CN109740522A - A kind of personnel's detection method, device, equipment and medium - Google Patents
A kind of personnel's detection method, device, equipment and medium Download PDFInfo
- Publication number
- CN109740522A CN109740522A CN201811645930.7A CN201811645930A CN109740522A CN 109740522 A CN109740522 A CN 109740522A CN 201811645930 A CN201811645930 A CN 201811645930A CN 109740522 A CN109740522 A CN 109740522A
- Authority
- CN
- China
- Prior art keywords
- radiofrequency signal
- thermal map
- personnel
- training
- sample
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000001514 detection method Methods 0.000 title claims abstract description 60
- 238000012549 training Methods 0.000 claims abstract description 94
- 238000013527 convolutional neural network Methods 0.000 claims abstract description 45
- 238000000034 method Methods 0.000 claims abstract description 35
- 238000012360 testing method Methods 0.000 claims abstract description 22
- 238000010586 diagram Methods 0.000 claims description 44
- 210000000988 bone and bone Anatomy 0.000 claims description 9
- 238000004590 computer program Methods 0.000 claims description 9
- 230000006399 behavior Effects 0.000 claims description 6
- 238000007689 inspection Methods 0.000 claims description 4
- 230000001537 neural effect Effects 0.000 claims description 4
- 238000012937 correction Methods 0.000 claims description 3
- 230000009286 beneficial effect Effects 0.000 abstract description 2
- 230000000875 corresponding effect Effects 0.000 description 16
- 230000002596 correlated effect Effects 0.000 description 7
- 230000008569 process Effects 0.000 description 7
- 238000000605 extraction Methods 0.000 description 6
- 238000004458 analytical method Methods 0.000 description 5
- 238000004364 calculation method Methods 0.000 description 2
- 230000006872 improvement Effects 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 238000012544 monitoring process Methods 0.000 description 2
- 230000000149 penetrating effect Effects 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 230000000007 visual effect Effects 0.000 description 2
- 230000004913 activation Effects 0.000 description 1
- 238000003491 array Methods 0.000 description 1
- 230000004888 barrier function Effects 0.000 description 1
- 230000015572 biosynthetic process Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000012512 characterization method Methods 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 210000001503 joint Anatomy 0.000 description 1
- 238000003475 lamination Methods 0.000 description 1
- 230000014759 maintenance of location Effects 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 230000008520 organization Effects 0.000 description 1
- 238000002360 preparation method Methods 0.000 description 1
- 230000000750 progressive effect Effects 0.000 description 1
- 238000009877 rendering Methods 0.000 description 1
- 238000003786 synthesis reaction Methods 0.000 description 1
Landscapes
- Image Analysis (AREA)
Abstract
The invention discloses a kind of personnel's detection method, device, equipment and media.The step of this method includes: to carry out convolutional neural networks training to the radiofrequency signal thermal map sample for including human body attitude information, generates radiofrequency signal identification model；Wherein, radiofrequency signal thermal map sample is to draw generation to human-body emitting radiofrequency signal and according to reflection signal of the human body to radiofrequency signal in advance；By transmitting radiofrequency signal to obtain the target radio frequency signal thermal map under target scene, and target radio frequency signal thermal map input radio frequency signal identification model is obtained into testing result to detect to target scene with the presence or absence of personnel.This method opposite can improve the order of accuarcy detected for personnel in space, and opposite can ensure the global reliability of personnel's detection.In addition, the present invention also provides a kind of personnel's detection device, equipment and medium, beneficial effect are same as above.
Description
Technical field
The present invention relates to personnel's detection fields, more particularly to a kind of personnel's detection method, device, equipment and medium.
Background technique
Human body attitude estimation is very important application in a computer vision technique, is widely used in monitoring, activity
The fields such as identification, game mainly extract arm, leg joint and the characteristic points such as trunk and head from human body image,
And two-dimensional skeleton is reformulated with these information.
It can be differentiated with the presence or absence of there is personnel in scene, therefore according to scene image in such a way that human body attitude is estimated
It is currently more widely used in personnel's detection in confined space, such as in train, aircraft public transport, it can with this
When public transport is stopped transport, whether public transport inside still have passenger, and then ensure if being detected by current mode
The personal safety of passenger.
Current to be detected in public transport with the presence or absence of personnel by human body attitude estimation, specific implementation is logical
The monitoring camera photographed scene image in public transport is crossed, and inputs convolutional Neural for scene image as judgment basis
Network model learns a series of individual features points that whether there is characterization organization of human body in scene image to be analyzed,
Determine to whether there is in public transport with this and not yet leaves personnel.But due to the shelter in public transport
Excessively, when personnel are obscured by an object, the camera that can not be monitored is taken, and then by convolutional neural networks model to scene
The analysis of image is also just difficult to learn the presence for having personnel in public transport, and when shelter may hide personnel
A part of body when, the difficulty of visual processes can be increased, increase the probability of erroneous judgement, therefore detect currently for personnel in space
Order of accuarcy it is lower, it is difficult to ensure personnel detection global reliability.
It can be seen that provide a kind of personnel's detection method, with the order of accuarcy that opposite raising detects personnel in space,
The opposite global reliability for ensuring personnel's detection, is those skilled in the art's urgent problem to be solved.
Summary of the invention
The object of the present invention is to provide a kind of personnel's detection method, device, equipment and media, with opposite raising for space
The order of accuarcy of interior personnel's detection, the opposite global reliability for ensuring personnel's detection.
