "基于监控视频的行人异常行为检测技术研究", 《电视技术》 * |
| 郭娅祥: "多目标跟踪系统的关键技术研究及实现", 《中国优秀博硕士学位论文全文数据库(硕士)信息科技辑》 * |
| 阿斯雅 等: "一种结合区域直方 图比较的单高斯背景建模算法的改进", 《计算机光盘软件与应用》 * |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113518201A (en) * | 2020-07-14 | 2021-10-19 | 阿里巴巴集团控股有限公司 | Video processing method, device and equipment |
| CN113158858A (en) * | 2021-04-09 | 2021-07-23 | 苏州爱可尔智能科技有限公司 | Behavior analysis method and system based on deep learning |
| CN113065500A (en) * | 2021-04-15 | 2021-07-02 | 电子科技大学 | Abnormal behavior control system for special actions |
| CN114187666A (en) * | 2021-12-23 | 2022-03-15 | 中海油信息科技有限公司 | Identification method and system for watching mobile phone while walking |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110414360A (en) | A kind of detection method and detection device of abnormal behaviour | |
| CN111666857B (en) | Human behavior recognition method, device and storage medium based on environment semantic understanding | |
| US10009579B2 (en) | Method and system for counting people using depth sensor | |
| CN112639873A (en) | Multi-object pose tracking device and method based on single-object pose estimator | |
| US20100278391A1 (en) | Apparatus for behavior analysis and method thereof | |
| CN109154976A (en) | Pass through the system and method for machine learning training object classifier | |
| CN108805900B (en) | Method and device for determining tracking target | |
| CN110807385A (en) | Target detection method and device, electronic equipment and storage medium | |
| CN106897658A (en) | The discrimination method and device of face live body | |
| JPWO2018025831A1 (en) | People flow estimation device, people flow estimation method and program | |
| US20150262068A1 (en) | Event detection apparatus and event detection method | |
| CN104933827A (en) | Rotary camera-based alarm method and terminal | |
| CN111696128A (en) | High-speed multi-target detection tracking and target image optimization method and storage medium | |
| CN111553326B (en) | Hand motion recognition method and device, electronic equipment and storage medium | |
| CN111914657B (en) | Pet behavior detection method and device, electronic equipment and storage medium | |
| CN112307868A (en) | Image recognition method, electronic device, and computer-readable medium | |
| CN109255360A (en) | A kind of objective classification method, apparatus and system | |
| KR20210062256A (en) | Method, program and system to judge abnormal behavior based on behavior sequence | |
| Zhu et al. | Egoobjects: A large-scale egocentric dataset for fine-grained object understanding | |
| CN110443179A (en) | It leaves the post detection method, device and storage medium | |
| CN114120242A (en) | Monitoring video behavior analysis method, system and terminal based on time sequence characteristics | |
| CN113780145A (en) | Sperm morphology detection method, sperm morphology detection device, computer equipment and storage medium | |
| CN113627334A (en) | Object behavior identification method and device | |
| CN116129523A (en) | Action recognition method, device, terminal and computer readable storage medium | |
| CN111753587A (en) | Method and device for detecting falling to ground |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20191105 |