CN109829392A - Examination hall cheating recognition methods, system, computer equipment and storage medium - Google Patents
Examination hall cheating recognition methods, system, computer equipment and storage medium Download PDFInfo
- Publication number
- CN109829392A CN109829392A CN201910026905.9A CN201910026905A CN109829392A CN 109829392 A CN109829392 A CN 109829392A CN 201910026905 A CN201910026905 A CN 201910026905A CN 109829392 A CN109829392 A CN 109829392A
- Authority
- CN
- China
- Prior art keywords
- examinee
- video frame
- video
- sub
- cheating
- 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
- 238000000034 method Methods 0.000 title claims abstract description 35
- 230000033001 locomotion Effects 0.000 claims abstract description 81
- 230000009471 action Effects 0.000 claims abstract description 67
- 239000000284 extract Substances 0.000 claims abstract description 8
- 238000013527 convolutional neural network Methods 0.000 claims description 40
- 210000002569 neuron Anatomy 0.000 claims description 18
- 230000015654 memory Effects 0.000 claims description 11
- 238000012545 processing Methods 0.000 claims description 10
- 238000013528 artificial neural network Methods 0.000 claims description 9
- 230000000306 recurrent effect Effects 0.000 claims description 8
- 238000000605 extraction Methods 0.000 claims description 6
- 238000005070 sampling Methods 0.000 claims description 6
- 230000001537 neural effect Effects 0.000 claims description 5
- 238000012163 sequencing technique Methods 0.000 claims 1
- 238000011156 evaluation Methods 0.000 abstract description 4
- 238000013473 artificial intelligence Methods 0.000 abstract 1
- 238000012549 training Methods 0.000 description 13
- 230000006870 function Effects 0.000 description 8
- 238000010586 diagram Methods 0.000 description 5
- 238000012544 monitoring process Methods 0.000 description 4
- 239000004615 ingredient Substances 0.000 description 3
- 230000008901 benefit Effects 0.000 description 2
- 238000009432 framing Methods 0.000 description 2
- 230000008569 process Effects 0.000 description 2
- 230000004913 activation Effects 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 238000004590 computer program Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000001815 facial effect Effects 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 230000007774 longterm Effects 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000004321 preservation Methods 0.000 description 1
- 230000006403 short-term memory Effects 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
Landscapes
- Processing Or Creating Images (AREA)
Abstract
This application involves artificial intelligence fields, provide examination hall cheating recognition methods, system, computer equipment and storage medium, method includes acquisition video frame, sub-video frame corresponding with each examinee is marked off from video frame, obtain sub-video frame group, extract motion characteristic, determine cheating movement, setting cheating probability value is acted according to cheating, calculate the average value for probability value of practising fraud corresponding to first examinee in each video-frequency band, if the average value is greater than early warning value, determine that there are cheatings by first examinee.Action recognition is carried out to several subgraph groups of each examinee, obtains cheating probability value, probability value of practising fraud in multiple video-frequency bands to any examinee counts, and realizes to the accurate positionin of examinee position in picture and to the accurate evaluation of examinee's cheating movement.
Description
Technical field
This application involves image identification technical field, especially examination hall cheating recognition methods, system, computer equipment and deposit
Storage media.
Background technique
Video monitoring is intuitive with it, accurate, timely abundant with the information content and is widely used in many occasions.In recent years,
With computer, network and image procossing, the rapid development of transmission technology, there has also been significant progresses for Video Supervision Technique.
Nowadays, more and more examination halls are mounted with video monitoring system, it is intended to prevent examinee from practising fraud, carry out evidence preservation etc..
Current examination video monitoring is to be moved by the way of artificial observation camera shooting and video to the target of the examinee in video
It is monitored, this needs very high human cost, it is difficult to which realization monitors the video content of big magnitude in real time, and evidence
Trouble is transferred, recognition accuracy is low.
Summary of the invention
Based on this, it is necessary to for the problem that current examination video monitoring human cost is high, cheating recognition efficiency is low, provide
A kind of cheating recognition methods of examination hall, system, computer equipment and storage medium.
A kind of examination hall cheating recognition methods, comprising: multiple video-frequency bands are acquired from the monitor video of examination hall;From the first video-frequency band
It is middle to obtain multiple video frames;First video-frequency band is any video-frequency band in the multiple video-frequency band;Identify each view
The head portrait region of each examinee in frequency frame is marked off with each examinee one by one from each video frame according to the head portrait region of examinee
Corresponding sub-video frame;Each sub-video frame for belonging to identical examinee is in chronological sequence sequentially combined, multiple sub- views are obtained
Frequency frame group;Each sub-video frame group and each examinee correspond;Extraction movement is special from sub-video frame group corresponding to the first examinee
Sign;First examinee is any examinee in each examinee;The motion characteristic is input to RNN action recognition model;
The RNN action recognition model carries out movement knowledge to sub-video frame group corresponding to first examinee according to the motion characteristic
Not, if there is the movement with preset posture Data Matching in sub-video frame group corresponding to first examinee, described in judgement
There are cheating movements by first examinee, and are acted according to the cheating, and corresponding cheating probability value is arranged for first examinee;If
The movement with preset posture Data Matching is not present in sub-video frame group corresponding to first examinee, then first examinee
Corresponding cheating probability value is 0;Finally respectively obtain cheating probability corresponding to first examinee in each video-frequency band
Value;The average value for probability value of practising fraud corresponding to first examinee in each video-frequency band is calculated, if the average value is greater than in advance
Alert value, then determine that there are cheatings by first examinee.
In one embodiment, the head portrait region for identifying each examinee in each video frame, comprising: pass through first
Convolutional neural networks CNN model extracts the head portrait characteristic information in each video frame respectively;Believed according to the head portrait feature
Breath identifies the head portrait region of each examinee in each video frame;The head portrait characteristic information is believed including at least head feature
One in breath, face feature information and five features information.