In order to solve the above technical problems, the present invention provides a kind of personnel's detection method, comprising:
Convolutional neural networks training is carried out to the radiofrequency signal thermal map sample for including human body attitude information, generates radio frequency letter
Number identification model；Wherein, radiofrequency signal thermal map sample is in advance to human-body emitting radiofrequency signal and according to human body to radiofrequency signal
Reflection signal draw generate；
By transmitting radiofrequency signal to obtain the target radio frequency signal thermal map under target scene, and target radio frequency signal is warm
Figure input radio frequency signal identification model obtains testing result to detect to target scene with the presence or absence of personnel.
Preferably, convolutional neural networks training is carried out to the radiofrequency signal thermal map sample for including human body attitude information, it is raw
Include: at radiofrequency signal identification model
Acquisition includes the radiofrequency signal thermal map sample of human body attitude information and corresponding with radiofrequency signal thermal map sample
Scene image sample；
Scene image sample is input to the preset human body attitude estimation model of convolutional neural networks, generates scene image sample
This corresponding standard joint point diagram；
By carrying out convolutional neural networks training to radiofrequency signal thermal map sample, training pattern is generated；
Training pattern is modified until training pattern training joint point diagram that radiofrequency signal thermal map is handled with
Deviation between the point diagram of standard joint is less than preset value, and training pattern is set as radiofrequency signal identification model.
Preferably, by carrying out convolutional neural networks training to radiofrequency signal thermal map sample, training pattern is generated specifically:
By carrying out the convolutional neural networks training based on space-time convolution to radiofrequency signal thermal map sample, training mould is generated
Type.
Preferably, by target radio frequency signal thermal map input radio frequency signal identification model whether there is personnel to target scene
It is detected, obtaining testing result includes:
Target radio frequency signal thermal map input radio frequency signal identification model is generated into human synovial point diagram；
Judge whether the target point in human synovial point diagram has artis；
If it is, will test result be set as characterize target scene there are the contents of personnel；
Otherwise, then it will test result to be set as characterizing the content that target scene does not have personnel.
Preferably, when judging whether the target point in human synovial point diagram has the result of artis to be, the party
Method further comprises:
The highest each artis of mutual information weighted value in human synovial point diagram is connected with each other, bone image is generated；
The behavior state of personnel under target scene is determined according to bone image.
Preferably, radiofrequency signal thermal map sample is specially the radiofrequency signal thermal map sample of rgb format.
In addition, the present invention also provides a kind of personnel's detection devices, comprising:
Model training module, for carrying out convolutional Neural net to the radiofrequency signal thermal map sample for including human body attitude information
Network training generates radiofrequency signal identification model；Wherein, radiofrequency signal thermal map sample is in advance to human-body emitting radiofrequency signal and root
Generation is drawn according to reflection signal of the human body to radiofrequency signal；
Model checking module, for by transmitting radiofrequency signal to obtain the target radio frequency signal thermal map under target scene,
And target radio frequency signal thermal map input radio frequency signal identification model is obtained with detecting to target scene with the presence or absence of personnel
Testing result.
Preferably, model training module includes:
Sample acquisition module, for obtain include human body attitude information radiofrequency signal thermal map sample and with radio frequency believe
Number corresponding scene image sample of thermal map sample；
Artis generation module is estimated for scene image sample to be input to the preset human body attitude of convolutional neural networks
Model generates the corresponding standard joint point diagram of scene image sample；
Training pattern generation module, for generating by carrying out convolutional neural networks training to radiofrequency signal thermal map sample
Training pattern；
Training pattern correction module, for being modified to training pattern until training pattern is to the processing of radiofrequency signal thermal map
Deviation between obtained training joint point diagram and standard joint point diagram is less than preset value, and training pattern is set as radio frequency letter
Number identification model.
In addition, the present invention also provides a kind of personnel inspection equipments, comprising:
Memory, for storing computer program；
Processor is realized when for executing computer program such as the step of above-mentioned personnel's detection method.
In addition, being stored with meter on computer readable storage medium the present invention also provides a kind of computer readable storage medium
Calculation machine program is realized when computer program is executed by processor such as the step of above-mentioned personnel's detection method.
Personnel's detection method provided by the present invention, by the radiofrequency signal thermal map sample for including human body attitude information
Convolutional neural networks training is carried out, generates corresponding radiofrequency signal identification model, wherein radiofrequency signal thermal map sample is by pre-
Generation first is drawn to human-body emitting radiofrequency signal and according to reflection signal of the human body to radiofrequency signal, and then passes through transmitting radio frequency
Signal to obtain the target radio frequency signal thermal map under target scene, and as the input picture of radiofrequency signal identification model with
Target radio frequency signal thermal map is handled by radiofrequency signal identification model, is realized to target scene with this with the presence or absence of personnel
It is detected, and then obtains testing result.This method is to first pass through convolutional neural networks to including in radiofrequency signal thermal map sample
The correlated characteristic of human body attitude be trained extraction, generating has the radiofrequency signal identification model for distinguishing characteristics of human body's ability,
And then radiofrequency signal thermal map under actual scene is analyzed by radiofrequency signal identification model, with this learn under actual scene whether
There are personnel.It is relayed since radiofrequency signal can penetrate shelter, shelter can be showed by radiofrequency signal thermal map
The object at rear, therefore even if personnel are obscured by an object, the presence of the personnel can be also learnt by radiofrequency signal thermal map, into
And this method carries out personnel's detection according to the radiofrequency signal thermal map generated by radiofrequency signal, opposite can improve for people in space
The order of accuarcy of member's detection, and opposite can ensure the global reliability of personnel's detection.In addition, the present invention also provides a kind of personnel
Detection device, equipment and medium, beneficial effect are same as above.
Detailed description of the invention
In order to illustrate the embodiments of the present invention more clearly, attached drawing needed in the embodiment will be done simply below
It introduces, it should be apparent that, drawings in the following description are only some embodiments of the invention, for ordinary skill people
For member, without creative efforts, it is also possible to obtain other drawings based on these drawings.
