CN110135259A - Silent formula living body image identification method, device, computer equipment and storage medium - Google Patents
Silent formula living body image identification method, device, computer equipment and storage medium Download PDFInfo
- Publication number
- CN110135259A CN110135259A CN201910298482.6A CN201910298482A CN110135259A CN 110135259 A CN110135259 A CN 110135259A CN 201910298482 A CN201910298482 A CN 201910298482A CN 110135259 A CN110135259 A CN 110135259A
- Authority
- CN
- China
- Prior art keywords
- picture
- image data
- verified
- multichannel
- living body
- 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 40
- 238000003860 storage Methods 0.000 title claims abstract description 14
- 238000004040 coloring Methods 0.000 claims abstract description 23
- 238000012549 training Methods 0.000 claims description 35
- 238000004590 computer program Methods 0.000 claims description 28
- 238000013527 convolutional neural network Methods 0.000 claims description 25
- 230000002708 enhancing effect Effects 0.000 claims description 23
- 230000006870 function Effects 0.000 claims description 19
- 230000008569 process Effects 0.000 claims description 10
- 238000010586 diagram Methods 0.000 claims description 8
- 238000010606 normalization Methods 0.000 claims description 6
- 238000010801 machine learning Methods 0.000 abstract 1
- 239000000284 extract Substances 0.000 description 4
- 241001269238 Data Species 0.000 description 3
- 230000006399 behavior Effects 0.000 description 3
- 230000008859 change Effects 0.000 description 3
- 238000013135 deep learning Methods 0.000 description 3
- 230000000694 effects Effects 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 3
- 238000000605 extraction Methods 0.000 description 3
- 238000013528 artificial neural network Methods 0.000 description 2
- 238000004422 calculation algorithm Methods 0.000 description 2
- 238000004364 calculation method Methods 0.000 description 2
- 238000006243 chemical reaction Methods 0.000 description 2
- 238000012790 confirmation Methods 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 230000004927 fusion Effects 0.000 description 2
- 238000013507 mapping Methods 0.000 description 2
- 239000000203 mixture Substances 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 238000012216 screening Methods 0.000 description 2
- 238000000926 separation method Methods 0.000 description 2
- 238000012360 testing method Methods 0.000 description 2
- 241001062009 Indigofera Species 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 238000009826 distribution Methods 0.000 description 1
- 230000005284 excitation Effects 0.000 description 1
- 238000003475 lamination Methods 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
- 230000001360 synchronised effect Effects 0.000 description 1
- 230000002123 temporal effect Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/40—Spoof detection, e.g. liveness detection
- G06V40/45—Detection of the body part being alive
Abstract
This application involves a kind of silent formula living body image identification method, device, computer equipment and storage medium based on machine learning.The described method includes: obtaining picture to be verified, according to the colouring information and luminance information of picture to be verified, construct the multichannel image data of picture to be verified, multichannel image data is inputted into pre-set depth convolutional network, obtain the corresponding feature tag of multichannel image data, when feature tag is matched with target labels, it is determined that picture to be verified is living body picture.It can be improved the efficiency of living body picture recognition using this method.
Description
Technical field
This application involves field of computer technology, more particularly to a kind of silent formula living body image identification method, device, meter
Calculate machine equipment and storage medium.
Background technique
With the development of computer technology, recognition of face has also obtained biggish development.When carrying out recognition of face, need
Face information is obtained by camera, then face information is identified, so that it is determined that the identity of people, but this mode
Under, can not confirm acquisition is the face information of living body, dangerous so as to cause recognition of face.
In traditional technology, in order to solve the problems, such as vivo identification, binocular camera can be used, obtains three-dimensional set letter
Breath, but this mode is high to hardware requirement, it can be achieved that property is poor, can realize vivo identification using software approach, utilize
When software approach is realized, need to obtain the face picture of multiframe under the cooperation of user, then by including in pictorial information
It whether is living body in the acquired picture of temporal information confirmation, however under this mode, complicated operation, and user is needed to cooperate
Just it is able to achieve.
Summary of the invention
Based on this, it is necessary in view of the above technical problems, provide a kind of problem complicated for operation for being able to solve vivo identification
Silent formula living body image identification method, device, computer equipment and storage medium.
A kind of silence formula living body image identification method, which comprises
Obtain picture to be verified;
According to the colouring information and luminance information of the picture to be verified, the multichannel picture of the picture to be verified is constructed
Data;
The multichannel image data is inputted into pre-set depth convolutional network, obtains the multichannel image data
Corresponding feature tag;
When the feature tag is matched with target labels, it is determined that the picture to be verified is living body picture.
In one of the embodiments, further include: according to the colouring information of the picture to be verified, obtain described to be verified
The RGB triple channel image data of picture;According to the luminance information of the picture to be verified, the HSV of the picture to be verified is obtained
Triple channel image data;According to the RGB triple channel image data and the HSV triple channel image data, the multi-pass is obtained
Road image data.
In one of the embodiments, further include: the multichannel image data is inputted into pre-set convolutional Neural
Network, by the convolutional layers of the convolutional neural networks to the RGB triple channel image data and HSV triple channel image data into
Row calculates, and obtains the corresponding picture feature of the multichannel image data;According to the picture feature, the multichannel figure is obtained
The corresponding feature tag of sheet data.
In one of the embodiments, further include: according to the full articulamentum of the depth convolutional network, obtain the picture
The probability of Feature Mapping to each default label is exported in the default label by pre-set normalization exponential function
One as the corresponding feature tag of the multichannel image data.