In one embodiment, the first convolutional neural networks CNN model that passes through extracts each video frame respectively
In head portrait characteristic information, comprising: sliding sampling is carried out to the first video frame by the convolution kernel in the first CNN model,
Obtain several small images of the first block;First video frame is any video frame;By each small figure of first block
As being input in the pond layer of the first CNN model;By the pond layer of the first CNN model from each first block
The head portrait characteristic information is extracted in small image;Finally extract the head portrait characteristic information in each video frame.
In one embodiment, described to extract motion characteristic from sub-video frame group corresponding to the first examinee, comprising: to be
The second convolutional neural networks CNN model is arranged in sub-video frame group corresponding to first examinee;Pass through the 2nd CNN model
Convolution kernel sliding sampling is carried out to each sub-video frame in sub-video frame group corresponding to first examinee respectively, obtain
The small image of several blocks of each sub-video frame;Successively the small image of block of same area in each sub-video frame is connected in series,
The small image group of several blocks is obtained, each small image group of block is input in the pond layer of the 2nd CNN model;It is logical
The pond layer for crossing the 2nd CNN model extracts the motion characteristic from each small image group of block.
In one embodiment, cheating probability value corresponding to first examinee calculated in each video-frequency band is flat
Mean value, comprising: one accumulator is set for first examinee;The value that the accumulator is deposited is initialized as 0;By described tired
Device is added to calculate the summation of the cheating probability value of sub-video frame group corresponding to first examinee in each video-frequency band;It will be described total
Division arithmetic is done with the quantity with video-frequency band, obtains the average value.
In one embodiment, the algorithmic formula of the RNN action recognition model are as follows:
Wherein, I is the dimension of input vector, and V is the character of vectorization or the dimension of character portion, and H is the mind of hidden layer
Through first number, K is the neuron number of output layer, and x is the motion characteristic that the 2nd CNN model extraction comes out, and v is recurrent neural net
The vector data that network is melted into motion characteristic recognition result,For in moment RNN action recognition model hidden layer neuron it is defeated
Enter,For the output of hidden layer neuron in moment RNN action recognition model;To be exported in moment RNN action recognition model
The input of layer neuron;For the output of output layer neuron in moment RNN action recognition model,For probability value of practising fraud.
In one embodiment, the motion characteristic includes at least action message of turning one's head, touch turn information, stands up to act
Information, arms swing action message and with one in other people contact action information.
Based on the same technical idea, the application also provides a kind of examination hall cheating identifying system, comprising:
Transceiver module, for acquiring multiple video-frequency bands from the monitor video of examination hall.
Processing module, for obtaining multiple video frames from the first video-frequency band;First video-frequency band is the multiple view
Any video-frequency band in frequency range;The head portrait region for identifying each examinee in each video frame, according to the head portrait region of examinee from
It is marked off in each video frame and the one-to-one sub-video frame of each examinee;To belong to each sub-video frame of identical examinee by
Chronological order is combined, and obtains multiple sub-video frame groups;Each sub-video frame group and each examinee correspond;It is examined from first
Motion characteristic is extracted in raw corresponding sub-video frame group;First examinee is any examinee in each examinee;By institute
It states motion characteristic and is input to RNN action recognition model;The RNN action recognition model is according to the motion characteristic to described first
Sub-video frame group corresponding to examinee carry out action recognition, if in sub-video frame group corresponding to first examinee exist with it is pre-
If the matched movement of attitude data, then determine that there are cheating movements by first examinee, and acted according to the cheating, is described
Corresponding cheating probability value is arranged in first examinee;If being not present and default appearance in sub-video frame group corresponding to first examinee
The movement of state Data Matching, then cheating probability value corresponding to first examinee is 0;It finally respectively obtains in each video-frequency band
First examinee corresponding to cheating probability value;Calculate cheating probability corresponding to first examinee in each video-frequency band
The average value of value determines that there are cheatings by first examinee if the average value is greater than early warning value.
Based on the same technical idea, the application also provides a kind of computer equipment, including transceiver, memory and processing
Device is stored with computer-readable instruction in the memory, when the computer-readable instruction is executed by the processor, so that
The processor executes the step of above-mentioned examination hall cheating recognition methods.
Based on the same technical idea, the application also provides a kind of storage medium for being stored with computer-readable instruction, institute
When stating computer-readable instruction and being executed by one or more processors, so that one or more processors execute above-mentioned examination hall such as and make
The step of disadvantage recognition methods.
Above-mentioned examination hall cheating recognition methods, system, computer equipment and storage medium, by being carried out to the examinee in image
Head portrait identification, identifies the accurate location of each examinee, then marks off subgraph corresponding with examinee position in the picture, right
Several corresponding subgraph groups of each examinee carry out action recognition, corresponding cheating probability value are obtained, to any examinee more
Probability value of practising fraud in a video-frequency band is counted, and is realized to the accurate positionin of examinee position in picture and is practised fraud to examinee dynamic
The accurate evaluation of work, reduces the human cost of examination hall video surveillance management, and improves dynamic to examinee's cheating in monitor video
The recognition efficiency of work.
Detailed description of the invention
It is various other a little with benefit for this field by reading the detailed description of following detailed description in detail
Interior those of ordinary skill will become clear.Attached drawing is only used for showing the purpose of specific embodiment, and is not considered as this
The limitation of application.
Fig. 1 is a kind of flow chart of examination hall cheating recognition methods in the application one embodiment.
Fig. 2 is the method flow diagram that the head portrait region of each examinee in each video frame is identified in the application one embodiment.
Fig. 3 is to obtain the method flow diagram of sub-video frame group in the application one embodiment.
Fig. 4 is the method flow diagram of cheating probability value assembly average in the application one embodiment.
Fig. 5 is a kind of schematic diagram of examination hall cheating identifying system in the application one embodiment.
Fig. 6 is the structural schematic diagram of computer equipment in the embodiment of the present application.