Fig. 1 is a kind of flow chart of personnel's detection method provided in an embodiment of the present invention；
Fig. 2 is the flow chart of another personnel's detection method provided in an embodiment of the present invention；
Fig. 3 is a kind of structure chart of personnel's detection device provided in an embodiment of the present invention.
Specific embodiment
Following will be combined with the drawings in the embodiments of the present invention, and technical solution in the embodiment of the present invention carries out clear, complete
Site preparation description, it is clear that described embodiments are only a part of the embodiments of the present invention, rather than whole embodiments.Based on this
Embodiment in invention, those of ordinary skill in the art are without making creative work, obtained every other
Embodiment belongs to the scope of the present invention.
Core of the invention is to provide a kind of personnel's detection method, and it is accurate that personnel in space are detected with opposite raising
Degree, the opposite global reliability for ensuring personnel's detection.Another core of the invention is to provide a kind of personnel's detection device, equipment
And medium.
In order to enable those skilled in the art to better understand the solution of the present invention, with reference to the accompanying drawings and detailed description
The present invention is described in further detail.
Embodiment one
Fig. 1 is a kind of flow chart of personnel's detection method provided in an embodiment of the present invention.Referring to FIG. 1, personnel detection side
The specific steps of method include:
Step S10: carrying out convolutional neural networks training to the radiofrequency signal thermal map sample for including human body attitude information, raw
At radiofrequency signal identification model.
Wherein, radiofrequency signal thermal map sample is in advance to human-body emitting radiofrequency signal and according to human body to the anti-of radiofrequency signal
It penetrates signal and draws generation.
It should be noted that focusing on carrying out the device signal thermal map for including human body attitude information in this step
Convolutional neural networks are trained, and the radiofrequency signal thermal map sample in this step is pre- first passes through to human-body emitting radiofrequency signal, in turn
The reflection signal reflected after radiofrequency signal is stopped to draw generation according to human body.Due to radiofrequency signal be through ovennodulation,
Possess the electric wave of certain tranmitting frequency, therefore there is the characteristic of wave in radiofrequency signal, during transmitting, such as barrier or
Human body Shi Zehui generates reflection while penetrating, and when the density contrast of the object penetrated is bigger, volume reflection is bigger, and base
In different objects reflection after reflection signal have specific velocity of wave and wavelength, therefore can according to reflection signal velocity of wave with
And wavelength distinguishes different objects, and since reflection signal can be reflected according to the shape of reflected object, in this step
Radiofrequency signal thermal map pass through in advance to human-body emitting radiofrequency signal, and then receive the reflection signal reflected through human body,
And according to the reflection signal carry out Image Rendering, generated with this include human body attitude information radiofrequency signal thermal map sample, tool
Body is to draw signal thermal map according to the intensity of the reflection signal on current scene direction.It is emphasized that signal thermal map is not
The shape of object is reflected, but according to the waveform diagram for the shape generation for reflecting object, reflect the bigger signal thermal map of intensity of signal
In waveform color it is redder, it is smaller then more blue, can in order to more comprehensively receive the signal that reflects back up of current scene each side
Signal reception is carried out so that a certain number of aerial arrays to be respectively set on horizontal and vertical different location.
Due to including the posture information of human body in radiofrequency signal thermal map sample, by convolutional neural networks to video
Signal thermal map is trained, and is learnt with this to the correlated characteristic in relation to human body attitude in radiofrequency signal thermal map sample, in turn
Generate the radiofrequency signal identification model that can analyze the correlated characteristic information of human body in radiofrequency signal thermal map.
Step S11: by transmitting radiofrequency signal to obtain the target radio frequency signal thermal map under target scene, and target is penetrated
Frequency signal thermal map input radio frequency signal identification model obtains testing result to detect to target scene with the presence or absence of personnel.
This step is the radiofrequency signal identification model by being previously obtained, and is made whether exist under true application scenarios
The detection of personnel.It needs to first pass through when being executed and emits radiofrequency signal in the scene, and then radio frequency is believed according to object under scene
The reflection signal of number reflection carries out the drafting of target radio frequency signal thermal map, and specific draw in mode and step S10 believes radio frequency
The method for drafting of number thermal map sample is close, but it needs to be emphasized that due to the purpose for generating radiofrequency signal thermal map sample be in order to
Convolution study is carried out to the feature of human body attitude, therefore radiofrequency signal thermal map sample is only generated by the reflection signal of human body, and this
Target radio frequency signal thermal map in step is generated under true usage scenario, often includes to remove under true usage scenario
Object other than human body, therefore after obtaining reflection signal, because of the specific frequency reflected by specific filter circuit with human body
Rate is filtered reflection signal, to remove the signal that other unrelated objects are reflected.
In addition, it is necessary to explanation, in this step implementation procedure, in order to reach transmitting radiofrequency signal and receive reflection
The purpose of signal needs the preparatory signal transmitting and receiving circuit installed in the scene for emitting radiofrequency signal and collecting reflection signal.
Signal transmitting and receiving circuit can be made of FMCW (CW with frequency modulation) signal generating circuit and signal receiving circuit, the former is using letter
Module number occurs, filter circuit, automatic gain module and antenna transceiving device are combined into signal projector, wherein signal occurs
The realization of module can be DDS (Direct Digital Synthesizer) and PLL (phaselocked loop) use in conjunction together, utilize
The high-resolution characteristic of DDS solves the contradiction of PLL frequency resolution and frequency switching time.The high frequency resolution of DDS changes
Into the stepped intervals of output frequency, while frequency multiplication is carried out using PLL and exports higher frequency signal.The frequency that this hybrid plan is realized
Rate synthesis system can obtain the frequency signal of broadband, high frequency resolution, switched at high frequency speed, meet Various Complex system
The requirement of system.