In one of the embodiments, further include: according to a pre-set picture, building correspondence and the primary figure
The secondary picture of piece;The secondary picture is the image data obtained by shooting a picture;According to the primary figure
Piece and the secondary picture establish the training set and verifying collection of the depth convolutional network;By the training set and in advance
The loss function of setting is trained initial convolutional neural networks, when the initial convolutional neural networks collect in the verifying
Accuracy rate when reaching threshold value, obtain the depth convolutional neural networks.
In one of the embodiments, further include: data enhancement operations are carried out to a picture, obtain multiple correspondences
In picture of enhancing of a picture;The data enhancement operations include: rotation process, zoom operations and overturning behaviour
Make;The data enhancement operations are carried out to the secondary picture, obtain multiple enhancing quadratic diagrams for corresponding to the secondary picture
Piece;According to picture of the enhancing and the secondary picture of enhancing, training set and the verifying of the depth convolutional network are established
Collection.
The feature tag includes 1 or 0 in one of the embodiments, and the target labels are 1, further includes: when described
It when feature tag is 1, determines that the feature tag is matched with the target labels, determines that the picture to be verified is living body figure
Piece.
A kind of silence formula living body picture recognition device, described device include:
Data acquisition module, for obtaining picture to be verified;
Characteristic extracting module constructs described to be tested for the colouring information and luminance information according to the picture to be verified
Demonstrate,prove the multichannel image data of picture;
Prediction module obtains described for the multichannel image data to be inputted pre-set depth convolutional network
The corresponding feature tag of multichannel image data;
Identification module, for when the feature tag is matched with target labels, it is determined that the picture to be verified is to live
Body picture.
A kind of computer equipment, including memory and processor, the memory are stored with computer program, the processing
Device performs the steps of when executing the computer program
Obtain picture to be verified;
According to the colouring information and luminance information of the picture to be verified, the multichannel picture of the picture to be verified is constructed
Data;
The multichannel image data is inputted into pre-set depth convolutional network, obtains the multichannel image data
Corresponding feature tag;
When the feature tag is matched with target labels, it is determined that the picture to be verified is living body picture.
A kind of computer readable storage medium, is stored thereon with computer program, and the computer program is held by processor
It is performed the steps of when row
Obtain picture to be verified;
According to the colouring information and luminance information of the picture to be verified, the multichannel picture of the picture to be verified is constructed
Data;
The multichannel image data is inputted into pre-set depth convolutional network, obtains the multichannel image data
Corresponding feature tag;
When the feature tag is matched with target labels, it is determined that the picture to be verified is living body picture.
Above-mentioned silence formula living body image identification method, device, computer equipment and storage medium, by obtaining silent formula
Image data carries out multichannel input to image data based on color characteristic and brightness, constructs multichannel image data, with
Input of the multichannel image data as depth convolutional network, depth convolutional network are to carry out depth by a large amount of image datas
Acquistion is arrived, and therefore, for multichannel image data, can complete low-level image feature extraction and low-level image feature to high-level characteristic
Conversion, since low-level image feature is to convert to obtain by brightness and color characteristic, high-level characteristic can and further be deepened
Connection between brightness and color characteristic, when detecting living body, full articulamentum maps to each label according to high-level characteristic
As a result, export corresponding feature tag, when the feature tag of output is matched with target labels, determine image data whether be
Living body picture.Because without obtaining the continuous picture with timing, so that it may determine whether image data is living body picture, the present invention
The technical solution of embodiment operated when realizing living body picture recognition it is more simple, to improve the efficiency of living body picture recognition.
Detailed description of the invention
Fig. 1 is the application scenario diagram of silent formula living body image identification method in one embodiment;
Fig. 2 is the flow diagram of silent formula living body image identification method in one embodiment;
Fig. 3 is the flow diagram that multichannel image data step is constructed in one embodiment;
Fig. 4 is the flow diagram of silent formula living body image identification method in another embodiment;
Fig. 5 is the structural block diagram of silent formula living body picture recognition device in one embodiment;
Fig. 6 is the internal structure chart of computer equipment in one embodiment.
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, not
For limiting the application.
Silence formula living body image identification method provided by the present application, can be applied in application environment as shown in Figure 1.Its
In, terminal 102 is communicated with server 104 by network by network.Wherein, terminal 102 can be, but not limited to be various
Photographic device, personal computer, laptop, smart phone, tablet computer and portable wearable device, server 104
It can be realized with the server cluster of the either multiple server compositions of independent server.
When terminal 102 is photographic device, terminal 102 passes through local area network or internet and 104 phase of server
Even, terminal 102 carries out shooting and the image data that shooting obtains is sent to service by network when receiving shooting instruction
Device 104.
In addition, server 104 includes the mode of two kinds of acquisition image datas when terminal 102 is personal computer, one is
By the camera function of personal computer, at this point, personal computer is equal to photographic device, the second is image data is stored in
In the storage medium of personal computer, so that image data be passed through on network by taking out image data from storage medium
Reach server 104.
By the above-mentioned means, server 104 gets image data, image data is defined as image data to be verified,
Server extracts the colouring information and luminance information of image data, constructs the multi-channel data of image data to be verified, thus will
In the depth convolutional network being arranged in multi-channel data input server 104, depth convolutional network data multichannel image data
Corresponding feature tag.Then server matches feature tag and target labels, if the two matches, determines to be tested
Card picture is living body picture.
In one embodiment, it as shown in Fig. 2, providing a kind of silent formula living body image identification method, answers in this way
For being illustrated for the server in Fig. 1, comprising the following steps:
Step 202, picture to be verified is obtained.
Picture to be verified indicates that server receives the image data with face, with face representation in picture region
At least there is a face.
