CN115578673A - Certificate authenticity verification method, system, computer device and storage medium - Google Patents
Certificate authenticity verification method, system, computer device and storage medium Download PDFInfo
- Publication number
- CN115578673A CN115578673A CN202211275484.1A CN202211275484A CN115578673A CN 115578673 A CN115578673 A CN 115578673A CN 202211275484 A CN202211275484 A CN 202211275484A CN 115578673 A CN115578673 A CN 115578673A
- Authority
- CN
- China
- Prior art keywords
- certificate
- dimensional
- learning
- feature data
- dimensional feature
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/40—Scenes; Scene-specific elements in video content
- G06V20/46—Extracting features or characteristics from the video content, e.g. video fingerprints, representative shots or key frames
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—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
- G06V10/77—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
- G06V10/80—Fusion, i.e. combining data from various sources at the sensor level, preprocessing level, feature extraction level or classification level
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Computing Systems (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Multimedia (AREA)
- General Health & Medical Sciences (AREA)
- Databases & Information Systems (AREA)
- Medical Informatics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Molecular Biology (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Image Analysis (AREA)
Abstract
The application relates to a certificate authenticity verification method, a certificate authenticity verification system, computer equipment and a storage medium, wherein the certificate authenticity verification method comprises the following steps: carrying out video acquisition on the certificate according to different preset angles to obtain a preprocessed video frame, and taking the preprocessed video frame as the input of an end-to-end learning algorithm; extracting two-dimensional feature data from the preprocessed video frames through a basic network, and fusing the two-dimensional feature data into corresponding three-dimensional feature data; acquiring three-dimensional characteristic data of a plurality of anti-counterfeiting areas from the three-dimensional characteristic data through a RoIAlign tool to establish a multi-task learning model of the end-to-end learning algorithm; outputting a verification result of the certificate through a multitask learning model of the end-to-end learning algorithm; and establishing a learning task through the three-dimensional characteristic data of each anti-counterfeiting area, thereby improving the accuracy and efficiency of certificate authenticity verification.
Description
Technical Field
The present application relates to the field of certificate verification, and in particular, to a method, a system, a computer device, and a storage medium for verifying authenticity of a certificate.
Background
In modern Electronic business, automated processes have been widely used and play an increasingly important role, such as EKYC (Electronic Know-based Customer) in financial systems, because of their advantages of efficiency improvement and cost reduction. The system automatically checks the authenticity of the identity document and extracts related information to determine whether to pass the next step of entering an automatic processing flow.
Counterfeiting of identity documents is not uncommon in reality, and there are copying, imitation, copying (i.e., taking a photograph of the document with an electronic device), and the like. On the other hand, most identity documents have anti-counterfeiting characteristics, which can be summarized into static and dynamic types: static characteristics refer to characteristics that are fixed and unchangeable, such as special patterns and colors; the dynamic characteristics refer to different color forms and the like, such as color-changing ink, gradient characters and the like, under different angles or illumination conditions. In practical application, the dynamic characteristics have stronger anti-counterfeiting property due to the difficulty in counterfeiting, and are the priority of the model algorithm.
The traditional machine learning method generally adopts a manual design (warped) feature, and then the feature is used as an input and is transmitted into a machine learning model such as a logistic regression (logistic regression), a Support Vector Machine (SVM), a random forest (random forest), a gradient boosting tree (gradient boosting tree) and the like to obtain a final result. However, the disadvantage of this pipeline approach is that the feature design learning generation is too manual (artificial), requiring many trial adjustments; and the feature learning and the model training are split, the optimal generated features do not always correspond to the optimal machine learning model, so that the optimal model is difficult to obtain, and the efficiency and the accuracy of the authenticity verification of the certificate are low.
Disclosure of Invention
Based on the method, the system, the computer equipment and the storage medium, the certificate authenticity verification method, the system, the computer equipment and the storage medium are provided, so that the efficiency and the accuracy of certificate authenticity verification are improved.
In one aspect, a method for verifying authenticity of a document is provided, the method comprising:
carrying out video acquisition on the certificate according to different preset angles to obtain a preprocessed video frame, and taking the preprocessed video frame as the input of an end-to-end learning algorithm;
extracting two-dimensional feature data from the preprocessed video frames through a basic network, and fusing the two-dimensional feature data into corresponding three-dimensional feature data;
acquiring three-dimensional characteristic data of a plurality of anti-counterfeiting areas from the three-dimensional characteristic data through a RoIAlign tool to establish a multi-task learning model of the end-to-end learning algorithm;
outputting a verification result of the certificate through a multitask learning model of the end-to-end learning algorithm;
wherein, a learning task is established by the three-dimensional characteristic data of the anti-counterfeiting area.
In one embodiment, the performing video capture on a certificate according to different preset angles to obtain a preprocessed video frame, and using the preprocessed video frame as an input of the end-to-end learning algorithm includes:
receiving a certificate turnover video shot by a user according to the indication information, and acquiring a video frame from the certificate turnover video;
identifying four corner coordinates of the certificate in the video frame so as to acquire the video frames of the certificate at different overturning angles according to the four corner coordinates;
detecting video frames of the certificate under different overturning angles to obtain the video frames meeting preset conditions under different overturning angles;
performing white balance processing on video frames meeting preset conditions at different turning angles to obtain a plurality of preprocessed video frames;
taking a plurality of the preprocessed video frames as input to the end-to-end learning algorithm.
In one embodiment, the acquiring two-dimensional feature data from the preprocessed video frames through the underlying network, and fusing the two-dimensional feature data into corresponding three-dimensional feature data, includes:
performing feature extraction on the plurality of preprocessed video frames according to a preset angle sequence through a basic network to obtain two-dimensional feature data of the preprocessed video frames according to the preset angle sequence;
performing feature stacking on the two-dimensional feature data according to the preset angle sequence to fuse the two-dimensional feature data into three-dimensional feature data;
and carrying out three-dimensional convolution operation on the three-dimensional characteristic data to obtain three-dimensional dynamic characteristic data.
