CN116959064B - Certificate verification method and device, computer equipment and storage medium - Google Patents
Certificate verification method and device, computer equipment and storage medium Download PDFInfo
- Publication number
- CN116959064B CN116959064B CN202310757031.0A CN202310757031A CN116959064B CN 116959064 B CN116959064 B CN 116959064B CN 202310757031 A CN202310757031 A CN 202310757031A CN 116959064 B CN116959064 B CN 116959064B
- Authority
- CN
- China
- Prior art keywords
- training
- face
- image
- verification
- certificate
- 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.)
- Active
Links
- 238000012795 verification Methods 0.000 title claims abstract description 166
- 238000000034 method Methods 0.000 title claims abstract description 47
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 82
- 238000001514 detection method Methods 0.000 claims abstract description 80
- 238000004364 calculation method Methods 0.000 claims abstract description 54
- 238000012549 training Methods 0.000 claims description 162
- 238000012360 testing method Methods 0.000 claims description 44
- 238000007781 pre-processing Methods 0.000 claims description 17
- 238000000605 extraction Methods 0.000 claims description 15
- 238000013528 artificial neural network Methods 0.000 claims description 14
- 238000004590 computer program Methods 0.000 claims description 14
- 238000012545 processing Methods 0.000 claims description 13
- 238000009825 accumulation Methods 0.000 claims description 12
- 230000008569 process Effects 0.000 claims description 11
- 230000000694 effects Effects 0.000 claims description 8
- 238000011156 evaluation Methods 0.000 claims description 7
- 238000010200 validation analysis Methods 0.000 claims description 6
- 102100029469 WD repeat and HMG-box DNA-binding protein 1 Human genes 0.000 claims description 5
- 101710097421 WD repeat and HMG-box DNA-binding protein 1 Proteins 0.000 claims description 5
- 238000013527 convolutional neural network Methods 0.000 claims description 5
- 238000013459 approach Methods 0.000 claims description 4
- 230000004927 fusion Effects 0.000 claims description 4
- 238000002372 labelling Methods 0.000 claims description 4
- 230000007246 mechanism Effects 0.000 claims description 4
- 238000010606 normalization Methods 0.000 claims description 4
- 238000005070 sampling Methods 0.000 claims description 4
- 238000013507 mapping Methods 0.000 claims description 3
- 238000005516 engineering process Methods 0.000 description 9
- 238000010586 diagram Methods 0.000 description 5
- 238000011176 pooling Methods 0.000 description 5
- 230000004044 response Effects 0.000 description 4
- 238000005242 forging Methods 0.000 description 2
- 238000011835 investigation Methods 0.000 description 2
- 238000012015 optical character recognition Methods 0.000 description 2
- 230000001360 synchronised effect Effects 0.000 description 2
- 230000009466 transformation Effects 0.000 description 2
- 208000037170 Delayed Emergence from Anesthesia Diseases 0.000 description 1
- 239000003086 colorant Substances 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 238000009826 distribution Methods 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 230000010365 information processing Effects 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 230000003068 static 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
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
- G06V40/16—Human faces, e.g. facial parts, sketches or expressions
- G06V40/172—Classification, e.g. identification
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F21/00—Security arrangements for protecting computers, components thereof, programs or data against unauthorised activity
- G06F21/30—Authentication, i.e. establishing the identity or authorisation of security principals
- G06F21/31—User authentication
- G06F21/32—User authentication using biometric data, e.g. fingerprints, iris scans or voiceprints
-
- 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
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/0464—Convolutional networks [CNN, ConvNet]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/74—Image or video pattern matching; Proximity measures in feature spaces
- G06V10/761—Proximity, similarity or dissimilarity measures
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/77—Processing image or video features in feature spaces; using data integration or data reduction, e.g. principal component analysis [PCA] or independent component analysis [ICA] or self-organising maps [SOM]; Blind source separation
- G06V10/774—Generating sets of training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks
-
- 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
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
- G06V40/16—Human faces, e.g. facial parts, sketches or expressions
- G06V40/161—Detection; Localisation; Normalisation
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Software Systems (AREA)
- Evolutionary Computation (AREA)
- Multimedia (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Artificial Intelligence (AREA)
- Computing Systems (AREA)
- Databases & Information Systems (AREA)
- Medical Informatics (AREA)
- General Engineering & Computer Science (AREA)
- Oral & Maxillofacial Surgery (AREA)
- Human Computer Interaction (AREA)
- Computer Security & Cryptography (AREA)
- Life Sciences & Earth Sciences (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Molecular Biology (AREA)
- Biophysics (AREA)
- Mathematical Physics (AREA)
- Biomedical Technology (AREA)
- Computer Hardware Design (AREA)
- Image Analysis (AREA)
Abstract
The application relates to a certificate verification method, a certificate verification device, computer equipment and a storage medium. The method comprises the following steps: responding to a certificate verification request, and acquiring a certificate image and a real acquisition image; carrying out validity verification on the certificate image by using a preset verification model, and outputting a verification result; when the verification result meets the preset condition, a first face in the certificate image and a second face in the actual acquisition image are obtained by using a first face detection algorithm; acquiring a first feature vector of the first face and a second feature vector of the second face by using a second face detection algorithm; obtaining a similarity calculation result between the first feature vector and the second feature vector by using a preset similarity calculation rule; and outputting a certificate verification result according to the similarity calculation result. By adopting the method, on the basis of replacing manual verification, whether the certificate is effective or not and whether uploading personnel are consistent with the certificate or not can be rapidly verified, and the single verification is avoided, so that the verification efficiency is improved.
Description
Technical Field
The present application relates to the field of information processing technologies, and in particular, to a certificate verification method, apparatus, computer device, and storage medium.
Background
With the increasing number of sites of application of identity documents, the phenomenon of counterfeiting of identity documents is expanding. The act of forging identity documents presents many social risks to the document parties.
Based on this, corresponding anti-fraud methods are currently offered on the market for the act of forging identity documents. Some application places utilize OCR optical character recognition technology or anti-counterfeiting characteristics of identity document to carry out true and false recognition on the identity document; some application sites conduct the true and false identification of identity documents by means of authorities; even some application sites are directly subjected to manual investigation.
However, the original copy of the identity document is not needed in many existing application places, only the identity verification is needed by providing the copy, the print, the copy or the document except the principal document, the existing anti-fraud method cannot effectively exclude the true or false of the copy, the print or the copy document, when verification equipment is needed, the additional cost of manpower and material resources is increased, when the assistance of an official institution is needed, the service can be developed after the effective communication with the official institution is needed in advance, and when the manual investigation is carried out, errors are easily generated due to the influence of artificial subjective factors, and the speed is low.