Specifically, server can receive the image data from photographic device, also can receive from personal computer
Etc. terminals image data, under a specific living body verifying scene, the face of target person appears in camera overlay area, takes the photograph
As the picture of device shooting camera overlay area, therefore the image data with face is uploaded to server, server should
Image data is labeled as picture to be verified.
Step 204, according to the colouring information and luminance information of picture to be verified, the multichannel picture of picture to be verified is constructed
Data.
Colouring information refers to the pixel data of picture to be verified, passes through the face of the available picture to be verified of colouring information
Color distribution situation can be indicated using the rgb format of standard, can also be indicated using other color formats;What luminance information referred to
It is the brightness performance information of picture to be verified, it, can be with by the brightness display effect of the available picture to be verified of luminance information
Luminance information is identified by HSV model;Multichannel image data refers to the image data for possessing many-sided information, this step
In specifically indicate simultaneously comprising colouring information and luminance information image data.
Specifically, can be obtained by the way of fusion after the colouring information and luminance information for extracting picture to be verified
To multichannel image data, multichannel image data can also be obtained by the way of fitting superposition.
Step 206, multichannel image data is inputted into pre-set depth convolutional network, obtains multichannel image data
Corresponding feature tag.
Depth convolutional network is to carry out deep learning to convolutional neural networks to obtain, multiple convolution in depth convolutional network
Layer establishes input image data and presets contacting between label by the deep learning to a large amount of image data, because
This can export the corresponding feature tag of multichannel image data when inputting multi-channel data, it is worth noting that, feature mark
Label are one of the labels of output layer of depth convolutional network.
Preliminary convolutional neural networks can be preset in server, then by the mass data collected, to convolution
Neural network is trained, the available depth convolutional network for meeting accuracy requirement.
Step 208, when feature tag is matched with target labels, it is determined that picture to be verified is living body picture.
Target labels are pre-set label in server, and target labels can be rolled up according to matched logic from depth
One is selected in label in the output layer of product network.Living body picture refers to that image data is by shooting true live subject
It obtains, the secondary picture that difference is obtained with the false face of shooting.
In above-mentioned silence formula living body image identification method, by obtaining the image data of silent formula, based on color characteristic and
Brightness carries out multichannel input to image data, multichannel image data is constructed, using multichannel image data as depth
The input of convolutional network, depth convolutional network is to carry out deep learning by a large amount of image datas to obtain, therefore, for multichannel
Image data can complete the conversion of low-level image feature extraction and low-level image feature to high-level characteristic, since low-level image feature is by bright
Degree feature and color characteristic convert to obtain, and therefore, high-level characteristic can and further be deepened between brightness and color characteristic
Connection, when detecting living body, full articulamentum according to high-level characteristic map to each label as a result, exporting corresponding feature mark
Label, when the feature tag of output is matched with target labels, determine whether image data is living body picture.Because without obtaining band
The continuous picture of timing, so that it may determine whether image data is living body picture, and the technical solution of the embodiment of the present invention is being realized
Operation is more simple when living body picture recognition, to improve the efficiency of living body picture recognition.
The technical solution of above-described embodiment, under each usage scenario, operation is very convenient, such as: on-line processing letter
When with card, need to enroll the face picture of applicant, and confirmation is the operation information of applicant, can pass through Shen at this time
Ask someone using terminal obtain applicant face picture, after face picture is uploaded to server, server passes through to picture number
According to handled, the sequence of operations such as data fusion, mode input, server exports result according to model and is confirmed whether it is application
People's operation behavior, operates conveniently.
In one embodiment, for step 202, server can also get video data, then from video data
Extract picture to be verified.
Specifically, video data is decomposed into multiple video frames, and by analyzing multiple video frames, analytic process packet
It includes, by carrying out noise analysis to the corresponding image data of video frame and by edge algorithms, calculating the corresponding figure of video frame
The size of face in sheet data, to select noise minimum, the maximum video frame of face area as picture to be verified.
In addition, video data can be obtained by single camera, to reduce the difficulty of data source data acquisition.
In one embodiment, as shown in figure 3, providing a kind of schematic flow chart for constructing multichannel image data, specifically
Steps are as follows:
Step 302, according to the colouring information of picture to be verified, the RGB triple channel image data of picture to be verified is obtained.
RGB triple channel image data indicate R (red) channel, (green) channel G and B (indigo plant) channel data, pass through by
Picture to be verified is inputted by RGB triple channel, available RGB triple channel image data.
Step 304, according to the luminance information of picture to be verified, the HSV triple channel image data of picture to be verified is obtained.
The number in H (tone) channel, (saturation degree) channel S and V (lightness) channel that HSV triple channel image data identifies
According to, by the way that picture to be verified is inputted by HSV triple channel, available HSV triple channel image data.
Step 306, according to RGB triple channel image data and HSV triple channel image data, multichannel image data is obtained.
In the present embodiment, by using the mode that multichannel inputs, bulk information in picture to be verified is extracted, therefore increase
The integrality of verifying picture description is treated, to improve model prediction training when carrying out model training and model prediction
Efficiency and the accuracy of model prediction.
For step 302, in one embodiment, RGB triple channel is referred to by the way that picture to be verified is inputted pixel separation
Tool, R value, G value and B value in isolated picture to be verified, such as one section of pixel-matrix rgb value be [(128,255,
255), (0,255,255), (128,0,255)], after the input of RGB triple channel, the data for obtaining the channel R are [128,0,128], are obtained
Data to the channel G are [225,225,0] and obtain the data of channel B to be [225,225,225].