In one embodiment, the obtaining three-dimensional feature data of a plurality of anti-counterfeiting areas from the three-dimensional feature data by using a roilign tool includes:
and extracting three-dimensional dynamic characteristic data at different anti-counterfeiting areas from the three-dimensional dynamic characteristic data through a RoIAlign tool according to the characteristic information of the anti-counterfeiting areas.
In one embodiment, the establishing a multi-task learning model of the end-to-end learning algorithm includes:
establishing different learning tasks for the three-dimensional dynamic characteristic data at different anti-counterfeiting areas so as to establish a multi-task learning model of the end-to-end learning algorithm;
the learning task also comprises a learning task established for the three-dimensional dynamic characteristic data of the whole certificate.
In one embodiment, the method further comprises:
the loss function of the end-to-end learning algorithm is the sum of the loss functions of different learning tasks;
wherein the loss functions of different learning tasks configure respective weights.
In one embodiment, the outputting the verification result of the certificate through the multitask learning model of the end-to-end learning algorithm comprises:
outputting a verification result of the certificate through a multitask learning model of the end-to-end learning algorithm;
and the verification result comprises a certificate authenticity verification result and a classification result of each learning task.
In another aspect, there is provided a document authentication system, including:
the algorithm input module is used for carrying out video acquisition on the certificate according to different preset angles so as to obtain a preprocessed video frame, and the preprocessed video frame is used as the input of an end-to-end learning algorithm;
the algorithm model module is used for acquiring two-dimensional feature data from the preprocessed video frames through a basic network and fusing the two-dimensional feature data into corresponding three-dimensional feature data; acquiring three-dimensional characteristic data of a plurality of anti-counterfeiting areas from the three-dimensional characteristic data through a RoIAlign tool to establish a multi-task learning model of the end-to-end learning algorithm, wherein a learning task is established through the three-dimensional characteristic data of each anti-counterfeiting area;
and the algorithm output module outputs the certificate verification result through the multi-task learning model of the end-to-end learning algorithm.
In another aspect, a computer device is provided, which includes a memory, a processor, and a computer program stored on the memory and executable on the processor, wherein the processor executes the computer program to implement the following steps:
carrying out video acquisition on the certificate according to different preset angles to obtain a preprocessed video frame, and taking the preprocessed video frame as the input of an end-to-end learning algorithm;
extracting two-dimensional feature data from the preprocessed video frames through a basic network, and fusing the two-dimensional feature data into corresponding three-dimensional feature data;
acquiring three-dimensional characteristic data of a plurality of anti-counterfeiting areas from the three-dimensional characteristic data through a RoIAlign tool to establish a multi-task learning model of the end-to-end learning algorithm;
and outputting the verification result of the certificate through a multitask learning model of the end-to-end learning algorithm.
In yet another aspect, a computer-readable storage medium is provided, which stores a program that, when executed by a processor, causes the processor to perform the steps of:
carrying out video acquisition on the certificate according to different preset angles to obtain a preprocessed video frame, and taking the preprocessed video frame as the input of an end-to-end learning algorithm;
extracting two-dimensional feature data from the preprocessed video frames through a basic network, and fusing the two-dimensional feature data into corresponding three-dimensional feature data;
acquiring three-dimensional characteristic data of a plurality of anti-counterfeiting areas from the three-dimensional characteristic data through a RoIAlign tool to establish a multi-task learning model of the end-to-end learning algorithm;
and outputting the verification result of the certificate through a multitask learning model of the end-to-end learning algorithm.
Compared with the prior art, the technical scheme of the application has the following advantages:
(1) The method adopts an end-to-end learning algorithm, only the input of original data to the output of a task result is concerned, namely, preprocessed video frames under different angles are used as the input of the end-to-end learning algorithm to output the verification result of certificate authenticity, the global optimization is realized, the accuracy of certificate authenticity verification is improved, in addition, the modeling and training of the end-to-end method are simpler and faster, and the final result is output by inputting an original picture;
(2) The method comprises the steps of combining a RoIAlign tool in an end-to-end learning algorithm, extracting three-dimensional characteristic data of an anti-counterfeiting area through the RoIAlign tool, and calculating a characteristic diagram (characteristic data) of each RoI by a bilinear interpolation method (bilinear interpolation), so that key area information can be fully reserved, and the accuracy of the characteristic data is improved;
(3) The end-to-end learning algorithm also establishes a multi-task learning mechanism aiming at a plurality of anti-counterfeiting areas of the certificate, can obtain the final authenticity of the certificate, can also obtain the characteristic scores of the anti-counterfeiting areas, is favorable for subsequent processing or manual analysis, and enables the algorithm model to be more robust and more generalized, namely, the model well represented on the training data set is also excellent in the test data set and actual production, thereby effectively avoiding the over-fitting problem.
(4) On the same production data set, the false certificate rejection rate of the traditional machine learning method is about 1%, and the true certificate passing rate is about 75%; and the scheme of this application can refuse all false certificates almost, and true certificate percent of pass exceeds 90% simultaneously, has very showing performance promotion.
The certificate authenticity verification method, the certificate authenticity verification system, the computer equipment and the storage medium effectively improve the accuracy of certificate authenticity verification.
Drawings
In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed to be used in the description of the embodiments are briefly introduced below, and it is obvious that the drawings in the following description are only some embodiments of the present application, and it is obvious for those skilled in the art to obtain other drawings based on these drawings without creative efforts.
FIG. 1 is a first method flow diagram of a method for verifying authenticity of a document provided by an embodiment of the application;
FIG. 2 is a flow chart of a second method of a method for verifying authenticity of a document provided by an embodiment of the application;
FIG. 3 is a flow chart of an end-to-end learning algorithm of a certificate authenticity verification method provided by an embodiment of the application;
FIG. 4 is a system diagram of a certificate authenticity verification system provided in an embodiment of the present application;
FIG. 5 is a diagram of an apparatus structure of a computer apparatus according to an embodiment of the present disclosure;
FIG. 6 is an illustration of a document provided in an embodiment of the application.
Detailed Description
In order to make the objects, technical solutions and advantages of the present application more apparent, the present application is described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the present application and are not intended to limit the present application.
As shown in FIG. 6, an example of a picture of an identity document having security features including static security features and dynamic security features. The dynamic security features include: selecting triangular color-changing ink (orange, green and other color changes under different angles) in the area 1; the character gradient feature in the frame selection area 2 (the character gradually changes from H to K under different angles); the sharpness of the small face image changes under different angles of the small face gradual change characteristics in the framing area 3). Static anti-counterfeiting characteristics such as material of the certificate, pattern of the certificate, color gradient of the certificate, etc. The application aims at verifying authenticity of the identity document with the dynamic anti-counterfeiting characteristic.
The first embodiment is as follows:
referring to fig. 1 and 3, fig. 1 is a flow chart of a first method of a certificate authenticity verification method provided by an embodiment of the application; fig. 3 is a flowchart of an end-to-end learning algorithm of a certificate authenticity verification method according to an embodiment of the present application.
The method comprises the following steps:
s101, carrying out video acquisition on the certificate according to different preset angles to obtain a preprocessed video frame, and taking the preprocessed video frame as the input of an end-to-end learning algorithm;
specifically, when the authenticity of the certificate is judged, information acquisition is required to be performed on the certificate at first, and mainly picture and/or video information acquisition is performed on the information on the surface of the certificate. When the video information of the certificate is collected, a user rotates the certificate at certain angles under the camera shooting and collecting device according to the indication to obtain videos of the certificate at different angles; preprocessing videos of the certificate under different angles to obtain preprocessed video frames, and then taking the preprocessed video frames under different angles as input of an end-to-end learning algorithm. The whole learning process of the end-to-end learning algorithm does not carry out artificial subproblem division, but completely passes through a deep learning model to directly learn mapping from original data to expected output, the model is input from the original to the final output as far as possible by reducing manual preprocessing and subsequent processing, more spaces capable of being automatically adjusted according to data are provided for the model, and the overall integrating degree of the model is increased to obtain a global optimal solution.
S102, extracting two-dimensional feature data from the preprocessed video frames through a basic network, and fusing the two-dimensional feature data into corresponding three-dimensional feature data;
specifically, after a preprocessed video frame is obtained, feature extraction is performed on the preprocessed video frame through a basic network such as eficientnet, resNet and the like, and two-dimensional feature data is extracted from the preprocessed video frame, wherein the two-dimensional feature data refers to two-dimensional plane feature data acquired by a certificate, namely a two-dimensional feature map; after the two-dimensional characteristic data is acquired, the two-dimensional characteristic data comprises two-dimensional characteristic data of the certificate under different overturning angles, the two-dimensional characteristic data under different overturning angles are fused to form three-dimensional characteristic data, the three-dimensional characteristic data reflects the mutual relation of the two-dimensional characteristic data under different overturning angles, for example, triangular color-changing ink in the frame selection area 1, and the three-dimensional characteristic data can reflect the change relation among different colors of ink, so that the authenticity of the certificate can be further judged through the three-dimensional characteristic data.
S103, acquiring three-dimensional feature data of a plurality of anti-counterfeiting areas from the three-dimensional feature data through a RoIAlign tool to establish a multi-task learning model of the end-to-end learning algorithm;
specifically, the three-dimensional feature data includes data of the entire certificate, and the authenticity verification of the certificate is to perform authenticity verification on an area with dynamic anti-counterfeiting features on the certificate, so after the three-dimensional feature data is acquired, three-dimensional feature data of the anti-counterfeiting region needs to be extracted from the three-dimensional feature data of the certificate through a roiign tool (region of interest alignment), and the certificate shown in fig. 6 includes three dynamic anti-counterfeiting regions, such as triangular color-changing ink of a frame selection region 1, character gradient features of a frame selection region 2, and small face gradient features of a frame selection region 3. After three-dimensional characteristic data of the three anti-counterfeiting areas are obtained, three learning tasks of an end-to-end learning algorithm are established by the three-dimensional characteristic data of the three anti-counterfeiting areas, and a multi-task learning model of the end-to-end learning algorithm is formed. Wherein, the three-dimensional characteristic data of one anti-counterfeiting area establishes a learning task. The roilign tool is used for extracting anti-counterfeiting area characteristics, roI (region of interest, roI) coordinates are changed into fractions under multiple sampling of a convolutional network, information loss is caused if the information is directly rounded, and model performance is easily reduced, so that the roilign tool is adopted for collecting characteristic data, the roilign tool adopts a bilinear interpolation method (bilinear interpolation) to calculate characteristic graphs (characteristic data) of all RoIs, key area information can be fully reserved, and the authenticity verification accuracy of certificates is improved.
S104, outputting a verification result of the certificate through a multitask learning model of the end-to-end learning algorithm;
specifically, after the multi-task learning model calculates the three-dimensional characteristic data of the anti-counterfeiting areas, the end-to-end learning algorithm can output the certificate authenticity verification result through the multi-task learning model.
