CN109871764A - A kind of abnormal behaviour recognition methods, device and storage medium - Google Patents
A kind of abnormal behaviour recognition methods, device and storage medium Download PDFInfo
- Publication number
- CN109871764A CN109871764A CN201910040080.6A CN201910040080A CN109871764A CN 109871764 A CN109871764 A CN 109871764A CN 201910040080 A CN201910040080 A CN 201910040080A CN 109871764 A CN109871764 A CN 109871764A
- Authority
- CN
- China
- Prior art keywords
- human body
- key point
- abnormal behaviour
- multiple image
- body key
- 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
Abstract
The present embodiments relate to nerual network technique fields, disclose a kind of abnormal behaviour recognition methods, comprising: obtain the multiple image comprising human body behavior, mark someone's body key point in multiple image；By in multiple image input shot and long term memory LSTM network model, the weighted value of each human body key point in multiple image is obtained；According to the corresponding relationship of the weighted value and abnormal behaviour of the weighted value of obtained each human body key point and pre-set each human body key point, the abnormal behaviour of human body in multiple image is determined.A kind of abnormal behaviour recognition methods, device and the storage medium that embodiment of the present invention provides can be improved the speed of identification abnormal behaviour, to improve recognition efficiency.
Description
Technical field
The present embodiments relate to nerual network technique field, in particular to a kind of abnormal behaviour recognition methods, device and
Storage medium.
Background technique
In recent years, the emergency event and anomalous event to take place frequently has seriously affected the public safety of society.For taking place frequently
Emergency event and the behavior that endangers public security such as anomalous event, countries in the world are using abnormal behaviour identification technology as peace
A kind of important means taken precautions against entirely.
Human body abnormal behaviour refers to the irregular improper behavior of human body, running, fight, fall ill, hesitate suddenly such as human body
It wanders, personnel largely assemble.Abnormal behaviour identification technology is that personnel's exception row is detected and identified from collected video sequence
For a kind of technology, which can effectively prevent the generation of the emergency events such as insurgent violence, tread event, can be monitoring
Personnel provide timely warning information, effectively related personnel are assisted to sentence in real time to the abnormal conditions occurred in monitoring scene
Break and takes measures.Currently, human body critical point detection field is usually that optical flow method is utilized to track human body key point, thus from continuous
The abnormal behaviour of human body is identified in the video image of frame.
However, it is found by the inventors that at least there are the following problems in the prior art: for more and more emergency events and different
Ordinary affair part when tracking human body abnormal behaviour using optical flow method in the prior art, is intended to each human body key point to carry out light stream
Method track identification, data processing amount is larger, causes abnormal behaviour recognition speed relatively slow and inefficient.
Summary of the invention
Embodiment of the present invention is designed to provide a kind of abnormal behaviour recognition methods, device and storage medium, energy
Accelerate the recognition speed of abnormal behaviour, enough to improve recognition efficiency.
In order to solve the above technical problems, embodiments of the present invention provide a kind of abnormal behaviour recognition methods, comprising: obtain
The multiple image comprising human body behavior is taken, marks someone's body key point in multiple image；By multiple image input shot and long term memory
In LSTM network model, the weighted value of each human body key point in multiple image is obtained；According to the power of obtained each human body key point
The corresponding relationship of the weighted value and abnormal behaviour of weight values and pre-set each human body key point, determines people in multiple image
The abnormal behaviour of body.
Embodiments of the present invention additionally provide a kind of abnormal behaviour identification device, comprising: at least one processor；With
And the memory being connect at least one processor communication；Wherein, memory is stored with and can be executed by least one processor
Instruction, instruction is executed by least one processor, so that at least one processor is able to carry out above-mentioned abnormal behaviour identification side
Method.
Embodiments of the present invention additionally provide a kind of computer readable storage medium, are stored with computer program, calculate
Machine program realizes above-mentioned abnormal behaviour recognition methods when being executed by processor.