For step 304, in one embodiment, HSV threeway Dow Jones index when by by picture to be verified input pixel separation
Tool, H value, S value and the V value of isolated picture to be verified, such as one section of pixel HSV value be [(1,0.5,0.5),
(2,0.3,0.3), (3,0.2,0.2)], wherein the unit of H value is angle, i.e., when H value is 1, needs to be converted into 1 corresponding
Angle, after the input of HSV triple channel, the H value for obtaining picture to be verified is [1,2,3], and obtaining S value is [0.5,0.3,0.2], is obtained
V value is [0.5,0.3,0.2].
For step 306, in one embodiment, multichannel image data can be by the value of triple channel RGB and triple channel HSV
Value be overlapped after, input same convolutional layer and carry out convolution algorithm, to establish the connection of each channel value.
In another embodiment, multichannel image data is inputted into depth convolutional network, has specifically carried out following operation: will
Multichannel image data inputs pre-set depth convolutional network, by the convolutional layer of depth convolutional network to RGB triple channel
Image data and HSV triple channel image data carry out convolutional calculation, obtain the corresponding picture feature of multichannel image data, according to
Picture feature obtains the corresponding feature tag of multichannel image data.
Specifically, RGB triple channel image data and HSV triple channel image data are low level feature, pass through multiple volumes
The convolutional calculation of lamination obtains high-level picture feature, therefore by depth convolutional network, can extract picture to be verified
High-level feature, to improve the accuracy of living body picture prediction.
In another embodiment, following operation has specifically been carried out to the process for exporting feature tag by picture feature: according to
The full articulamentum of depth convolutional network obtains the probability that picture feature maps to each default label, to be referred to by normalization
Number function exports one in default label as the corresponding feature tag of multichannel image data.
Specifically, by full articulamentum, the node established in picture feature and the full connection relationship for connecting node layer, then
Regression forecasting is carried out using normalization exponential function (softmax layer), to export the corresponding spy of multichannel image data
Levy label.The excitation function of full articulamentum can choose Relu function and carry out Nonlinear Mapping.
In one embodiment, as shown in figure 4, providing a kind of schematic flow body of the mode of trained depth convolutional network,
Specific step is as follows:
Step 402, according to a pre-set picture, building corresponds to the secondary picture of a picture.
Secondary picture is the image data obtained by shooting a picture, and one time picture refers to living body picture.
It can be shot by internet or entity, obtain an a large amount of picture, data supporting has been provided.
Step 404, according to a picture and secondary picture, the training set and verifying collection of depth convolutional network are established.
It include the secondary picture of a large amount of pictures and corresponding number in training set, it also includes suitable one that verifying, which is concentrated,
The secondary picture of secondary picture and corresponding number.Data in training set are responsible for being trained initial convolutional neural networks, and test
Card collection is responsible for verifying training effect.
Step 406, by training set and pre-set loss function, initial convolutional neural networks are trained,
When initial convolutional neural networks verifying collection accuracy rate reach threshold value when, obtain depth convolutional neural networks.
The default output valve of loss function is set in server, when the not up to default output valve of the output valve of loss function
When, according to the value that loss function exports, the parameter of parameter and full articulamentum to convolutional layer is adjusted, to carry out initial
The training of convolutional neural networks.Accuracy rate is referred to by that will verify the picture concentrated or the input training of secondary picture
After initial convolutional neural networks afterwards, obtained statistics accuracy rate.
In the present embodiment, collected by a picture and secondary picture project training collection and verifying, to reach to initial volume
The purpose that product neural network is trained, so as to improve the accuracy of depth convolutional network prediction.
For step 402, in one embodiment, the data source of a picture can be video data, thus video data
In extract video frame, preliminary screening is carried out to video frame, i.e. the excessive video frame of screening noise, so as to regard according to one section
Frequency is according to multiple pictures are obtained, so that data volume is greatly expanded, so as to further increase depth convolutional network
Training degree.
It is worth noting that being concentrated in a training set or verifying, the quantity phase of the quantity of a picture and secondary picture
Deng to guarantee that each picture is predicted in training, there is higher accuracy.
For step 404, in another embodiment, establishing training set and verifying collection, specific step is as follows: to primary figure
Piece carries out data enhancement operations, obtains multiple pictures of enhancing for corresponding to a picture;Data enhancement operations include: rotation
Operation, zoom operations and turning operation carry out data enhancement operations to secondary picture, obtain multiple corresponding to secondary picture
Enhance secondary picture, according to picture of enhancing and enhance secondary picture, establishes the training set and verifying collection of depth convolutional network.
In the present embodiment, the method for proposing a kind of outward bound collection and verifying collection sample, therefore depth convolution can be improved
The training degree of network, further increases the forecasting accuracy of depth convolutional network.
Further, rotation process can carry out duplication operation, then carry out to duplicate using a former picture as original part
It rotates by a certain angle to obtain a new picture, multiple available pictures of multiple rotary, same mode of operation passes through
Multiple available secondary pictures of one secondary picture are as sample.
Further, zoom operations refer to zooming in and out pixel size, such as a picture of a 1920*1080
It is scaled a picture of a 1280*720, so that being expanded by a secondary picture is two, carries out different degrees of scaling,
Multiple available secondary pictures, although that row zoom operations will not change its display effect, when carrying out feature extraction,
There is change to the dimension of input data.Similarly, it can also be carried out by quantity of the turning operation to a picture and secondary picture
It expands.
In one embodiment, depth convolutional network output default label include 1 and 0, therefore feature tag may be 1 or
Person 0, when setting 1 for target labels, when feature tag and 1, it is determined that feature tag and target labels matching, determine to
Verifying picture is living body picture.It is worth noting that when depth convolutional network output label be 1 when, then it represents that input to
Verifying picture is living body picture.