The certificate authenticity verification is realized by an end-to-end learning algorithm, only the input of original data to task result output is concerned, namely, preprocessed video frames under different angles are used as the input of the end-to-end learning algorithm, the certificate authenticity verification result is output, the global optimization is realized, the certificate authenticity verification accuracy is improved, in addition, the modeling and training of the end-to-end method are simpler and faster, and the final result is output by inputting an original picture; the end-to-end learning algorithm is combined with a RoIAlign tool, three-dimensional characteristic data of the anti-counterfeiting area are extracted through the RoIAlign tool, a bilinear interpolation method (bilinear interpolation) is adopted to calculate characteristic graphs (characteristic data) of all RoIs, key area information can be fully reserved, and the accuracy of the characteristic data is improved; in addition, the end-to-end learning algorithm also establishes a multi-task learning mechanism aiming at a plurality of anti-counterfeiting areas of the certificate, so that the authenticity of the final certificate can be obtained, the characteristic scores of the anti-counterfeiting areas can be obtained, the follow-up processing or manual analysis is facilitated, the multi-task learning method enables an algorithm model to be more robust and more generalized, namely the model well represented on a training data set is excellent in test data set and actual production, and the over-fitting problem is effectively avoided. On the same production data set, the false certificate rejection rate of the traditional machine learning method is about 1%, and the true certificate passing rate is about 75%; the scheme of the application can almost reject all false certificates, and meanwhile, the passing rate of the true certificates exceeds 90%, so that the performance is remarkably improved. The scheme of the application is not limited to specific identity documents, and can be applied to identity document scenes with dynamic anti-counterfeiting characteristics.
In one embodiment, the video capturing the certificate according to different preset angles to obtain a preprocessed video frame, and using the preprocessed video frame as an input of the end-to-end learning algorithm includes:
receiving a certificate turnover video shot by a user according to the indication information, and acquiring a video frame from the certificate turnover video;
specifically, the user rotates the certificate by certain angles under the camera shooting and collecting device according to the indication to acquire videos of the certificate under different angles, namely the certificate overturning video. The end-to-end learning algorithm receives the certificate turnover video shot by the user according to the indication information, and then each frame of video, namely video frame, is obtained from the certificate turnover video.
Identifying four corner coordinates of the certificate in the video frame so as to acquire the video frames of the certificate at different overturning angles according to the four corner coordinates;
specifically, after a video frame is acquired, the video frame can be identified so as to identify four corner coordinates of a certificate on the video frame, the turning angle of the certificate in the video frame can be obtained by calculating the four corner coordinates of the certificate, images of the certificate at different angles are required when the authenticity of the certificate is judged, and therefore after the video frame is acquired, the angular coordinates of the video frame are identified and calculated so as to obtain video frames at a plurality of turning angles, the characteristic data of the video frames at different turning angles is extracted, and the authenticity of the certificate is further verified.
Detecting the video frames of the certificate under different turnover angles to acquire the video frames meeting preset conditions under different turnover angles;
specifically, after video frames at different turning angles are acquired, the video frames at different turning angles are detected to detect whether the acquired video frames at different turning angles meet preset conditions, that is, whether quality requirements, such as definition and completeness, are met. The method comprises the following steps that a plurality of video frames under different overturning angles are sampled so as to select the video frames meeting preset conditions; for example, 10 sheets are sampled, ensuring that they have various spatial orientations and are arranged at flip angles.
Performing white balance processing on video frames meeting preset conditions under different turnover angles to obtain a plurality of preprocessed video frames;
specifically, after video frames meeting preset conditions at different turning angles are acquired, the video frames need to be preprocessed. Since the certificate may be in different environmental scenes, such as different environments of illumination and color tone, during the actual shooting of the certificate, the environments may affect the extraction of subsequent data. In AI algorithms, improving the quality of data is a faster and easier method than adjusting the network architecture, because preprocessing the data limits the scope of the input data and thus simplifies the algorithm problem. Therefore, the image (video frame) is subjected to white balance processing, and the influence of the illumination environment on the algorithm model is eliminated to obtain the preprocessed video frame.
Taking a plurality of the preprocessed video frames as input to the end-to-end learning algorithm.
Specifically, after the pre-processing video frame is acquired, the acquired pre-processing video frame is used as the input of the end-to-end learning algorithm, so that the verification result is output through the model of the end-to-end learning algorithm.
In one embodiment, the acquiring two-dimensional feature data from the preprocessed video frames through the infrastructure network, and fusing the two-dimensional feature data into corresponding three-dimensional feature data includes:
performing feature extraction on the plurality of preprocessed video frames through a basic network according to a preset angle sequence to obtain two-dimensional feature data of the preprocessed video frames according to the preset angle sequence;
specifically, the pre-processing video frames comprise images with different turning angles, after the pre-processing video frames are obtained, the pre-processing video frames are sequenced according to the turning angles, and then feature extraction is carried out on the sequenced pre-processing video frames through a basic network so as to obtain two-dimensional feature data of the pre-processing video frames according to a preset angle sequence.
Performing feature stacking on the two-dimensional feature data according to the preset angle sequence to fuse the two-dimensional feature data into three-dimensional feature data;
specifically, after two-dimensional feature data of the preprocessed video frames according to a preset angle sequence are acquired, the two-dimensional feature data need to be converted into three-dimensional feature data, that is, the two-dimensional feature data according to the preset angle sequence are subjected to feature stacking to be fused into the three-dimensional feature data. The two-dimensional feature data includes two-dimensional feature data of different flip angles, and therefore the three-dimensional feature data reflects the relationship between the two-dimensional feature data of different flip angles.
And carrying out three-dimensional convolution operation on the three-dimensional characteristic data to obtain three-dimensional dynamic characteristic data.
Specifically, after the three-dimensional feature data is obtained, the dynamic relationship between two-dimensional data at different angles needs to be effectively reflected through the three-dimensional feature data, so that the three-dimensional feature data is obtained through three-dimensional convolution operation, and the change and the mutual relationship between adjacent video frames can be effectively reflected through the three-dimensional dynamic feature data. In fact, adjacent pictures (video frames) can be seen as a sequence, and naturally we can also consider learning these features using a Recurrent Neural Network (RNN), or a recently popular attention (attention) mechanism.
In one embodiment, the obtaining three-dimensional feature data of a plurality of anti-counterfeiting areas from the three-dimensional feature data by a RoIAlign tool includes:
and extracting three-dimensional dynamic characteristic data at different anti-counterfeiting areas from the three-dimensional dynamic characteristic data through a RoIAlign tool according to the characteristic information of the anti-counterfeiting areas.