Personnel's detection method provided by the present invention, by the radiofrequency signal thermal map sample for including human body attitude information
Convolutional neural networks training is carried out, generates corresponding radiofrequency signal identification model, wherein radiofrequency signal thermal map sample is by pre-
Generation first is drawn to human-body emitting radiofrequency signal and according to reflection signal of the human body to radiofrequency signal, and then passes through transmitting radio frequency
Signal to obtain the target radio frequency signal thermal map under target scene, and as the input picture of radiofrequency signal identification model with
Target radio frequency signal thermal map is handled by radiofrequency signal identification model, is realized to target scene with this with the presence or absence of personnel
It is detected, and then obtains testing result.This method is to first pass through convolutional neural networks to including in radiofrequency signal thermal map sample
The correlated characteristic of human body attitude be trained extraction, generating has the radiofrequency signal identification model for distinguishing characteristics of human body's ability,
And then radiofrequency signal thermal map under actual scene is analyzed by radiofrequency signal identification model, with this learn under actual scene whether
There are personnel.It is relayed since radiofrequency signal can penetrate shelter, shelter can be showed by radiofrequency signal thermal map
The object at rear, therefore even if personnel are obscured by an object, the presence of the personnel can be also learnt by radiofrequency signal thermal map, into
And this method carries out personnel's detection according to the radiofrequency signal thermal map generated by radiofrequency signal, opposite can improve for people in space
The order of accuarcy of member's detection, and opposite can ensure the global reliability of personnel's detection.
Embodiment two
On the basis of the above embodiments, the present invention also provides a series of preferred embodiments.
Fig. 2 is the flow chart of another personnel's detection method provided in an embodiment of the present invention.Step S11 and Fig. 1 phase in Fig. 2
Together, details are not described herein.
As shown in Fig. 2, as a preferred embodiment, to the radiofrequency signal thermal map sample for including human body attitude information
The training of this progress convolutional neural networks, generating radiofrequency signal identification model includes:
Step S20: acquisition include human body attitude information radiofrequency signal thermal map sample and with radiofrequency signal thermal map sample
This corresponding scene image sample.
It is emphasized that the scene image sample obtained in this step be it is corresponding with signal thermal map sample, also
It is to say, the scene image sample and radiofrequency signal thermal map sample obtained in this step includes identical human body attitude letter
Breath.In addition, the scene image sample in this step refers to the visual pattern shot by camera.
Step S21: scene image sample is input to the preset human body attitude of convolutional neural networks and estimates model, generates field
The corresponding standard joint point diagram of scape image pattern.
It should be noted that being identified due to being currently based on deep learning to the human body attitude in vision level, generate
Joint point diagram belongs to that well known to a person skilled in the art contents, therefore specific execute details this will not be repeated here.The mesh of this step
Be scene image sample to be input to convolutional neural networks, and then currently existing human body attitude is estimated by convolutional neural networks
Model is counted, the extraction of artis is carried out to human body attitude in scene image sample, and then artis is collected and is generated as corresponding
Standard joint point diagram, human body attitude is able to reflect out by the artis in the point diagram of standard joint.
Step S22: by carrying out convolutional neural networks training to radiofrequency signal thermal map sample, training pattern is generated.
This step is by carrying out convolutional neural networks training to radiofrequency signal thermal map sample, it is therefore an objective to radiofrequency signal thermal map
Feature in sample is extracted and is clustered, and ultimately generates training pattern.Training pattern is in radiofrequency signal thermal map sample
Feature extract and generated after clustering learning, it is therefore an objective to can according to the feature learnt in advance in radiofrequency signal thermal map
Analyze the information about human body attitude.It is those skilled in the art due to carrying out convolutional neural networks training to image pattern
Well known technology contents, therefore this will not be repeated here.
Step S23: being modified training pattern until training pattern closes the training that radiofrequency signal thermal map is handled
Deviation between node diagram and standard joint point diagram is less than preset value, and training pattern is set as radiofrequency signal identification model.
This step is the committed step of present embodiment, and the core concept of this step is that existing human body attitude is estimated mould
Type carries out human body attitude to identify standard of the standard joint generated point diagram as adjusting training model.Since training pattern is used
In a series of artis for being analyzed and characterized human body attitude in radiofrequency signal thermal map, and with the position of artis and quantitative determination
Under scene whether there is personnel, therefore theoretically for, training pattern should be with people for the analysis result of radiofrequency signal thermal map
Body Attitude estimation model is consistent to the analysis result of scene image, therefore for the precision of analysis of opposite raising training pattern,
Present embodiment is modified to training pattern until the training joint point diagram that training pattern handles radiofrequency signal thermal map
Deviation between the point diagram of standard joint is less than preset value, that is, thinks the precision of analysis and existing human body appearance of training pattern
State estimates that model is close, and then is set as can be applied to the radiofrequency signal identification model under real scene for training pattern.
The parameter in convolutional neural networks training process is repaired in addition, the amendment in this step refers in itself
Change, achievees the purpose that correct training pattern with this, modified and reached to gained model due to the parameter to convolutional neural networks
Amendment, be content known to those skilled in the art, this will not be repeated here.
Present embodiment estimates that model carries out the training of training pattern as standard by relatively accurate human body attitude
The amendment of relative efficiency.
On the basis of the above embodiment, as a preferred embodiment, by radiofrequency signal thermal map sample
Convolutional neural networks training is carried out, training pattern is generated specifically:
By carrying out the convolutional neural networks training based on space-time convolution to radiofrequency signal thermal map sample, training mould is generated
Type.