Embodiment of the present invention in terms of existing technologies, provides a kind of abnormal behaviour recognition methods, comprising: obtains
Multiple image comprising human body behavior marks someone's body key point in multiple image；By multiple image input shot and long term memory
In LSTM network model, the weighted value of each human body key point in multiple image is obtained；According to the power of obtained each human body key point
The corresponding relationship of the weighted value and abnormal behaviour of weight values and pre-set each human body key point, determines people in multiple image
The abnormal behaviour of body.Light stream is carried out to human body key point each in sequential frame image using optical flow method in compared with the prior art
The way of method track identification passes through power of the LSTM network model directly to the human body key point in multiple image in present embodiment
Weight values are predicted, to determine people according to the weighted value of prediction and preset weighted value and the corresponding relationship of abnormal behaviour
The abnormal behaviour of body avoids a large amount of data processing without being tracked identification to each human body key point, greatly plus
The fast speed of abnormal behaviour identification, to improve the efficiency of abnormal behaviour identification.
In addition, LSTM network model is trained by following steps: obtaining the training set figure for being marked with human body key point
The label weighted value of each human body key point in picture and training set image；Training set image is inputted into the LSTM network model
In, predict the weighted value of each human body key point；The weighted value and label weighted value obtained according to prediction, calculates LSTM network mould
The loss function value of type；According to the parameter of obtained loss function value adjustment LSTM network model, so that loss function value meets
First preset condition.The concrete mode of trained LSTM network model is shown in the program.
In addition, the step of obtaining the multiple image comprising human body abnormal behaviour, someone's body key point is marked in multiple image,
It specifically includes: obtaining the multiple image for containing human body abnormal behaviour to be detected；Multiple image is inputted in critical point detection model,
It identifies each human body key point in multiple image, and marks each one body key point in multiple image.Key is utilized in the program
Point detection model detects each human body key point in multiple image automatically, can quickly obtain each frame image in multiple image
Human body key point, and it is more complete compared to obtained human body key point for human body key point is manually marked in the picture
Face.
In addition, each human body key point in identification multiple image, specifically includes: using similarity mode method in multiple image
Each frame image in track human body；Identify the key point of the human body tracked in each frame image.Using similar in the program
Degree matching method tracks human body in the picture, so that it is guaranteed that the human body key point detected belongs to same human body, avoids due to not
Key point with human body is obscured and the recognition accuracy of abnormal behaviour is caused to reduce.
In addition, critical point detection model is trained by following steps: obtaining the frame for being marked with the true key point of human body
Image set；Frame image set is inputted in critical point detection model, predicts human body key point；According to the human body key point of prediction and people
The true key point of body calculates the loss function value of prediction model；Critical point detection model is adjusted according to obtained loss function value
Parameter make loss function value meet the second preset condition.The training method of critical point detection model is given in the program.
Detailed description of the invention
One or more embodiments are illustrated by the picture in corresponding attached drawing, these exemplary theorys
The bright restriction not constituted to embodiment, the element in attached drawing with same reference numbers label are expressed as similar element, remove
Non- to have special statement, composition does not limit the figure in attached drawing.
Fig. 1 is the flow diagram of the abnormal behaviour recognition methods of first embodiment according to the present invention；
Fig. 2 is the flow diagram of the abnormal behaviour recognition methods of second embodiment according to the present invention；
Fig. 3 is the structural schematic diagram of the abnormal behaviour identification device of third embodiment according to the present invention.
Specific embodiment
In order to make the object, technical scheme and advantages of the embodiment of the invention clearer, below in conjunction with attached drawing to the present invention
Each embodiment be explained in detail.However, it will be understood by those skilled in the art that in each embodiment party of the present invention
In formula, in order to make the reader understand this application better, many technical details are proposed.But even if without these technical details
And various changes and modifications based on the following respective embodiments, the application technical solution claimed also may be implemented.
The first embodiment of the present invention is related to a kind of abnormal behaviour recognition methods, and the core of present embodiment is, mention
A kind of abnormal behaviour recognition methods is supplied, comprising: obtain the multiple image comprising human body behavior, mark someone's body in multiple image
Key point；By in multiple image input shot and long term memory LSTM network model, the power of each human body key point in multiple image is obtained
Weight values；According to the weighted value of obtained each human body key point and the weighted value and exception of pre-set each human body key point
The corresponding relationship of behavior determines the abnormal behaviour of human body in multiple image.Using optical flow method to continuous in compared with the prior art
Each human body key point carries out optical flow method track identification in frame image, and it is directly right to pass through LSTM network model in present embodiment
The weighted value of human body key point in multiple image predicted, thus according to the weighted value of prediction and preset weighted value with
The corresponding relationship of abnormal behaviour determines the abnormal behaviour of human body, greatly increases the speed of abnormal behaviour identification, to mention
The high efficiency of abnormal behaviour identification.