Specifically, in order to effectively reflect the change and the mutual relationship between adjacent video frames, the three-dimensional dynamic feature data is required to be used for representing, and therefore, the three-dimensional dynamic feature data of the anti-counterfeiting area needs to be extracted from the three-dimensional dynamic feature data of the certificate through a roiign tool. The document shown in figure 6 requires three-dimensional dynamic feature data for three security areas.
In one embodiment, the establishing a multitask learning model of the end-to-end learning algorithm comprises:
establishing different learning tasks for the three-dimensional dynamic characteristic data at different anti-counterfeiting areas so as to establish a multi-task learning model of the end-to-end learning algorithm;
the learning task also comprises a learning task established for the three-dimensional dynamic characteristic data of the whole certificate.
Specifically, after three-dimensional dynamic characteristic data of three anti-counterfeiting areas are obtained, three learning tasks of an end-to-end learning algorithm are established by the three-dimensional dynamic characteristic data of the three anti-counterfeiting areas, and a multi-task learning model of the end-to-end learning algorithm is formed. And establishing the learning task by the three-dimensional characteristic data of each anti-counterfeiting area. In addition, the learning task also comprises a learning task established for the three-dimensional dynamic characteristic data of the whole certificate, and for the certificate shown in fig. 6, four learning tasks can be established, such as the four learning tasks shown in fig. 3, including the learning task of a triangular color-changing ink area, the learning task of a character gradient area, the learning task of a small face gradient area and the learning task of the whole certificate.
In one embodiment, the method further comprises:
the loss function of the end-to-end learning algorithm is the sum of the loss functions of different learning tasks;
wherein the loss functions of different learning tasks configure respective weights.
Specifically, each learning task corresponds to one loss function, the four learning tasks include four loss functions, the loss function of the whole end-to-end learning algorithm is embodied by the sum of the four loss functions, and in addition, the loss function of each learning task can be further provided with corresponding weights to adjust the influence of the four loss functions.
In one embodiment, the outputting the verification result of the certificate through the multitask learning model of the end-to-end learning algorithm comprises:
outputting a verification result of the certificate through a multitask learning model of the end-to-end learning algorithm;
and the verification result comprises a certificate authenticity verification result and a classification result of each learning task.
In particular, the verification result of the certificate can be output through a multitask learning model of an end-to-end learning algorithm, and the verification result comprises the verification result of the whole certificate and the classification result of each anti-counterfeiting area for later result analysis. Wherein, the binary result, i.e. true or false, is a number between 0 and 1 for model prediction, and the larger the number is, the higher the probability of true is, so they can be called score. For example, a score of more than 0.7 represents a true certificate, a verification is passed, a score of less than 0.4 represents a false certificate, a verification is rejected, a manual review is submitted in the middle of 0.4 to 0.7, and a final verification result is determined according to the result of the manual review.
Example two:
referring to fig. 2 and 3, fig. 2 is a flowchart of a second method of a certificate authenticity verification method provided by an embodiment of the present application. In the method shown in fig. 2, the same or similar contents as those in the methods shown in fig. 1 and fig. 3 may refer to the description in the methods shown in fig. 1 and fig. 3, and are not repeated herein.
S201, receiving a certificate turnover video shot by a user according to indication information, and acquiring a video frame from the certificate turnover video;
and the user rotates the certificate at certain angles under the camera shooting and collecting device according to the indication so as to acquire the video of the certificate at different angles, namely the certificate overturning video. The end-to-end learning algorithm receives the certificate turnover video shot by the user according to the indication information, and then each frame of video, namely video frame, is obtained from the certificate turnover video.
S202, identifying four corner coordinates of the certificate in the video frame, and acquiring video frames of the certificate at different overturning angles according to the four corner coordinates;
after the video frame is acquired, the video frame can be identified so as to identify four corner coordinates of the certificate on the video frame, the turning angle of the certificate in the video frame can be obtained by calculating the four corner coordinates of the certificate, and images of the certificate at different angles are required when the authenticity of the certificate is judged.
S203, detecting the video frames of the certificate at different turnover angles to obtain the video frames meeting preset conditions at different turnover angles;
after video frames at different turning angles are acquired, the video frames at different turning angles are detected to detect whether the acquired video frames at different turning angles meet preset conditions, namely whether quality requirements, such as requirements on definition, completeness and the like, are met.
S204, performing white balance processing on the video frames meeting the preset conditions under different turnover angles to obtain a plurality of preprocessed video frames;
after video frames meeting preset conditions at different turning angles are obtained, the video frames need to be preprocessed. In the actual shooting of the certificate, the certificate may be in different environmental scenes, such as different environments of illumination, color tones and the like, which may affect the extraction of subsequent data, so that the white balance processing is performed on the image (video frame), and the influence of the illumination environment on the algorithm model is eliminated to obtain the preprocessed video frame.
S205, taking a plurality of the preprocessed video frames as input of the end-to-end learning algorithm;
after the pre-processing video frame is obtained, the obtained pre-processing video frame is used as the input of an end-to-end learning algorithm, so that the verification result is output through the model of the end-to-end learning algorithm.
S206, performing feature extraction on the plurality of preprocessed video frames according to a preset angle sequence through a basic network to obtain two-dimensional feature data of the preprocessed video frames according to the preset angle sequence;
the method comprises the steps that a plurality of preprocessed video frames comprise images with different turning angles, after the preprocessed video frames are obtained, the preprocessed video frames are sequenced according to the turning angles, and then feature extraction is carried out on the sequenced preprocessed video frames through a basic network so as to obtain two-dimensional feature data of the preprocessed video frames according to a preset angle sequence.