Specific embodiment
It is with reference to the accompanying drawings and embodiments, right in order to which the objects, technical solutions and advantages of the application are more clearly understood
The application is further elaborated.It should be appreciated that specific embodiment described herein is only used to explain the application, and
It is not used in restriction the application.
Those skilled in the art of the present technique are appreciated that unless expressly stated, singular " one " used herein, " one
It is a ", " described " and "the" also may include plural form.It is to be further understood that used in the description of the present application
Wording " comprising " refers to that there are the feature, program, step, operation, element and/or component, but it is not excluded that in the presence of or add
Add other one or more features, program, step, operation, element, component and/or their group.
Fig. 1 is a kind of flow chart of examination hall cheating recognition methods in the application one embodiment, as shown in Figure 1, described examine
Field cheating recognition methods, may include steps of S1-S7:
Step S1, multiple video-frequency bands are acquired from the monitor video of examination hall, and multiple video frames are obtained from the first video-frequency band.
First video-frequency band is any video-frequency band in the multiple video-frequency band.
Video frame is the image frame of a frame.Examination hall monitor video can continuously be adopted with preset sample frequency
Sample obtains multiple video frames.Video frame carries the action message of period examination hall examinee, by respectively examining in multiple video frames
Raw action message is identified, so as to judge that examinee acts with the presence or absence of cheating.
Step S2, the head portrait region for identifying each examinee in each video frame, according to the head portrait region of examinee from each
It is marked off in the video frame and the one-to-one sub-video frame of each examinee.
Multiple examinees are had in each video frame, to judge that each examinee acts with the presence or absence of cheating, it is necessary to will respectively examine
Life distinguishes, and individually identifies to each examinee.Head portrait feature is to discriminate between the maximally efficient feature of each examinee, and the present embodiment is logical
Cross the head portrait region that head portrait identification determines each examinee, it is determined that the head portrait region of examinee has also determined that the position of examinee, then
Mark off the certain area size for belonging to each examinee in the video frame, i.e., with the one-to-one sub-video frame of each examinee.Antithetical phrase view
Movement in frequency frame region is identified that have identified corresponding examinee acts with the presence or absence of cheating.
As shown in Fig. 2, in one embodiment, the head portrait region of each examinee in each video frame of identification described in step S2
Include the following steps S211-S212:
Step S211, pass through the first convolutional neural networks (Convolutional Neural Networks, CNN) model
The head portrait characteristic information in each video frame is extracted respectively.
Head portrait characteristic information includes that head feature information, face feature information, five features information, head-mount product etc. hold
Easily pick out the relevant information of head portrait.Each head portrait characteristic information includes the information such as color, profile, relative position.
In one embodiment, step S211 includes the following steps S2111-S2112:
Step S2111, sliding sampling is carried out to the first video frame by the convolution kernel in the first CNN model, obtained
Several small images of the first block;Each small image of first block is input in the pond layer of the first CNN model.
First video frame is any video frame.
The convolution kernel of 3*3 width is set, and through convolution kernel since the initial position of video frame, the stride of convolution kernel is 1
A pixel, gradually traverses video requency frame data, runs convolution algorithm, is sampled to video frame.
Step S2112, described in being extracted from each small image of first block by the pond layer of the first CNN model
Head portrait characteristic information.
Convolution kernel extract video frame different characteristic, the feature extracted is sampled by pond layer, remove due to
There are spatial redundancy informations caused by stronger correlation between adjacent pixel inside image, obtain head portrait characteristic information.
Step S212, the head portrait region of each examinee in each video frame is identified according to the head portrait characteristic information.
Head portrait identification model is according to head portrait characteristic information, by related to head feature, facial characteristics, five features etc.
Information recognized, identify the head portrait of each examinee and its position in video frame.
Head portrait identification model is trained in advance.Specifically, acquisition head portrait identification in trained video is monitored from examination hall
Trained video frame will do each head portrait recognition training manually marked according to various head portrait characteristic informations and use video frame as just
Sample will not do each head portrait recognition training marked and use video frame as negative sample.By manually selecting with head portrait spy
The head portrait recognition training video frame of reference breath, is labeled as positive sample;It will be without the head portrait recognition training of head portrait characteristic information
Negative sample is labeled as with video frame.First CNN model samples head portrait recognition training with video frame, and sample information is given
To head portrait identification model.Head portrait identification model carries out recognition training to each head portrait characteristic information, is modified and is remembered each according to mark
The weighted value of head portrait characteristic information, to realize the identification to head portrait.
Behind the head portrait region for identifying each examinee, similarity knowledge is carried out to the image in preset range around head portrait region
Not, the relevant picture of each examinee is spliced according to similarity, forms the corresponding sub-video frame of each examinee.
Since the head portrait of examinee in video is likely to be shaking, in different video frames, same examinee is drawn
The position of the sub-video frame separated is possible to different.For example, the head portrait of a examinee appears in the position b1, t2 frame figure in t1 frame image
As in, the head portrait of a examinee appears in the position b2, according to the head portrait regional assignment of a examinee sub- view corresponding with its band of position
Frequency frame obtained the movement string picture of the examinee by arranging sub-video frame each before and after a examinee, i.e. a examinee
Sub-video frame group.
Step S3, each sub-video frame for belonging to identical examinee is in chronological sequence sequentially combined, obtains multiple sub- views
Frequency frame group.
Each sub-video frame group and each examinee correspond.
There is examinee corresponding with a same examinee band of position in each width video frame, will belong to same examinee's
Band of position image zooming-out comes out, and sequentially combines again, forms sub-video frame group corresponding with the examinee, i.e. a sub-video
Frame group corresponds to a corresponding examinee, the motion characteristic with corresponding examinee in each sub-video frame group.
As shown in figure 3, in one embodiment, step S3 includes the following steps S31-S32:
Step S31: the identical mark of sub-video frame flag of identical examinee will be belonged in each video frame.