It should be understood that although each step in the flow chart of Fig. 2-4 is successively shown according to the instruction of arrow,
These steps are not that the inevitable sequence according to arrow instruction successively executes.Unless expressly stating otherwise herein, these steps
Execution there is no stringent sequences to limit, these steps can execute in other order.Moreover, at least one in Fig. 2-4
Part steps may include that perhaps these sub-steps of multiple stages or stage are not necessarily in synchronization to multiple sub-steps
Completion is executed, but can be executed at different times, the execution sequence in these sub-steps or stage is also not necessarily successively
It carries out, but can be at least part of the sub-step or stage of other steps or other steps in turn or alternately
It executes.
In one embodiment, as shown in figure 5, providing a kind of silent formula living body picture recognition device, comprising: data obtain
Modulus block 502, characteristic extracting module 504, prediction module 506 and identification module 508, in which:
Data acquisition module 502, for obtaining picture to be verified.
Characteristic extracting module 504, for the colouring information and luminance information according to the picture to be verified, building it is described to
Verify the multichannel image data of picture.
Prediction module 506 obtains institute for the multichannel image data to be inputted pre-set depth convolutional network
State the corresponding feature tag of multichannel image data.
Identification module 508, for when the feature tag is matched with target labels, it is determined that the picture to be verified is
Living body picture.
Characteristic extracting module 504 is also used to the colouring information according to the picture to be verified in one of the embodiments,
Obtain the RGB triple channel image data of the picture to be verified;According to the luminance information of the picture to be verified, obtain it is described to
Verify the HSV triple channel image data of picture;According to the RGB triple channel image data and the HSV triple channel image data,
Obtain the multichannel image data.
Characteristic extracting module 504 is also used to input the multichannel image data preparatory in one of the embodiments,
The depth convolutional network of setting, by the convolutional layer of the depth convolutional network to the RGB triple channel image data and HSV tri-
Channel image data is calculated, and the corresponding picture feature of the multichannel image data is obtained;According to the picture feature, obtain
To the corresponding feature tag of the multichannel image data.
Characteristic extracting module 504 is also used to the full connection according to the depth convolutional network in one of the embodiments,
Layer, obtains the probability that the picture feature maps to each default label, passes through pre-set normalization exponential function, output
One in the default label is used as the corresponding feature tag of the multichannel image data.
It in one of the embodiments, further include model training module, for according to a pre-set picture, building
Secondary picture corresponding to a picture;The secondary picture is the picture number obtained by shooting a picture
According to;According to a picture and the secondary picture, the training set and verifying collection of the depth convolutional network are established;Pass through institute
Training set and pre-set loss function are stated, initial convolutional neural networks are trained, when the initial convolutional Neural
Network obtains the depth convolutional neural networks when the accuracy rate of the verifying collection reaches threshold value.
Model training module in one of the embodiments, is also used to carry out data enhancement operations to a picture,
Obtain multiple pictures of enhancing for corresponding to a picture;The data enhancement operations include: rotation process, scaling behaviour
Work and turning operation;The data enhancement operations are carried out to the secondary picture, obtain multiple corresponding to the secondary picture
The secondary picture of enhancing;According to picture of the enhancing and the secondary picture of enhancing, the depth convolutional network is established
Training set and verifying collection.
The feature tag includes 1 or 0 in one of the embodiments,;The target labels are 1;Identification module 508 is also
For determining that the feature tag is matched with the target labels, determining the picture to be verified when the feature tag is 1
For living body picture.
Specific restriction about silent formula living body picture recognition device may refer to above for silent formula living body picture
The restriction of recognition methods, details are not described herein.Modules in above-mentioned silence formula living body picture recognition device can whole or portion
Divide and is realized by software, hardware and combinations thereof.Above-mentioned each module can be embedded in the form of hardware or independently of computer equipment
In processor in, can also be stored in a software form in the memory in computer equipment, in order to processor calling hold
The corresponding operation of the above modules of row.
In one embodiment, a kind of computer equipment is provided, which can be server, internal junction
Composition can be as shown in Figure 6.The computer equipment include by system bus connect processor, memory, network interface and
Database.Wherein, the processor of the computer equipment is for providing calculating and control ability.The memory packet of the computer equipment
Include non-volatile memory medium, built-in storage.The non-volatile memory medium is stored with operating system, computer program and data
Library.The built-in storage provides environment for the operation of operating system and computer program in non-volatile memory medium.The calculating
The database of machine equipment is for storing image data.The network interface of the computer equipment is used to pass through network with external terminal
Connection communication.To realize a kind of silent formula living body image identification method when the computer program is executed by processor.
It will be understood by those skilled in the art that structure shown in Fig. 6, only part relevant to application scheme is tied
The block diagram of structure does not constitute the restriction for the computer equipment being applied thereon to application scheme, specific computer equipment
It may include perhaps combining certain components or with different component layouts than more or fewer components as shown in the figure.
In one embodiment, a kind of computer equipment, including memory and processor are provided, which is stored with
Computer program, the processor perform the steps of when executing computer program
Obtain picture to be verified;
According to the colouring information and luminance information of the picture to be verified, the multichannel picture of the picture to be verified is constructed
Data;
The multichannel image data is inputted into pre-set depth convolutional network, obtains the multichannel image data
Corresponding feature tag;
When the feature tag is matched with target labels, it is determined that the picture to be verified is living body picture.
In one embodiment, it also performs the steps of when processor executes computer program according to the figure to be verified
The colouring information of piece obtains the RGB triple channel image data of the picture to be verified;Believed according to the brightness of the picture to be verified
Breath, obtains the HSV triple channel image data of the picture to be verified;According to the RGB triple channel image data and the HSV tri-
Channel image data obtains the multichannel image data.