S207, stacking the characteristics of the two-dimensional characteristic data according to the preset angle sequence to fuse the two-dimensional characteristic data into three-dimensional characteristic data;
after the two-dimensional feature data of the preprocessed video frames according to the preset angle sequence are obtained, the two-dimensional feature data need to be converted into three-dimensional feature data, namely, the two-dimensional feature data according to the preset angle sequence are subjected to feature stacking to be fused into the three-dimensional feature data.
S208, performing three-dimensional convolution operation on the three-dimensional characteristic data to obtain three-dimensional dynamic characteristic data;
after the three-dimensional characteristic data is acquired, the dynamic relationship between two-dimensional data at different angles needs to be effectively reflected through the three-dimensional characteristic data, so that the three-dimensional characteristic data is subjected to three-dimensional convolution operation to acquire the three-dimensional dynamic characteristic data, and the change and the mutual relationship between adjacent video frames can be effectively reflected through the three-dimensional dynamic characteristic data.
S209, extracting three-dimensional dynamic characteristic data at different anti-counterfeiting areas from the three-dimensional dynamic characteristic data through a RoIAlign tool according to the characteristic information of the anti-counterfeiting areas;
in order to effectively reflect the change and the mutual relation between adjacent video frames, the change and the mutual relation need to be reflected through three-dimensional dynamic characteristic data, so that the three-dimensional dynamic characteristic data of the anti-counterfeiting area needs to be extracted from the three-dimensional dynamic characteristic data of the certificate through a RoIAlign tool.
S210, establishing different learning tasks for the three-dimensional dynamic characteristic data in different anti-counterfeiting areas to establish a multi-task learning model of the end-to-end learning algorithm;
after three-dimensional dynamic characteristic data of the three anti-counterfeiting areas are obtained, three learning tasks of an end-to-end learning algorithm are established by the three-dimensional dynamic characteristic data of the three anti-counterfeiting areas, and a multi-task learning model of the end-to-end learning algorithm is formed. And establishing the learning task by the three-dimensional characteristic data of each anti-counterfeiting area. In addition, the learning tasks also comprise learning tasks established for three-dimensional dynamic characteristic data of the whole certificate, and for the certificate as shown in fig. 6, four learning tasks can be established, such as the four learning tasks shown in fig. 3, including the learning task of a triangular color-changing ink area, the learning task of a character gradient area, the learning task of a small face gradient area and the learning task of the whole certificate.
S211, outputting the verification result of the certificate through the multitask learning model of the end-to-end learning algorithm.
The certificate verification result can be output through a multitask learning model of an end-to-end learning algorithm, and the verification result comprises the verification result of the whole certificate and the binary classification result of each anti-counterfeiting area for later result analysis.
It should be understood that although the steps in the flowcharts of fig. 1-2 are shown in order as indicated by the arrows, the steps are not necessarily performed in order as indicated by the arrows. The steps are not limited to being performed in the exact order illustrated and, unless explicitly stated herein, may be performed in other orders. Moreover, at least some of the steps in fig. 1-2 may include multiple sub-steps or multiple stages that are not necessarily performed at the same time, but may be performed at different times, and the order of performing the sub-steps or stages is not necessarily sequential, but may be performed in turn or alternately with other steps or at least some of the sub-steps or stages of other steps.
Example three:
referring to fig. 4, fig. 4 is a system structure diagram of a certificate authenticity verification system provided in an embodiment of the present application.
The certificate authenticity verification system of the embodiment includes:
the algorithm input module is used for carrying out video acquisition on the certificate according to different preset angles so as to obtain a preprocessed video frame, and the preprocessed video frame is used as the input of an end-to-end learning algorithm;
when the authenticity of the certificate is judged, firstly, information acquisition is required to be carried out on the certificate, and mainly picture and/or video information acquisition is carried out on the information on the surface of the certificate. When the video information of the certificate is collected, a user rotates the certificate at certain angles under the camera shooting and collecting device according to the indication to obtain videos of the certificate at different angles; preprocessing videos of the certificate under different angles through an algorithm input module to obtain preprocessed video frames, and then taking the preprocessed video frames under different angles as input of an end-to-end learning algorithm.
The algorithm model module is used for acquiring two-dimensional feature data from the preprocessed video frames through a basic network and fusing the two-dimensional feature data into corresponding three-dimensional feature data; acquiring three-dimensional characteristic data of a plurality of anti-counterfeiting areas from the three-dimensional characteristic data through a RoIAlign tool to establish a multi-task learning model of the end-to-end learning algorithm, wherein a learning task is established through the three-dimensional characteristic data of each anti-counterfeiting area;
after the preprocessed video frame is obtained, feature extraction is carried out on the preprocessed video frame through an algorithm model module, two-dimensional feature data are extracted from the preprocessed video frame, and the two-dimensional feature data refer to two-dimensional plane feature data acquired by a certificate, namely a two-dimensional feature map; after the two-dimensional feature data are obtained through the algorithm model module, the two-dimensional feature data comprise two-dimensional feature data of the certificate under different overturning angles, the two-dimensional feature data under the different overturning angles are fused to form three-dimensional feature data, and the three-dimensional feature data reflect the mutual relation of the two-dimensional feature data under the different overturning angles. The three-dimensional characteristic data comprises data of the whole certificate, and the authenticity verification of the certificate is to perform authenticity verification on an area with dynamic anti-counterfeiting characteristics on the certificate, so that after the three-dimensional characteristic data is obtained, the three-dimensional characteristic data of the anti-counterfeiting area needs to be extracted from the three-dimensional characteristic data of the certificate through an algorithm model module.
And the algorithm output module outputs the certificate verification result through the multitask learning model of the end-to-end learning algorithm.
After the multi-task learning model calculates the three-dimensional characteristic data of the anti-counterfeiting areas, the end-to-end learning algorithm can output the result of certificate authenticity verification through the algorithm output module.