Classified according to examinee to the sub-video frame in video frame, it is identical to the sub-video frame flag for belonging to same examinee
Unique identification, to distinguish each sub-video frame.
Step S32: mark sub-video frame having the same in each video frame is in chronological sequence sequentially combined, is obtained
To sub-video frame group.
Step S4, motion characteristic is extracted from sub-video frame group corresponding to the first examinee.
First examinee is any examinee in each examinee.
One sub- video frame group contains complete movement string information of the examinee within the period, i.e. motion characteristic, from
The movement string information of corresponding examinee is extracted in sub-video frame group.
Motion characteristic include turn one's head action message, touch turn information, action message of standing up, arms swing action message,
With the recognizable relevant information that whether examinee practises fraud out such as other people contact action information.
In one embodiment, step S4 includes the following steps S41-S45:
Step S41, the second convolutional neural networks CNN model is set for sub-video frame group corresponding to first examinee.
Step S42, by the convolution kernel of the 2nd CNN model respectively to sub-video frame corresponding to first examinee
Each sub-video frame in group carries out sliding sampling, obtains the small image of several blocks of each sub-video frame.
Step S43, successively the small image of block of same area in each sub-video frame is connected in series, obtains several blocks
Each small image group of block is input in the pond layer of the 2nd CNN model by small image group.
Step S44, the movement is extracted from each small image group of block by the pond layer of the 2nd CNN model
Feature.
RNN action recognition model is trained in advance.Acquisition action recognition training in trained video is monitored from examination hall
With sub-video frame group;The action recognition training is made of with sub-video frame group continuous multiple sub-video frames.By artificial basis
Various motion characteristics are labeled with each action recognition training and use sub-video frame group as positive sample, do not mark each action recognition training
Use sub-video frame group as negative sample.By manually selecting the action recognition training sub-video frame group with cheating movement,
It is labeled as positive sample;The action recognition acted without cheating training is labeled as negative sample with sub-video frame group.It and is each work
Corresponding cheating probability value is arranged in disadvantage movement, and the cheating probability value of regular event is set as 0;Known by the 2nd CNN model from movement
The static motion characteristic of extraction in sub-video frame group Xun Lian not be used, and the motion characteristic of extraction is given to RNN action recognition model.
RNN action recognition model carries out recognition training to each motion characteristic, and the weighted value of each motion characteristic is modified and remembered according to mark,
To realize the identification acted to examinee;In turn, by the identification to various movements, the corresponding cheating probability of each movement is identified
Value.
Length memory network (LSTM, Long Short-Term Memory) is one kind of Recognition with Recurrent Neural Network (RNN), energy
Enough solve the problems, such as that long-term " memory " in conventional recycle neural network is unserviceable, the ability with distance study.
LSTM has a succession of form for repeating neural network module, and replicated blocks have different structures.It has four layers
Neural net layer interacts in a particular manner.Horizontal line representative unit state, linear interaction, it is ensured that information
It transmits down.Selectively information is allowed to pass through, is made of sigmoid neural net layer and point-by-point multiplying.
Sigmoid layers by variable mappings between 0 and 1, describing whether each ingredient should pass through thresholding.sigmoid
Layer, sigmoid are that neural network algorithm selects sigmoid function as activation primitive, are used to describe neural network,
Sigmoid executes operation to each input data, using sigmoid function, also can choose hyperbolic tangent function (tanh).In advance
If 0 represents " any ingredient is not allowed to pass through ", and 1 represents " ingredient is allowed to pass through ".There are three types of similar thresholdings by LSTM, determine respectively
Which fixed information needs from " the forgeing thresholding sigmoid layers " given up in location mode, determines to need to store in location mode
" the input threshold layer " of which new information and value is added to " tanh layers " of state, which part needs in determining means state
The sigmoid layer and tanh function of output.
Step S5, the motion characteristic is input to Recognition with Recurrent Neural Network (Recurrent Neural Network, RNN)
Action recognition model;The RNN action recognition model is according to the motion characteristic to sub-video corresponding to first examinee
Frame group carries out action recognition, if existing in sub-video frame group corresponding to first examinee dynamic with preset posture Data Matching
Make, then determine that there are cheating movements by first examinee, and acted according to the cheating, is arranged for first examinee corresponding
Cheating probability value;If the movement with preset posture Data Matching is not present in sub-video frame group corresponding to first examinee,
Then cheating probability value corresponding to first examinee is 0.Finally respectively obtain the first examinee institute in each video-frequency band
Corresponding cheating probability value.
In one embodiment, the algorithmic formula of RNN action recognition model is as follows:
Wherein, I is the dimension of input vector, and V is the character of vectorization or the dimension of character portion, and H is the mind of hidden layer
Through first number, K is the neuron number of output layer, and x is the motion characteristic that the 2nd CNN model extraction comes out, and v is recurrent neural net
The vector data that network is melted into motion characteristic recognition result,For in moment RNN action recognition model hidden layer neuron it is defeated
Enter,For the output of hidden layer neuron in moment RNN action recognition model;To be exported in moment RNN action recognition model
The input of layer neuron;For the output of output layer neuron in moment RNN action recognition model,For probability value of practising fraud, body
Having showed the moment corresponds to ratio of the neuron output value relative to the adduction of all neuron output values of output layer.
A possibility that same movement of examinee, amplitude of fluctuation is different, then practises fraud are also different.For example, be all movement of turning one's head,
It is left and right to wave within the scope of 45 °, it is not necessarily cheating, when left and right swing angle is greater than 45 °, then has very big cheating can
It can property.Therefore, a threshold value is preset to each target action by RNN action recognition model first, it is dynamic for dividing the target
As regular event or cheating movement.When target action is more than preset threshold, then it is determined as movement of practising fraud, otherwise it is assumed that
It is regular event.