The realization details of the abnormal behaviour recognition methods of present embodiment is specifically described below, the following contents is only
For convenience of the realization details provided is understood, not implement the necessary of this programme.
The flow diagram of abnormal behaviour recognition methods in present embodiment is as shown in Figure 1:
Step 101: obtaining the multiple image comprising human body behavior, mark someone's body key point in multiple image.
Specifically, obtaining the multiple image comprising single behavior or the multiple image comprising more people's behaviors, abnormal behaviour
Such as: it makes a phone call, hover, fight, wherein making a phone call and hovering belongs to single abnormal behaviour, fights, and it is abnormal to belong to more people
Behavior.Someone's body key point is marked in multiple image in each frame image.Wherein, so-called human body key point includes at least: face,
The body joint point coordinate location information at the positions such as hand, foot, leg, human body key point can also include: human body face characteristic point
Co-ordinate position information.
Step 102: multiple image being inputted in LSTM network model, the weight of each human body key point in multiple image is obtained
Value.
Specifically, LSTM network model is used to predict the weighted value of human body key point in each frame image, it will be with someone
The LSTM network model that the multiple image input training of body key point is completed, it is crucial just can to obtain each human body in each frame image
The weighted value of point.Optical flow method is carried out to human body key point each in sequential frame image using optical flow method in compared with the prior art
Track identification directly carries out the weighted value of the human body key point in multiple image by LSTM network model in present embodiment
Prediction, to determine the exception of human body according to the weighted value of prediction and preset weighted value and the corresponding relationship of abnormal behaviour
Behavior, greatly increases the speed of abnormal behaviour identification, to improve the efficiency of abnormal behaviour identification
In addition, LSTM network model is trained by following steps: obtaining the training set figure for being marked with human body key point
The label weighted value of each human body key point in picture and training set image；Training set image is inputted into the LSTM network model
In, predict the weighted value of each human body key point；The weighted value and label weighted value obtained according to prediction, calculates LSTM network mould
The loss function value of type；According to the parameter of obtained loss function value adjustment LSTM network model, so that loss function value meets
First preset condition.
Specifically, LSTM network model is before the weighted value of each human body key point for predicting each frame image,
It needs to be trained LSTM network model to be embodied as each human body key point and distribute different importance.In training LSTM network
When model, need to be marked with the training set image of human body key point as the input of LSTM network model, LSTM network model
The weighted value and preparatory mark predicted the weighted value of each human body key point in each frame image in training set image, and prediction is obtained
In the loss function of label weighted value input LSTM network model of the note in training set image, loss function value is calculated, according to
The parameter for the loss function value adjustment LSTM network model being calculated, so that loss function value meets the first preset condition.Its
In, the first preset condition is that loss function value is less than a certain setting value.When setting value is sufficiently small, then show that loss function value is enough
Small, characterization LSTM network model training is completed.Specifically, setting value can be quasi- according to prediction of the user for LSTM network model
Exactness requires to be set, herein without specifically limiting.
It should be noted that the process of above-mentioned trained LSTM network model is the process of a circulation, loss is having been calculated
After function, after loss function value and the first preset condition comparison adjustment LSTM network model parameter, continuing will be in training set
Image input LSTM network model in predict the weighted value of each human body key point in training set image, and by the weighted value of prediction
It is again inputted into the loss function of LSTM network model with the label weighted value marked in advance in corresponding image, calculates loss
Functional value, until loss function value meets the first preset condition, i.e. LSTM network model training is completed.
Step 103: according to the weighted value of obtained each human body key point and the power of pre-set each human body key point
The corresponding relationship of weight values and abnormal behaviour determines the abnormal behaviour of human body in multiple image.
Specifically, being previously provided with the weighted value of each human body key point and the corresponding relationship of abnormal behaviour.For example, when different
Chang Hangwei is when making a phone call, and the movement at the positions such as elbow, wrist, shoulder of human body is larger, the corresponding key point power in these positions
Weight values are larger, and the position at other positions such as leg, foot etc. and this behavior relation of making a phone call are weaker, therefore, the power of distribution
Weight is also corresponding smaller.In addition, the hand of human body and the amplitude of variation of face are larger, these portions when abnormal behaviour is to smoke
The weighted value of the corresponding key point in position is larger, and this is different for the position of other positions such as chest, leg and foot etc. and smoking
The relationship of Chang Hangwei is weaker, and therefore, the weight of distribution is also corresponding smaller.It is corresponding for different abnormal behaviours in advance in present embodiment
The key point of partes corporis humani position distribute different weighted values, each human body that can be predicted according to LSTM network model is crucial
The weighted value of point determines the abnormal behaviour of human body in the image of input LSTM network model by pre-set corresponding relationship.
