CN117611923B - Identification method and system for identity document authenticity - Google Patents
Identification method and system for identity document authenticity Download PDFInfo
- Publication number
- CN117611923B CN117611923B CN202410024890.3A CN202410024890A CN117611923B CN 117611923 B CN117611923 B CN 117611923B CN 202410024890 A CN202410024890 A CN 202410024890A CN 117611923 B CN117611923 B CN 117611923B
- Authority
- CN
- China
- Prior art keywords
- data set
- image
- model
- training data
- training
- 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
- 238000000034 method Methods 0.000 title claims abstract description 40
- 238000012549 training Methods 0.000 claims abstract description 101
- 238000013145 classification model Methods 0.000 claims abstract description 65
- 238000003708 edge detection Methods 0.000 claims abstract description 46
- 238000007781 pre-processing Methods 0.000 claims abstract description 17
- 238000012545 processing Methods 0.000 claims abstract description 17
- 238000010276 construction Methods 0.000 claims description 11
- 238000005457 optimization Methods 0.000 claims description 5
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 claims description 3
- 230000006872 improvement Effects 0.000 description 9
- 230000008569 process Effects 0.000 description 8
- 238000005516 engineering process Methods 0.000 description 6
- 238000013527 convolutional neural network Methods 0.000 description 5
- 230000000694 effects Effects 0.000 description 5
- 238000012795 verification Methods 0.000 description 4
- 230000004927 fusion Effects 0.000 description 3
- 230000007246 mechanism Effects 0.000 description 3
- RWSOTUBLDIXVET-UHFFFAOYSA-N Dihydrogen sulfide Chemical compound S RWSOTUBLDIXVET-UHFFFAOYSA-N 0.000 description 2
- 230000001815 facial effect Effects 0.000 description 2
- 230000006870 function Effects 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 239000004384 Neotame Substances 0.000 description 1
- 230000006399 behavior Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000009977 dual effect Effects 0.000 description 1
- 230000002349 favourable effect Effects 0.000 description 1
- 238000005242 forging Methods 0.000 description 1
- 238000007689 inspection Methods 0.000 description 1
- HLIAVLHNDJUHFG-HOTGVXAUSA-N neotame Chemical compound CC(C)(C)CCN[C@@H](CC(O)=O)C(=O)N[C@H](C(=O)OC)CC1=CC=CC=C1 HLIAVLHNDJUHFG-HOTGVXAUSA-N 0.000 description 1
- 235000019412 neotame Nutrition 0.000 description 1
- 108010070257 neotame Proteins 0.000 description 1
- 238000011524 similarity measure Methods 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
Classifications
-
- 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
- G06V10/764—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using classification, e.g. of video objects
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/13—Edge detection
-
- 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
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/56—Extraction of image or video features relating to colour
-
- 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
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Physics & Mathematics (AREA)
- Multimedia (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Databases & Information Systems (AREA)
- Computing Systems (AREA)
- Artificial Intelligence (AREA)
- Image Analysis (AREA)
Abstract
The invention provides an identification document authenticity identification method and system, which relate to the technical field of image processing and comprise the following steps: constructing a training data set comprising real images and fake images of the identity document, the identity document copy and the passport; preprocessing the training data set, detecting edges, and highlighting edges of the image; based on the boundary classification model, adjusting a model structure to construct an initial classification model; reading the preprocessed training data set into an initial two-class model, setting training parameters, and performing iterative training on the initial two-class model to obtain a final two-class model; and carrying out pre-processing on subpoena images on the user, and transmitting the images into a classification model to obtain a prediction result. According to the invention, the edge line of the portrait and the background in the certificate image is better identified by training the two classification models based on edge detection through the data set, so that the authenticity of the certificate is accurately identified.
Description
Technical Field
The invention relates to the technical field of image processing, in particular to an identification method and an identification system for authenticity of an identity document.