There are many kinds of the suspicious action meetings of cheating, such as: it makes a gesture;It stands, movement of squatting down;It turns one's head, touch turn;Frequently
It bows movement;Arm substantially wobbling action etc..These biggish movements of cheating probability can be preset as target action by us.No
Same target action, cheating probability may also be different;Same target action is because of the factors such as its movement range difference, probability of practising fraud
It may also be different.For this purpose, being that corresponding probability value is arranged in various target actions previously according to experience.In this way, to various target actions
It is finely divided, improves the accuracy and science of cheating action recognition.
Sub-video frame group be containing same examinee act string information image, compared to only by video frame act into
Row identification, can more by identifying to a series of continuous action of examinee come for confirming method that whether examinee practises fraud
Whether the movement for accurately judging examinee is really cheating.
Step S6, the average value for calculating probability value of practising fraud corresponding to first examinee in each video-frequency band, if described
Average value is greater than early warning value, then determines that there are cheatings by first examinee.
It may will recognise that multiple cheating movements in one sub- video frame group, will also correspond to multiple cheating probability values, example
It such as turns one's head and makes a gesture simultaneously.At this moment, maximum cheating probability value can be selected from multiple cheating probability values to be counted;?
The combination settings one larger cheating probability value that multiple cheatings can be acted, the statistics for probability value of practising fraud.Every height view
The cheating probability value that frequency frame group identifies is possible to different, therefore, within a preset time, it is possible to need multiple groups sub-video frame group
After the cheating probability value superposition identified, early warning value just can exceed that, it is also possible to which the cheating that a framing subgraph group identifies is general
Rate value will be more than early warning value.Obviously, in the case that the cheating probability value that a framing subgraph group identifies is more than early warning value, table
Show that the cheating movement of the examinee is especially big.
As shown in figure 4, probability value of practising fraud in one embodiment, in step S6 carries out assembly average and includes the following steps
S61-S63:
Step S61, one accumulator is set for first examinee.
Step S62, the value that the accumulator is deposited is initialized as 0.
Step S63, sub-video frame group corresponding to first examinee in each video-frequency band is calculated by the accumulator
Cheating probability value summation.
Step S64, the quantity of the summation and video-frequency band is done into division arithmetic, obtains the average value.
The superposition result of accumulator within a preset time is compared with early warning value;If superposition result is greater than
Early warning value then determines that the first examinee practises fraud.
Step S7, if it is determined that first examinee practises fraud, then enter the early warning stage, carry out warning note, and to described pre-
If the video in the time carries out storage processing.Early warning of practising fraud further includes adjusting to the direction of specific camera head in examination hall
It is whole, and camera is focused on to the first examinee of cheating, with the movement of the sharp examinee for preferably observing and storing cheating.
Above-described embodiment identifies the accurate location of each examinee, then by carrying out head portrait identification to the examinee in image
Subgraph corresponding with examinee position is marked off in the picture, several subgraph groups corresponding to each examinee act
Identification, obtains cheating probability value, and probability value of practising fraud in multiple video-frequency bands to any examinee is counted, realized in picture
The accurate positionin of examinee position and the accurate evaluation that examinee's cheating is acted.
Based on the same technical idea, present invention also provides a kind of examination hall cheating identifying systems, as shown in figure 5, this is
System includes including transceiver module 1 and processing module 2.The processing module 2 is used to control the transmitting-receiving operation of the transceiver module 1.
The transceiver module 1, for acquiring multiple video-frequency bands from the monitor video of examination hall.
The processing module 2, for obtaining multiple video frames from the first video-frequency band;First video-frequency band is described more
Any video-frequency band in a video-frequency band;The head portrait region for identifying each examinee in each video frame, according to the head portrait area of examinee
Domain marks off and the one-to-one sub-video frame of each examinee from each video frame;Each sub-video of identical examinee will be belonged to
Frame is in chronological sequence sequentially combined, and obtains multiple sub-video frame groups;Each sub-video frame group and each examinee correspond;From
Motion characteristic is extracted in sub-video frame group corresponding to one examinee;First examinee is any examinee in each examinee;
The motion characteristic is input to RNN action recognition model;The RNN action recognition model is according to the motion characteristic to described
Sub-video frame group corresponding to first examinee carries out action recognition, if existing in sub-video frame group corresponding to first examinee
With the movement of preset posture Data Matching, then determine that there are cheating movements by first examinee, and acted according to the cheating, is
Corresponding cheating probability value is arranged in first examinee;If in sub-video frame group corresponding to first examinee there is no with it is pre-
If the matched movement of attitude data, then cheating probability value corresponding to first examinee is 0;Finally respectively obtain each video
Cheating probability value corresponding to first examinee in section;Calculate cheating corresponding to first examinee in each video-frequency band
The average value of probability value determines that there are cheatings by first examinee if the average value is greater than early warning value.
Above-described embodiment identifies the accurate location of each examinee, then by carrying out head portrait identification to the examinee in image
Subgraph corresponding with examinee position is marked off in the picture, several subgraph groups corresponding to each examinee act
Identification, obtains cheating probability value, and probability value of practising fraud in multiple video-frequency bands to any examinee is counted, realized in picture
The accurate positionin of examinee position and the accurate evaluation that examinee's cheating is acted.
Based on the same technical idea, the application also proposed a kind of computer equipment, as shown in fig. 6, the computer
Equipment includes transceiver 901, processor 902 and memory 903, is stored with computer-readable instruction in the memory 903, institute
When stating computer-readable instruction and being executed by the processor 902, so that the processor executes the institute in the respective embodiments described above
The step of examination hall cheating recognition methods stated.
The corresponding entity device of transceiver module 1 shown in Fig. 5 is transceiver 901 shown in fig. 6,901 energy of transceiver
It enough realizes all or part of function of transceiver module 1, or realizes and the same or similar function of transceiver module 1.