Further, when abnormal behaviour is more similar, the weighted value of relative each human body key point is also roughly the same
When, for such more similar abnormal behaviour, it is crucial the human body obtained by LSTM network model can be tracked by optical flow method
Weighted value distributes biggish key point to identify the type of abnormal behaviour in point, tracks all human body key points compared to optical flow method
For, the treating capacity of data is substantially reduced, the speed of abnormal behaviour identification is accelerated.Certainly, those skilled in the art can be with
Understand, for above situation, such more similar abnormal behaviour can also be distinguished using logic judgment, it is no longer superfluous herein
It states.
Compared with prior art, embodiment of the present invention provides a kind of abnormal behaviour recognition methods, comprising: acquisition includes
The multiple image of human body behavior marks someone's body key point in multiple image；Multiple image is inputted in LSTM network model, is obtained
The weighted value of each human body key point into multiple image；According to the weighted value of obtained each human body key point and preset
Each human body key point weighted value and abnormal behaviour corresponding relationship, determine the abnormal behaviour of human body in multiple image.It compares
In carrying out optical flow method track identification, this implementation to human body key point each in sequential frame image using optical flow method in the prior art
Directly the weighted value of the human body key point in multiple image is predicted by LSTM network model in mode, thus according to pre-
The weighted value of survey and preset weighted value and the corresponding relationship of abnormal behaviour determine the abnormal behaviour of human body, greatly improve
The speed of abnormal behaviour identification, to improve the efficiency of abnormal behaviour identification.
Second embodiment of the present invention is related to a kind of abnormal behaviour recognition methods.Second embodiment and the first embodiment party
The improvement of formula is identical, mainly thes improvement is that, obtain include human body abnormal behaviour multiple image, label has in multiple image
It the step of human body key point, specifically includes: obtaining the multiple image for containing human body abnormal behaviour to be detected；Multiple image is inputted
In critical point detection model, each human body key point in multiple image is identified, and each one body key point is marked in multiple image.
Flow diagram such as Fig. 2 of abnormal behaviour recognition methods in present embodiment shows, specifically includes:
Step 201: obtaining the multiple image for containing human body abnormal behaviour to be detected.
Specifically, multiframe image to be detected containing human body behavior is obtained, wherein whether has human body abnormal for detecting
Behavior.Such as: it, can be by monitor video when checking monitoring video needs to lock and has the suspicious suspect of a certain abnormal behaviour
Image making wherein whether there is human body abnormal behaviour for detecting at the image of successive frame, to help to lock abnormal behaviour
Suspicious suspect.
Step 202: multiple image is inputted in critical point detection model, identifies each human body key point in multiple image,
And each one body key point is marked in multiple image.
Specifically, multiple image to be detected is inputted in critical point detection model, it is each in automatic identification multiple image
The key point of human body, to be marked each human body key point in multiple image, so that LSTM network model is according to input
The label human body key point that there is the multiple image of key point to come in forecast image.It is examined in present embodiment using key point
It surveys model and detects each human body key point in multiple image automatically, can quickly obtain the people of each frame image in multiple image
Body key point, and it is more comprehensive compared to obtained human body key point for human body key point is manually marked in the picture.
In addition, critical point detection model is trained by following steps: obtaining frame image set to be detected and described
The true key point of human body in frame image set；Frame image set is inputted in critical point detection model, predicts human body key point；According to
The human body key point and the true key point of human body of prediction, calculate the loss function value of prediction model；According to obtained loss function
The parameter of value adjustment critical point detection model makes loss function value meet the second preset condition.
Specifically, critical point detection model is more convolutional neural networks.Critical point detection model is each for predicting
Before each human body key point of frame image, need to be trained critical point detection model to realize that each human body of Accurate Prediction is crucial
Point.Training critical point detection model when, need include human body to be detected frame image set as critical point detection model
Input, in critical point detection model prediction training set image in each frame image each human body key point weighted value, and will be pre-
The weighted value measured and the loss letter for marking the label weighted value in training set image to input critical point detection model in advance
In number, loss function value is calculated, according to the parameter for the loss function value adjustment critical point detection model being calculated, so that loss
Functional value meets the second preset condition.Wherein, the second preset condition is that loss function value is less than a certain setting value.When setting value foot
It is enough small, then show that loss function value is sufficiently small, characterization critical point detection model training is completed.Specifically, setting value can basis
User requires to set for the prediction accuracy of critical point detection model, herein without specifically limiting.