Background
In digital banking, mobile wallet and other applications, the technology means of OCR recognition of user identity document information, and comparison of document facial photo and living photo facial photo is used, so that the user can conveniently and remotely open an account by self-service eKYC (electronically confirming the identity of the user). However, the user also has the common means of stealing identity documents of other people through forging the identity document, maliciously opening accounts, developing illegal actions and the like, covering the face of the original document through a face photo of the user, or replacing the face of the user by the face of the user through the face-matting operation in the face photo of the identity document, and copying even after covering the face of the original document, thereby obtaining the synthesized identity document. Therefore, in the process eKYC, the capability of identifying the authenticity of the identity document is added to identify whether the face photo in the document photo is falsified or not.
Currently, the largest set of deep face counterfeits ForgeryNet is disclosed, mainly by using some newer methods to extract features, such as attention mechanisms, vision Transformer model and CNN in combination. For example: an improved Xception with dual attention mechanisms and feature fusion for detecting face counterfeits; the method captures different high semantic features of the face image in the middle stream by using convolution of different levels, and introduces a convolution block attention module and feature fusion to refine and reorganize the high semantic features; in the exit stream, a self-attention mechanism and a depth separable convolution are adopted to respectively learn global information and local information of the fusion characteristics, so that the classification capability of the model is improved. Still another is authentication of an identity document based on Guilloche (neotame) pattern forgery detection, learning the similarity of a pair of identity documents through a convolutional neural network and extracting the highly discriminating features between them, and then determining the authenticity of the identity document through a similarity measure. Unlike the first, the convolutional neural network of the second type requires an input of a pair of identity documents, and one requires the model to know the authenticity to compare the authenticity of the second document. Specifically, the actual dataset used in the second method is MIDV-2020, where the document background pattern of this dataset is a Guilloche (guilloche) geometric pattern of unique shape formed by computer generated fine lines interlaced, very sharp. Whereas the counterfeit dataset FMIDV 2022 was generated from the real dataset, a region block that visually contained only the pattern and did not have any foreground was selected as a candidate region for generating the counterfeit dataset by applying the copy-move operation. The model is mainly used for judging the authenticity of the certificate by checking the consistency of the pattern on the background of the certificate and the similarity of the pattern with the actual pattern of the same country. However, in practical applications, the Guilloche patterns in many national identity documents (including identity cards, passports, etc.) are not very clear, and the Guilloche geometric patterns are more blurred after photographing. So this method of detecting Guilloche the pattern background does not identify the authenticity of the document well.
Therefore, the existing model and algorithm are not completely suitable for the actual eKYC remote account opening scene, and three fake certificate modes exist in the eKYC remote account opening scene, wherein the first mode is to use the PS technology P to attach the identity certificate to the certificate, the second mode is to directly cover the face part of the identity certificate by using the big head photo, and the third mode is to obtain a new identity certificate through copying on the basis of the second mode; how to propose an identity document authenticity identification method suitable for eKYC remote account opening scenes, and identify the counterfeit document behaviors of the three scenes becomes a problem to be solved urgently.
Disclosure of Invention
In view of the above problems, the present invention provides a method and a system for identifying authenticity of an identity document, which are used for training a CNN model after image edge detection preprocessing, identifying a document copy and a method for identifying authenticity of an identity document based on the CNN model, and identifying a counterfeit document forged by a plurality of methods in eKYC remote account opening scenarios.
In order to achieve the above object, the present invention provides a method for identifying authenticity of an identity document, comprising:
constructing a training data set comprising real images and fake images of the identity document, the identity document copy and the passport;
preprocessing the training data set, detecting edges, and highlighting edges of the image;
based on the boundary classification model, adjusting a model structure to construct an initial classification model;
Reading the preprocessed training data set into the initial two-class model, setting training parameters, and performing iterative training on the initial two-class model to obtain a final two-class model;
And carrying out pretreatment and edge detection on subpoena images on the user, and transmitting the images into the classification model to obtain a prediction result.
As a further improvement of the invention, the training data set is provided with a label, and the label is marked by each image in the training data set, so that authenticity of each identity document, identity document copy and passport is marked.