Disclosure of Invention
Based on the above, it is necessary to provide a certificate verification method, device, computer equipment and storage medium to solve the technical problems of low manual verification speed and large error in the existing verification singleization.
A method of credential validation, the method comprising:
responding to a certificate verification request, and acquiring a certificate image and a real acquisition image;
Carrying out validity verification on the certificate image by using a preset verification model, and outputting a verification result;
When the verification result meets the preset condition, a first face in the certificate image and a second face in the actual acquisition image are obtained by using a first face detection algorithm;
Acquiring a first feature vector of the first face and a second feature vector of the second face by using a second face detection algorithm;
Obtaining a similarity calculation result between the first feature vector and the second feature vector by using a preset similarity calculation rule;
And outputting a certificate verification result according to the similarity calculation result.
In one embodiment, prior to the step of responding to the credential verification request, the method comprises:
Responding to the model training request, acquiring image data related to a plurality of certificates to construct a data set;
Performing first preprocessing on partial image data in the data set to obtain first training data;
Inputting the first training data into a pre-built neural network architecture to perform model pre-training to obtain a pre-training result;
Performing second preprocessing on the image data remained in the data set according to the pre-training result to obtain second training data;
and inputting the second training data into the pre-trained model, and performing model training according to a preset training rule to obtain a verification model.
In one embodiment, the step of obtaining the first training data after performing the first preprocessing on the partial image data in the dataset includes:
labeling part of image data in the data set as effective credentials and ineffective credentials;
randomly sampling the marked image data, and constructing a training set and a testing set according to the proportion to serve as first training data.
In one embodiment, the step of inputting the first training data into the pre-built neural network architecture to perform model pre-training to obtain a pre-training result includes:
Performing model pre-training on the pre-built neural network architecture by using image data in a training set, and performing model testing by using the image data in a testing set to obtain a testing result;
And when the evaluation index exceeds the preset ratio in the test result, acquiring the weight value of the current pre-training model to be used as a pre-training result.
In one embodiment, the step of performing a second preprocessing on the image data remaining in the dataset according to the pre-training result to obtain second training data includes:
predicting the residual image data in the data set by using the weight value to obtain a prediction result;
and marking the rest image data by taking the predicted result after manual verification as a label to obtain second training data, and dividing the second training data into a training set and a testing set according to a proportion.
In one embodiment, the step of inputting the second training data into the pre-trained model and training the model according to a preset training rule to obtain the verification model includes:
dividing the training set into a plurality of sub-batches as a whole batch size so as to perform model training;
carrying out gradient iterative computation on each sub-batch, and then carrying out gradient superposition;
Based on a preset accumulation factor, after the number of sub-batches is accumulated to a preset count value, gradient updating processing is carried out, a gradient accumulator is emptied, the next round of accumulation is started, and training is stopped until the preset training round number is reached;
And similarly, training continuously by using training data in the training set, storing weight values obtained by training, and evaluating the image description effect by using the test set until the index of the test set meets the requirement, thereby obtaining a verification model meeting the requirement.
In one embodiment, the step of verifying the validity of the certificate image by using a preset verification model and outputting a verification result includes:
performing third preprocessing on the certificate image according to a preset image processing rule;
Inputting the third preprocessed certificate image into the verification model, and outputting a verification score of the certificate image so as to judge that the certificate image meets the preset condition according to the fact that the verification score is higher than a preset score threshold.
In one embodiment, the first face detection algorithm is a SCRFD face detection algorithm, and the SCRFD face detection algorithm at least comprises a sparse convolution algorithm, an FPN structure, an Anchor-Free algorithm and a IoU-Net algorithm;
the method for acquiring the first face in the certificate image and the second face in the actual acquisition image by using the first face detection algorithm comprises the following steps:
Carrying out multi-scale face feature extraction on the certificate image by using a sparse convolution algorithm to obtain face features with different scales in the certificate image;
Carrying out multi-scale fusion on face features with different scales by using an FPN structure, and emphasizing the features by using an attention mechanism to obtain a feature set;
Carrying out face frame detection on the feature set by using an Anchor-Free algorithm, predicting the position and the size of the face frame by using density regression points, and predicting the intersection ratio between the face frame and a true value by using a IoU-Net algorithm so as to correct the position and the size of the face frame and obtain a first face;
And similarly, carrying out face detection on the actual acquisition image through SCRFD face detection algorithm to obtain a second face in the actual acquisition image.
In one embodiment, the second face detection algorithm adopts FaceNet algorithm, faceNet algorithm includes a convolutional neural network of Inception structure and European space;
The step of obtaining a first feature vector of the first face and a second feature vector of the second face by using a second face detection algorithm comprises the following steps:
Based on a face coordinate system, carrying out normalization processing by utilizing coordinate data of each key point in the first face;
extracting depth features of the first face by utilizing Inception structures;
Mapping the extracted depth features into an European space to obtain a first feature vector of a first face;
And so on, obtaining a second feature vector of the second face.
In one embodiment, the similarity calculation rule uses a cosine similarity relation, where the cosine similarity relation is:
Wherein a i represents the corresponding element value of the first feature vector a, and B i represents the corresponding element value of the second feature vector B;
the step of obtaining a similarity calculation result between the first feature vector and the second feature vector by using a preset similarity calculation rule comprises the following steps:
and obtaining similarity values between-1 and-1 through cosine similarity relational expression calculation so that the similarity values approach to 1, and representing that the certificate image is similar to the actual acquired image.
In one embodiment, the step of outputting the certificate verification result according to the similarity calculation result includes:
comparing the similarity value with a preset similarity threshold value, and outputting a comparison result;
if the similarity value is smaller than the similarity threshold value in the comparison result, judging that the certificate image and the actual acquisition image are different people.
A credential verification device, the device comprising: the device comprises an image acquisition module, an effective verification module, a face detection module, a feature extraction module, a similar calculation module and a result output module, wherein,
The image acquisition module is used for responding to the certificate verification request to acquire a certificate image and an actual acquisition image;
The effective verification module is used for verifying the effectiveness of the certificate image by utilizing a preset verification model and outputting a verification result;
the face detection module is used for acquiring a first face in the certificate image and a second face in the actual acquisition image by utilizing a first face detection algorithm when the verification result meets a preset condition;
the feature extraction module is used for acquiring a first feature vector of the first face and a second feature vector of the second face by using a second face detection algorithm;
The similarity calculation module is used for obtaining a similarity calculation result between the first feature vector and the second feature vector by utilizing a preset similarity calculation rule;
the result output module is used for outputting a certificate verification result according to the similarity calculation result.