It should be noted that the process of above-mentioned trained critical point detection model is the process of a circulation, it is damaged calculating
After losing function, after loss function value and the first preset condition comparison adjustment critical point detection model parameter, continue to train
The weighted value of each human body key point in training set image is predicted in the image input critical point detection model of concentration, and by prediction
The label weighted value marked in advance in weighted value and corresponding image is again inputted into the loss function of critical point detection model,
Loss function value is calculated, until loss function value meets the second preset condition, i.e. critical point detection model training is completed.
Step 203: multiple image being inputted in LSTM network model, the weight of each human body key point in multiple image is obtained
Value.
Step 204: according to the weighted value of obtained each human body key point and the power of pre-set each human body key point
The corresponding relationship of weight values and abnormal behaviour determines the abnormal behaviour of human body in multiple image.
Above-mentioned steps 203 and step 204 in first embodiment step 102 and step 103 it is roughly the same, herein not
It repeats again.
Further, it identifies each human body key point in multiple image, specifically includes: using similarity mode method in multiframe
Human body is tracked in each frame image of image；Identify the key point of the human body tracked in each frame image.
Specifically, due to that can guarantee multiple human bodies in a frame image, in order to ensure obtaining same human body not
Human body key point in image at same frame, while avoiding since the identification of abnormal behaviour is obscured and caused to the key point of different human body
Accuracy rate reduces, and it is crucial to ensure to obtain human body of the human body in each frame image to need to track human body in multiple image
Point tracks human body using similarity mode method in the present embodiment, so that it is guaranteed that the human body key point detected in the picture
Belong to same human body.Specific how the method for tracing of similarity mode method belongs to the prior art, herein without excessively illustrating.
Alternatively it is also possible to human body is tracked in multiple image using optimal/minimum distance match method, it is any can be real
The mode of human body is tracked in present multiple image within the protection scope of present embodiment.
Compared with prior art, a kind of abnormal behaviour recognition methods is provided in embodiment of the present invention, obtaining includes people
The multiple image of body abnormal behaviour, is specifically included: being obtained containing to be detected the step of marking someone's body key point in multiple image
The multiple image of human body abnormal behaviour；Multiple image is inputted in critical point detection model, identifies each human body in multiple image
Key point, and each one body key point is marked in multiple image.Multiframe is detected automatically using critical point detection model in the program
Each human body key point in image, can quickly obtain the human body key point of each frame image in multiple image, and compared to
It is more comprehensive that the human body key point obtained for human body key point is manually marked in the picture.
The step of various methods divide above, be intended merely to describe it is clear, when realization can be merged into a step or
Certain steps are split, multiple steps are decomposed into, as long as including identical logical relation, all in the protection scope of this patent
It is interior；To adding inessential modification in algorithm or in process or introducing inessential design, but its algorithm is not changed
Core design with process is all in the protection scope of the patent.
Third embodiment of the invention is related to a kind of abnormal behaviour identification device, as shown in figure 3, including at least one processing
Device 301；And the memory 302 with the communication connection of at least one processor 301；Wherein, be stored with can be by extremely for memory 302
The instruction that a few processor 01 executes, instruction is executed by least one processor 301, so that at least one 301 energy of processor
Enough execute the abnormal behaviour recognition methods of any of the above-described embodiment.
Wherein, memory 302 is connected with processor 301 using bus mode, and bus may include any number of interconnection
Bus and bridge, bus is by one or more processors together with the various circuit connections of memory 302.Bus can also incite somebody to action
Together with various other circuit connections of management circuit or the like, these are all abilities for such as peripheral equipment, voltage-stablizer
Well known to domain, therefore, it will not be further described herein.Bus interface is provided between bus and transceiver and is connect
Mouthful.Transceiver can be an element, is also possible to multiple element, such as multiple receivers and transmitter, provides for passing
The unit communicated on defeated medium with various other devices.The data handled through processor are passed on the radio medium by antenna
Defeated, further, antenna also receives data and transfers data to processor 301.