In one embodiment, it also performs the steps of when processor executes computer program by the multichannel picture
Data input pre-set depth convolutional network, by the convolutional layer of the depth convolutional network to the RGB triple channel figure
Sheet data and HSV triple channel image data are calculated, and the corresponding picture feature of the multichannel image data is obtained;According to institute
Picture feature is stated, the corresponding feature tag of the multichannel image data is obtained.
In one embodiment, it also performs the steps of when processor executes computer program according to the depth convolution
The full articulamentum of network obtains the probability that the picture feature maps to each default label, passes through pre-set normalization
Exponential function exports one in the default label as the corresponding feature tag of the multichannel image data.
In one embodiment, it also performs the steps of when processor executes computer program according to pre-set one
Secondary picture, building correspond to the secondary picture of a picture;The secondary picture is obtained by shooting a picture
The image data arrived;According to a picture and the secondary picture, establishes the training set of the depth convolutional network and test
Card collection;By the training set and pre-set loss function, initial convolutional neural networks are trained, when described first
Beginning convolutional neural networks obtain the depth convolutional neural networks when the accuracy rate of the verifying collection reaches threshold value.
In one embodiment, processor execute computer program when also perform the steps of to a picture into
Row data enhancement operations obtain multiple pictures of enhancing for corresponding to a picture;The data enhancement operations include:
Rotation process, zoom operations and turning operation;The data enhancement operations are carried out to the secondary picture, obtain multiple correspondences
In the secondary picture of enhancing of the secondary picture;According to picture of the enhancing and the secondary picture of the enhancing, described in foundation
The training set and verifying collection of depth convolutional network.
In one embodiment, the feature tag includes 1 or 0;The target labels are 1;Processor executes computer
It is also performed the steps of when program when the feature tag is 1, determines that the feature tag is matched with the target labels,
Determine that the picture to be verified is living body picture.
In one embodiment, a kind of computer readable storage medium is provided, computer program is stored thereon with, is calculated
Machine program performs the steps of when being executed by processor
Obtain picture to be verified;
According to the colouring information and luminance information of the picture to be verified, the multichannel picture of the picture to be verified is constructed
Data;
The multichannel image data is inputted into pre-set depth convolutional network, obtains the multichannel image data
Corresponding feature tag;
When the feature tag is matched with target labels, it is determined that the picture to be verified is living body picture.
In one embodiment, it also performs the steps of when computer program is executed by processor according to described to be verified
The colouring information of picture obtains the RGB triple channel image data of the picture to be verified;According to the brightness of the picture to be verified
Information obtains the HSV triple channel image data of the picture to be verified;According to the RGB triple channel image data and the HSV
Triple channel image data obtains the multichannel image data.
In one embodiment, it is also performed the steps of when computer program is executed by processor by the multichannel figure
Sheet data inputs pre-set depth convolutional network, by the convolutional layer of the depth convolutional network to the RGB triple channel
Image data and HSV triple channel image data are calculated, and the corresponding picture feature of the multichannel image data is obtained;According to
The picture feature obtains the corresponding feature tag of the multichannel image data.
In one embodiment, it also performs the steps of when computer program is executed by processor and is rolled up according to the depth
The full articulamentum of product network, obtains the probability that the picture feature maps to each default label, passes through pre-set normalizing
Change exponential function, exports one in the default label as the corresponding feature tag of the multichannel image data.
In one embodiment, it also performs the steps of when computer program is executed by processor according to pre-set
Picture, building correspond to the secondary picture of a picture;The secondary picture is by shooting a picture
Obtained image data;According to a picture and the secondary picture, establish the depth convolutional network training set and
Verifying collection;By the training set and pre-set loss function, initial convolutional neural networks are trained, when described
Initial convolutional neural networks obtain the depth convolutional neural networks when the accuracy rate of the verifying collection reaches threshold value.
In one embodiment, it is also performed the steps of when computer program is executed by processor to a picture
Data enhancement operations are carried out, multiple pictures of enhancing for corresponding to a picture are obtained;The data enhancement operations packet
It includes: rotation process, zoom operations and turning operation;The data enhancement operations are carried out to the secondary picture, obtain multiple
The secondary picture of enhancing corresponding to the secondary picture;According to picture of the enhancing and the secondary picture of enhancing, establish
The training set and verifying collection of the depth convolutional network.
In one embodiment, the feature tag includes 1 or 0;The target labels are 1;Computer program is processed
Device is also performed the steps of when executing when the feature tag is 1, determines the feature tag and the target labels
Match, determines that the picture to be verified is living body picture.
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, the computer program can be stored in a non-volatile computer
In read/write memory medium, the computer program is when being executed, it may include such as the process of the embodiment of above-mentioned each method.Wherein,
To any reference of memory, storage, database or other media used in each embodiment provided herein,
Including non-volatile and/or volatile memory.Nonvolatile memory may include read-only memory (ROM), programming ROM
(PROM), electrically programmable ROM (EPROM), electrically erasable ROM (EEPROM) or flash memory.Volatile memory may include
Random access memory (RAM) or external cache.By way of illustration and not limitation, RAM is available in many forms,
Such as static state RAM (SRAM), dynamic ram (DRAM), synchronous dram (SDRAM), double data rate sdram (DDRSDRAM), enhancing
Type SDRAM (ESDRAM), synchronization link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM
(RDRAM), direct memory bus dynamic ram (DRDRAM) and memory bus dynamic ram (RDRAM) etc..