As a further improvement of the invention, the training data set is preprocessed to highlight the edges of the image; comprising the following steps:
Carrying out graying treatment on each image in the training data set;
And (3) carrying out edge detection processing on the image subjected to the graying processing by adopting a Canny edge detection algorithm, and highlighting the edge of the image.
As a further improvement of the invention, the threshold value of the Canny edge detection algorithm is adjusted for the training data set after the graying treatment, and the edge detection treatment is carried out on the image after the graying treatment according to the Canny edge detection algorithm after the threshold value adjustment.
As a further improvement of the invention, the structure of the model EDGECLASSIFIER is adjusted according to the Canny edge detection algorithm after threshold adjustment, so as to obtain an initial classification model.
As a further improvement of the present invention, performing iterative training on the initial two-classification model to obtain a final two-classification model, including:
the loss function adopts cross entropy loss and adopts an Adam optimization algorithm as an optimizer;
Calculating the prediction result of the initial two-classification model and the loss of the label of the image in the training data set through cross entropy loss;
and (3) performing iterative training, and storing training parameters with minimum loss to obtain a final classification model.
As a further improvement of the invention, the certificate image uploaded by the user is judged whether to be a certificate copy or not through the color standard deviation, and if so, the authenticity prediction is not needed.
As a further improvement of the invention, if the judgment result is negative, preprocessing and edge detection are carried out, and the two classification models are transmitted to obtain a prediction result.
As a further improvement of the present invention, preprocessing and edge detection are performed on the document image, including:
Graying the certificate image;
And (3) carrying out edge detection processing on the certificate image subjected to graying processing by adopting a Canny edge detection algorithm, and highlighting the edge of the image.
The invention also provides an identity document authenticity identification system, which comprises: the system comprises a training data set construction module, a training data set processing module, a model construction module, a model training module and a certificate authenticity identification module;
the training data set construction module is used for:
constructing a training data set comprising real images and fake images of the identity document, the identity document copy and the passport;
the training data set processing module is used for:
preprocessing the training data set, detecting edges, and highlighting edges of the image;
The model construction module is used for:
based on the boundary classification model, adjusting a model structure to construct an initial classification model;
The model training module is used for:
Reading the preprocessed training data set into the initial two-class model, setting training parameters, and performing iterative training on the initial two-class model to obtain a final two-class model;
the certificate authenticity identification module is used for:
And carrying out pretreatment and edge detection on subpoena images on the user, and transmitting the images into the classification model to obtain a prediction result.
Compared with the prior art, the invention has the beneficial effects that:
According to the invention, the preprocessing and edge detection are carried out on the data set, so that the edge lines of the portrait and the background in the certificate image are more prominent, and the image characteristics of the initial two-classification model can be better learned in the training process; further, the document image input by the user is preprocessed, so that the edge line of the portrait and the background in the document image can be better identified by the classification model, authenticity can be better identified, an accurate prediction result can be made, and compared with the traditional OCR identification, the method is more suitable for a user autonomous remote eKYC account opening scene, and the method effectively identifies and checks the condition that the document photo P is on the identity document image and the condition that the main photo is used for shielding the identity document primary photo through the PS technology, and effectively prevents malicious remote account opening actions.
Before the identification of the certificate image is carried out, the invention judges whether the image is a certificate copy or not by utilizing the color standard deviation of the image, judges whether the image is a color image or a black-and-white image by utilizing the color standard deviation, and can judge that the certificate is a copy if the image is the black-and-white image, the invention directly refuses to open a remote account without continuing the identification of the certificate.
According to the invention, a boundary classification model EDGECLASSIFIER is adopted to construct a two-class model, the boundary classification model EDGECLASSIFIER comprises two layers of convolutions, recognition features can be rapidly mastered through training to obtain a final two-class model, and meanwhile, prediction results can be rapidly calculated and given out when certificate images are subjected to true and false two-class, so that the recognition efficiency is improved.