Processor 301 is responsible for management bus and common processing, can also provide various functions, including timing, periphery connects
Mouthful, voltage adjusting, power management and other control functions.And memory 302 can be used for storage processor and execute behaviour
Used data when making.
Embodiments of the present invention additionally provide a kind of computer readable storage medium, are stored with computer program, described
The abnormal behaviour recognition methods of any of the above-described embodiment is realized when computer program is executed by processor.
That is, it will be understood by those skilled in the art that implement the method for the above embodiments be can be with
Relevant hardware is instructed to complete by program, which is stored in a storage medium, including some instructions are to make
It obtains an equipment (can be single-chip microcontroller, chip etc.) or processor (processor) executes side described in each embodiment of the application
The all or part of the steps of method.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.
It will be understood by those skilled in the art that the respective embodiments described above are to realize specific embodiments of the present invention,
And in practical applications, can to it, various changes can be made in the form and details, without departing from the spirit and scope of the present invention.
Claims (9)
1. a kind of abnormal behaviour recognition methods characterized by comprising
The multiple image comprising human body behavior is obtained, marks someone's body key point in the multiple image；
The multiple image is inputted in shot and long term Memory Neural Networks model, each human body key point in the multiple image is obtained
Weighted value；
According to the weighted value of obtained each human body key point and the weighted value of pre-set each human body key point with it is different
The corresponding relationship of Chang Hangwei determines the abnormal behaviour of human body in the multiple image.
2. abnormal behaviour recognition methods according to claim 1, which is characterized in that the LSTM network model passes through following
Step is trained:
Obtain the label power for being marked with each human body key point in the training set image and the training set image of human body key point
Weight values；
The training set image is inputted in the LSTM network model, predicts the weighted value of each human body key point；
According to the weighted value and the label weighted value that prediction obtains, the loss function of the LSTM network model is calculated
Value；
The parameter of the LSTM network model is adjusted according to the obtained loss function value, so that the loss function value meets
First preset condition.
3. abnormal behaviour recognition methods according to claim 1, which is characterized in that described obtain includes human body abnormal behaviour
Multiple image, the step of someone's body key point is marked in multiple image, specifically include:
Obtain the multiple image for containing human body abnormal behaviour to be detected；
The multiple image is inputted in critical point detection model, identifies each human body key point in the multiple image, and
Each human body key point is marked in the multiple image.
4. abnormal behaviour recognition methods according to claim 3, which is characterized in that in the identification multiple image
Each human body key point, specifically includes:
Human body is tracked in each frame image of the multiple image using similarity mode method；
Identify the key point of the human body tracked in each frame image.
5. abnormal behaviour recognition methods according to claim 3, which is characterized in that the critical point detection model by with
Lower step is trained:
Obtain the frame image set for being marked with the true key point of human body；
The frame image set is inputted in the critical point detection model, predicts the human body key point；
According to the human body key point of prediction and the true key point of the human body, the loss function of the prediction model is calculated
Value；
Make the loss function value full according to the parameter that the obtained loss function value adjusts the critical point detection model
The second preset condition of foot.
6. abnormal behaviour recognition methods according to claim 1, which is characterized in that described obtain includes the more of human body behavior
Frame image specifically:
Obtain the multiple image comprising single behavior or the multiple image comprising more people's behaviors.
7. abnormal behaviour recognition methods according to claim 1, which is characterized in that the human body key point includes at least:
The body joint point coordinate location information at the positions such as face, hand, foot, leg.
8. a kind of abnormal behaviour identification device characterized by comprising
At least one processor；And
The memory being connect at least one described processor communication；Wherein,
The memory is stored with the instruction that can be executed by least one described processor, and described instruction is by described at least one
It manages device to execute, so that at least one described processor is able to carry out the identification of the abnormal behaviour as described in any in claim 1 to 7
Method.