The corresponding entity device of processing module 2 shown in Fig. 5 is processor 902 shown in fig. 6,902 energy of processor
It enough realizes all or part of function of processing module 2, or realizes and the same or similar function of processing module 2.
Based on the same technical idea, present invention also provides a kind of storage medium for being stored with computer-readable instruction,
When the computer-readable instruction is executed by one or more processors, so that one or more processors execute above-mentioned each implementation
Example in the examination hall practise fraud recognition methods the step of.
Those of ordinary skill in the art will appreciate that realizing all or part of the process in above-described embodiment method, being can be with
Relevant hardware is instructed to complete by computer program, which can be stored in a computer-readable storage and be situated between
In matter, the program is when being executed, it may include such as the process of the embodiment of above-mentioned each method.Wherein, storage medium above-mentioned can be
The non-volatile memory mediums such as magnetic disk, CD, read-only memory (Read-Only Memory, ROM) or random storage note
Recall body (Random Access Memory, RAM) etc..
Each technical characteristic of embodiment described above can be combined arbitrarily, for simplicity of description, not to above-mentioned reality
It applies all possible combination of each technical characteristic in example to be all described, as long as however, the combination of these technical characteristics is not deposited
In contradiction, all should be considered as described in this specification.
The several embodiments of the application above described embodiment only expresses, the description thereof is more specific and detailed, but simultaneously
The limitation to the application the scope of the patents therefore cannot be interpreted as.It should be pointed out that for those of ordinary skill in the art
For, without departing from the concept of this application, various modifications and improvements can be made, these belong to the guarantor of the application
Protect range.Therefore, the scope of protection shall be subject to the appended claims for the application patent.
Claims (10)
- The recognition methods 1. a kind of examination hall is practised fraud characterized by comprisingMultiple video-frequency bands are acquired from the monitor video of examination hall;Multiple video frames are obtained from the first video-frequency band;First video-frequency band is any video in the multiple video-frequency band Section;The head portrait region for identifying each examinee in each video frame, according to the head portrait region of examinee from each video frame It marks off and the one-to-one sub-video frame of each examinee;Each sub-video frame for belonging to identical examinee is in chronological sequence sequentially combined, multiple sub-video frame groups are obtained;Each son Video frame group and each examinee correspond;Motion characteristic is extracted from sub-video frame group corresponding to the first examinee;First examinee is appointing in each examinee One examinee;The motion characteristic is input to Recognition with Recurrent Neural Network RNN action recognition model;The RNN action recognition model is according to institute It states motion characteristic and action recognition is carried out to sub-video frame group corresponding to first examinee, if corresponding to first examinee There is movement with preset posture Data Matching in sub-video frame group, then determine first examinee there are cheating movement, and root It is acted according to the cheating, corresponding cheating probability value is set for first examinee;If son corresponding to first examinee regards There is no the movements with preset posture Data Matching in frequency frame group, then cheating probability value corresponding to first examinee is 0;Most Cheating probability value corresponding to first examinee in each video-frequency band is respectively obtained eventually;The average value for probability value of practising fraud corresponding to first examinee in each video-frequency band is calculated, if the average value is greater than in advance Alert value, then determine that there are cheatings by first examinee.
- The recognition methods 2. examination hall according to claim 1 is practised fraud, which is characterized in thatThe head portrait region of each examinee in each video frame of identification, comprising:Extract the head portrait characteristic information in each video frame respectively by the first convolutional neural networks CNN model;The head portrait region of each examinee in each video frame is identified according to the head portrait characteristic information;The head portrait characteristic information includes at least one in head feature information, face feature information and five features information.
- The recognition methods 3. examination hall according to claim 2 is practised fraud, which is characterized in thatIt is described that head portrait characteristic information in each video frame is extracted by the first convolutional neural networks CNN model respectively, packet It includes:Sliding sampling is carried out to the first video frame by the convolution kernel in the first CNN model, obtains several the first blocks Small image;First video frame is any video frame;Each small image of first block is input to the first CNN In the pond layer of model;The head portrait characteristic information is extracted from each small image of first block by the pond layer of the first CNN model;Finally extract the head portrait characteristic information in each video frame.
- The recognition methods 4. examination hall according to claim 1 is practised fraud, which is characterized in thatIt is described to extract motion characteristic from sub-video frame group corresponding to the first examinee, comprising:For sub-video frame group corresponding to first examinee, the second convolutional neural networks CNN model is set;By the convolution kernel of the 2nd CNN model respectively to every height in sub-video frame group corresponding to first examinee Video frame carries out sliding sampling, obtains the small image of several blocks of each sub-video frame;Successively the small image of block of same area in each sub-video frame is connected in series, obtains the small image group of several blocks, it will Each small image group of the block is input in the pond layer of the 2nd CNN model;The motion characteristic is extracted from each small image group of block by the pond layer of the 2nd CNN model.
- The recognition methods 5. examination hall according to claim 1 is practised fraud, which is characterized in thatThe average value of cheating probability value corresponding to first examinee calculated in each video-frequency band, comprising:For first examinee, one accumulator is set;The value that the accumulator is deposited is initialized as 0;The cheating probability value of sub-video frame group corresponding to first examinee in each video-frequency band is calculated by the accumulator Summation;The quantity of the summation and video-frequency band is done into division arithmetic, obtains the average value.
- The recognition methods 6. examination hall according to claim 1 is practised fraud, which is characterized in thatThe algorithmic formula of the RNN action recognition model are as follows:Wherein, I is the dimension of input vector, and V is the character of vectorization or the dimension of character portion, and H is the neuron of hidden layer Number, K are the neuron number of output layer, and x is the motion characteristic that the 2nd CNN model extraction comes out, and v is recurrent neural net The vector data that network is melted into the motion characteristic recognition result,For hidden layer neuron in moment RNN action recognition model Input,For the output of hidden layer neuron in moment RNN action recognition model;For in moment RNN action recognition model The input of output layer neuron;For the output of output layer neuron in moment RNN action recognition model,For probability of practising fraud Value.