According to the method, the structure of a boundary classification model EDGECLASSIFIER is adjusted according to the threshold value of the adjusted Canny edge detection algorithm, so that an initial classification model is obtained; in the process, a model structure with more obvious edge lines after being processed in the boundary classification model EDGECLASSIFIER can be selected as an initial classification model, so that the initial classification model can conveniently identify images in a training dataset, and the training effect of the initial classification model is improved.
Drawings
FIG. 1 is a flow chart of a method for verifying the authenticity of an identity document according to one embodiment of the present invention;
fig. 2 is a schematic diagram of an identification document authentication system according to an embodiment of the present invention.
Detailed Description
For the purpose of making the objects, technical solutions and advantages of the embodiments of the present invention more apparent, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention, and it is apparent that the described embodiments are some embodiments of the present invention, but not all embodiments of the present invention. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
The invention is described in further detail below with reference to the attached drawing figures:
as shown in fig. 1, the identification document authenticity identification method provided by the invention aims at a user self-service remote eKYC account opening scene, and performs authenticity identification classification through a boundary classification model (EDGECLASSIFIER), so that not only can the authenticity of an identification document and a passport be effectively detected, but also a copy of the identification document can be detected from the identification document and the passport, and the copy of the identification document of the user is returned to be inconsistent with the uploading requirement of the identification document; the method specifically comprises the following steps:
s1, constructing a training data set comprising an identity document, an identity document copy and a real image and a fake image of a passport;
wherein,
Aiming at the certificate types required by the user self-help remote eKYC account opening scene, including an identity card and a passport, a training data set is constructed;
Further, the method comprises the steps of,
The real images of the identity document, the identity document copy and the passport adopt the real license existing in the user self-service remote eKYC account opening system;
the counterfeit image of an identity document, a copy of an identity document, and a passport includes: the method comprises the steps of adopting a PS technology to irradiate an identity document P onto a primary identity document or a passport, adopting a big head photo to shade a face part on the primary identity document or the passport, adopting the big head photo to shade the primary identity document and then copying, and adopting the PS technology to irradiate the identity document P onto the primary identity document or the passport and then copying.
Still further, the method comprises the steps of,
The training data set is provided with labels, which are unified into image marking labels in the training data set, and the serial numbers and authenticity of the identity documents, the identity document copies and the passports are marked. The label is as follows: 00001: true, 00002: false, 00003: copying;
s2, preprocessing a training data set, detecting edges, and highlighting edges of the image;
wherein,
Carrying out graying treatment on each image in the training data set;
And the Canny edge detection algorithm is adopted to carry out edge detection processing on the image subjected to the graying processing, the processed image edge line is more obvious, the edge of the image is highlighted, the model training stage is convenient, and the edge is identified by the initial classification model.
Further, the method comprises the steps of,
According to the training data set after the graying treatment, the threshold value of the Canny edge detection algorithm is adjusted in a targeted manner, so that the inspection effect of the Canny edge detection algorithm after adjustment is better;
and then carrying out edge detection processing on the image subjected to the graying processing according to a Canny edge detection algorithm after the threshold value is adjusted, so that the edge detection effect and efficiency are improved.
S3, based on the boundary classification model, adjusting a model structure, and constructing an initial classification model;
wherein,
According to a Canny edge detection algorithm after threshold adjustment, adjusting the structure of a boundary classification model, wherein the boundary classification model comprises two layers of convolutions, in the process of adjusting the structure of the boundary classification model, testing an adjusting effect of an input image, and selecting a model structure with more obvious edge lines after the input image is processed as an initial classification model; the method is favorable for subsequent training to obtain a better classification model.