9. a kind of computer readable storage medium, is stored with computer program, which is characterized in that the computer program is processed
The abnormal behaviour recognition methods as described in any in claim 1 to 7 is realized when device executes.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910040080.6A CN109871764A (en) | 2019-01-16 | 2019-01-16 | A kind of abnormal behaviour recognition methods, device and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910040080.6A CN109871764A (en) | 2019-01-16 | 2019-01-16 | A kind of abnormal behaviour recognition methods, device and storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109871764A true CN109871764A (en) | 2019-06-11 |
Family
ID=66917711
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910040080.6A Pending CN109871764A (en) | 2019-01-16 | 2019-01-16 | A kind of abnormal behaviour recognition methods, device and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109871764A (en) |
Cited By (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110378515A (en) * | 2019-06-14 | 2019-10-25 | 平安科技（深圳）有限公司 | A kind of prediction technique of emergency event, device, storage medium and server |
| CN110472573A (en) * | 2019-08-14 | 2019-11-19 | 北京思图场景数据科技服务有限公司 | A kind of human body behavior analysis method, equipment and computer storage medium based on body key point |
| CN110493809A (en) * | 2019-08-16 | 2019-11-22 | 惠州Tcl移动通信有限公司 | Mobile terminal and its communication data method for detecting abnormality, computer-readable medium |
| CN111079536A (en) * | 2019-11-18 | 2020-04-28 | 高新兴科技集团股份有限公司 | Behavior analysis method based on human body key point time sequence, storage medium and equipment |
| CN111144368A (en) * | 2019-12-31 | 2020-05-12 | 重庆和贯科技有限公司 | Student behavior detection method based on long-time and short-time memory neural network |
| CN111464780A (en) * | 2020-03-12 | 2020-07-28 | 北京邮电大学 | Risk behavior early warning system, method and device based on edge cloud cooperation |
| CN111754669A (en) * | 2020-06-24 | 2020-10-09 | 桂林理工大学 | College student management system based on face recognition technology |
| CN112180357A (en) * | 2020-09-15 | 2021-01-05 | 珠海格力电器股份有限公司 | Safety protection method and system |
| CN112381068A (en) * | 2020-12-25 | 2021-02-19 | 四川长虹电器股份有限公司 | Method and system for detecting 'playing mobile phone' of person |
| CN113128383A (en) * | 2021-04-07 | 2021-07-16 | 杭州海宴科技有限公司 | Recognition method for campus student cheating behavior |
| CN114518801A (en) * | 2022-02-18 | 2022-05-20 | 美的集团(上海)有限公司 | Device control method, computer program product, control device, and storage medium |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106815639A (en) * | 2016-12-27 | 2017-06-09 | 东软集团股份有限公司 | The abnormal point detecting method and device of flow data |
| CN108038469A (en) * | 2017-12-27 | 2018-05-15 | 百度在线网络技术（北京）有限公司 | Method and apparatus for detecting human body |
| CN108205655A (en) * | 2017-11-07 | 2018-06-26 | 北京市商汤科技开发有限公司 | A kind of key point Forecasting Methodology, device, electronic equipment and storage medium |
| CN108229355A (en) * | 2017-12-22 | 2018-06-29 | 北京市商汤科技开发有限公司 | Activity recognition method and apparatus, electronic equipment, computer storage media, program |
-
2019
- 2019-01-16 CN CN201910040080.6A patent/CN109871764A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106815639A (en) * | 2016-12-27 | 2017-06-09 | 东软集团股份有限公司 | The abnormal point detecting method and device of flow data |
| CN108205655A (en) * | 2017-11-07 | 2018-06-26 | 北京市商汤科技开发有限公司 | A kind of key point Forecasting Methodology, device, electronic equipment and storage medium |
| CN108229355A (en) * | 2017-12-22 | 2018-06-29 | 北京市商汤科技开发有限公司 | Activity recognition method and apparatus, electronic equipment, computer storage media, program |
| CN108038469A (en) * | 2017-12-27 | 2018-05-15 | 百度在线网络技术（北京）有限公司 | Method and apparatus for detecting human body |
Non-Patent Citations (2)
| Title |
|---|
| 兰翠玲: "《"行为识别：让机器学会‘察言观色’第一步》", 《微软亚洲研究院》 * |
| 兰翠玲: "《行为识别：让机器学会‘察言观色’第一步》", 《微软亚洲研究院》 * |
Cited By (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110378515A (en) * | 2019-06-14 | 2019-10-25 | 平安科技（深圳）有限公司 | A kind of prediction technique of emergency event, device, storage medium and server |
| CN110472573A (en) * | 2019-08-14 | 2019-11-19 | 北京思图场景数据科技服务有限公司 | A kind of human body behavior analysis method, equipment and computer storage medium based on body key point |
| CN110493809B (en) * | 2019-08-16 | 2023-07-18 | 惠州Tcl移动通信有限公司 | Mobile terminal, communication data anomaly detection method thereof and computer readable medium |
| CN110493809A (en) * | 2019-08-16 | 2019-11-22 | 惠州Tcl移动通信有限公司 | Mobile terminal and its communication data method for detecting abnormality, computer-readable medium |
| CN111079536A (en) * | 2019-11-18 | 2020-04-28 | 高新兴科技集团股份有限公司 | Behavior analysis method based on human body key point time sequence, storage medium and equipment |
| CN111079536B (en) * | 2019-11-18 | 2023-08-29 | 高新兴科技集团股份有限公司 | Behavior analysis method, storage medium and device based on human body key point time sequence |
| CN111144368A (en) * | 2019-12-31 | 2020-05-12 | 重庆和贯科技有限公司 | Student behavior detection method based on long-time and short-time memory neural network |
| CN111464780A (en) * | 2020-03-12 | 2020-07-28 | 北京邮电大学 | Risk behavior early warning system, method and device based on edge cloud cooperation |
| CN111754669A (en) * | 2020-06-24 | 2020-10-09 | 桂林理工大学 | College student management system based on face recognition technology |
| CN112180357A (en) * | 2020-09-15 | 2021-01-05 | 珠海格力电器股份有限公司 | Safety protection method and system |
| CN112381068B (en) * | 2020-12-25 | 2022-05-31 | 四川长虹电器股份有限公司 | Method and system for detecting 'playing mobile phone' of person |
| CN112381068A (en) * | 2020-12-25 | 2021-02-19 | 四川长虹电器股份有限公司 | Method and system for detecting 'playing mobile phone' of person |
| CN113128383A (en) * | 2021-04-07 | 2021-07-16 | 杭州海宴科技有限公司 | Recognition method for campus student cheating behavior |
| CN114518801A (en) * | 2022-02-18 | 2022-05-20 | 美的集团(上海)有限公司 | Device control method, computer program product, control device, and storage medium |
| CN114518801B (en) * | 2022-02-18 | 2023-10-27 | 美的集团(上海)有限公司 | Device control method, control device, and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109871764A (en) | A kind of abnormal behaviour recognition methods, device and storage medium | |
| CN106412501B (en) | A kind of the construction safety behavior intelligent monitor system and its monitoring method of video | |
| CN110619657B (en) | Multi-camera linkage multi-target tracking method and system for intelligent communities | |
| CN109271961B (en) | Behavior pattern judgment method based on figure identity feature recognition | |
| CN109002801B (en) | Face shielding detection method and system based on video monitoring | |
| CN107808502B (en) | A kind of image detection alarm method and device | |
| CN108924333A (en) | Fraudulent call recognition methods, device and system | |
| CN104200466B (en) | A kind of method for early warning and video camera | |
| CN110414400B (en) | Automatic detection method and system for wearing of safety helmet on construction site | |
| CN109376639B (en) | Accompanying personnel early warning system and method based on portrait recognition | |
| CN109117721A (en) | A kind of pedestrian hovers detection method | |
| CN103942850A (en) | Medical staff on-duty monitoring method based on video analysis and RFID (radio frequency identification) technology | |
| CN109446936A (en) | A kind of personal identification method and device for monitoring scene | |
| CN110490126B (en) | Safe deposit box safety control system based on artificial intelligence | |
| CN112183317A (en) | Live working field violation behavior detection method based on space-time diagram convolutional neural network | |
| CN103164315A (en) | Computer using time prompting method and system based on intelligent video analysis | |
| CN102254394A (en) | Antitheft monitoring method for poles and towers in power transmission line based on video difference analysis | |
| CN110070155A (en) | A kind of comprehensive Activity recognition method and system of the inmate based on wearable device | |
| CN107016361A (en) | Recognition methods and device based on video analysis | |
| CN109615256A (en) | Wisdom garden security protection risk control method, storage medium, electronic equipment and system | |
| CN112491849B (en) | Power terminal vulnerability attack protection method based on flow characteristics | |
| CN114937237A (en) | Construction site safety monitoring method and system based on AI intelligent identification | |
| CN111914673A (en) | Target behavior detection method and device and computer readable storage medium | |
| CN106507037A (en) | Intelligent control method and system that personnel invade/cross the border | |
| CN111291728A (en) | Detection system, detection equipment and detection method for illegal crossing of transmission belt behavior |
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: 20190611 |