A computer device comprising a memory storing a computer program and a processor which when executing the computer program performs the steps of:
responding to a certificate verification request, and acquiring a certificate image and a real acquisition image;
Carrying out validity verification on the certificate image by using a preset verification model, and outputting a verification result;
When the verification result meets the preset condition, a first face in the certificate image and a second face in the actual acquisition image are obtained by using a first face detection algorithm;
Acquiring a first feature vector of the first face and a second feature vector of the second face by using a second face detection algorithm;
Obtaining a similarity calculation result between the first feature vector and the second feature vector by using a preset similarity calculation rule;
And outputting a certificate verification result according to the similarity calculation result.
A computer readable storage medium having stored thereon a computer program which when executed by a processor performs the steps of:
responding to a certificate verification request, and acquiring a certificate image and a real acquisition image;
Carrying out validity verification on the certificate image by using a preset verification model, and outputting a verification result;
When the verification result meets the preset condition, a first face in the certificate image and a second face in the actual acquisition image are obtained by using a first face detection algorithm;
Acquiring a first feature vector of the first face and a second feature vector of the second face by using a second face detection algorithm;
Obtaining a similarity calculation result between the first feature vector and the second feature vector by using a preset similarity calculation rule;
And outputting a certificate verification result according to the similarity calculation result.
The certificate verification method, the certificate verification device, the computer equipment and the storage medium comprise the following technical effects:
(1) Because the step of verifying the validity of the certificate image by using a preset verification model and outputting a verification result is adopted, when the certificate image is judged to be an invalid certificate, the output result of the certificate verification can be that the certificate fraud risk exists.
(2) Because the step of judging whether the certificate image and the actual image are the same person or not is adopted when the certificate image is the effective certificate, when the output result of the certificate verification is judged to be normal, namely no certificate fraud operation is carried out; when the certificate image and the actual image are different people, the output result of certificate verification can be that certificate fraud exists.
(3) The combination of the three technologies of effective certificate detection, face detection and face comparison is adopted, so that the application of multiple detection technologies is combined, certificate fraud is further reduced, authenticity of a certificate can be detected and verified more comprehensively, detection evaluation indexes and safety are improved, and errors and labor cost in effective certificate identification due to a single technology are reduced.
(4) The adopted verification model is updated continuously by using training data, so that the verification model can keep the effect improved; in the actual use process, diversified samples are continuously added, and the verification model can flexibly cope with various conditions.
(5) The method can replace manual verification, avoid single verification, greatly reduce manual subjective influence, improve verification speed, and treat about 8.8 anti-evidence fraud tasks per second in the actual operation process using household hardware equipment, which is far higher than manual processing speed.
Drawings
FIG. 1 is a diagram of an application environment for a credential verification method in one embodiment;
FIG. 2 is a block diagram of the architecture of a credential verification device in one embodiment;
FIG. 3 is an internal block diagram of a computer device in one embodiment.
Detailed Description
The present application will be described in further detail with reference to the drawings and examples, in order to make the objects, technical solutions and advantages of the present application more apparent. It should be understood that the specific embodiments described herein are for purposes of illustration only and are not intended to limit the scope of the application.
In one embodiment, as shown in FIG. 1, a method of credential verification is provided, comprising the steps of:
step S100, in response to the certificate verification request, acquiring a certificate image and an actual acquisition image.
Step S200, verifying the validity of the certificate image by using a preset verification model, and outputting a verification result.
Step S300, when the verification result meets the preset condition, a first face in the certificate image and a second face in the actual acquisition image are obtained by using a first face detection algorithm.
Step S400, a first feature vector of the first face and a second feature vector of the second face are obtained by using a second face detection algorithm.
Step S500, obtaining a similarity calculation result between the first feature vector and the second feature vector by using a preset similarity calculation rule.
Step S600, outputting a certificate verification result according to the similarity calculation result.
In one embodiment, in step S100, in the step of acquiring a document image and an actual image in response to a document authentication request, the document image may be represented as a document image representing an identity. In the actual operation process, one certificate image can be adopted for model training, and multiple certificate images can also be adopted for model training. The actual acquisition image is shown as a live photo containing a human face acquired in the actual operation process.
In one embodiment, step S100, prior to the step of responding to the credential verification request, includes steps 110-150.
Step S110, responding to the model training request, acquiring a plurality of certificate related image data to construct a data set. The image data comprises the image data of the effective certificate and the image data of the ineffective certificate which are acquired in the active uploading and internet public data of the user, and the acquired image data are integrated together to form a data set. In one embodiment, the image data in the dataset includes two types, valid credentials and invalid credentials, 3000 sheets of valid credentials and 3500 sheets of invalid credentials. The image data of the invalid credentials comprise a copy, a print and a reproduction, wherein 800 copies, 900 prints and 1000 reproduction are printed, the image data of the invalid credentials are marked as the invalid credentials in the early stage of training, and the invalid credentials do not need to be marked independently, and in the training process, the training data set is subjected to image enhancement processing.
Step S120, after performing first preprocessing on part of image data in the dataset, obtaining first training data.
In one embodiment, step S120, after performing a first preprocessing on a part of image data in the dataset, obtains first training data, including:
and step S121, labeling part of image data in the data set as valid credentials and invalid credentials.
Step S122, randomly sampling the marked image data, and constructing a training set and a testing set according to the proportion to serve as first training data.
In one embodiment, the image data in the training set is subjected to a random transformation of color, brightness, contrast, rotation, flipping, and random cropping, and then the transformed image data is padded to the size required by the model training input.
In one embodiment, the data set is divided into a training set and a test set by random sampling, wherein the training set accounts for 70% -80% of the data set, the test set accounts for 20% -30% of the data set, for example, the training set accounts for 80% of the data set, and the test set accounts for 20% of the data set. Further illustratively, the data set is typically divided into a training set, a validation set, and a test set. The training set is used for training the model, the verification set is used for adjusting parameters and selecting the model, the test set is used for reporting the performance of the final model, but in actual operation, the number of invalid certificates, especially copies, prints and turnups, which can be acquired is small, and in order to more fully utilize data resources, the test set in the embodiment is used as both the verification set and the test set.
In one embodiment, the test set performs as shown in Table 1 below.
| Precision rate | Recall rate of recall | F 1-fraction | |
| Invalidation of | 0.98 | 0.93 | 0.95 |
| Effective and effective | 0.94 | 0.99 | 0.96 |
TABLE 1
In the actual operation process, the face comparison method after the face detection of the open source is tested by using the test set, and the evaluation indexes are all up to 0.94.
Further, in this embodiment, the first preprocessing is performed on a portion of the image data in the data set to label the image data of the valid certificate and the invalid certificate pair, and then the labeled image data is randomly sampled in proportion, so as to obtain the training set and the test set to be trained.
Step S130, inputting the first training data into a pre-built neural network architecture to perform model pre-training, and obtaining a pre-training result.