Each technical characteristic of above embodiments can be combined arbitrarily, for simplicity of description, not to above-described embodiment
In each technical characteristic it is all possible combination be all described, as long as however, the combination of these technical characteristics be not present lance
Shield 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
It cannot therefore be construed as limiting the scope of the patent.It should be pointed out that coming for those of ordinary skill in the art
It says, without departing from the concept of this application, various modifications and improvements can be made, these belong to the protection of the application
Range.Therefore, the scope of protection shall be subject to the appended claims for the application patent.
Claims (10)
1. a kind of silence formula living body image identification method, which comprises
Obtain picture to be verified;
According to the colouring information and luminance information of the picture to be verified, the multichannel picture number of the picture to be verified is constructed
According to;
The multichannel image data is inputted into pre-set depth convolutional network, it is corresponding to obtain the multichannel image data
Feature tag;
When the feature tag is matched with target labels, it is determined that the picture to be verified is living body picture.
2. the method according to claim 1, wherein being believed according to the colouring information of the picture to be verified and brightness
Breath constructs the multichannel image data of the picture to be verified, comprising:
According to the colouring information of the picture to be verified, the RGB triple channel image data of the picture to be verified is obtained;
According to the luminance information of the picture to be verified, the HSV triple channel image data of the picture to be verified is obtained;
According to the RGB triple channel image data and the HSV triple channel image data, the multichannel image data is obtained.
3. according to the method described in claim 2, it is characterized in that, the multichannel image data is inputted pre-set depth
Convolutional network is spent, the corresponding feature tag of the multichannel image data is obtained, comprising:
The multichannel image data is inputted into pre-set depth convolutional network, passes through the convolution of the depth convolutional network
Layer calculates the RGB triple channel image data and HSV triple channel image data, obtains the multichannel image data pair
The picture feature answered;
According to the picture feature, the corresponding feature tag of the multichannel image data is obtained.
4. according to the method described in claim 3, it is characterized in that, obtaining the multichannel picture according to the picture feature
The corresponding feature tag of data, comprising:
According to the full articulamentum of the depth convolutional network, the probability that the picture feature maps to each default label is obtained,
By pre-set normalization exponential function, one exported in the default label is used as the multichannel image data pair
The feature tag answered.
5. the method according to claim 1, wherein the mode of training depth convolutional network, comprising:
According to a pre-set picture, building corresponds to the secondary picture of a picture;The secondary picture is logical
It crosses and shoots the image data that a picture obtains;
According to a picture and the secondary picture, the training set and verifying collection of the depth convolutional network are established;
By the training set and pre-set loss function, initial convolutional neural networks are trained, when described first
Beginning convolutional neural networks obtain the depth convolutional neural networks when the accuracy rate of the verifying collection reaches threshold value.
6. according to the method described in claim 5, it is characterized in that, being established according to a picture and the secondary picture
The training set and verifying collection of the depth convolutional network, comprising:
Data enhancement operations are carried out to a picture, obtain multiple pictures of enhancing for corresponding to a picture;
The data enhancement operations include: rotation process, zoom operations and turning operation;
The data enhancement operations are carried out to the secondary picture, obtain multiple enhancing quadratic diagrams for corresponding to the secondary picture
Piece;
According to picture of the enhancing and the secondary picture of enhancing, training set and the verifying of the depth convolutional network are established
Collection.
7. method according to any one of claims 1 to 6, which is characterized in that the feature tag includes 1 or 0;The mesh
Marking label is 1;
When the feature tag is matched with target labels, it is determined that the picture to be verified is living body picture, comprising:
When the feature tag is 1, determines that the feature tag is matched with the target labels, determine the picture to be verified
For living body picture.
8. a kind of silence formula living body picture recognition device, which is characterized in that described device includes:
Data acquisition module, for obtaining picture to be verified;
Characteristic extracting module constructs the figure to be verified for the colouring information and luminance information according to the picture to be verified
The multichannel image data of piece;
Prediction module obtains the multi-pass for the multichannel image data to be inputted pre-set depth convolutional network
The corresponding feature tag of road image data;
Identification module, for when the feature tag is matched with target labels, it is determined that the picture to be verified is living body figure
Piece.
9. a kind of computer equipment, including memory and processor, the memory are stored with computer program, feature exists
In the processor realizes method described in any one of claims 1 to 7 when executing computer program the step of.
10. a kind of computer readable storage medium, is stored thereon with computer program, which is characterized in that the computer program
The step of method described in any one of claims 1 to 7 is realized when being executed by processor.