It should be noted that space-time convolution refers to input multiframe (L frame) image, then once with step-by-step movement (in time side
To) to certain amount frame (D frame) carry out convolution (result of multiple (primary handled image) 2D convolution kernels is stacked up,
Result as 3D convolution), D < L, thus retention time information, rather than convolution is carried out to all frames in synchronization, it keeps away
Exempt from time to rupture information.
In specific operation process, convolutional neural networks can be set to being made of ten layers of 9*5*5 space-time convolutional layer, made
Convolution is carried out with the convolution kernel that step-length is 1*2*2.And batch normalization is carried out simultaneously using ReLU activation primitive after each layer
Output, and then by the output of the two-way of coding network (vertically, level is each all the way) data, use the convolutional network (strided that strides
Convolutional networks) it is combined, the characteristic information in figure is summarized to remove Spatial Dimension, and then take
The deconvolution (fractionally strided convolutions) for the 3*6*6 that 4 layers of convolution step-length are 1* (1/2) * (1/2)
The warp lamination composition for the 3*6*6 that layer and level 1 volume product step-length are 1* (1/4) * (1/4), is decoded as characteristic information to train joint
Point diagram, with this according to the discrepancy adjustment training pattern between training joint point diagram and standard joint point diagram.
Present embodiment is to guarantee the invariance of space-time using the purpose of space-time convolution, and then guarantees the whole of training pattern
Body accuracy.
In addition, as a preferred embodiment, by target radio frequency signal thermal map input radio frequency signal identification model with
Target scene is detected with the presence or absence of personnel, obtaining testing result includes:
Target radio frequency signal thermal map input radio frequency signal identification model is generated into human synovial point diagram；
Judge whether the target point in human synovial point diagram has artis；
If it is, will test result be set as characterize target scene there are the contents of personnel；
Otherwise, then it will test result to be set as characterizing the content that target scene does not have personnel.
It should be noted that due to consideration that artis is to generate after extracting human body attitude feature, and close
Relative positional relationship between node is fixed, therefore present embodiment is by determining that positional relationship is opposite in human synovial point diagram
Whether fixed target point has artis, to determine in target scene with the presence or absence of personnel.Present embodiment is according only to people
Relatively-stationary minority point determines with the presence or absence of personnel in target scene, relatively with the presence or absence of artis in the point diagram of body joint
Overall overhead needed for reducing testing result generating process.In addition, the selection of the target point in present embodiment and whole
Body quantity should be not specifically limited herein depending on the degree of accuracy of detection.
On the basis of the above embodiment, as a preferred embodiment, when judging in human synovial point diagram
It is when being that whether target point, which has the result of artis, and this method further comprises:
The highest each artis of mutual information weighted value in human synovial point diagram is connected with each other, bone image is generated；
The behavior state of personnel under target scene is determined according to bone image.
It should be noted that mutual information (Mutual Information) is a kind of measure information mode in information theory, it
The information content about another stochastic variable for including in a stochastic variable can be regarded as.Mutual information in present embodiment
What weighted value characterized is the relevance degree in human synovial point diagram between each artis, since human body is in active procedure,
There is linkage between each artis, therefore the relevance between artis, Jin Erben can be characterized in a manner of mutual information
The highest artis of correlation degree is connected with each other by embodiment, i.e., mutual information weighted value in human synovial point diagram is highest each
Artis is connected with each other, and generates bone image with this, and can be according to skeletal form in bone image after generating bone image
Infer the behavior state of the personnel under target scene.Present embodiment can further infer the behavior state of personnel,
And then personnel can be analyzed with the presence or absence of danger according to behavior state, further improve the comprehensive of personnel's detection.
On the basis of a series of above-mentioned embodiments, as a preferred embodiment, radiofrequency signal thermal map sample
The specially radiofrequency signal thermal map sample of rgb format.
It should be noted that due to consideration that the image of rgb format has the advantages that color expression range is wide, therefore this reality
Mode is applied to be trained the radiofrequency signal thermal map sample of rgb format, it can be according to radiofrequency signal thermal map sample relative abundance
Color expression range improves the rich of in radiofrequency signal thermal map sample extracted feature, and then with respect to improving training sample
Overall accuracy.
Embodiment three
Hereinbefore the embodiment of personnel's detection is described in detail, the present invention also provides a kind of and this method
Corresponding personnel's detection device, since the embodiment of device part is corresponded to each other with the embodiment of method part, device portion
The embodiment divided refers to the description of the embodiment of method part, wouldn't repeat here.
Fig. 3 is a kind of structure chart of personnel's detection device provided in an embodiment of the present invention.People provided in an embodiment of the present invention
Member's detection device, comprising:
Model training module 10, for carrying out convolutional Neural to the radiofrequency signal thermal map sample for including human body attitude information
Network training generates radiofrequency signal identification model；Wherein, radiofrequency signal thermal map sample be in advance to human-body emitting radiofrequency signal simultaneously
Generation is drawn according to reflection signal of the human body to radiofrequency signal.
Model checking module 11, for obtaining the target radio frequency signal heat under target scene by transmitting radiofrequency signal
Figure, and by target radio frequency signal thermal map input radio frequency signal identification model to be detected to target scene with the presence or absence of personnel,
Obtain testing result.