S4, reading the preprocessed training data set into an initial two-class model, setting training parameters, and performing iterative training on the initial two-class model to obtain a final two-class model;
wherein,
In the training process, the training data set is divided into a training set and a verification set according to the sequence numbers in each image label, the training set is used for training the initial two-class model, the verification set is used for verifying the training result, and images contained in the training set and the verification set are adjusted in the training data set for continuous iterative optimization;
Further, the method comprises the steps of,
The loss function adopts cross entropy loss, and the cross entropy loss is suitable for classification problems; in the process of calculating the loss, calling the label of the current verification set to acquire the authenticity of the image, and calculating the loss;
adopting an Adam optimization algorithm as an optimizer;
and calculating the prediction result of the initial two-class model and the Loss of the label of the image in the training data set through cross entropy Loss, and finally storing the model parameters with minimum training Loss and optimal training parameters through continuous iterative optimization to obtain the final two-class model.
S5, carrying out pretreatment and edge detection on subpoena images on the user, and transmitting the images into a classification model to obtain a prediction result.
Wherein,
When subpoena images are uploaded by a user, firstly, judging whether the images are certificate copies or not according to the color standard deviation before preprocessing, wherein the size of the color standard deviation can reflect whether the images are black and white or color, and the default certificate copies are black and white, so that the images with smaller color standard deviation are regarded as the certificate copies; because the user self-service remote eKYC opens an account and is always not allowed to use the certificate copy, the system can remind the user that the uploaded certificate does not meet the requirements, and preprocessing and the authenticity prediction through a two-classification model are not needed; the link can also effectively save the system calculation force and improve the processing efficiency of the system.
Further, the method comprises the steps of,
If the judgment result is negative, preprocessing and edge detection are carried out, and the two classification models are transmitted to obtain a prediction result.
Further, preprocessing and edge detection are performed on the certificate image, including:
graying treatment is carried out on the certificate image;
And the edge detection algorithm is adopted to carry out edge detection treatment on the certificate image subjected to the graying treatment, the edge line of the treated certificate image is more obvious, the edge of the certificate image is highlighted, and the identification and the prediction of the authenticity of the two-classification model are facilitated.
Finally, the predicted result of subpoena images on the user is output.
As shown in fig. 2, the present invention further provides an identity document authenticity identification system, including: the system comprises a training data set construction module, a training data set processing module, a model construction module, a model training module and a certificate authenticity identification module;
the training data set construction module is used for:
constructing a training data set comprising real images and fake images of the identity document, the identity document copy and the passport;
A training data set processing module for:
preprocessing the training data set, detecting edges, and highlighting edges of the image;
the model building module is used for:
based on the boundary classification model, adjusting a model structure to construct an initial classification model;
Model training module for:
Reading the preprocessed training data set into an initial two-class model, setting training parameters, and performing iterative training on the initial two-class model to obtain a final two-class model;
the certificate authenticity identification module is used for:
and carrying out pretreatment and edge detection on subpoena images on the user, and transmitting the images into a classification model to obtain a prediction result.
Examples
(1) Uploading a certificate image forged by PS to a system when a user opens an account independently and remotely eKYC, and judging that the certificate image is a color image according to the color standard deviation, and not a certificate copy; carrying out image graying on a certificate image uploaded by a user, identifying edge lines of an identity certificate by using a Canny edge detection algorithm, obtaining an identified image, inputting the identified image into a pre-trained classification model (boundary classification model EDGECLASSIFIER), detecting the edge lines of a face certificate by the classification model, judging the authenticity of the certificate according to the learned characteristics, and giving a prediction result, namely:
{
"result: Fake"
}
As can be seen, the document image is a fake document image, and the system will remind the user that there is a problem in uploading the content and does not transact account opening.
(2) When a user opens an account from a remote eKYC, uploading a certificate image which is blocked and forged by the big head photo to the system, and judging that the certificate image is a color image according to the color standard deviation, so that the certificate image is not a certificate copy; carrying out image graying on a certificate image uploaded by a user, identifying edge lines of an identity document by using a Canny edge detection algorithm, obtaining an identified image, inputting the identified image into a pre-trained two-class model (EDGECLASSIFIER model), detecting the edge lines of a big head photo and the edge lines of the big head photo to cover the edge of an original document by the two-class model, judging the authenticity of the document according to the learned characteristics, and giving a prediction result, namely:
{
"result: Fake"
}
As can be seen, the document image is a fake document image, and the system will remind the user that there is a problem in uploading the content and does not transact account opening.