Priority Applications (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910298482.6A CN110135259A (en) | 2019-04-15 | 2019-04-15 | Silent formula living body image identification method, device, computer equipment and storage medium |
PCT/CN2019/122920 WO2020211396A1 (en) | 2019-04-15 | 2019-12-04 | Silent living body image recognition method and apparatus, computer device and storage medium |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910298482.6A CN110135259A (en) | 2019-04-15 | 2019-04-15 | Silent formula living body image identification method, device, computer equipment and storage medium |
Publications (1)
Publication Number | Publication Date |
---|---|
CN110135259A true CN110135259A (en) | 2019-08-16 |
Family
ID=67569940
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910298482.6A Pending CN110135259A (en) | 2019-04-15 | 2019-04-15 | Silent formula living body image identification method, device, computer equipment and storage medium |
Country Status (2)
Country | Link |
---|---|
CN (1) | CN110135259A (en) |
WO (1) | WO2020211396A1 (en) |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2020211396A1 (en) * | 2019-04-15 | 2020-10-22 | 深圳壹账通智能科技有限公司 | Silent living body image recognition method and apparatus, computer device and storage medium |
CN112257685A (en) * | 2020-12-08 | 2021-01-22 | 成都新希望金融信息有限公司 | Face copying recognition method and device, electronic equipment and storage medium |
CN113111750A (en) * | 2021-03-31 | 2021-07-13 | 智慧眼科技股份有限公司 | Face living body detection method and device, computer equipment and storage medium |
CN116259091A (en) * | 2023-01-18 | 2023-06-13 | 北京飞腾时光信息科技有限公司 | Method and device for detecting silent living body |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107818313A (en) * | 2017-11-20 | 2018-03-20 | 腾讯科技(深圳)有限公司 | Vivo identification method, device, storage medium and computer equipment |
CN107992842A (en) * | 2017-12-13 | 2018-05-04 | 深圳云天励飞技术有限公司 | Biopsy method, computer installation and computer-readable recording medium |
CN109101925A (en) * | 2018-08-14 | 2018-12-28 | 成都智汇脸卡科技有限公司 | Biopsy method |
CN109376518A (en) * | 2018-10-18 | 2019-02-22 | 深圳壹账通智能科技有限公司 | Privacy leakage method and relevant device are prevented based on recognition of face |
Family Cites Families (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109034102B (en) * | 2018-08-14 | 2023-06-16 | 腾讯科技(深圳)有限公司 | Face living body detection method, device, equipment and storage medium |
CN109271863B (en) * | 2018-08-15 | 2022-03-18 | 北京小米移动软件有限公司 | Face living body detection method and device |
CN110135259A (en) * | 2019-04-15 | 2019-08-16 | 深圳壹账通智能科技有限公司 | Silent formula living body image identification method, device, computer equipment and storage medium |
-
2019
- 2019-04-15 CN CN201910298482.6A patent/CN110135259A/en active Pending
- 2019-12-04 WO PCT/CN2019/122920 patent/WO2020211396A1/en active Application Filing
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107818313A (en) * | 2017-11-20 | 2018-03-20 | 腾讯科技(深圳)有限公司 | Vivo identification method, device, storage medium and computer equipment |
CN107992842A (en) * | 2017-12-13 | 2018-05-04 | 深圳云天励飞技术有限公司 | Biopsy method, computer installation and computer-readable recording medium |
CN109101925A (en) * | 2018-08-14 | 2018-12-28 | 成都智汇脸卡科技有限公司 | Biopsy method |
CN109376518A (en) * | 2018-10-18 | 2019-02-22 | 深圳壹账通智能科技有限公司 | Privacy leakage method and relevant device are prevented based on recognition of face |
Non-Patent Citations (1)
Title |
---|
刘呈云: "基于纹理分析的活体人脸检测算法研究", 中国优秀硕士学位论文全文数据库信息科技辑, vol. 2018, no. 1, pages 138 - 1453 * |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2020211396A1 (en) * | 2019-04-15 | 2020-10-22 | 深圳壹账通智能科技有限公司 | Silent living body image recognition method and apparatus, computer device and storage medium |
CN112257685A (en) * | 2020-12-08 | 2021-01-22 | 成都新希望金融信息有限公司 | Face copying recognition method and device, electronic equipment and storage medium |
CN113111750A (en) * | 2021-03-31 | 2021-07-13 | 智慧眼科技股份有限公司 | Face living body detection method and device, computer equipment and storage medium |
CN116259091A (en) * | 2023-01-18 | 2023-06-13 | 北京飞腾时光信息科技有限公司 | Method and device for detecting silent living body |
CN116259091B (en) * | 2023-01-18 | 2023-11-10 | 北京飞腾时光信息科技有限公司 | Method and device for detecting silent living body |
Also Published As
Publication number | Publication date |
---|---|
WO2020211396A1 (en) | 2020-10-22 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110135259A (en) | Silent formula living body image identification method, device, computer equipment and storage medium | |
CN111368342B (en) | Image tampering identification model training method, image tampering identification method and device | |
CN110334585A (en) | Table recognition method, apparatus, computer equipment and storage medium | |
CN110502986A (en) | Identify character positions method, apparatus, computer equipment and storage medium in image | |
CN108830208A (en) | Method for processing video frequency and device, electronic equipment, computer readable storage medium | |
CN108229575A (en) | For detecting the method and apparatus of target | |
CN108463823B (en) | Reconstruction method and device of user hair model and terminal | |
CN108198177A (en) | Image acquiring method, device, terminal and storage medium | |
CN112446302B (en) | Human body posture detection method, system, electronic equipment and storage medium | |
CN111753782B (en) | False face detection method and device based on double-current network and electronic equipment | |
CN108537115B (en) | Image recognition method and device and electronic equipment | |
CN108399052A (en) | Picture compression method, apparatus, computer equipment and storage medium | |
CN108734214A (en) | Image-recognizing method and device, electronic equipment, storage medium | |
CN108184075A (en) | For generating the method and apparatus of image | |
CN108764321A (en) | Image-recognizing method and device, electronic equipment, storage medium | |
WO2020181706A1 (en) | Plant species identification method and apparatus | |
CN113269149A (en) | Living body face image detection method and device, computer equipment and storage medium | |
CN113689382B (en) | Tumor postoperative survival prediction method and system based on medical images and pathological images | |
WO2021249006A1 (en) | Method and apparatus for identifying authenticity of facial image, and medium and program product | |
CN114565955A (en) | Face attribute recognition model training and community personnel monitoring method, device and equipment | |
CN113066074A (en) | Visual saliency prediction method based on binocular parallax offset fusion | |
CN111428740A (en) | Detection method and device for network-shot photo, computer equipment and storage medium | |
CN113947568B (en) | Image processing method and device, electronic equipment and storage medium | |
CN109766832A (en) | A kind of image pre-processing method, device, system, equipment and storage medium | |
CN115375965A (en) | Preprocessing method for target scene recognition and target scene recognition method |
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 |