In one embodiment, step S130, inputting the first training data into the pre-built neural network architecture to perform model pre-training, and obtaining the pre-training result includes:
Step S131, pre-training the model of the pre-built neural network architecture by using the image data in the training set, and performing a model test by using the image data in the testing set to obtain a test result.
In one embodiment, the pre-built neural network architecture includes an input layer, a convolution layer, a pooling layer, a convolution block in the convolution layer, a global average pooling layer, a full connection layer, an output layer, and the like, as shown in table 2 below, a schematic structural diagram of the neural network architecture is provided.
TABLE 2
Conv1 represents a convolution layer for extracting low-level features of an input picture; maxPooling denotes a pooling layer for downsampling and feature dimension reduction; conv2_x, conv3_x, conv4_x, conv5_x represent convolution blocks composed of several residual blocks, each convolution block containing multiple convolution layers for progressively extracting advanced features of the image data; global average pooling layer: the method comprises the steps of compressing the height and width dimensions of a feature map to be 1, so that global feature representation of each feature channel is obtained; full tie layer: for mapping the global feature representation onto the output class; softmax output for outputting probability distributions for each category.
Step S132, when the evaluation index exceeds the preset ratio in the test result, the weight value of the current pre-training model is obtained as the pre-training result.
In one embodiment, 600 valid certificates and 600 invalid certificates are selected and distributed in a training set and a testing set, for example, 400 training sets and 200 testing sets, model pre-training is performed on a pre-built neural network architecture by using the training set, meanwhile, the currently generated weight is evaluated by using the testing set, training is stopped when the evaluation indexes of the training set and the testing set reach a preset ratio, for example, the preset ratio is 85%, and the weight value of the currently trained model is obtained so as to label the rest unlabeled image data.
And step S140, performing second preprocessing on the image data remained in the data set according to the pre-training result to obtain second training data.
In one embodiment, step S140, performing a second preprocessing on the image data remaining in the dataset according to the pre-training result, to obtain second training data, includes:
And step S141, predicting the residual image data in the dataset by using the weight value to obtain a prediction result. In one embodiment, the remaining unlabeled image data is predicted using the weight values of the current model.
And S142, marking the rest image data by taking the predicted result after manual verification as a label to obtain second training data, and dividing the second training data into a training set and a testing set according to a proportion.
Further, because of uncertainty of the prediction result, in this embodiment, the remaining image data is further labeled by means of a manual verification process, for example, after the prediction process is performed on the remaining image data by using a prediction technology, the remaining image data is labeled by using the verified prediction result as a label, so that the labeling time is reduced.
Step S150, inputting second training data into the pre-trained model, and training the model according to a preset training rule to obtain a verification model.
In one embodiment, step S150, inputting the second training data into the pre-trained model, and training the model according to the preset training rule to obtain the verification model, includes:
In step S151, the training set is divided into a plurality of sub-batches as the whole batch size, so as to perform model training.
Step S152, performing gradient iterative computation on each sub-batch, and then performing gradient superposition.
Step S153, based on a preset accumulation factor, after the number of sub-batches is accumulated to a preset count value, gradient updating processing is performed, a gradient accumulator is emptied, the next round of accumulation is started, and training is stopped until the preset number of training rounds is reached.
Step S154, and the like, training is continuously carried out by utilizing training data in a training set, weight values obtained by training are stored, and the effect of image description is evaluated by utilizing a testing set until the index of the testing set meets the requirement, and then a verification model meeting the requirement is obtained.
Further, in this embodiment, the image data in the training set is used as a whole batch, then the gradient calculation of the whole batch is split into the gradient calculation of a plurality of sub-batches, then the gradients of the sub-batches are added, and the gradients are updated after being accumulated to a certain number, so that the memory requirement can be effectively reduced, and meanwhile, the stability and accuracy of the model are improved.
The specific method can comprise the following steps: (1) An accumulation factor is preset, and in one embodiment, is set to 4 to indicate the number of sub-lots to be accumulated. (2) The whole training set is divided into a plurality of sub-batches, and the number of each sub-batch is the original batch number/accumulation factor. In one embodiment, the number of batches is preset to 64. (3) And iteratively training the sub-batches, and accumulating the gradient of each sub-batch. (4) When the accumulated number of sub-batches reaches the accumulation factor, gradient updating operation is carried out, the gradient accumulator is emptied, and the next round of accumulation is started. (5) stopping training when the designated training round number is reached.
Further, in this embodiment, the training set is used to train continuously, and the neural network weight value obtained by training is stored at the same time, and the test set is used to evaluate the image description effect until the index of the test set meets the requirement, and the corresponding weight value is applied to the model to obtain the verification model meeting the requirement. Inputting the certificate image into a verification model in real time, and outputting the score of the image data through the verification model to judge that the certificate image is a valid certificate.
Step S200, verifying the validity of the certificate image by using a preset verification model, and outputting a verification result.
In one embodiment, step S200, performing validity verification on the document image by using a preset verification model, and outputting a verification result, includes:
step S210, third preprocessing is carried out on the certificate image according to preset image processing rules.
In one embodiment, the document image is subjected to a random transformation of rotation, flipping, and random cropping, and then the transformed document image is trimmed to the size required for verification model input.
Step S220, inputting the third preprocessed certificate image into the verification model, and outputting a verification score of the certificate image so as to judge that the certificate image meets the preset condition according to the fact that the verification score is higher than a preset score threshold.
Step S300, when the verification result meets the preset condition, a first face in the certificate image and a second face in the actual acquisition image are obtained by using a first face detection algorithm.
In one embodiment, the first face detection algorithm employs SCRFD face detection algorithms, SCRFD face detection algorithms include at least a sparse convolution algorithm, FPN structure, anchor-Free algorithm, and IoU-Net algorithm.
In one embodiment, step S300, when the verification result meets the preset condition, uses a first face detection algorithm to obtain a first face in the certificate image and a second face in the actual acquired image, includes:
Step S310, carrying out multi-scale face feature extraction on the certificate image by using a sparse convolution algorithm to obtain face features with different scales in the certificate image.
Step S320, performing multi-scale fusion on the face features with different scales by using the FPN structure, and emphasizing the features by using an attention mechanism to obtain a feature set.
And S330, carrying out face frame detection on the feature set by using an Anchor-Free algorithm, predicting the position and the size of the face frame by using density regression points, and predicting the intersection ratio between the face frame and the true value by using a IoU-Net algorithm so as to correct the position and the size of the face frame and obtain the first face.
Step S340, and the like, performing face detection on the actual acquisition image through SCRFD face detection algorithm to obtain a second face in the actual acquisition image.
Further describing, in this embodiment, an open source SCRFD face detection algorithm (SCRFD, full scale Sparse Convolutional Regression Face Detector) is used to perform face detection and key point detection on the document image and the actual image, and screenshot is performed on the face region. SCRFD the face detection algorithm is a highly efficient face detection algorithm, which includes the following aspects: feature extraction, feature fusion and detection head.