Personnel's detection device provided by the present invention, by the radiofrequency signal thermal map sample for including human body attitude information
Convolutional neural networks training is carried out, generates corresponding radiofrequency signal identification model, wherein radiofrequency signal thermal map sample is by pre-
Generation first is drawn to human-body emitting radiofrequency signal and according to reflection signal of the human body to radiofrequency signal, and then passes through transmitting radio frequency
Signal to obtain the target radio frequency signal thermal map under target scene, and as the input picture of radiofrequency signal identification model with
Target radio frequency signal thermal map is handled by radiofrequency signal identification model, is realized to target scene with this with the presence or absence of personnel
It is detected, and then obtains testing result.The present apparatus is to first pass through convolutional neural networks to including in radiofrequency signal thermal map sample
The correlated characteristic of human body attitude be trained extraction, generating has the radiofrequency signal identification model for distinguishing characteristics of human body's ability,
And then radiofrequency signal thermal map under actual scene is analyzed by radiofrequency signal identification model, with this learn under actual scene whether
There are personnel.It is relayed since radiofrequency signal can penetrate shelter, shelter can be showed by radiofrequency signal thermal map
The object at rear, therefore even if personnel are obscured by an object, the presence of the personnel can be also learnt by radiofrequency signal thermal map, into
And the present apparatus carries out personnel's detection according to the radiofrequency signal thermal map generated by radiofrequency signal, opposite can improve for people in space
The order of accuarcy of member's detection, and opposite can ensure the global reliability of personnel's detection.
On the basis of embodiment three, model training module includes:
Sample acquisition module, for obtain include human body attitude information radiofrequency signal thermal map sample and with radio frequency believe
Number corresponding scene image sample of thermal map sample.
Artis generation module is estimated for scene image sample to be input to the preset human body attitude of convolutional neural networks
Model generates the corresponding standard joint point diagram of scene image sample.
Training pattern generation module, for generating by carrying out convolutional neural networks training to radiofrequency signal thermal map sample
Training pattern.
Training pattern correction module, for being modified to training pattern until training pattern is to the processing of radiofrequency signal thermal map
Deviation between obtained training joint point diagram and standard joint point diagram is less than preset value, and training pattern is set as radio frequency letter
Number identification model.
Example IV
The present invention also provides a kind of personnel inspection equipments, comprising:
Memory, for storing computer program；
Processor is realized when for executing computer program such as the step of above-mentioned personnel's detection method.
Personnel inspection equipment provided by the present invention, by the radiofrequency signal thermal map sample for including human body attitude information
Convolutional neural networks training is carried out, generates corresponding radiofrequency signal identification model, wherein radiofrequency signal thermal map sample is by pre-
Generation first is drawn to human-body emitting radiofrequency signal and according to reflection signal of the human body to radiofrequency signal, and then passes through transmitting radio frequency
Signal to obtain the target radio frequency signal thermal map under target scene, and as the input picture of radiofrequency signal identification model with
Target radio frequency signal thermal map is handled by radiofrequency signal identification model, is realized to target scene with this with the presence or absence of personnel
It is detected, and then obtains testing result.This equipment is to first pass through convolutional neural networks to including in radiofrequency signal thermal map sample
The correlated characteristic of human body attitude be trained extraction, generating has the radiofrequency signal identification model for distinguishing characteristics of human body's ability,
And then radiofrequency signal thermal map under actual scene is analyzed by radiofrequency signal identification model, with this learn under actual scene whether
There are personnel.It is relayed since radiofrequency signal can penetrate shelter, shelter can be showed by radiofrequency signal thermal map
The object at rear, therefore even if personnel are obscured by an object, the presence of the personnel can be also learnt by radiofrequency signal thermal map, into
And this equipment carries out personnel's detection according to the radiofrequency signal thermal map generated by radiofrequency signal, opposite can improve for people in space
The order of accuarcy of member's detection, and opposite can ensure the global reliability of personnel's detection.
In addition, being stored with meter on computer readable storage medium the present invention also provides a kind of computer readable storage medium
Calculation machine program is realized when computer program is executed by processor such as the step of above-mentioned personnel's detection method.
Computer readable storage medium provided by the present invention, by the radiofrequency signal heat for including human body attitude information
The training of pattern this progress convolutional neural networks, generates corresponding radiofrequency signal identification model, wherein radiofrequency signal thermal map sample is
By what is generated in advance to human-body emitting radiofrequency signal and according to human body to the reflection signal drafting of radiofrequency signal, and then pass through hair
Radiofrequency signal is penetrated to obtain the target radio frequency signal thermal map under target scene, and as the input of radiofrequency signal identification model
Image is realized with this to whether target scene is deposited with being handled by radiofrequency signal identification model target radio frequency signal thermal map
It is detected in personnel, and then obtains testing result.This computer readable storage medium is to first pass through convolutional neural networks to penetrating
The correlated characteristic for the human body attitude for including in frequency signal thermal map sample is trained extraction, and generating has discrimination characteristics of human body's ability
Radiofrequency signal identification model, and then radiofrequency signal thermal map under actual scene is analyzed by radiofrequency signal identification model, with
This is learned under actual scene with the presence or absence of personnel.It is relayed since radiofrequency signal can penetrate shelter, passes through radiofrequency signal
Thermal map can show the object at shelter rear, therefore even if personnel are obscured by an object, and can also pass through radiofrequency signal thermal map
Learn the presence of the personnel, so this computer readable storage medium according to the radiofrequency signal thermal map generated by radiofrequency signal into
Administrative staff's detection opposite can improve the order of accuarcy detected for personnel in space, and opposite can ensure personnel's detection
Global reliability.
A kind of personnel's detection method provided by the present invention, device, equipment and medium are described in detail above.It says
Each embodiment is described in a progressive manner in bright book, and the highlights of each of the examples are the differences with other embodiments
Place, the same or similar parts in each embodiment may refer to each other.For device, equipment disclosed in embodiment and medium
Speech, since it is corresponded to the methods disclosed in the examples, so being described relatively simple, related place is referring to method part illustration
?.It should be pointed out that for those skilled in the art, without departing from the principle of the present invention, also
Can be with several improvements and modifications are made to the present invention, these improvement and modification also fall into the protection scope of the claims in the present invention
It is interior.