In one embodiment, the document authenticity verification system further comprises:
the loss function module is used for configuring a loss function of the whole end-to-end learning algorithm;
each learning task corresponds to one loss function, the four learning tasks comprise four loss functions, the loss function of the whole end-to-end learning algorithm is embodied by the sum of the four loss functions, in addition, the loss function of each learning task can also be provided with corresponding weight to adjust the influence of the four loss functions, and the loss function module is used for realizing the configuration of the loss functions.
For the specific definition of the certificate authenticity verification system, reference may be made to the above definition of the method, which is not described in detail here. All or part of each module in the certificate authenticity verification system can be realized by software, hardware and a combination thereof. The modules can be embedded in a hardware form or independent of a processor in the computer device, and can also be stored in a memory in the computer device in a software form, so that the processor can call and execute operations corresponding to the modules.
Example four:
the embodiment provides a computer device, which comprises a memory, a processor and a computer program stored on the memory and capable of running on the processor, and is characterized in that the processor implements the steps of the certificate authenticity verification method when executing the computer program.
The computer device may be a terminal, and its internal structure diagram may be as shown in fig. 5. The computer device comprises a processor, a memory, a network interface, a display screen and an input device which are connected through a system bus. Wherein the processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device comprises a nonvolatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of an operating system and computer programs in the non-volatile storage medium. 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 method of verifying authenticity of a document. The display screen of the computer equipment can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer equipment can be a touch layer covered on the display screen, a key, a track ball or a touch pad arranged on a shell of the computer equipment, an external keyboard, a touch pad or a mouse and the like.
It should be understood by those skilled in the art that the structure shown in fig. 5 is a block diagram of only a portion of the structure associated with the present application, and does not constitute a limitation on the computing devices to which the present application may be applied, and a particular computing device may include more or less components than shown, or may combine certain components, or have a different arrangement of components.
In one embodiment, a computer device is provided, which includes a memory, a processor, and a computer program stored on the memory and executable on the processor, the processor implementing the following steps when executing the computer program:
carrying out video acquisition on the certificate according to different preset angles to obtain a preprocessed video frame, and taking the preprocessed video frame as the input of an end-to-end learning algorithm;
extracting two-dimensional feature data from the preprocessed video frames through a basic network, and fusing the two-dimensional feature data into corresponding three-dimensional feature data;
acquiring three-dimensional characteristic data of a plurality of anti-counterfeiting areas from the three-dimensional characteristic data through a RoIAlign tool to establish a multi-task learning model of the end-to-end learning algorithm;
and outputting the verification result of the certificate through a multitask learning model of the end-to-end learning algorithm.
Example five:
the present embodiments provide a computer-readable storage medium having stored thereon a computer program which, when executed by a processor, performs the steps of:
carrying out video acquisition on the certificate according to different preset angles to obtain a preprocessed video frame, and taking the preprocessed video frame as the input of an end-to-end learning algorithm;
extracting two-dimensional feature data from the preprocessed video frames through a basic network, and fusing the two-dimensional feature data into corresponding three-dimensional feature data;
acquiring three-dimensional characteristic data of a plurality of anti-counterfeiting areas from the three-dimensional characteristic data through a RoIAlign tool to establish a multi-task learning model of the end-to-end learning algorithm;
and outputting the verification result of the certificate through a multitask learning model of the end-to-end learning algorithm.
It will be understood by those skilled in the art that all or part of the processes of the methods of the embodiments described above may be implemented by hardware instructions of a computer program, which may be stored in a non-volatile computer-readable storage medium, and when executed, may include the processes of the embodiments of the methods described above. Any reference to memory, storage, database or other medium used in the embodiments provided herein can include non-volatile and/or volatile memory. Non-volatile 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 DRAM (SLDRAM), rambus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).
All possible combinations of the technical features in the above embodiments may not be described for the sake of brevity, but should be considered as being within the scope of the present disclosure as long as there is no contradiction between the combinations of the technical features.
The above-mentioned embodiments only express several embodiments of the present application, and the description thereof is more specific and detailed, but not construed as limiting the scope of the invention. It should be noted that, for a person skilled in the art, several variations and modifications can be made without departing from the concept of the present application, which falls within the scope of protection of the present application. Therefore, the protection scope of the present patent shall be subject to the appended claims.
Claims (10)
1. A certificate authenticity verification method is characterized by comprising the following steps:
carrying out video acquisition on the certificate according to different preset angles to obtain a preprocessed video frame, and taking the preprocessed video frame as the input of an end-to-end learning algorithm;
extracting two-dimensional feature data from the preprocessed video frames through a basic network, and fusing the two-dimensional feature data into corresponding three-dimensional feature data;
acquiring three-dimensional characteristic data of a plurality of anti-counterfeiting areas from the three-dimensional characteristic data through a RoIAlign tool to establish a multi-task learning model of the end-to-end learning algorithm;
outputting a verification result of the certificate through a multitask learning model of the end-to-end learning algorithm;
wherein, the three-dimensional characteristic data of one anti-counterfeiting area establishes a learning task.
2. The method for verifying authenticity of a document according to claim 1, wherein said capturing a video of the document at different predetermined angles to obtain a pre-processed video frame, and using the pre-processed video frame as an input of the end-to-end learning algorithm comprises:
receiving a certificate turnover video shot by a user according to the indication information, and acquiring a video frame from the certificate turnover video;
identifying four corner coordinates of the certificate in the video frame so as to acquire the video frames of the certificate under different turnover angles according to the four corner coordinates;
detecting the video frames of the certificate under different turnover angles to acquire the video frames meeting preset conditions under different turnover angles;
performing white balance processing on video frames meeting preset conditions at different turning angles to obtain a plurality of preprocessed video frames;
taking a plurality of the preprocessed video frames as input to the end-to-end learning algorithm.
3. The method for verifying authenticity of a document according to claim 2, wherein the obtaining of two-dimensional feature data from the preprocessed video frames via the basic network and the fusing of the two-dimensional feature data into corresponding three-dimensional feature data comprises:
performing feature extraction on the plurality of preprocessed video frames according to a preset angle sequence through a basic network to obtain two-dimensional feature data of the preprocessed video frames according to the preset angle sequence;
performing feature stacking on the two-dimensional feature data according to the preset angle sequence to fuse the two-dimensional feature data into three-dimensional feature data;
and carrying out three-dimensional convolution operation on the three-dimensional characteristic data to obtain three-dimensional dynamic characteristic data.
4. The method for verifying authenticity of a document according to claim 3, wherein the obtaining three-dimensional feature data of a plurality of anti-counterfeiting areas from the three-dimensional feature data by a RoIAlign tool comprises:
and extracting three-dimensional dynamic characteristic data at different anti-counterfeiting areas from the three-dimensional dynamic characteristic data through a RoIAlign tool according to the characteristic information of the anti-counterfeiting area.
5. The method of claim 4, wherein the establishing a multitask learning model of the end-to-end learning algorithm comprises:
establishing different learning tasks for the three-dimensional dynamic characteristic data at different anti-counterfeiting areas so as to establish a multi-task learning model of the end-to-end learning algorithm;
the learning task also comprises a learning task established for the three-dimensional dynamic characteristic data of the whole certificate.
6. A method of verifying authenticity of a document as claimed in any one of claims 1, further comprising:
the loss function of the end-to-end learning algorithm is the sum of the loss functions of different learning tasks;
wherein the loss functions of different learning tasks configure respective weights.
7. The method for verifying authenticity of a document according to claim 1, wherein said outputting a result of said document verification by a multitask learning model of said end-to-end learning algorithm comprises:
outputting a verification result of the certificate through a multitask learning model of the end-to-end learning algorithm;
and the verification result comprises a certificate authenticity verification result and a classification result of each learning task.
8. A document authenticity verification system, comprising:
the algorithm input module is used for carrying out video acquisition on the certificate according to different preset angles so as to obtain a preprocessed video frame, and the preprocessed video frame is used as the input of an end-to-end learning algorithm;
the algorithm model module is used for acquiring two-dimensional feature data from the preprocessed video frames through a basic network and fusing the two-dimensional feature data into corresponding three-dimensional feature data; acquiring three-dimensional characteristic data of a plurality of anti-counterfeiting areas from the three-dimensional characteristic data through a RoIAlign tool to establish a multi-task learning model of the end-to-end learning algorithm, wherein a learning task is established through the three-dimensional characteristic data of each anti-counterfeiting area;
and the algorithm output module outputs the certificate verification result through the multitask learning model of the end-to-end learning algorithm.
9. A computer device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, characterized in that the steps of the method according to any of claims 1 to 7 are implemented by the processor when executing the computer program.
10. A computer-readable storage medium characterized by: the computer readable storage medium stores a program which, when executed by a processor, causes the processor to perform the steps of the method of any one of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211275484.1A CN115578673A (en) | 2022-10-18 | 2022-10-18 | Certificate authenticity verification method, system, computer device and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211275484.1A CN115578673A (en) | 2022-10-18 | 2022-10-18 | Certificate authenticity verification method, system, computer device and storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115578673A true CN115578673A (en) | 2023-01-06 |
Family
ID=84585435
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211275484.1A Pending CN115578673A (en) | 2022-10-18 | 2022-10-18 | Certificate authenticity verification method, system, computer device and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115578673A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117237682A (en) * | 2023-10-17 | 2023-12-15 | 支付宝(杭州)信息技术有限公司 | Certificate verification method and device, storage medium and electronic equipment |
-
2022
- 2022-10-18 CN CN202211275484.1A patent/CN115578673A/en active Pending
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117237682A (en) * | 2023-10-17 | 2023-12-15 | 支付宝(杭州)信息技术有限公司 | Certificate verification method and device, storage medium and electronic equipment |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| KR102138082B1 (en) | Method, system, device and readable storage medium to realize insurance claim fraud prevention based on multiple image consistency | |
| CN110163193B (en) | Image processing method, image processing device, computer-readable storage medium and computer equipment | |
| CN109886077B (en) | Image recognition method and device, computer equipment and storage medium | |
| CN112215180B (en) | Living body detection method and device | |
| CN111950424B (en) | Video data processing method and device, computer and readable storage medium | |
| KR20200118842A (en) | Identity authentication method and device, electronic device and storage medium | |
| CN110516541B (en) | Text positioning method and device, computer readable storage medium and computer equipment | |
| CN112801057B (en) | Image processing method, image processing device, computer equipment and storage medium | |
| CN107316029B (en) | A kind of living body verification method and equipment | |
| CN108833785A (en) | Fusion method, device, computer equipment and the storage medium of multi-view image | |
| CN112418278A (en) | Multi-class object detection method, terminal device and storage medium | |
| Passos et al. | A review of deep learning‐based approaches for deepfake content detection | |
| CN111881740B (en) | Face recognition method, device, electronic equipment and medium | |
| CN111666932A (en) | Document auditing method and device, computer equipment and storage medium | |
| CN111582155B (en) | Living body detection method, living body detection device, computer equipment and storage medium | |
| CN108921131A (en) | A kind of method and device generating Face datection model, three-dimensional face images | |
| CN115578673A (en) | Certificate authenticity verification method, system, computer device and storage medium | |
| CN113111880A (en) | Certificate image correction method and device, electronic equipment and storage medium | |
| CN114331946A (en) | Image data processing method, device and medium | |
| CN109145703A (en) | Intelligent identification method, device, equipment and medium | |
| CN115082941A (en) | Form information acquisition method and device for form document image | |
| CN112651333B (en) | Silence living body detection method, silence living body detection device, terminal equipment and storage medium | |
| CN113642639B (en) | Living body detection method, living body detection device, living body detection equipment and storage medium | |
| CN112686847B (en) | Identification card image shooting quality evaluation method and device, computer equipment and medium | |
| CN113034433B (en) | Data authentication method, device, equipment and 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 |