When the features are extracted, the SCRFD face detection algorithm adopts a sparse convolution algorithm to extract the face features on multiple scales. The sparse convolution algorithm has few non-zero elements in the convolution kernel, so that the calculated amount can be remarkably reduced. When the features are fused, the SCRFD face detection algorithm integrates the features with different scales by adopting an FPN structure (Feature Pyramid Network), and uses an attention mechanism to emphasize important features so as to fuse the features extracted on different scales. In the detection head, SCRFD face detection algorithm adopts Anchor-Free algorithm to detect the face, does not need to predefine Anchor Box, and predicts the position and the size of the face frame through dense density regression points. Meanwhile, the SCRFD face detection algorithm also uses IoU-Net algorithm to predict the cross-over ratio between the detection frame and the true value, and is used for correcting the position and the size of the detection frame. In actual operation, the SCRFD face detection pre-training trained by using WIDERFACE data sets performs well on the self-owned certificate data sets, so that the pre-training model can be directly used for carrying out face screenshot on the certificate images and the actual acquired images.
In one embodiment, the second face detection algorithm employs FaceNet algorithm, faceNet algorithm comprising a convolutional neural network of Inception structure and European space. Step S400, a step of acquiring a first feature vector of a first face and a second feature vector of a second face by using a second face detection algorithm, including:
In step S410, normalization processing is performed by using the coordinate data of each key point in the first face based on the face coordinate system.
Step S420, depth feature extraction is performed on the first face by utilizing Inception structures.
Step S430, the extracted feature information is mapped to European space to obtain a first feature vector of the first face.
Step S440, and so on, obtain the second feature vector of the second face.
Further describing, in this embodiment, the face feature extraction is performed on the first face of the evidence image and the second face of the actual acquired image by using an open-source FaceNet algorithm. The principle of FaceNet algorithm is that the face image is mapped to a low-dimensional Euclidean space, so that the face images of the same person are closer in the same Euclidean space, and the face images of different persons are farther. Implementation of FaceNet algorithm includes: face detection and alignment, depth feature extraction and feature vector output.
And when the faces are detected and aligned, detecting the face key points of the first face/the second face, correcting the rotation of the first face/the second face by utilizing the coordinates of the face key points, scaling to a specified size, and normalizing the image. When the depth features are extracted, the depth feature extraction is performed by using a Inception-structure convolutional neural network, wherein the Inception-structure convolutional neural network comprises a plurality of convolutional layers, pooling layers, full-connection layers, normalization layers and the like, and the features such as edges, textures, colors and the like in the first face/the second face can be extracted. When the feature vector is output, the extracted depth feature is mapped to a low-dimensional European space through the full-connection layer, and a first feature vector of the first face and a second feature vector of the second face are obtained.
In one embodiment, the difference between the first feature vector and the second feature vector is calculated by a cosine similarity relation. Namely, the similarity calculation rule adopts a cosine similarity relation, and the cosine similarity relation is as follows:
Where a i represents the corresponding element value of the first feature vector a and B i represents the corresponding element value of the second feature vector B.
Step S500, a step of obtaining a similarity calculation result between the first feature vector and the second feature vector by using a preset similarity calculation rule, including:
And obtaining similarity values between-1 and-1 through cosine similarity relational computation, and obtaining similarity values between-1 and-1 through cosine similarity relational computation so that the similarity values approach 1, and representing that the certificate image is similar to the actual acquired image.
Further, a similarity value between-1 and-1 is obtained through cosine similarity relation calculation, wherein-1 represents that the directions of the first feature vector and the second feature vector are opposite, 1 represents that the directions of the first feature vector and the second feature vector are the same, and when the similarity value approaches to 1, the more similar the certificate image and the actual acquired image are.
In one embodiment, step S600, the step of outputting the certificate verification result according to the similarity calculation result, includes:
In step S610, the similarity value is compared with a preset similarity threshold, and a comparison result is output.
In step S620, if the similarity value is smaller than the similarity threshold in the comparison result, it is determined that the document image and the actual image are different people.
Further, the similarity threshold is t, and if the similarity threshold is smaller than t, it is determined that the document image and the actual image are not the same person.
In this embodiment, from step 100 to step 600, functions of verification model, face detection, feature extraction, similarity comparison, etc. are integrated, and after each functional module is connected in series, an end-to-end output from inputting an image to obtaining a final result is formed.
In the certificate verification method, a certificate image and a real acquisition image are acquired in response to a certificate verification request; and carrying out validity verification on the certificate image by using a preset verification model, and outputting a verification result. If the document image is determined to be an invalid document, the output of the document verification may be that there is a risk of document fraud.
If the certificate image is a valid certificate and the certificate image and the actual image are the same person, the output result of certificate verification is judged to be normal, namely no certificate fraud operation is performed; when the certificate image and the actual image are different people, the output result of certificate verification can be that certificate fraud exists.
The combination of three technologies of effective certificate detection, face detection and face comparison enables the application of multiple detection technologies to be combined, further reduces certificate fraud, can detect and verify authenticity of a certificate more comprehensively, improves detection evaluation indexes and safety, and reduces errors and labor cost in effective certificate identification due to a single technology.
The adopted verification model is updated continuously by using training data, so that the verification model can keep the effect improved; in the actual use process, diversified samples are continuously added, and the verification model can flexibly cope with various conditions.
The method can replace manual verification, avoid single verification, greatly reduce manual subjective influence, improve verification speed, and treat about 8.8 anti-evidence fraud tasks per second in the actual operation process using household hardware equipment, which is far higher than manual processing speed.
It should be understood that the steps in the flowchart are shown in order as indicated by the arrows, but the steps are not necessarily performed in order as indicated by the arrows. The steps are not strictly limited to the order of execution unless explicitly recited herein, and the steps may be executed in other orders. Moreover, at least some of the steps in fig. 1 may include multiple sub-steps or stages that are not necessarily performed at the same time, but may be performed at different times, nor do the order in which the sub-steps or stages are performed necessarily performed in sequence, but may be performed alternately or alternately with at least a portion of other steps or sub-steps of other steps.
In one embodiment, as shown in FIG. 2, there is provided a credential verification device comprising: an image acquisition module 100, a validation module 200, a face detection module 300, a feature extraction module 400, a similarity calculation module 500, and a result output module 600, wherein,
The image acquisition module 100 is configured to acquire a document image and an actual image in response to a document authentication request.
The validity verification module 200 is configured to perform validity verification on the document image by using a preset verification model, and output a verification result.
The face detection module 300 is configured to acquire a first face in the document image and a second face in the actual acquired image by using a first face detection algorithm when the verification result meets a preset condition.