It should also be noted that, in the present specification, relational terms such as first and second and the like be used merely to by
One entity or operation are distinguished with another entity or operation, without necessarily requiring or implying these entities or operation
Between there are any actual relationship or orders.Moreover, the terms "include", "comprise" or its any other variant meaning
Covering non-exclusive inclusion, so that the process, method, article or equipment for including a series of elements not only includes that
A little elements, but also including other elements that are not explicitly listed, or further include for this process, method, article or
The intrinsic element of equipment.In the absence of more restrictions, the element limited by sentence "including a ...", is not arranged
Except there is also other identical elements in the process, method, article or apparatus that includes the element.
Claims (10)
1. a kind of personnel's detection method characterized by comprising
Convolutional neural networks training is carried out to the radiofrequency signal thermal map sample for including human body attitude information, radiofrequency signal is generated and knows
Other model；Wherein, the radiofrequency signal thermal map sample is in advance to human-body emitting radiofrequency signal and according to the human body to described
The reflection signal of radiofrequency signal draws generation；
By transmitting radiofrequency signal to obtain the target radio frequency signal thermal map under target scene, and the target radio frequency signal is warm
Figure inputs the radiofrequency signal identification model to detect to the target scene with the presence or absence of personnel, obtains testing result.
2. the method according to claim 1, wherein described pair include human body attitude information radiofrequency signal heat
The training of pattern this progress convolutional neural networks, generating radiofrequency signal identification model includes:
Acquisition include human body attitude information the radiofrequency signal thermal map sample and with the radiofrequency signal thermal map sample pair
The scene image sample answered；
The scene image sample is input to the preset human body attitude estimation model of convolutional neural networks, generates the scene figure
Decent corresponding standard joint point diagram；
By carrying out convolutional neural networks training to the radiofrequency signal thermal map sample, training pattern is generated；
The training pattern is modified until the training pattern closes the training that the radiofrequency signal thermal map is handled
Deviation between node diagram and standard joint point diagram is less than preset value, and the training pattern is set as the radio frequency and is believed
Number identification model.
3. according to the method described in claim 2, it is characterized in that, described by being rolled up to the radiofrequency signal thermal map sample
Product neural metwork training, generates training pattern specifically:
It is trained by carrying out the convolutional neural networks based on space-time convolution to the radiofrequency signal thermal map sample, described in generation
Training pattern.
4. according to the method described in claim 2, it is characterized in that, described will penetrate described in target radio frequency signal thermal map input
To detect to the target scene with the presence or absence of personnel, obtain testing result includes: frequency signal identification model
The target radio frequency signal thermal map is inputted into the radiofrequency signal identification model and generates human synovial point diagram；
Judge whether the target point in the human synovial point diagram has artis；
If it is, being set as the testing result to characterize the target scene, there are the contents of personnel；
Otherwise, then the testing result is set as characterizing the content that the target scene does not have personnel.
5. according to the method described in claim 4, it is characterized in that, when the target point in the judgement human synovial point diagram
Whether the result with artis is when being for position, and this method further comprises:
The highest each artis of mutual information weighted value in the human synovial point diagram is connected with each other, bone image is generated；
The behavior state of personnel under the target scene is determined according to the bone image.
6. according to claim 1 to method described in 5 any one, which is characterized in that the radiofrequency signal thermal map sample is specific
For the radiofrequency signal thermal map sample of rgb format.
7. a kind of personnel's detection device characterized by comprising
Model training module, for carrying out convolutional neural networks instruction to the radiofrequency signal thermal map sample for including human body attitude information
Practice, generates radiofrequency signal identification model；Wherein, the radiofrequency signal thermal map sample is in advance to human-body emitting radiofrequency signal and root
Generation is drawn according to reflection signal of the human body to the radiofrequency signal；
Model checking module, for obtaining the target radio frequency signal thermal map under target scene by transmitting radiofrequency signal, and will
The target radio frequency signal thermal map inputs the radiofrequency signal identification model to carry out to the target scene with the presence or absence of personnel
Detection obtains testing result.
8. device according to claim 7, which is characterized in that the model training module includes:
Sample acquisition module includes the radiofrequency signal thermal map sample of human body attitude information and penetrates with described for obtaining
The corresponding scene image sample of frequency signal thermal map sample；
Artis generation module is estimated for the scene image sample to be input to the preset human body attitude of convolutional neural networks
Model generates the corresponding standard joint point diagram of the scene image sample；
Training pattern generation module, for generating by carrying out convolutional neural networks training to the radiofrequency signal thermal map sample
Training pattern；
Training pattern correction module, for being modified to the training pattern until the training pattern is to the radiofrequency signal
Deviation between the training joint point diagram that thermal map is handled and standard joint point diagram is less than preset value, by the training
Model specification is the radiofrequency signal identification model.
9. a kind of personnel inspection equipment characterized by comprising
Memory, for storing computer program；
Processor realizes such as personnel's detection method as claimed in any one of claims 1 to 6 when for executing the computer program
The step of.