- The recognition methods 7. examination hall according to claim 1 is practised fraud, which is characterized in thatThe motion characteristic includes at least action message of turning one's head, touch turn information, action message of standing up, arms swing movement letter Breath and with one in other people contact action information.
- The identifying system 8. a kind of examination hall is practised fraud characterized by comprisingTransceiver module, for acquiring multiple video-frequency bands from the monitor video of examination hall;Processing module, for obtaining multiple video frames from the first video-frequency band;First video-frequency band is the multiple video-frequency band In any video-frequency band;The head portrait region for identifying each examinee in each video frame, according to the head portrait region of examinee from each It is marked off in the video frame and the one-to-one sub-video frame of each examinee;Each sub-video frame of identical examinee will be belonged to temporally Sequencing is combined, and obtains multiple sub-video frame groups;Each sub-video frame group and each examinee correspond;From the first examinee institute Motion characteristic is extracted in corresponding sub-video frame group;First examinee is any examinee in each examinee;It will be described dynamic Recognition with Recurrent Neural Network RNN action recognition model is input to as feature;The RNN action recognition model is according to the motion characteristic pair Sub-video frame group corresponding to first examinee carries out action recognition, if in sub-video frame group corresponding to first examinee In the presence of the movement with preset posture Data Matching, then determine that there are cheating movements by first examinee, and dynamic according to the cheating Make, corresponding cheating probability value is set for first examinee;If not deposited in sub-video frame group corresponding to first examinee In the movement with preset posture Data Matching, then cheating probability value corresponding to first examinee is 0;It finally respectively obtains every Cheating probability value corresponding to first examinee in a video-frequency band;It calculates corresponding to first examinee in each video-frequency band Cheating probability value average value, if the average value be greater than early warning value, determine that there are cheatings by first examinee.
- 9. a kind of computer equipment, which is characterized in that including transceiver, memory and processor, be stored in the memory Computer-readable instruction, when the computer-readable instruction is executed by the processor, so that the processor executes such as right It is required that the step in 1 to 7 in any examination hall cheating recognition methods.
- 10. a kind of storage medium for being stored with computer-readable instruction, which is characterized in that the computer-readable instruction is by one Or multiple processors are when executing, so that one or more processors execute the examination hall as described in any in claim 1 to 7 and practise fraud Step in recognition methods.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910026905.9A CN109829392A (en) | 2019-01-11 | 2019-01-11 | Examination hall cheating recognition methods, system, computer equipment and storage medium |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910026905.9A CN109829392A (en) | 2019-01-11 | 2019-01-11 | Examination hall cheating recognition methods, system, computer equipment and storage medium |
Publications (1)
Publication Number | Publication Date |
---|---|
CN109829392A true CN109829392A (en) | 2019-05-31 |
Family
ID=66860894
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910026905.9A Pending CN109829392A (en) | 2019-01-11 | 2019-01-11 | Examination hall cheating recognition methods, system, computer equipment and storage medium |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109829392A (en) |
Cited By (13)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110290354A (en) * | 2019-07-10 | 2019-09-27 | 十里泉科技(重庆)有限公司 | A kind of with no paper examination hall examination monitoring data management method |
CN110852247A (en) * | 2019-11-07 | 2020-02-28 | 北京云迹科技有限公司 | Abnormality detection method, abnormality detection device, electronic apparatus, and computer-readable storage medium |
CN111242819A (en) * | 2020-01-20 | 2020-06-05 | 重庆强大锐智科技服务有限公司 | Online learning examination system |
CN111241985A (en) * | 2020-01-08 | 2020-06-05 | 腾讯科技(深圳)有限公司 | Video content identification method and device, storage medium and electronic equipment |
CN111249742A (en) * | 2020-01-21 | 2020-06-09 | 腾讯科技(深圳)有限公司 | Cheating user detection method and device, storage medium and electronic equipment |
CN111553203A (en) * | 2020-04-10 | 2020-08-18 | 北京明略软件***有限公司 | Method and device for monitoring examination fraud behaviors |
CN111611865A (en) * | 2020-04-23 | 2020-09-01 | 平安国际智慧城市科技股份有限公司 | Examination cheating behavior identification method, electronic equipment and storage medium |
CN112380951A (en) * | 2020-11-10 | 2021-02-19 | 浙江大华技术股份有限公司 | Method and device for identifying abnormal behavior, computer equipment and storage medium |
CN113065490A (en) * | 2021-04-12 | 2021-07-02 | 合肥越明交通电子科技有限公司 | Detection method and device for intelligent driving test |
CN113095968A (en) * | 2019-12-23 | 2021-07-09 | 深圳云天励飞技术有限公司 | Cheating identification method and device and electronic equipment |
CN113392920A (en) * | 2021-06-25 | 2021-09-14 | 北京百度网讯科技有限公司 | Method, apparatus, device, medium, and program product for generating cheating prediction model |
CN114140751A (en) * | 2021-12-13 | 2022-03-04 | 江苏商贸职业学院 | Examination room monitoring method and system |
CN115147872A (en) * | 2022-08-30 | 2022-10-04 | 北京竞业达数字***科技有限公司 | Method and system for detecting teacher behaviors in examination room |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105654130A (en) * | 2015-12-30 | 2016-06-08 | 成都数联铭品科技有限公司 | Recurrent neural network-based complex image character sequence recognition system |
CN107491717A (en) * | 2016-06-13 | 2017-12-19 | 科大讯飞股份有限公司 | The detection method that cheats at one's exam and device |
CN108985259A (en) * | 2018-08-03 | 2018-12-11 | 百度在线网络技术(北京)有限公司 | Human motion recognition method and device |
-
2019
- 2019-01-11 CN CN201910026905.9A patent/CN109829392A/en active Pending
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105654130A (en) * | 2015-12-30 | 2016-06-08 | 成都数联铭品科技有限公司 | Recurrent neural network-based complex image character sequence recognition system |
CN107491717A (en) * | 2016-06-13 | 2017-12-19 | 科大讯飞股份有限公司 | The detection method that cheats at one's exam and device |
CN108985259A (en) * | 2018-08-03 | 2018-12-11 | 百度在线网络技术(北京)有限公司 | Human motion recognition method and device |
Non-Patent Citations (1)
Title |
---|
李慧: "基于视频的考场异常行为识别研究", 《中国优秀博硕士学位论文全文数据库(硕士)信息科技辑》 * |
Cited By (18)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110290354A (en) * | 2019-07-10 | 2019-09-27 | 十里泉科技(重庆)有限公司 | A kind of with no paper examination hall examination monitoring data management method |
CN110852247A (en) * | 2019-11-07 | 2020-02-28 | 北京云迹科技有限公司 | Abnormality detection method, abnormality detection device, electronic apparatus, and computer-readable storage medium |
CN113095968A (en) * | 2019-12-23 | 2021-07-09 | 深圳云天励飞技术有限公司 | Cheating identification method and device and electronic equipment |
CN111241985A (en) * | 2020-01-08 | 2020-06-05 | 腾讯科技(深圳)有限公司 | Video content identification method and device, storage medium and electronic equipment |
US11983926B2 (en) | 2020-01-08 | 2024-05-14 | Tencent Technology (Shenzhen) Company Limited | Video content recognition method and apparatus, storage medium, and computer device |
CN111242819A (en) * | 2020-01-20 | 2020-06-05 | 重庆强大锐智科技服务有限公司 | Online learning examination system |
CN111242819B (en) * | 2020-01-20 | 2023-09-19 | 重庆强大锐智科技服务有限公司 | Online learning examination system |
CN111249742B (en) * | 2020-01-21 | 2022-05-17 | 腾讯科技(深圳)有限公司 | Cheating user detection method and device, storage medium and electronic equipment |
CN111249742A (en) * | 2020-01-21 | 2020-06-09 | 腾讯科技(深圳)有限公司 | Cheating user detection method and device, storage medium and electronic equipment |
CN111553203A (en) * | 2020-04-10 | 2020-08-18 | 北京明略软件***有限公司 | Method and device for monitoring examination fraud behaviors |
CN111611865A (en) * | 2020-04-23 | 2020-09-01 | 平安国际智慧城市科技股份有限公司 | Examination cheating behavior identification method, electronic equipment and storage medium |
CN112380951A (en) * | 2020-11-10 | 2021-02-19 | 浙江大华技术股份有限公司 | Method and device for identifying abnormal behavior, computer equipment and storage medium |
CN112380951B (en) * | 2020-11-10 | 2023-04-07 | 浙江大华技术股份有限公司 | Method and device for identifying abnormal behavior, computer equipment and storage medium |
CN113065490A (en) * | 2021-04-12 | 2021-07-02 | 合肥越明交通电子科技有限公司 | Detection method and device for intelligent driving test |
CN113392920A (en) * | 2021-06-25 | 2021-09-14 | 北京百度网讯科技有限公司 | Method, apparatus, device, medium, and program product for generating cheating prediction model |
CN114140751A (en) * | 2021-12-13 | 2022-03-04 | 江苏商贸职业学院 | Examination room monitoring method and system |
CN114140751B (en) * | 2021-12-13 | 2024-02-09 | 江苏商贸职业学院 | Examination room monitoring method and system |
CN115147872A (en) * | 2022-08-30 | 2022-10-04 | 北京竞业达数字***科技有限公司 | Method and system for detecting teacher behaviors in examination room |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109829392A (en) | Examination hall cheating recognition methods, system, computer equipment and storage medium | |
Fan et al. | Video anomaly detection and localization via gaussian mixture fully convolutional variational autoencoder | |
Lv et al. | Simultaneously localize, segment and rank the camouflaged objects | |
Lu et al. | Deep learning for fall detection: Three-dimensional CNN combined with LSTM on video kinematic data | |
CN110837784B (en) | Examination room peeping and cheating detection system based on human head characteristics | |
Oliva et al. | Top-down control of visual attention in object detection | |
CN107924472B (en) | Image classification method and system based on brain computer interface | |
CN108062574B (en) | Weak supervision target detection method based on specific category space constraint | |
CN109858393A (en) | Rule-breaking vehicle recognition methods, system, computer equipment and storage medium | |
Joshi et al. | Robust sports image classification using InceptionV3 and neural networks | |
CN103517042B (en) | A kind of nursing house old man's hazardous act monitoring method | |
CN106022220A (en) | Method for performing multi-face tracking on participating athletes in sports video | |
CN109886241A (en) | Driver fatigue detection based on shot and long term memory network | |
Gatt et al. | Detecting human abnormal behaviour through a video generated model | |
David | An intellectual individual performance abnormality discovery system in civic surroundings | |
Omarov et al. | A Skeleton-based Approach for Campus Violence Detection. | |
CN109918996A (en) | The illegal action identification method of personnel, system, computer equipment and storage medium | |
EP4123592A3 (en) | Human-object interaction detection method, neural network and training method therefor, device, and medium | |
Tomas et al. | Human activity recognition using combined deep architectures | |
CN109086717B (en) | Violent behavior detection system and method based on human skeleton and motion signal characteristics | |
Miao et al. | Abnormal Behavior Learning Based on Edge Computing toward a Crowd Monitoring System | |
Khan et al. | Comparative study of various crowd detection and classification methods for safety control system | |
Zhang et al. | Key frame extraction based on quaternion Fourier transform with multiple features fusion | |
CN113822198A (en) | Peanut growth monitoring method, system and medium based on UAV-RGB image and deep learning | |
CN112487926A (en) | Scenic spot feeding behavior identification method based on space-time diagram convolutional network |
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: 20190531 |