The feature extraction module 400 is configured to obtain a first feature vector of a first face and a second feature vector of a second face by using a second face detection algorithm.
The similarity calculation module 500 is configured to obtain a similarity calculation result between the first feature vector and the second feature vector by using a preset similarity calculation rule.
The result output module 600 is configured to output a certificate verification result according to the similarity calculation result.
For specific limitations on the credential verification device, reference may be made to the limitations of the credential verification method hereinabove, and will not be described in detail herein. The various modules in the credential verification device described above can be implemented in whole or in part in software, hardware, and combinations thereof. The above modules may be embedded in hardware or may be independent of a processor in the computer device, or may be stored in software in a memory in the computer device, so that the processor may call and execute operations corresponding to the above modules.
In one embodiment, a computer device is provided, which may be a server, the internal structure of which may be as shown in fig. 3. The computer device includes a processor, a memory, a network interface, and a database connected by a system bus. Wherein the processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, computer programs, and a database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The database of the computer device is for storing credential verification data. The network interface of the computer device is used for communicating with an external terminal through a network connection. The computer program is executed by a processor to implement a credential verification method.
It will be appreciated by those skilled in the art that the structure shown in FIG. 3 is merely a block diagram of some of the structures associated with the present inventive arrangements and is not limiting of the computer device to which the present inventive arrangements may be applied, and that a particular computer device may include more or fewer components than shown, or may combine some of the components, or have a different arrangement of components.
In one embodiment, a computer device is provided comprising a memory and a processor, the memory having stored therein a computer program, the processor when executing the computer program performing the steps of:
responding to a certificate verification request, and acquiring a certificate image and a real acquisition image; carrying out validity verification on the certificate image by using a preset verification model, and outputting a verification result; when the verification result meets the preset condition, a first face in the certificate image and a second face in the actual acquisition image are obtained by using a first face detection algorithm; acquiring a first feature vector of the first face and a second feature vector of the second face by using a second face detection algorithm; obtaining a similarity calculation result between the first feature vector and the second feature vector by using a preset similarity calculation rule; and outputting a certificate verification result according to the similarity calculation result.
In one embodiment, a computer readable storage medium is provided having a computer program stored thereon, which when executed by a processor, performs the steps of:
responding to a certificate verification request, and acquiring a certificate image and a real acquisition image; carrying out validity verification on the certificate image by using a preset verification model, and outputting a verification result; when the verification result meets the preset condition, a first face in the certificate image and a second face in the actual acquisition image are obtained by using a first face detection algorithm; acquiring a first feature vector of the first face and a second feature vector of the second face by using a second face detection algorithm; obtaining a similarity calculation result between the first feature vector and the second feature vector by using a preset similarity calculation rule; and outputting a certificate verification result according to the similarity calculation result.
Those skilled in the art will appreciate that implementing all or part of the above described methods may be accomplished by way of a computer program stored on a non-transitory computer readable storage medium, which when executed, may comprise the steps of the embodiments of the methods described above. Any reference to memory, storage, database, or other medium used in embodiments provided herein may include non-volatile and/or volatile memory. The nonvolatile memory can include Read Only Memory (ROM), programmable ROM (PROM), electrically Programmable ROM (EPROM), electrically Erasable Programmable ROM (EEPROM), or flash memory. Volatile memory can include Random Access Memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms such as Static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double Data Rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (SYNCHLINK) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), among others.
The technical features of the above embodiments may be arbitrarily combined, and all possible combinations of the technical features in the above embodiments are not described for brevity of description, however, as long as there is no contradiction between the combinations of the technical features, they should be considered as the scope of the description.
The above examples illustrate only a few embodiments of the application, which are described in detail and are not to be construed as limiting the scope of the application. It should be noted that it will be apparent to those skilled in the art that several variations and modifications can be made without departing from the spirit of the application, which are all within the scope of the application. Accordingly, the scope of protection of the present application is to be determined by the appended claims.
Claims (12)
1. A method of validating a document, the method comprising:
responding to a certificate verification request, and acquiring a certificate image and a real acquisition image;
carrying out validity verification on the certificate image by using a preset verification model, and outputting a verification result;
when the verification result meets a preset condition, a first face in the certificate image and a second face in the actual acquisition image are obtained by using a first face detection algorithm; wherein, the first face detection algorithm adopts SCRFD face detection algorithm;
Acquiring a first feature vector of the first face and a second feature vector of the second face by using a second face detection algorithm; wherein, the second face detection algorithm adopts FaceNet algorithm;
obtaining a similarity calculation result between the first feature vector and the second feature vector by using a preset similarity calculation rule;
Outputting a certificate verification result according to the similarity calculation result;
Wherein, prior to the step of responding to the credential verification request, comprising:
Responding to the model training request, acquiring image data related to a plurality of certificates to construct a data set;
After carrying out first preprocessing on part of the image data in the data set, constructing a training set and a testing set, and obtaining first training data;
inputting the first training data into a pre-built neural network architecture to perform model pre-training to obtain a pre-training result;
Performing second preprocessing on the image data remained in the data set according to the pre-training result to obtain second training data;
Inputting the second training data into the pre-trained model, and performing model training according to a preset training rule to obtain a verification model;
the step of inputting the second training data into the pre-trained model and training the model according to a preset training rule to obtain a verification model comprises the following steps:
Dividing the training set into a plurality of sub-batches as a whole batch size so as to perform model training;
Carrying out gradient iterative computation on each sub-batch, and then carrying out gradient superposition;
Based on a preset accumulation factor, after the number of sub-batches is accumulated to a preset count value, gradient updating processing is carried out, a gradient accumulator is emptied, the next round of accumulation is started, and training is stopped until the preset training round number is reached;
And similarly, training continuously by using training data in the training set, storing weight values obtained by training, and evaluating the image description effect by using the test set until the test set index meets the requirement, thereby obtaining a verification model meeting the requirement.
2. The method of claim 1, wherein the step of obtaining first training data after performing a first preprocessing on a portion of the image data in the dataset includes:
Labeling part of the image data in the data set as effective credentials and ineffective credentials;
Randomly sampling the noted image data, and constructing a training set and a testing set according to proportion to serve as first training data.
3. The method of claim 2, wherein the step of inputting the first training data into a pre-built neural network architecture for model pre-training to obtain a pre-training result comprises:
Performing model pre-training on a pre-built neural network architecture by using the image data in the training set, and performing model testing by using the image data in the testing set to obtain a testing result;
And when the evaluation index exceeds the preset ratio in the test result, acquiring the weight value of the current pre-training model to be used as a pre-training result.
4. A method of validating a document according to claim 3, wherein the step of performing a second pre-process on the image data remaining in the dataset based on the pre-training results to obtain second training data comprises:
Predicting the rest image data in the dataset by using the weight value to obtain a prediction result;
and marking the rest image data by taking the predicted result after manual verification as a label to obtain second training data, and dividing the second training data into the training set and the testing set according to a proportion.