10. a kind of computer readable storage medium, which is characterized in that be stored with computer on the computer readable storage medium
Program is realized when the computer program is executed by processor such as personnel's detection method as claimed in any one of claims 1 to 6
Step.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811645930.7A CN109740522B (en) | 2018-12-29 | 2018-12-29 | Personnel detection method, device, equipment and medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811645930.7A CN109740522B (en) | 2018-12-29 | 2018-12-29 | Personnel detection method, device, equipment and medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109740522A true CN109740522A (en) | 2019-05-10 |
| CN109740522B CN109740522B (en) | 2023-05-02 |
Family
ID=66362805
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811645930.7A Active CN109740522B (en) | 2018-12-29 | 2018-12-29 | Personnel detection method, device, equipment and medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109740522B (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110175554A (en) * | 2019-05-23 | 2019-08-27 | 南方科技大学 | Detection method, device, equipment and the storage medium of indoor arrangement |
| CN110287774A (en) * | 2019-05-14 | 2019-09-27 | 广东康云科技有限公司 | Object method for detecting, system and storage medium based on WIFI |
| CN111291692A (en) * | 2020-02-17 | 2020-06-16 | 咪咕文化科技有限公司 | Video scene recognition method and device, electronic equipment and storage medium |
| CN111652040A (en) * | 2020-04-14 | 2020-09-11 | 广西万云科技有限公司 | Method and system for recognizing human body activity based on radio frequency and signal self-adaptive convolutional neural network |
| CN113671199A (en) * | 2020-07-28 | 2021-11-19 | 国网江苏省电力有限公司南通供电分公司 | Sampling system for sealed sample submission |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105787439A (en) * | 2016-02-04 | 2016-07-20 | 广州新节奏智能科技有限公司 | Depth image human body joint positioning method based on convolution nerve network |
| CN106236061A (en) * | 2015-06-04 | 2016-12-21 | 松下知识产权经营株式会社 | Human body detection device |
| CN107886069A (en) * | 2017-11-10 | 2018-04-06 | 东北大学 | A kind of multiple target human body 2D gesture real-time detection systems and detection method |
| CN108229389A (en) * | 2017-12-29 | 2018-06-29 | 努比亚技术有限公司 | Facial image processing method, apparatus and computer readable storage medium |
| 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 |
-
2018
- 2018-12-29 CN CN201811645930.7A patent/CN109740522B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106236061A (en) * | 2015-06-04 | 2016-12-21 | 松下知识产权经营株式会社 | Human body detection device |
| CN105787439A (en) * | 2016-02-04 | 2016-07-20 | 广州新节奏智能科技有限公司 | Depth image human body joint positioning method based on convolution nerve network |
| CN107886069A (en) * | 2017-11-10 | 2018-04-06 | 东北大学 | A kind of multiple target human body 2D gesture real-time detection systems and detection method |
| CN108229389A (en) * | 2017-12-29 | 2018-06-29 | 努比亚技术有限公司 | Facial image processing method, apparatus and computer readable storage medium |
| 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 |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110287774A (en) * | 2019-05-14 | 2019-09-27 | 广东康云科技有限公司 | Object method for detecting, system and storage medium based on WIFI |
| CN110175554A (en) * | 2019-05-23 | 2019-08-27 | 南方科技大学 | Detection method, device, equipment and the storage medium of indoor arrangement |
| CN110175554B (en) * | 2019-05-23 | 2024-01-12 | 南方科技大学 | Indoor layout detection method, device, equipment and storage medium |
| CN111291692A (en) * | 2020-02-17 | 2020-06-16 | 咪咕文化科技有限公司 | Video scene recognition method and device, electronic equipment and storage medium |
| CN111291692B (en) * | 2020-02-17 | 2023-10-20 | 咪咕文化科技有限公司 | Video scene recognition method and device, electronic equipment and storage medium |
| CN111652040A (en) * | 2020-04-14 | 2020-09-11 | 广西万云科技有限公司 | Method and system for recognizing human body activity based on radio frequency and signal self-adaptive convolutional neural network |
| CN113671199A (en) * | 2020-07-28 | 2021-11-19 | 国网江苏省电力有限公司南通供电分公司 | Sampling system for sealed sample submission |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109740522B (en) | 2023-05-02 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109740522A (en) | A kind of personnel's detection method, device, equipment and medium | |
| CN105872477B (en) | video monitoring method and video monitoring system | |
| EP3869459B1 (en) | Target object identification method and apparatus, storage medium and electronic apparatus | |
| CN105518709B (en) | The method, system and computer program product of face for identification | |
| CN107330437B (en) | Feature extraction method based on convolutional neural network target real-time detection model | |
| CN108319901B (en) | Biopsy method, device, computer equipment and the readable medium of face | |
| CN109447169A (en) | The training method of image processing method and its model, device and electronic system | |
| CN110110617A (en) | Medical image dividing method, device, electronic equipment and storage medium | |
| CN109271888A (en) | Personal identification method, device, electronic equipment based on gait | |
| CN110222787A (en) | Multiscale target detection method, device, computer equipment and storage medium | |
| CN107240122A (en) | Video target tracking method based on space and time continuous correlation filtering | |
| CN109214366A (en) | Localized target recognition methods, apparatus and system again | |
| CN106056089B (en) | A kind of 3 d pose recognition methods and system | |
| CN106104573A (en) | Hough processor | |
| CN110462633A (en) | A kind of method, apparatus and electronic equipment of recognition of face | |
| CN109685768A (en) | Lung neoplasm automatic testing method and system based on lung CT sequence | |
| CN114862837A (en) | Human body security check image detection method and system based on improved YOLOv5s | |
| CN108256404A (en) | Pedestrian detection method and device | |
| CN105740780A (en) | Method and device for human face in-vivo detection | |
| CN110263768A (en) | A kind of face identification method based on depth residual error network | |
| CN106033601A (en) | Method and apparatus for detecting abnormal situation | |
| CN108960015A (en) | A kind of vehicle system automatic identifying method and device based on deep learning | |
| CN110490052A (en) | Face datection and face character analysis method and system based on cascade multi-task learning | |
| CN110070029A (en) | A kind of gait recognition method and device | |
| CN109961029A (en) | A kind of dangerous goods detection method, device and computer readable storage medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |