CN110503091A - Certificate verification method, device and storage medium based on neural network model - Google Patents
Certificate verification method, device and storage medium based on neural network model Download PDFInfo
- Publication number
- CN110503091A CN110503091A CN201910658721.4A CN201910658721A CN110503091A CN 110503091 A CN110503091 A CN 110503091A CN 201910658721 A CN201910658721 A CN 201910658721A CN 110503091 A CN110503091 A CN 110503091A
- Authority
- CN
- China
- Prior art keywords
- certificate
- neural network
- region
- staff
- network model
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/22—Matching criteria, e.g. proximity measures
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/60—Type of objects
- G06V20/62—Text, e.g. of license plates, overlay texts or captions on TV images
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V30/00—Character recognition; Recognising digital ink; Document-oriented image-based pattern recognition
- G06V30/10—Character recognition
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Multimedia (AREA)
- Biophysics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Computing Systems (AREA)
- Molecular Biology (AREA)
- General Health & Medical Sciences (AREA)
- Computational Linguistics (AREA)
- Biomedical Technology (AREA)
- Health & Medical Sciences (AREA)
- Image Analysis (AREA)
Abstract
The present invention relates to a kind of artificial intelligence technologys, disclose a kind of certificate verification method based on neural network model, it include: to obtain the staff's image for entering a place, and detect the certificate region in staff's image, utilize neural network model trained in advance, the certificate character is compared by the certificate character for identifying the certificate region with staff's identity database, determines the identity information for entering the staff in the place.The present invention also proposes a kind of certificate verifying device based on neural network model and a kind of computer readable storage medium.The present invention realizes the verifying of personnel identity.
Description
Technical field
The present invention relates to artificial intelligence field more particularly to it is a kind of based on intelligentized staff's certificate based on nerve
Certificate verification method, device and the computer readable storage medium of network model.
Background technique
Nowadays the intelligentized certificate verification mode such as face, fingerprint, sound emerges one after another, and improves verifying effect really
Rate.But face, fingerprint, voice authentication are also required for corresponding environment, and not all environment be suitable for face, fingerprint and
Voice authentication.For example, shop personnel require wearing cleanness clothing, cap, mask and hand for the workshop of dustlessization
Then set generates workshop by entering after Airshower chamber.In the generation workshop of dustlessization, since staff is by the big model of clothing
Encirclement is enclosed, and the workshop environment of dustlessization is noisy, wants through verification methods such as recognition of face, fingerprint recognition, voice authentications
It is really difficult into the generation workshop of dustlessization.So for some work for needing to enter the verification environments such as noisy, dustlessization
For personnel, the above verification method and the use that do not sound feasible.
Summary of the invention
The present invention provides a kind of certificate verification method based on neural network model, device and computer-readable storage medium
Matter, main purpose are when user scans in knowledge base, show accurately search result to user.
To achieve the above object, a kind of certificate verification method based on neural network model provided by the invention, comprising:
The staff's image for entering place is obtained, and detects the certificate region in staff's image;
Using neural network model trained in advance, the certificate character in the certificate region is identified;
The information retrieved in the database and the certificate character matches, determines according to matching result and enters the field
Staff identity information.
Optionally, described to detect that the certificate region in staff's image includes:
The doubtful certificate region in staff's image is detected using digital image processing techniques, determines described doubt
Like the profile in certificate region, candidate certificate region is obtained;
The certificate region is filtered out from the candidate certificate region using mathematical model trained in advance.
Optionally, the digital processing technology includes:
Using staff's image described in Gaussian Blur technology and gray processing technical treatment, and to staff's image
It carries out Sobel Operator operation and obtains the doubtful certificate region
Optionally, using neural network model trained in advance, the certificate character in the certificate region is identified, comprising:
Certificate region is switched to the feature vector of neural network;
Trained neural network passes through the feature vector in the doubtful certificate region of non-linear conversion equation calculation in advance, and passes through
The certificate character recognition of loss function output prediction goes out the character in certificate region.
Optionally, the information retrieved in the database and the certificate character matches, it is true according to matching result
Surely include: into the step of identity information of the staff in the place
The certificate character is switched into binary number;
The binary number that called data library has had both judges whether there is identical binary number, when having identical two
When system number, the identity information for entering the staff in place is determined according to the identical binary number.
In addition, to achieve the above object, the present invention also provides a kind of certificate verifying devices based on neural network model, it should
Device includes memory and processor, and the certificate verification program that can be run on the processor is stored in the memory,
The certificate verification program realizes following steps when being executed by the processor:
The staff's image for entering place is obtained, and detects the certificate region in staff's image;
Using neural network model trained in advance, the certificate character in the certificate region is identified;
The information retrieved in the database and the certificate character matches, determines according to matching result and enters the field
Staff identity information.
Optionally, described to detect that the certificate region in staff's image includes:
The doubtful certificate region in staff's image is detected using digital image processing techniques, and by finding out
The profile in the doubtful certificate region, obtains candidate certificate region;
The certificate region is filtered out from the candidate certificate region using mathematical model trained in advance.
Optionally, the digital processing technology includes:
Using staff's image described in Gaussian Blur technology and gray processing technical treatment, and to staff's image
It carries out Sobel Operator operation and obtains the doubtful certificate region.
Optionally, using neural network model trained in advance, identify that the certificate character in the certificate region includes:
Certificate region is switched to the feature vector of neural network;
Trained neural network passes through the feature vector in the doubtful certificate region of non-linear conversion equation calculation in advance, and passes through
The certificate character recognition of loss function output prediction goes out the character in certificate region.
In addition, to achieve the above object, it is described computer-readable the present invention also provides a kind of computer readable storage medium
Certificate verification program is stored on storage medium, the certificate verification program can be executed by one or more processor, with reality
Now as described above certificate verification method based on neural network model the step of.
Certificate verification method based on neural network model, device and computer readable storage medium proposed by the present invention,
The staff's image for entering a place is being obtained, and is detecting the certificate region in staff's image, is utilizing mind
Through network model, the certificate character in the certificate region is identified, and certificate character and staff's identity database are compared
Compared with the identity information of the determining staff for entering place, to show accurately verification result.
Detailed description of the invention
Fig. 1 is the flow diagram for the certificate verification method based on neural network model that one embodiment of the invention provides;
Fig. 2 is the internal structure signal for the certificate verifying device based on neural network model that one embodiment of the invention provides
Figure;
Certificate verification program in the certificate verifying device based on neural network model that Fig. 3 provides for one embodiment of the invention
Module diagram.
The embodiments will be further described with reference to the accompanying drawings for the realization, the function and the advantages of the object of the present invention.
Specific embodiment
It should be appreciated that the specific embodiments described herein are merely illustrative of the present invention, it is not intended to limit the present invention.
The present invention provides a kind of certificate verification method based on neural network model.It is real for the present invention one shown in referring to Fig.1
The flow diagram of the certificate verification method based on neural network model of example offer is provided.This method can be held by a device
Row, which can be by software and or hardware realization.
In the present embodiment, the certificate verification method based on neural network model includes:
S1, the staff's image for entering place is obtained, and detects the certificate region in staff's image.
Including being certificate positioning, support vector machines (SVM) is trained to be judged with certificate for the detection in the certificate region in the present invention
Two processes.In the detection process in entire certificate region, the segment that many may be certificate can be obtained, the present invention is by these segments
It carries out whether manual sort is certificate, is put into training in SVM model, obtains a judgment models of SVM.In actual certificate mistake
Cheng Zhong, then the segment of be likely to be certificate is inputted SVM judgment models, it is automatically selected by SVM model actually true
The exactly segment of certificate completes certificate detecting step.
Preferably, the S1 further comprises:
Firstly, carrying out Gaussian Blur processing to the staff's image being installed with comprising certificate that camera takes.It is described
A kind of image processing techniques of Gaussian Blur, effect is that the work clothes image that will be taken thickens.The present invention is using average
Gaussian Blur method.The value of each pixel of the mean-Gaussian fuzz method takes the average value of all pixels (totally 8) around, left
The pixel value of side red point is 2 originally, after fuzzy, just at 1 (mean value for taking all pixels around).Thus it completes to packet
The Gaussian Blur of staff's image containing certificate is handled.
After Gaussian Blur is handled, the present invention further carries out gray processing processing to image, by color image by ash
After degreeization processing, become gray scale picture, it is meant that subsequent all operations are based not on color information, reduce calculation amount.
Further, the present invention is to by Gaussian Blur processing and gray processing, treated that image carries out Sobel operator again
Operation obtains the single order horizontal direction derivative of image.The Soble operator principle is to ask image the level and Vertical Square of single order
To derivative, according to the size of derivative value to determine whether being edge, and edge described herein, it is exactly the possibility of doubtful certificate
Picture region.For convenience of calculation, Soble operator does not go derivation really, but has used the side of the weighted sum of all boundary values
Method is academicly referred to as " convolution ".Weight is known as " convolution mask ".After Sobel operation, the picture region of doubtful certificate is obvious
Distinguish.
Binarization operation is carried out to the picture region containing doubtful certificate.In gray level image, the value of each pixel is 0-
Number between 255 represents gloomy degree.Adaptive threshold T is set at this time, if the picture region containing certificate and doubtful certificate
The value of domain pixel is greater than threshold value T, is set as 1;If pixel value is less than threshold value T, it is set as 0.It is eventually converted into bianry image, i.e., often
A pixel only has 1 and 0 two value may.After above all operations, the wheel of certificate and doubtful certificate is finally found out
It is wide.Since certificate size is typically small and rectangular, therefore setting length is 7 centimetres, and wide is 3 centimetres.It is all to meet this size
Profile all be retained, ungratified rejecting.By the double selection of Soble operator and size, candidate certificate figure is finally obtained
Piece.
Further, the picture of candidate certificate is input in SVM model by the present invention one by one, then utilizes SVM mould
The judgement of type, if the answer of SVM model is not certificate picture, then continuing to next judgement, if it is certificate region, then
Picture is put into an output listing.Output listing is prepared for next step certificate character recognition.Support vector machines is whole
The thing that a step is done is to find an interface, is put between two classifications of certificate and non-certificate, thus distinguishes and comes to testify
Part and non-certificate.So whole process is need to look for the process of interface, wherein interface is also known as hyperplane, with following formula table
Show certificate:
wTX+b=0
W is adjustable weight vector;T is the transposition for turning certificate vector;B represents biasing, offset of the hyperplane relative to origin
Amount.If interface has optimal solution, referred to as optimal hyperlane, and the discriminant function of optimal solution and optimal solution difference is as follows:
Wherein, the distance of two categorization vectors of certificate and non-certificate to optimal hyperlane is r, then:
It can be obtained in conjunction with the discriminant function of optimal solution before:
So can obtain, if it is desired that obtaining, r is maximum, then w must just make minimum.It is convex by Lagrange's equation building one
Function:
s.t y(i)(ωTx(i)+ b) >=1, i=1 ..., m
There is the convex function of above formula to construct Lagrangian are as follows:
It is as follows to w and b derivation respectively after Lagrangian:
Last interface formula is obtained in conjunction with Lagrangian and derivation formula:
Based on the above, positioning and judgement to certificate are completed, correct certificate region picture is filtered out.
S2, using neural network model trained in advance, identify the certificate character in the certificate region.
In order to effectively identify certificate, it is necessary first to design effective neural network structure.Due to certificate character
Several combinatorics on words mostly, thus depth it is suitable and and uncomplicated neural network structure can generate and accurate predict to tie
Fruit.Therefore it only needs to construct input layer, hidden layer and output layer.Input layer is the data entry of entire neural network, main
It is used to define different types of data input, other parts is facilitated to carry out quantification treatment;Hidden layer is for defeated to input layer
The data entered carry out nonlinear processing, and carrying out nonlinear fitting to input data based on excitation function can be effectively ensured
The predictive ability of model;Output layer is after hidden layer, is unique output of entire model, for the knot handled hidden layer
Fruit carries out output expression, i.e., final certificate number.
Certificate region picture is switched to the feature vector of neural network first, sets neural network threshold value.Because image is
It is made of pixel, so certificate region picture can be switched to orderly aligned bivector, the feature vector as neural network.
Determine the loss function and non-linear conversion equation of neural network.Loss function (loss function) be for
Estimate the inconsistent degree of the predicted value f (x) and true value Y of neural network model, it is a non-negative real-valued function.Lose letter
Number is smaller, and the robustness of model is better.It is bigger to being required in speed since digital recognition accuracy is very high, so of the invention
Using in quadratic loss function.Derive that quadratic loss function is first by central-limit theorem Sum Maximum Likelihood Estimate (MLE) method
Then a kind of loss function for assuming that sample and noise all Gaussian distributeds fits optimal straight line (i.e. each point to recurrence
The distance of straight line and the smallest straight line).Non-linear transfer function is added, is in order to allow deep-neural-network to become significant, such as
Fruit neural network does not have non-linear transfer function, then each layer network output is all linear function, such neural network, even if
What single layer network also can be achieved on.So non-linear transfer function is to make simulation more complicated, robustness is higher.It is non-linear
Transfer equation uses Sigmoid function:
Wherein, IiFor the input value of i-th of implicit unit.Based on the loss function and non-linear conversion for determining neural network
The characteristic value of certificate can be calculated in the determination of equation, deep learning network, so as to realize the character recognition of certificate.
S3, the information retrieved in the database and the certificate character matches, determine according to matching result and enter institute
State the identity information of the staff in place.
When staff enters the place for needing authentication, learning model captures and identifies the certificate of staff
Number, by comparing the database of typing, obtains staff corresponding to the certificate number, complete authentication.
In detail, in S3, the certificate character is switched to the binary number of database storage requirement, and called data library
The binary number having had, judges whether the two has identical binary number, when there is identical binary number, according to the phase
Same binary number determines the identity information for entering the staff in place.
Invention also provides a kind of certificate verifying device based on neural network model.It is real for the present invention one referring to shown in Fig. 2
The schematic diagram of internal structure of the certificate verifying device based on neural network model of example offer is provided.
In the present embodiment, the certificate verifying device 1 based on neural network model can be PC (Personal
Computer, PC) or the terminal devices such as smart phone, tablet computer, portable computer, it is also possible to one kind
Server etc..The certificate verifying device 1 based on neural network model includes at least memory 11, processor 12, communication bus
13 and network interface 14.
Wherein, memory 11 include at least a type of readable storage medium storing program for executing, the readable storage medium storing program for executing include flash memory,
Hard disk, multimedia card, card-type memory (for example, SD or DX memory etc.), magnetic storage, disk, CD etc..Memory 11
It can be the internal storage unit of the certificate verifying device 1 based on neural network model in some embodiments, such as this is based on
The hard disk of the certificate verifying device 1 of neural network model.Memory 11 is also possible in further embodiments based on nerve net
It is equipped on the External memory equipment of the certificate verifying device 1 of network model, such as the certificate verifying device 1 based on neural network model
Plug-in type hard disk, intelligent memory card (Smart Media Card, SMC), secure digital (Secure Digital, SD) card,
Flash card (Flash Card) etc..Further, memory 11 can also both include the certificate verification based on neural network model
The internal storage unit of device 1 also includes External memory equipment.Memory 11 can be not only used for storage and be installed on based on nerve
Application software and Various types of data, such as the code of certificate verification program 01 of certificate verifying device 1 of network model etc., can be with
For temporarily storing the data that has exported or will export.
Processor 12 can be in some embodiments a central processing unit (Central Processing Unit,
CPU), controller, microcontroller, microprocessor or other data processing chips, the program for being stored in run memory 11
Code or processing data, such as execute certificate verification program 01 etc..
Communication bus 13 is for realizing the connection communication between these components.
Network interface 14 optionally may include standard wireline interface and wireless interface (such as WI-FI interface), be commonly used in
Communication connection is established between the device 1 and other electronic equipments.
Optionally, which can also include user interface, and user interface may include display (Display), input
Unit such as keyboard (Keyboard), optional user interface can also include standard wireline interface and wireless interface.It is optional
Ground, in some embodiments, display can be light-emitting diode display, liquid crystal display, touch-control liquid crystal display and OLED
(Organic Light-Emitting Diode, Organic Light Emitting Diode) touches device etc..Wherein, display can also be appropriate
Referred to as display screen or display unit, for be shown in the information handled in the certificate verifying device 1 based on neural network model with
And for showing visual user interface.
Fig. 2 illustrates only the certificate based on neural network model with component 11-14 and certificate verification program 01 and tests
Card device 1, it will be appreciated by persons skilled in the art that structure shown in fig. 1 is not constituted to based on neural network model
The restriction of certificate verifying device 1 may include perhaps combining certain components or not than illustrating less perhaps more components
Same component layout.
In 1 embodiment of device shown in Fig. 2, certificate verification program 01 is stored in memory 11;Processor 12 executes
Following steps are realized when the certificate verification program 01 stored in memory 11:
Step 1: obtaining the staff's image for entering place, and detect the certificate area in staff's image
Domain.
For certificate detecting step including being certificate positioning, support vector machines (SVM) is trained to judge two processes with certificate.Entirely
In certificate detection process, the segment that many may be certificate can be obtained, whether is certificate by these segments progress manual sort, puts
Enter training in SVM model, obtains a judgment models of SVM.During actual certificate, we again be likely to be card
The segment of part inputs SVM judgment models, automatically selected by SVM model be actually actually certificate segment, complete card
Part detecting step.
It is handled firstly, for the work clothes comprising certificate that is installed with that camera takes by Gaussian Blur.Gaussian Blur
A kind of image processing techniques, effect is that the work clothes image that will be taken thickens.The present invention is fuzzy using mean-Gaussian
Method.
The value of each pixel of mean-Gaussian fuzz method takes the average value of all pixels (totally 8) around, and the left side is red
The pixel value of point was 2 originally, after fuzzy, just at 1 (mean value for taking all pixels around).Thus it completes to comprising certificate
Cleanness clothing Gaussian Blur processing.After Gaussian Blur is handled, immediately to image carry out gray processing processing, will comprising card
The color image of part becomes gray scale picture after gray processing is handled, it is meant that subsequent all operations are based not on color
Information reduces calculation amount.
Further, it to by Gaussian Blur processing and gray processing, treated that image carries out Sobel operator operation again, obtains
To the single order horizontal direction derivative of image.Soble operator principle is to ask image the level and vertical direction derivative of single order, according to
The size of derivative value is to determine whether be edge, and edge described herein, is exactly the possibility picture region of doubtful certificate.For
Convenience of calculation, Soble operator academicly claim there is no the method really gone derivation, but used the weighted sum of all boundary values
Make " convolution ".Weight is known as " convolution mask ".After Sobel operation, the picture region of certificate and doubtful certificate is by apparent area
It branches away.
Binarization operation is carried out to containing the picture region of certificate and doubtful certificate.In gray level image, each pixel
Value is the number between 0-255, represents gloomy degree.Adaptive threshold T is set at this time, if containing certificate and doubtful certificate
The value of picture region pixel is greater than threshold value T, is set as 1;If pixel value is less than threshold value T, it is set as 0.It is eventually converted into binary map
Picture, i.e., each pixel only has 1 and 0 two value may.After above all operations, certificate and doubtful card are finally found out
The profile of part.Since certificate size is typically small and rectangular, therefore setting length is 7 centimetres, and wide is 3 centimetres.It is all to meet this
The profile of a size is all retained, ungratified rejecting.By the double selection of Soble operator and size, certificate is finally obtained
Region and doubtful certificate region picture.
Certificate region and doubtful certificate region can be collectively referred to as candidate certificate picture, this step is by the picture one of candidate certificate
It is input in SVM model with opening, then utilizes the judgement of SVM model, if the answer of SVM model is not certificate picture, that
Next judgement is continued to, if it is certificate region, then picture is put into an output listing.Output listing be next
Step certificate character recognition is prepared.Support vector machines is to find an interface in the thing that entire step is done, and is put into card
Between two classifications of part and doubtful certificate, certificate and doubtful certificate are thus distinguished.So whole process need to look for interface
Process, wherein interface is also known as hyperplane, indicates certificate with following formula:
wTX+b=0
W is adjustable weight vector;T is the transposition for turning certificate vector;B represents biasing, offset of the hyperplane relative to origin
Amount.If interface has optimal solution, referred to as optimal hyperlane, and the discriminant function of optimal solution and optimal solution difference is as follows:
Wherein, the distance of two categorization vectors of certificate and non-certificate to optimal hyperlane is r, then:
It can be obtained in conjunction with the discriminant function of optimal solution before:
So can obtain, if it is desired that obtaining, r is maximum, then w must just make minimum.It is convex by Lagrange's equation building one
Function:
s.t y(i)(ωTx(i)+ b) >=1, i=1 ..., m
There is the convex function of above formula to construct Lagrangian are as follows:
It is as follows to w and b derivation respectively after Lagrangian:
Last interface formula is obtained in conjunction with Lagrangian and derivation formula:
Based on the above, positioning and judgement to certificate are completed, correct certificate region picture is filtered out.
Step 2: identifying the certificate character in the certificate region using neural network model trained in advance.
In order to effectively identify certificate, it is necessary first to design effective neural network structure.Due to certificate character
Several combinatorics on words mostly, thus depth it is suitable and and uncomplicated neural network structure can generate and accurate predict to tie
Fruit.Therefore it only needs to construct input layer, hidden layer and output layer.Input layer is the data entry of entire neural network, main
It is used to define different types of data input, other parts is facilitated to carry out quantification treatment;Hidden layer is for defeated to input layer
The data entered carry out nonlinear processing, and carrying out nonlinear fitting to input data based on excitation function can be effectively ensured
The predictive ability of model;Output layer is after hidden layer, is unique output of entire model, for the knot handled hidden layer
Fruit carries out output expression, i.e., final certificate number.
Certificate region picture is switched to the feature vector of neural network first, sets neural network threshold value.Because image is
It is made of pixel, so certificate region picture can be switched to orderly aligned bivector, the feature vector as neural network.
Determine the loss function and non-linear conversion equation of neural network.Loss function (loss function) be for
Estimate the inconsistent degree of the predicted value f (x) and true value Y of neural network model, it is a non-negative real-valued function.Lose letter
Number is smaller, and the robustness of model is better.It is bigger to being required in speed since digital recognition accuracy is very high, so of the invention
Using in quadratic loss function.Derive that quadratic loss function is first by central-limit theorem Sum Maximum Likelihood Estimate (MLE) method
Then a kind of loss function for assuming that sample and noise all Gaussian distributeds fits optimal straight line (i.e. each point to recurrence
The distance of straight line and the smallest straight line).Non-linear transfer function is added, is in order to allow deep-neural-network to become significant, such as
Fruit neural network does not have non-linear transfer function, then each layer network output is all linear function, such neural network, even if
What single layer network also can be achieved on.So non-linear transfer function is to make simulation more complicated, robustness is higher.It is non-linear
Transfer equation uses Sigmoid function:
Wherein, IiFor the input value of i-th of implicit unit.Based on the loss function and non-linear conversion for determining neural network
The characteristic value of certificate can be calculated in the determination of equation, deep learning network, so as to realize the character recognition of certificate.
Step 3: retrieve in the database with information that the certificate character matches, according to matching result determine into
Enter the identity information of the staff in the place.
When staff enters the place for needing authentication, deep learning model captures and identifies staff's
Certificate number obtains staff corresponding to the certificate number, completes authentication by comparing the database of typing.
In detail, in S3, the certificate character is switched to the binary number of database storage requirement, and called data library
The binary number having had, judges whether the two has identical binary number, when there is identical binary number, according to the phase
Same binary number determines the identity information for entering the staff in place.
Optionally, in other embodiments, certificate verification program can also be divided into one or more module, and one
Or multiple modules are stored in memory 11, and performed by one or more processors (the present embodiment is processor 12)
To complete the present invention, the so-called module of the present invention is the series of computation machine program instruction section for referring to complete specific function, is used
In implementation procedure of the description certificate verification program in the certificate verifying device based on neural network model.
For example, referring to shown in Fig. 3, for the present invention is based on the cards in one embodiment of certificate verifying device of neural network model
The program module schematic diagram of part proving program, in the embodiment, the certificate verification program can be divided into positioning certificate area
Domain module 10, identification certificate character module 20, verifying certificate character module 30, illustratively:
The positioning certificate regions module 10 is used for: being obtained the staff's image for entering place, and is detected the work
Make the certificate region in personnel's image.
The identification certificate character module 20 is used for: using neural network model trained in advance, identifying the certificate area
Certificate character in domain.
The verifying certificate character module module 30 is used for: retrieving in the database and the certificate character matches
Information determines the identity information for entering the staff in the place according to matching result.
The programs moulds such as above-mentioned positioning certificate regions module 10, identification certificate character module 20, verifying certificate character module 30
Block is performed realized functions or operations step and is substantially the same with above-described embodiment, and details are not described herein.
In addition, the embodiment of the present invention also proposes a kind of computer readable storage medium, the computer readable storage medium
On be stored with certificate verification program, the certificate verification program can be executed by one or more processors, to realize following operation:
The staff's image for entering place is obtained, and detects the certificate region in staff's image;
Using neural network model trained in advance, the certificate character in the certificate region is identified;
The information retrieved in the database and the certificate character matches, determines according to matching result and enters the field
Staff identity information.
Computer readable storage medium specific embodiment of the present invention and the above-mentioned certificate verification based on neural network model
Each embodiment of device and method is essentially identical, does not make tired state herein.
It should be noted that the serial number of the above embodiments of the invention is only for description, do not represent the advantages or disadvantages of the embodiments.And
The terms "include", "comprise" herein or any other variant thereof is intended to cover non-exclusive inclusion, so that packet
Process, device, article or the method for including a series of elements not only include those elements, but also including being not explicitly listed
Other element, or further include for this process, device, article or the intrinsic element of method.Do not limiting more
In the case where, the element that is limited by sentence "including a ...", it is not excluded that including process, device, the article of the element
Or there is also other identical elements in method.
Through the above description of the embodiments, those skilled in the art can be understood that above-described embodiment side
Method can be realized by means of software and necessary general hardware platform, naturally it is also possible to by hardware, but in many cases
The former is more preferably embodiment.Based on this understanding, technical solution of the present invention substantially in other words does the prior art
The part contributed out can be embodied in the form of software products, which is stored in one as described above
In storage medium (such as ROM/RAM, magnetic disk, CD), including some instructions are used so that terminal device (it can be mobile phone,
Computer, server or network equipment etc.) execute method described in each embodiment of the present invention.
The above is only a preferred embodiment of the present invention, is not intended to limit the scope of the invention, all to utilize this hair
Equivalent structure or equivalent flow shift made by bright specification and accompanying drawing content is applied directly or indirectly in other relevant skills
Art field, is included within the scope of the present invention.
Claims (10)
1. a kind of certificate verification method based on neural network model, which is characterized in that the described method includes:
The staff's image for entering place is obtained, and detects the certificate region in staff's image;
Using neural network model trained in advance, the certificate character in the certificate region is identified;
The information retrieved in the database and the certificate character matches, determines according to matching result and enters the place
The identity information of staff.
2. as described in claim 1 based on the certificate verification method of neural network model, which is characterized in that described to detect institute
The certificate region stated in staff's image includes:
The doubtful certificate region in staff's image is detected using digital image processing techniques, determines the doubtful card
The profile in part region obtains candidate certificate region;
The certificate region is filtered out from the candidate certificate region using mathematical model trained in advance.
3. as claimed in claim 2 based on the certificate verification method of neural network model, which is characterized in that the digital processing
Technology includes:
It is carried out using staff's image described in Gaussian Blur technology and gray processing technical treatment, and to staff's image
Sobel Operator operation obtains the doubtful certificate region.
4. the certificate verification method based on neural network model as described in any one of claims 1 to 3, feature exist
In identifying the certificate character in the certificate region using neural network model trained in advance, comprising:
Certificate region is switched to the feature vector of neural network;
Trained neural network passes through the feature vector in the doubtful certificate region of non-linear conversion equation calculation in advance, and passes through loss
The certificate character recognition of function output prediction goes out the character in certificate region.
5. as described in claim 1 based on the certificate verification method of neural network model, which is characterized in that described in database
Middle retrieval and the information that the certificate character matches, the body for entering the staff in the place is determined according to matching result
The step of part information includes:
The certificate character is switched into binary number;
The binary number that called data library has had both judges whether there is identical binary number, when there is identical binary system
When number, the identity information for entering the staff in place is determined according to the identical binary number.
6. a kind of certificate verifying device based on neural network model, which is characterized in that described device includes memory and processing
Device is stored with the certificate verification program that can be run on the processor on the memory, and the certificate verification program is by institute
It states when processor executes and realizes following steps:
The staff's image for entering place is obtained, and detects the certificate region in staff's image;
Using neural network model trained in advance, the certificate character in the certificate region is identified;
The information retrieved in the database and the certificate character matches, determines according to matching result and enters the place
The identity information of staff.
7. as claimed in claim 6 based on the certificate verifying device of neural network model, which is characterized in that described to detect institute
The certificate region stated in staff's image includes:
The doubtful certificate region in staff's image is detected using digital image processing techniques, determines the doubtful card
The profile in part region obtains candidate certificate region;
The certificate region is filtered out from the candidate certificate region using mathematical model trained in advance.
8. as claimed in claim 7 based on the certificate verifying device of neural network model, which is characterized in that the digital processing
Technology includes:
It is carried out using staff's image described in Gaussian Blur technology and gray processing technical treatment, and to staff's image
Sobel Operator operation obtains the doubtful certificate region.
9. the certificate verifying device based on neural network model as described in any one of claim 6 to 9, feature exist
In identifying that the certificate character in the certificate region includes: using neural network model trained in advance
Certificate region is switched to the feature vector of neural network;
Trained neural network passes through the feature vector in the doubtful certificate region of non-linear conversion equation calculation in advance, and passes through loss
The certificate character recognition of function output prediction goes out the character in certificate region.
10. a kind of computer readable storage medium, which is characterized in that be stored with certificate on the computer readable storage medium and test
Program is demonstrate,proved, the certificate verification program can be executed by one or more processor, to realize as any in claim 1 to 5
The step of certificate verification method based on neural network model described in item.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910658721.4A CN110503091A (en) | 2019-07-19 | 2019-07-19 | Certificate verification method, device and storage medium based on neural network model |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910658721.4A CN110503091A (en) | 2019-07-19 | 2019-07-19 | Certificate verification method, device and storage medium based on neural network model |
Publications (1)
Publication Number | Publication Date |
---|---|
CN110503091A true CN110503091A (en) | 2019-11-26 |
Family
ID=68586667
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910658721.4A Pending CN110503091A (en) | 2019-07-19 | 2019-07-19 | Certificate verification method, device and storage medium based on neural network model |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110503091A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111007064A (en) * | 2019-12-13 | 2020-04-14 | 常州大学 | Intelligent logging lithology identification method based on image identification |
-
2019
- 2019-07-19 CN CN201910658721.4A patent/CN110503091A/en active Pending
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111007064A (en) * | 2019-12-13 | 2020-04-14 | 常州大学 | Intelligent logging lithology identification method based on image identification |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN107633204B (en) | Face occlusion detection method, apparatus and storage medium | |
US10936973B1 (en) | Adversarial example detection method and apparatus, computing device, and non-volatile computer-readable storage medium | |
CN110728209B (en) | Gesture recognition method and device, electronic equipment and storage medium | |
CN107808120B (en) | Glasses localization method, device and storage medium | |
CN110097051A (en) | Image classification method, device and computer readable storage medium | |
CN106358444B (en) | Method and system for face verification | |
CN110363231A (en) | Abnormality recognition method, device and storage medium based on semi-supervised deep learning | |
EP4099217A1 (en) | Image processing model training method and apparatus, device, and storage medium | |
CN107633205B (en) | lip motion analysis method, device and storage medium | |
US9575566B2 (en) | Technologies for robust two-dimensional gesture recognition | |
CN107274543B (en) | A kind of recognition methods of bank note, device, terminal device and computer storage medium | |
CN111414888A (en) | Low-resolution face recognition method, system, device and storage medium | |
US20190228209A1 (en) | Lip movement capturing method and device, and storage medium | |
CN108564040B (en) | Fingerprint activity detection method based on deep convolution characteristics | |
Weinman et al. | Sign detection in natural images with conditional random fields | |
CN110135889A (en) | Method, server and the storage medium of intelligent recommendation book list | |
CN113139536A (en) | Text verification code identification method and equipment based on cross-domain meta learning and storage medium | |
CN108268641A (en) | Invoice information recognition methods and invoice information identification device, equipment and storage medium | |
CN111488798B (en) | Fingerprint identification method, fingerprint identification device, electronic equipment and storage medium | |
CN114448664B (en) | Method and device for identifying phishing webpage, computer equipment and storage medium | |
CN110503091A (en) | Certificate verification method, device and storage medium based on neural network model | |
CN110321883A (en) | Method for recognizing verification code and device, readable storage medium storing program for executing | |
CN109033797A (en) | A kind of authority setting method and device | |
TW201941018A (en) | Control method of fingerprint identification module being suitable for the fingerprint identification module and a control module including a classifier | |
TWI770947B (en) | Verification method and verification apparatus based on attacking image style transfer |
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 |