5. The method for verifying the certificate as set forth in claim 1, wherein the step of verifying the validity of the certificate image by using a preset verification model and outputting a verification result includes:
Performing third preprocessing on the certificate image according to a preset image processing rule;
Inputting the third preprocessed certificate image into the verification model, and outputting a verification score of the certificate image so as to judge that the certificate image meets a preset condition when the verification score is higher than a preset score threshold.
6. The credential validation method according to claim 1, wherein the SCRFD face detection algorithm includes at least a sparse convolution algorithm, FPN structure, anchor-Free algorithm, and IoU-Net algorithm;
the step of acquiring the first face in the certificate image and the second face in the actual acquired image by using a first face detection algorithm comprises the following steps:
carrying out multi-scale face feature extraction on the certificate image by using the sparse convolution algorithm to obtain face features with different scales in the certificate image;
carrying out multi-scale fusion on the face features with different scales by using the FPN structure, and emphasizing the features by using an attention mechanism to obtain a feature set;
performing face frame detection on the feature set by using the Anchor-Free algorithm, predicting the position and the size of the face frame by using density regression points, and predicting the intersection ratio between the face frame and a true value by using the IoU-Net algorithm so as to correct the position and the size of the face frame and obtain a first face;
And by analogy, performing face detection on the actual acquisition image through the SCRFD face detection algorithm to acquire a second face in the actual acquisition image.
7. The credential validation method according to claim 1, wherein the FaceNet algorithm comprises a convolutional neural network of Inception structure and a euclidean space;
The step of obtaining the first feature vector of the first face and the second feature vector of the second face by using a second face detection algorithm includes:
Based on a face coordinate system, carrying out normalization processing by utilizing coordinate data of each key point in the first face;
Performing depth feature extraction on the first face by using the Inception structures;
mapping the extracted depth features into an European space to obtain a first feature vector of the first face;
And so on, obtaining a second feature vector of the second face.
8. The method of claim 1, wherein the similarity calculation rule uses a cosine similarity relation, the cosine similarity relation being:
Wherein a i represents the corresponding element value of the first feature vector a, and B i represents the corresponding element value of the second feature vector B;
the step of obtaining a similarity calculation result between the first feature vector and the second feature vector by using a preset similarity calculation rule includes:
and calculating and obtaining similarity values between-1 and-1 through the cosine similarity relational expression so that the similarity values approach 1, and representing that the certificate image is similar to the actual acquired image.
9. The method of claim 8, wherein the step of outputting the certificate verification result based on the similarity calculation result comprises:
comparing the similarity value with a preset similarity threshold value, and outputting a comparison result;
and if the similarity value is smaller than the similarity threshold in the comparison result, judging that the certificate image and the actual acquired image are different people.
10. A credential verification device, the device comprising:
the image acquisition module is used for responding to the certificate verification request and acquiring a certificate image and an actual acquisition image;
The effective verification module is used for verifying the validity of the certificate image by using a preset verification model and outputting a verification result;
The face detection module is used for acquiring a first face in the certificate image and a second face in the actual acquisition image by utilizing a first face detection algorithm when the verification result meets a preset condition;
the feature extraction module is used for acquiring a first feature vector of the first face and a second feature vector of the second face by using a second face detection algorithm;
The similarity calculation module is used for obtaining a similarity calculation result between the first feature vector and the second feature vector by utilizing a preset similarity calculation rule;
the result output module is used for outputting a certificate verification result according to the similarity calculation result;
wherein the image acquisition module, prior to the step of responding to the credential validation request, comprises:
Responding to the model training request, acquiring image data related to a plurality of certificates to construct a data set;
After carrying out first preprocessing on part of the image data in the data set, constructing a training set and a testing set, and obtaining first training data;
inputting the first training data into a pre-built neural network architecture to perform model pre-training to obtain a pre-training result;
Performing second preprocessing on the image data remained in the data set according to the pre-training result to obtain second training data;
Inputting the second training data into the pre-trained model, and performing model training according to a preset training rule to obtain a verification model;
the step of inputting the second training data into the pre-trained model and training the model according to a preset training rule to obtain a verification model comprises the following steps:
Dividing the training set into a plurality of sub-batches as a whole batch size so as to perform model training;
Carrying out gradient iterative computation on each sub-batch, and then carrying out gradient superposition;
Based on a preset accumulation factor, after the number of sub-batches is accumulated to a preset count value, gradient updating processing is carried out, a gradient accumulator is emptied, the next round of accumulation is started, and training is stopped until the preset training round number is reached;
And similarly, training continuously by using training data in the training set, storing weight values obtained by training, and evaluating the image description effect by using the test set until the test set index meets the requirement, thereby obtaining a verification model meeting the requirement.
11. A computer device comprising a memory and a processor, the memory storing a computer program, characterized in that the processor implements the steps of the method of any one of claims 1 to 9 when the computer program is executed.