(3) When a user opens an account from a remote eKYC, uploading a certificate image covered by a big head sticker and copied to the system, after reading the certificate image, firstly calculating the color standard deviation of the image to judge whether the image is a black-and-white copy, and outputting the following information:
{
"Warning": "The image is black and white, and it is a photocopy of the document, which does not meet the requirements.
}
the document image is known to be a copy of the document, and the system will alert the user that the uploaded content is unsatisfactory.
The invention has the advantages that:
According to the invention, the preprocessing and edge detection are carried out on the data set, so that the edge lines of the portrait and the background in the certificate image are more prominent, and the image characteristics of the initial two-classification model can be better learned in the training process; further, the document image input by the user is preprocessed, so that the edge line of the portrait and the background in the document image can be better identified by the classification model, authenticity can be better identified, an accurate prediction result can be made, and compared with the traditional OCR identification, the method is more suitable for a user autonomous remote eKYC account opening scene, and the method effectively identifies and checks the condition that the document photo P is on the identity document image and the condition that the main photo is used for shielding the identity document primary photo through the PS technology, and effectively prevents malicious remote account opening actions.
Before the identification of the certificate image is carried out, the invention judges whether the image is a certificate copy or not by utilizing the color standard deviation of the image, judges whether the image is a color image or a black-and-white image by utilizing the color standard deviation, and can judge that the certificate is a copy if the image is the black-and-white image, the invention directly refuses to open a remote account without continuing the identification of the certificate.
The model EDGECLASSIFIER is adopted to construct the two-class model, the model EDGECLASSIFIER comprises two layers of convolution, the recognition characteristics can be quickly mastered through training to obtain the final two-class model, and meanwhile, the prediction result can be quickly calculated and given out when the authenticity of the certificate image is classified, so that the recognition efficiency is improved.
According to the structure of the threshold adjustment model EDGECLASSIFIER of the adjusted Canny edge detection algorithm, an initial classification model is obtained; in the process, a model structure with more obvious processed edge lines in the model EDGECLASSIFIER can be selected as an initial classification model, so that the initial classification model can conveniently identify images in a training dataset, and the training effect of the initial classification model is improved.
The above is only a preferred embodiment of the present invention, and is not intended to limit the present invention, but various modifications and variations can be made to the present invention by those skilled in the art. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.
Claims (4)
1. A method for verifying the authenticity of an identity document, comprising:
constructing a training data set comprising real images and fake images of the identity document, the identity document copy and the passport;
Preprocessing the training data set, detecting edges, and highlighting edges of the image; wherein, include: carrying out graying treatment on each image in the training data set; adjusting a threshold value of a Canny edge detection algorithm aiming at the training data set subjected to the graying treatment, and carrying out edge detection treatment on the image subjected to the graying treatment by adopting the Canny edge detection algorithm after the threshold value is adjusted to highlight the edge of the image;
Based on the boundary classification model, adjusting a model structure to construct an initial classification model, wherein the structure of the boundary classification model is adjusted according to a Canny edge detection algorithm after threshold adjustment to obtain the initial classification model;
Reading the preprocessed training data set subjected to edge detection into the initial two-class model, setting training parameters, and performing iterative training on the initial two-class model to obtain a final two-class model;
and carrying out subpoena images on the user, judging whether the certificate image uploaded by the user is a certificate copy through the color standard deviation, if so, carrying out no true-false prediction, and if not, carrying out pretreatment and edge detection on the certificate image and transmitting the certificate image into the two-classification model to obtain a prediction result.
2. The identity document authentication method of claim 1, wherein: the training data set is provided with labels, the labels are uniformly marked by the images in the training data set, and authenticity of each identity document, identity document copy and passport is marked.
3. The identity document authentication method of claim 1, wherein: performing iterative training on the initial two-classification model to obtain a final two-classification model, wherein the iterative training comprises the following steps:
the loss function adopts cross entropy loss and adopts an Adam optimization algorithm as an optimizer;
Calculating the prediction result of the initial two-classification model and the loss of the label of the image in the training data set through cross entropy loss;
and (3) performing iterative training, and storing training parameters with minimum loss to obtain a final classification model.
4. An identity document authenticity identification system applying an identity document authenticity identification method according to any one of claims 1 to 3, comprising: the system comprises a training data set construction module, a training data set processing module, a model construction module, a model training module and a certificate authenticity identification module;
the training data set construction module is used for:
constructing a training data set comprising real images and fake images of the identity document, the identity document copy and the passport;
the training data set processing module is used for:
Preprocessing the training data set, detecting edges, and highlighting edges of the image; wherein, include: carrying out graying treatment on each image in the training data set; adjusting a threshold value of a Canny edge detection algorithm aiming at the training data set subjected to the graying treatment, and carrying out edge detection treatment on the image subjected to the graying treatment by adopting the Canny edge detection algorithm after the threshold value is adjusted to highlight the edge of the image;
The model construction module is used for:
Based on the boundary classification model, adjusting a model structure to construct an initial classification model; the structure of the boundary classification model is adjusted according to a Canny edge detection algorithm after threshold adjustment, so that an initial classification model is obtained;
The model training module is used for:
Reading the preprocessed training data set subjected to edge detection into the initial two-class model, setting training parameters, and performing iterative training on the initial two-class model to obtain a final two-class model;
the certificate authenticity identification module is used for:
And carrying out pretreatment and edge detection on subpoena images on the user, and transmitting the images into the classification model to obtain a prediction result.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410024890.3A CN117611923B (en) | 2024-01-08 | 2024-01-08 | Identification method and system for identity document authenticity |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410024890.3A CN117611923B (en) | 2024-01-08 | 2024-01-08 | Identification method and system for identity document authenticity |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN117611923A CN117611923A (en) | 2024-02-27 |
| CN117611923B true CN117611923B (en) | 2024-08-13 |
Family
ID=89951853
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202410024890.3A Active CN117611923B (en) | 2024-01-08 | 2024-01-08 | Identification method and system for identity document authenticity |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117611923B (en) |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| 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 |
| CN112163511A (en) * | 2020-09-25 | 2021-01-01 | 天津大学 | Method for identifying authenticity of image |
Family Cites Families (18)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111783505A (en) * | 2019-05-10 | 2020-10-16 | 北京京东尚科信息技术有限公司 | Method and device for identifying forged faces and computer-readable storage medium |
| CN111222434A (en) * | 2019-12-30 | 2020-06-02 | 深圳市爱协生科技有限公司 | Method for obtaining evidence of synthesized face image based on local binary pattern and deep learning |
| CN112132812B (en) * | 2020-09-24 | 2023-06-30 | 平安科技（深圳）有限公司 | Certificate verification method and device, electronic equipment and medium |
| EP4141830A1 (en) * | 2021-08-31 | 2023-03-01 | Thales Dis France SAS | Method for detecting a forgery of an identity document |
| KR102410588B1 (en) * | 2021-09-15 | 2022-06-22 | 주식회사 풀스택 | Device and method for detecting counterfeit indentification card |
| CN114170655A (en) * | 2021-11-29 | 2022-03-11 | 西安电子科技大学 | Knowledge distillation-based face counterfeiting cue migration method |
| CN114444566B (en) * | 2021-12-16 | 2024-07-23 | 厦门市美亚柏科信息股份有限公司 | Image forgery detection method and device and computer storage medium |
| CN114332865B (en) * | 2022-03-11 | 2022-06-03 | 北京锐融天下科技股份有限公司 | Certificate OCR recognition method and system |
| CN114821825B (en) * | 2022-06-30 | 2022-12-06 | 广州中平智能科技有限公司 | Multi-granularity face forgery detection method, system, equipment and medium |
| CN115346278A (en) * | 2022-08-26 | 2022-11-15 | 北京字跳网络技术有限公司 | Image detection method, device, readable medium and electronic equipment |
| CN115588226A (en) * | 2022-11-01 | 2023-01-10 | 重庆邮电大学 | High-robustness deep-forged face detection method |
| CN116228644A (en) * | 2022-12-15 | 2023-06-06 | 北京迈格威科技有限公司 | Image detection method, electronic device and storage medium |
| CN116310754A (en) * | 2023-02-16 | 2023-06-23 | 中国科学院计算技术研究所 | Incremental learning method and system for detecting fake face image video |
| CN116188439A (en) * | 2023-03-06 | 2023-05-30 | 巧智绘科技(浙江)有限公司 | False face-changing image detection method and device based on identity recognition probability distribution |
| CN116612355A (en) * | 2023-06-02 | 2023-08-18 | 支付宝(杭州)信息技术有限公司 | Training method and device for face fake recognition model, face recognition method and device |
| CN116645562A (en) * | 2023-06-12 | 2023-08-25 | 华南理工大学 | Detection method for fine-grained fake image and model training method thereof |
| CN117079355A (en) * | 2023-08-29 | 2023-11-17 | 中国信息通信研究院 | Object image fake identifying method and device and electronic equipment |
| CN117275063A (en) * | 2023-09-19 | 2023-12-22 | 中国科学院计算技术研究所 | Face depth counterfeiting detection method and system based on three-dimensional information time sequence consistency |
-
2024
- 2024-01-08 CN CN202410024890.3A patent/CN117611923B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| 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 |
| CN112163511A (en) * | 2020-09-25 | 2021-01-01 | 天津大学 | Method for identifying authenticity of image |
Also Published As
| Publication number | Publication date |
|---|---|
| CN117611923A (en) | 2024-02-27 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US20220406029A1 (en) | Tamper detection for identification documents | |
| US11087125B2 (en) | Document authenticity determination | |
| US11023708B2 (en) | Within document face verification | |
| CN110570549B (en) | Intelligent unlocking method and corresponding device | |
| Abramova et al. | Detecting copy–move forgeries in scanned text documents | |
| KR102410588B1 (en) | Device and method for detecting counterfeit indentification card | |
| CN117496583B (en) | Deep fake face detection positioning method capable of learning local difference | |
| Rajan et al. | An extensive study on currency recognition system using image processing | |
| CN117611923B (en) | Identification method and system for identity document authenticity | |
| Shinde et al. | Identification of fake currency using soft computing | |
| CN115775409A (en) | Face image tamper-proof fusion detection method | |
| EP3982289A1 (en) | Method for validation of authenticity of an image present in an object, object with increased security level and method for preparation thereof, computer equipment, computer program and appropriate reading means | |
| CN114694196A (en) | Living body classifier establishing method, human face living body detection method and device | |
| Koti et al. | Netrahitha: Currency Detector Jacket for visually impaired people | |
| Kumar et al. | Currency Authentication using Color Based Processing | |
| Samarasinghe et al. | Sri lanka driving license forgery detection | |
| Shelke et al. | Fake Currency Detection Using Image Processing and Random Forest Algorithm | |
| Brahmapurikar et al. | Currency Recognization for visually impaired People using Tensorflow | |
| CN115116147B (en) | Image recognition, model training, living body detection method and related device | |
| Rathore et al. | Indian Currency Recognition for Visually Impaired Individuals | |
| Al-Hila et al. | Fuzzy Logic Weighted Averaging Algorithm for Malaysian Banknotes Reader Featuring Counterfeit Detection: Manuscript Received: 27 February 2023, Accepted: 28 March 2023, Published: 15 September 2023, ORCiD: 0000-0003-1477-8449, https://doi. org/10.33093/jetap. 2023.5. 2.3 | |
| Naseem et al. | Counterfeit Recognition of Pakistani Currency | |
| Ananth et al. | Fake Currency Detection Using Pattern Recognition Algorithm | |
| Shinde et al. | Fake Currency Detection using Image Processing | |
| Ali et al. | Forgery Localization in Images Using Deep Learning |
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 |