12. A computer readable storage medium, on which a computer program is stored, characterized in that the computer program, when being executed by a processor, implements the steps of the method of any of claims 1 to 9.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310757031.0A CN116959064B (en) | 2023-06-25 | 2023-06-25 | Certificate verification method and device, computer equipment and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310757031.0A CN116959064B (en) | 2023-06-25 | 2023-06-25 | Certificate verification method and device, computer equipment and storage medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN116959064A CN116959064A (en) | 2023-10-27 |
| CN116959064B true CN116959064B (en) | 2024-04-26 |
Family
ID=88445352
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310757031.0A Active CN116959064B (en) | 2023-06-25 | 2023-06-25 | Certificate verification method and device, computer equipment and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116959064B (en) |
Citations (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105956578A (en) * | 2016-05-23 | 2016-09-21 | 深圳华中科技大学研究院 | Face verification method based on identity document information |
| CN107818308A (en) * | 2017-10-31 | 2018-03-20 | 平安科技（深圳）有限公司 | A kind of recognition of face intelligence comparison method, electronic installation and computer-readable recording medium |
| CN107844748A (en) * | 2017-10-17 | 2018-03-27 | 平安科技（深圳）有限公司 | Auth method, device, storage medium and computer equipment |
| CN108229427A (en) * | 2018-01-29 | 2018-06-29 | 南京华科和鼎信息科技有限公司 | A kind of identity-based certificate and the identity security verification method and system of recognition of face |
| CN109255299A (en) * | 2018-08-13 | 2019-01-22 | 北京市商汤科技开发有限公司 | Identity identifying method and device, electronic equipment and storage medium |
| CN109359502A (en) * | 2018-08-13 | 2019-02-19 | 北京市商汤科技开发有限公司 | False-proof detection method and device, electronic equipment, storage medium |
| CN109472208A (en) * | 2018-10-12 | 2019-03-15 | 平安科技（深圳）有限公司 | Certificates handling method, apparatus, computer equipment and storage medium based on recognition of face |
| WO2019128367A1 (en) * | 2017-12-26 | 2019-07-04 | 广州广电运通金融电子股份有限公司 | Face verification method and apparatus based on triplet loss, and computer device and storage medium |
| CN111191539A (en) * | 2019-12-20 | 2020-05-22 | 江苏常熟农村商业银行股份有限公司 | Certificate authenticity verification method and device, computer equipment and storage medium |
| CN112115748A (en) * | 2019-06-21 | 2020-12-22 | 腾讯科技（深圳）有限公司 | Certificate image identification method, certificate image identification device, terminal and storage medium |
| KR20210040313A (en) * | 2020-06-30 | 2021-04-13 | 베이징 바이두 넷컴 사이언스 앤 테크놀로지 코., 엘티디. | Image recognition method, apparatus, device, computer storage medium and program |
| CN113591603A (en) * | 2021-07-09 | 2021-11-02 | 北京旷视科技有限公司 | Certificate verification method and device, electronic equipment and storage medium |
| CN113723238A (en) * | 2021-08-18 | 2021-11-30 | 北京深感科技有限公司 | Human face lightweight network model construction method and human face recognition method |
-
2023
- 2023-06-25 CN CN202310757031.0A patent/CN116959064B/en active Active
Patent Citations (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105956578A (en) * | 2016-05-23 | 2016-09-21 | 深圳华中科技大学研究院 | Face verification method based on identity document information |
| CN107844748A (en) * | 2017-10-17 | 2018-03-27 | 平安科技（深圳）有限公司 | Auth method, device, storage medium and computer equipment |
| CN107818308A (en) * | 2017-10-31 | 2018-03-20 | 平安科技（深圳）有限公司 | A kind of recognition of face intelligence comparison method, electronic installation and computer-readable recording medium |
| WO2019128367A1 (en) * | 2017-12-26 | 2019-07-04 | 广州广电运通金融电子股份有限公司 | Face verification method and apparatus based on triplet loss, and computer device and storage medium |
| CN108229427A (en) * | 2018-01-29 | 2018-06-29 | 南京华科和鼎信息科技有限公司 | A kind of identity-based certificate and the identity security verification method and system of recognition of face |
| CN109255299A (en) * | 2018-08-13 | 2019-01-22 | 北京市商汤科技开发有限公司 | Identity identifying method and device, electronic equipment and storage medium |
| CN109359502A (en) * | 2018-08-13 | 2019-02-19 | 北京市商汤科技开发有限公司 | False-proof detection method and device, electronic equipment, storage medium |
| CN109472208A (en) * | 2018-10-12 | 2019-03-15 | 平安科技（深圳）有限公司 | Certificates handling method, apparatus, computer equipment and storage medium based on recognition of face |
| CN112115748A (en) * | 2019-06-21 | 2020-12-22 | 腾讯科技（深圳）有限公司 | Certificate image identification method, certificate image identification device, terminal and storage medium |
| CN111191539A (en) * | 2019-12-20 | 2020-05-22 | 江苏常熟农村商业银行股份有限公司 | Certificate authenticity verification method and device, computer equipment and storage medium |
| KR20210040313A (en) * | 2020-06-30 | 2021-04-13 | 베이징 바이두 넷컴 사이언스 앤 테크놀로지 코., 엘티디. | Image recognition method, apparatus, device, computer storage medium and program |
| CN113591603A (en) * | 2021-07-09 | 2021-11-02 | 北京旷视科技有限公司 | Certificate verification method and device, electronic equipment and storage medium |
| CN113723238A (en) * | 2021-08-18 | 2021-11-30 | 北京深感科技有限公司 | Human face lightweight network model construction method and human face recognition method |
Non-Patent Citations (4)
| Title |
|---|
| Sample and Computation Redistribution for Efficient Face Detection;Jia Guo等;《arXiv:2105.04714》;1-10 * |
| 二代身份证应用中的人脸验证算法与系统设计;陈鹏旭;《中国优秀硕士学位论文全文数据库 (信息科技辑)》(第1期);I138-1626 * |
| 基于特征智能识别的业扩报装申请材料校验技术研究;潘明明等;《供用电》;第37卷(第06期);9-12+32 * |
| 快速人脸检测技术在证件照中的应用与研究;陈史政;《泰山学院学报》;第36卷(第03期);57-60 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN116959064A (en) | 2023-10-27 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111860674B (en) | Sample category identification method, sample category identification device, computer equipment and storage medium | |
| CN111667011B (en) | Damage detection model training and vehicle damage detection method, device, equipment and medium | |
| US10410292B2 (en) | Method, system, apparatus, and storage medium for realizing antifraud in insurance claim based on consistency of multiple images | |
| CN111191539B (en) | Certificate authenticity verification method and device, computer equipment and storage medium | |
| CN109816200B (en) | Task pushing method, device, computer equipment and storage medium | |
| CN110738236B (en) | Image matching method and device, computer equipment and storage medium | |
| CN111191568A (en) | Method, device, equipment and medium for identifying copied image | |
| CN112418009A (en) | Image quality detection method, terminal device and storage medium | |
| CN109271957B (en) | Face gender identification method and device | |
| CN111275070B (en) | Signature verification method and device based on local feature matching | |
| CN113642639B (en) | Living body detection method, living body detection device, living body detection equipment and storage medium | |
| CN112232971A (en) | Anti-fraud detection method, anti-fraud detection device, computer equipment and storage medium | |
| EP4141830A1 (en) | Method for detecting a forgery of an identity document | |
| CN111445058A (en) | Data analysis method, device, equipment and computer readable storage medium | |
| CN115223022A (en) | Image processing method, device, storage medium and equipment | |
| CN117558011B (en) | Image text tampering detection method based on self-consistency matrix and multi-scale loss | |
| Al Qudah et al. | Using Artificial Intelligence applications for E-Government services as iris recognition | |
| CN117315678B (en) | Digital image tampering identification method, system, equipment and storage medium | |
| CN112041847A (en) | Providing images with privacy tags | |
| CN116959064B (en) | Certificate verification method and device, computer equipment and storage medium | |
| CN117314713A (en) | License anti-counterfeiting data processing method based on digital watermark | |
| CN116612486A (en) | Risk group identification method and system based on image identification and graph calculation | |
| Alherbawi et al. | JPEG image classification in digital forensic via DCT coefficient analysis | |
| CN113077355B (en) | Insurance claim settlement method and device, electronic equipment and storage medium | |
| CN113744046A (en) | Service data processing method and device, computer equipment and storage medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |