CN111324874A

CN111324874A - Certificate authenticity identification method and device

Info

Publication number: CN111324874A
Application number: CN202010069702.0A
Authority: CN
Inventors: 陈志军; 徐崴
Original assignee: Alipay Labs Singapore Pte Ltd
Current assignee: Alipay Labs Singapore Pte Ltd
Priority date: 2020-01-21
Filing date: 2020-01-21
Publication date: 2020-06-23
Anticipated expiration: 2040-01-21
Also published as: CN111324874B

Abstract

One or more embodiments of the present specification provide a method and an apparatus for identifying authenticity of a certificate, wherein the method includes: acquiring multi-frame certificate image information of a target certificate, wherein the target certificate comprises: at least one certificate anti-counterfeiting mark. Identifying the authenticity of a target certificate by using a pre-trained certificate authenticity identification model based on certificate anti-counterfeiting marks in the acquired multi-frame certificate image information to obtain a model output result; the certificate authenticity identification model is obtained by performing multi-frame image fusion training on a preset sample certificate image set by utilizing a deep learning method. And determining the authenticity identification result aiming at the target certificate according to the obtained model output result.

Description

Certificate authenticity identification method and device

Technical Field

The document relates to the technical field of internet, in particular to a certificate authenticity identification method and device.

Background

At present, with the coming of the internet era and the rapid development of the mobile internet technology, the internet is widely applied to daily study, work and life of people. Various daily transactions can be processed and presented by people through the internet using smartphones. The user can install corresponding application programs in the smart phone according to respective actual requirements, for example, a payment application, a financing application, an instant messaging application, a shopping application and the like.

Currently, if a user needs to open a certain service under a certain application program, the user is required to upload certificate image information of a target certificate so as to perform identity verification based on the certificate image information, and open a corresponding internet service for the user after the identity verification is passed (for example, open an electronic wallet service for the user). However, there may be a case where the document image information uploaded by the user is obtained by shooting a counterfeit document that is maliciously forged, so as to achieve the purpose of falsifying the document with false or falsely, and therefore, it is necessary to identify the authenticity of the target document so as to intercept the service opening request of the target document as the forged document in time.

Therefore, a technical scheme for more reliable certificate authenticity identification with high identification accuracy is needed.

Disclosure of Invention

One or more embodiments of the present specification aim to provide a method of authenticating a document. The certificate authenticity identification method comprises the following steps:

acquiring multi-frame certificate image information of a target certificate, wherein the target certificate comprises: at least one certificate anti-counterfeiting mark. Identifying the authenticity of the target certificate by using a pre-trained certificate authenticity identification model based on the certificate anti-counterfeiting mark in the multi-frame certificate image information to obtain a model output result; the certificate authenticity identification model is obtained by performing multi-frame image fusion training on a preset sample certificate image set by using a deep learning method. And determining the authenticity identification result aiming at the target certificate according to the model output result.

An object of one or more embodiments of the present specification is to provide a document authenticity identification device. This certificate true and false recognition device includes:

the certificate image acquisition module acquires multi-frame certificate image information of a target certificate, wherein the target certificate comprises: at least one certificate anti-counterfeiting mark. The certificate authenticity identification module is used for identifying the authenticity of the target certificate based on the certificate anti-counterfeiting mark in the multi-frame certificate image information by utilizing a pre-trained certificate authenticity identification model to obtain a model output result; the certificate authenticity identification model is obtained by performing multi-frame image fusion training on a preset sample certificate image set by using a deep learning method. And the authenticity result determining module is used for determining an authenticity identification result aiming at the target certificate according to the model output result.

An object of one or more embodiments of the present specification is to provide a certificate authenticity identifying apparatus including: a processor; and a memory arranged to store computer executable instructions.

The computer executable instructions, when executed, cause the processor to acquire a plurality of frames of document image information of a target document, wherein the target document comprises: at least one certificate anti-counterfeiting mark. Identifying the authenticity of the target certificate by using a pre-trained certificate authenticity identification model based on the certificate anti-counterfeiting mark in the multi-frame certificate image information to obtain a model output result; the certificate authenticity identification model is obtained by performing multi-frame image fusion training on a preset sample certificate image set by using a deep learning method. And determining the authenticity identification result aiming at the target certificate according to the model output result.

It is an object of one or more embodiments of the present specification to provide a storage medium for storing computer-executable instructions. The executable instructions, when executed by a processor, obtain a plurality of frames of credential image information of a target credential, wherein the target credential includes: at least one certificate anti-counterfeiting mark. Identifying the authenticity of the target certificate by using a pre-trained certificate authenticity identification model based on the certificate anti-counterfeiting mark in the multi-frame certificate image information to obtain a model output result; the certificate authenticity identification model is obtained by performing multi-frame image fusion training on a preset sample certificate image set by using a deep learning method. And determining the authenticity identification result aiming at the target certificate according to the model output result.

Drawings

In order to more clearly illustrate one or more embodiments or prior art solutions of the present specification, the drawings that are needed in the description of the embodiments or prior art will be briefly described below, it is obvious that the drawings in the following description are only some of the embodiments described in one or more of the specification, and that other drawings can be obtained by those skilled in the art without inventive exercise.

FIG. 1 is a first schematic flow chart of a method for authenticating a document according to one or more embodiments of the present disclosure;

FIG. 2 is a second flowchart of a method for authenticating a document according to one or more embodiments of the present disclosure;

FIG. 3 is a third flowchart of a method for authenticating documents according to one or more embodiments of the present disclosure;

FIG. 4 is a fourth flowchart of a method for authenticating documents according to one or more embodiments of the present disclosure;

FIG. 5 is a schematic flow chart of a fifth method for authenticating a document according to one or more embodiments of the present disclosure;

FIG. 6 is a schematic diagram illustrating a specific implementation principle of a certificate authenticity identification method according to one or more embodiments of the present disclosure;

fig. 7 is a schematic diagram illustrating a first module of the certificate authenticity identifying apparatus according to one or more embodiments of the present disclosure;

FIG. 8 is a schematic diagram illustrating a second module of the document authenticity identification apparatus according to one or more embodiments of the present disclosure;

fig. 9 is a schematic structural diagram of a certificate authenticity identification device provided in one or more embodiments of the present specification.

Detailed Description

In order to make the technical solutions in one or more embodiments of the present disclosure better understood, the technical solutions in one or more embodiments of the present disclosure will be clearly and completely described below with reference to the accompanying drawings in one or more embodiments of the present disclosure, and it is obvious that the described embodiments are only a part of one or more embodiments of the present disclosure, but not all embodiments. All other embodiments that can be derived by a person skilled in the art from the embodiments described in one or more of the present specification without inventive step should be considered within the scope of protection of this document.

It should be noted that one or more embodiments and features of the embodiments in the present description may be combined with each other without conflict. Reference will now be made in detail to one or more embodiments of the disclosure, examples of which are illustrated in the accompanying drawings.

One or more embodiments of the present disclosure provide a method and an apparatus for identifying authenticity of a certificate, where after obtaining information of a multi-frame certificate image of a target certificate, a pre-trained certificate authenticity identification model is used to perform fusion identification on the multi-frame certificate image of the target certificate in a multi-frame image fusion manner, so as to obtain an authenticity identification result for the target certificate, and thus, the information change condition of a certificate anti-counterfeit mark in the multi-frame certificate image can be accurately identified automatically based on the multi-frame certificate image, thereby improving the accuracy of certificate authenticity identification, so as to intercept the situation that the target certificate is counterfeit in time, and further improve the accuracy of subsequent service response.

Fig. 1 is a first schematic flow chart of a certificate authenticity identification method provided in one or more embodiments of the present disclosure, where the method in fig. 1 can be executed by a client and also can be executed by a server, where the client can be a mobile terminal such as a smart phone and also can be a terminal device such as an internet of things device, and specifically, the client can be used to collect and acquire multi-frame certificate image information of a target certificate and identify authenticity of the target certificate based on the multi-frame certificate image information, and when the target certificate is determined to be a real certificate, identify whether to execute a corresponding control operation based on the certificate image information, or upload the certificate image information to the server, so that the server continues to verify valid information in the certificate image information; the server may be a background server or a cloud server, and specifically, the server is configured to receive multi-frame certificate image information uploaded by a client, identify whether a target certificate is true or false based on the multi-frame certificate image information, verify valid information in the certificate image information when the target certificate is determined to be a real certificate, and provide a certain service for a user when the valid information is verified, for example, open an account or an authority of a payment application for the user.

As shown in fig. 1, the certificate authenticity identification method at least includes the following steps:

s102, obtaining multi-frame certificate image information of a target certificate, wherein the target certificate comprises: at least one certificate security feature;

the target certificate can be an entity certificate such as an identity card, a passport, a social security card or a medical security card, and the certificate anti-counterfeiting mark can be a mark with an anti-counterfeiting function, wherein attribute information of the target certificate changes when the target certificate is in different postures; for the forged certificate, when the target certificate is positioned at different postures, the anti-counterfeiting mark on the target certificate cannot change, or the change of the anti-counterfeiting mark does not accord with a preset change rule, so that the change information of the certificate anti-counterfeiting mark in the multi-frame certificate image information can be identified by adopting a multi-frame image fusion mode to determine the certificate authenticity identification result of the target certificate;

specifically, after detecting a certificate image acquisition request, a client acquires a certificate video stream of a target certificate by using a camera device, can randomly extract multi-frame certificate image information from the certificate video stream, and can also extract multi-frame certificate image information according to a preset certificate selection rule;

preferably, in order to further improve the identification accuracy of the authenticity of the certificate, the certificate video stream may be acquired when the target certificate has a posture change state, and the certificate posture difference between the pieces of multi-frame certificate image information is greater than a preset posture difference threshold.

S104, identifying the authenticity of the target certificate by using a pre-trained certificate authenticity identification model based on the certificate anti-counterfeiting mark in the acquired multi-frame certificate image information to obtain a model output result; the certificate authenticity identification model is obtained by performing multi-frame image fusion training on a preset sample certificate image set by utilizing a deep learning method;

specifically, after the multiframe certificate image information of the target certificate is obtained, the multiframe certificate image information is input to a certificate authenticity identification model trained in advance, the certificate authenticity identification model is used for carrying out image fusion processing on the multiframe certificate image information, and image fusion feature vectors are obtained, wherein the image fusion feature vectors comprise: the certificate authenticity identification model is used for identifying the authenticity of the target certificate based on the image fusion feature vector and obtaining a model output result, wherein the output of the certificate authenticity identification model is a first probability that the target certificate is a real certificate and/or a second probability that the target certificate is a forged certificate;

wherein, to the training process of certificate true and false recognition model, above-mentioned sample certificate image set includes: a plurality of model training samples, each model training sample comprising: the corresponding relation between the multi-frame sample certificate image of the sample certificate and the certificate authenticity result; wherein, this sample certificate includes: the multi-frame sample certificate image is acquired when the sample certificate is positioned at different postures, and contains the change information of the at least one certificate anti-counterfeiting mark;

specifically, in the training process of the certificate authenticity identification model, the selected sample certificate comprises: two certificates, a true certificate and a counterfeit certificate; and performing deep learning on the corresponding relation between the change information of the certificate anti-counterfeiting mark in the multi-frame sample certificate image in each model training sample and the authenticity of the certificate by using a deep learning method and adopting an image fusion mode so as to perform optimization training on model parameters in a preset deep learning network model to obtain a certificate authenticity identification model.

S106, determining an authenticity identification result aiming at the target certificate according to the obtained model output result;

wherein, the model output result may include: the method includes the steps that a first probability that a target certificate is a real certificate and/or a second probability that the target certificate is a fake certificate are/is determined, and specifically, if the first probability is larger than a first preset threshold value, the target certificate is determined to be the real certificate, or if the second probability is larger than a second preset threshold value, the target certificate is determined to be the fake certificate.

The method comprises the steps that in the process of identifying the authenticity of a target certificate by using a trained certificate authenticity identification model, single-frame identification is not carried out on certificate image information of each frame of target certificate to obtain attribute information of a certificate anti-counterfeiting mark, and then the attribute information of the certificate anti-counterfeiting mark corresponding to a plurality of frames of certificate images is compared to identify the authenticity of the target certificate; but directly carries out fusion processing on multi-frame certificate images to obtain image fusion characteristic vectors, and then identifies the authenticity of the target certificate based on the image fusion characteristic vectors, thereby avoiding the problem that the false judgment rate of the authenticity of the certificate is increased due to the identification error of the anti-counterfeiting mark of a single-frame certificate image, and further improving the identification accuracy of the authenticity of the certificate.

In one or more embodiments of the present specification, after obtaining information of multiple frames of certificate images of a target certificate, the multi-frame certificate images of the target certificate are fused and identified by using a pre-trained certificate authenticity identification model in a multi-frame image fusion manner, so as to obtain an authenticity identification result for the target certificate, which can accurately identify information change conditions of a certificate anti-counterfeiting mark in the multi-frame certificate images based on the multi-frame certificate images, thereby improving accuracy of certificate authenticity identification, so as to intercept a situation that the target certificate is a counterfeit certificate in time, and further improve accuracy of subsequent service response.

Specifically, for the case that the client identifies the authenticity of the target certificate, the client collects a certificate video stream of the target certificate, and determines the authenticity identification result for the target certificate based on the certificate video stream and the steps S102 to S106; when the target certificate is determined to be a real certificate, identifying whether to execute corresponding control operation or not based on effective information in the certificate image information, or uploading the certificate image information to a server so that the server verifies the effective information in the certificate image information; in addition, the certificate authenticity identification model can be obtained by training the server based on the sample certificate image set in advance, and the trained certificate authenticity identification model is stored to the local client.

Correspondingly, aiming at the situation that the server identifies the authenticity of the target certificate, the client acquires the certificate video stream of the target certificate, the client directly uploads the certificate video stream to the server, the server extracts multi-frame certificate image information from the certificate video stream, or the client extracts multi-frame certificate image information from the certificate video stream and uploads the multi-frame certificate image information to the server, so that the server determines the authenticity identification result of the target certificate based on the steps S102 to S106; when the target certificate is determined to be a real certificate, verifying effective information in the certificate image information; the process of extracting multi-frame certificate image information according to the preset certificate selection rule in the certificate video stream collected from the camera device can be executed by a client or a server, and the client can directly upload the collected certificate video stream to the server under the condition of execution by the server.

Wherein, need in advance based on sample certificate image collection and adopt the mode of image fusion, carry out the parameter training to predetermineeing the degree of depth learning network model, obtain the certificate true and false recognition model that trains, this certificate true and false recognition model includes: a deep learning network model; correspondingly, as shown in fig. 2, in the foregoing step S104, the pre-trained certificate authenticity identifying model is used, and based on the certificate anti-counterfeit identifier in the acquired multi-frame certificate image information, the authenticity of the target certificate is identified, so as to obtain a model output result, which specifically includes:

s1041, carrying out image fusion processing on multi-frame certificate image information by using a pre-trained deep learning network model to obtain an image fusion feature vector, wherein the image fusion feature vector comprises: the characteristic data is used for representing the change information of the at least one certificate anti-counterfeiting mark;

specifically, after multi-frame certificate image information of a target certificate is acquired, the multi-frame certificate image information is input into a pre-trained deep learning network model, and the deep learning network model performs image fusion processing on the multi-frame certificate image information to obtain an image fusion feature vector, wherein a network layer used for image fusion can be a certain preset feature layer in the deep learning network model, for example, the deep learning network model is taken as a convolutional neural network model, and the network layer used for image fusion can be any one of an input layer, a convolutional layer, a pooling layer or a full-link layer.

S1042, identifying the authenticity of the target certificate based on the obtained image fusion feature vector to obtain a model output result;

specifically, after an image fusion feature vector of the target certificate is obtained, a final model output result is obtained based on the image fusion feature vector, wherein the model output result includes: a first probability that the target document is a genuine document and/or a second probability that the target document is a counterfeit document.

Taking the last network layer of the deep learning network model as an example of a full connection layer, performing full connection operation on the image fusion feature vector by using the full connection layer, wherein the output of the full connection network is 2-dimensional output (namely the position of 0 and the position of 1), namely a certificate two-classification recognition result, and the probability value corresponding to the position of 0 represents a second probability that the target certificate is a forged certificate; correspondingly, the probability value corresponding to the position of 1 represents the first probability that the target certificate is the real certificate.

In a specific implementation, the deep learning network model includes: a convolutional neural network model, wherein the convolutional neural network model may be a CNN model, the convolutional neural network model comprising: at least one of an input layer, a convolution layer, a pooling layer, and a full-link layer;

wherein, can choose for use different network layers to carry out image fusion according to actual demand and handle, based on this, to carrying out the process of fusing to multiframe certificate image information, can adopt following three kinds of image fusion modes, specifically do:

the first image fusion mode, S1041, performs image fusion processing on the multi-frame certificate image information by using a pre-trained deep learning network model to obtain an image fusion feature vector, and specifically includes:

firstly, carrying out image fusion processing on multi-frame certificate image information by using an input layer in a pre-trained convolutional neural network model to obtain certificate image fusion data;

secondly, carrying out convolution processing on the obtained certificate image fusion data by utilizing a convolution layer in a pre-trained convolutional neural network model to obtain convolution characteristics of the fusion data;

and step three, performing pooling processing on the obtained convolution features of the fusion data by using a pooling layer in a pre-trained convolution neural network model to obtain image fusion feature vectors.

In one or more embodiments of the present description, an input layer in a convolutional neural network model may be used as a network layer for image fusion, that is, an image fusion process is performed on multiple frames of certificate image information at the input layer to obtain certificate image fusion data, for example, the multiple frames of certificate image information are subjected to a superposition process, and then the obtained certificate image fusion data are sequentially input to a convolutional layer and a pooling layer to perform a convolution and pooling process on the certificate image fusion data to obtain a final image fusion feature vector;

then, aiming at the condition that the last network layer of the convolutional neural network model is a full connection layer, inputting the obtained image fusion characteristic vector to the full connection layer, and performing full connection operation on the image fusion characteristic vector by the full connection layer to obtain a final model output result;

in order to further improve the identification accuracy of the authenticity of the certificate, a time sequence neural network model can be introduced, the obtained image fusion characteristic vectors are input into the time sequence neural network model, the time sequence neural network model is used for carrying out corresponding processing on the image fusion characteristic vectors corresponding to a plurality of moments to obtain comprehensive time sequence characteristic vectors, the last network layer of the time sequence neural network model is a full connection layer, and the full connection layer carries out full connection operation on the comprehensive time sequence characteristic vectors to obtain a final model output result.

The second fusion mode, S1041, performs image fusion processing on the multi-frame certificate image information by using a pre-trained deep learning network model to obtain an image fusion feature vector, specifically including:

inputting multi-frame certificate image information into a convolutional layer by utilizing an input layer in a pre-trained convolutional neural network model;

secondly, carrying out convolution processing and image fusion processing on multi-frame certificate image information by using the convolution layer to obtain certificate image fusion data;

in order to extract characteristic data which can represent the change of attribute information of the anti-counterfeiting mark better from multi-frame certificate image information, the convolutional neural network model can comprise a plurality of convolutional layers;

in specific implementation, the first convolutional layer can be used as a network layer for image fusion, specifically, the first convolutional layer is used for carrying out image fusion processing on multi-frame certificate image information to obtain initial certificate image fusion data, then the initial certificate image fusion data is input to the next convolutional layer to carry out convolution processing on the initial certificate image fusion data until the last convolutional layer to obtain final certificate image fusion data, and the final certificate image fusion data is input to the pooling layer;

correspondingly, the middle convolutional layer can also be used as a network layer for image fusion, specifically, the previous convolutional layers are used for respectively carrying out convolution processing on multiple frames of certificate image information to obtain certificate image convolution characteristics respectively corresponding to each frame of certificate image information, the middle convolutional layers are used for carrying out characteristic fusion processing on the obtained multiple certificate image convolution characteristics to obtain initial certificate image fusion data, the initial certificate image fusion data are input into the next convolutional layer to carry out convolution processing on the initial certificate image fusion data until the last convolutional layer to obtain final certificate image fusion data, and the final certificate image fusion data are input into the pooling layer;

correspondingly, the tail convolutional layer can also be used as a network layer for image fusion, specifically, the former convolutional layer is used for respectively carrying out convolution processing on multi-frame certificate image information to obtain certificate image convolution characteristics respectively corresponding to each frame of certificate image information, the tail convolutional layer is used for carrying out characteristic fusion processing on the obtained plurality of certificate image convolution characteristics to obtain certificate image fusion data, and the certificate image fusion data are input into the pooling layer.

And step three, performing pooling processing on the obtained certificate image fusion data by using a pooling layer in a pre-trained convolutional neural network model to obtain an image fusion characteristic vector.

In one or more embodiments of the present description, not only the input layer may be used as a network layer for image fusion, but also a convolutional layer in a convolutional neural network model may be used as a network layer for image fusion, that is, image fusion and convolution processing are performed on multi-frame certificate image information at the convolutional layer to obtain certificate image fusion data, and then the obtained certificate image fusion data is input to the pooling layer to perform pooling processing on the certificate image fusion data to obtain a final image fusion feature vector;

The third fusion mode, S1041, performs image fusion processing on the multi-frame certificate image information by using a pre-trained deep learning network model to obtain an image fusion feature vector, specifically including:

secondly, carrying out convolution processing on a plurality of frames of certificate image information respectively by using the convolution layer to obtain the certificate image convolution characteristics of each frame of the certificate image information;

step three, performing pooling processing on the obtained certificate image convolution characteristics by using a pooling layer in a pre-trained convolution neural network model to obtain certificate image pooling characteristics of each frame of certificate image information;

and step four, carrying out fusion processing on the pooling features of the multiple certificate images by using a fully connected layer in a pre-trained convolutional neural network model to obtain image fusion feature vectors.

In one or more embodiments of the present description, not only the input layer or the convolutional layer may be used as a network layer for image fusion, but also the full connection layer in the convolutional neural network model may be used as a network layer for image fusion, that is, multi-frame certificate image information is sequentially input to the convolutional layer and the pooling layer to perform convolution and pooling on the multi-frame certificate image information, so as to obtain certificate image pooling features corresponding to each frame of certificate image information, and then the full connection layer is used to perform feature fusion processing on the multiple certificate image pooling features, so as to obtain an image fusion feature vector;

then, aiming at the condition that the last network layer of the convolutional neural network model is a full connection layer, the full connection layer is also used for performing full connection operation on the image fusion characteristic vector to obtain a final model output result;

In addition, when the method is specifically implemented, a pooling layer in the convolutional neural network model can be used as a network layer for image fusion, namely, multi-frame certificate image information is input into a convolutional layer firstly, so that convolution processing is carried out on the multi-frame certificate image information, and certificate image convolution characteristics corresponding to all the frames of certificate image information are obtained; and performing feature fusion and pooling on the convolution features of the multiple certificate images by using the pooling layer to obtain image fusion feature vectors.

Wherein, in order to further improve the degree of accuracy of certificate true and false recognition result, can introduce the time sequence neural network model, after utilizing convolution neural network model to obtain image fusion eigenvector, not directly utilize the full tie layer to carry out the full tie operation to this image fusion eigenvector, obtain two categorised recognition results of certificate, but input this image fusion eigenvector into time sequence neural network model, obtain synthesizing the time sequence eigenvector, again based on this synthesize the time sequence eigenvector, obtain two categorised recognition results of certificate, based on this, above-mentioned certificate true and false recognition model still includes: a time-series neural network model;

correspondingly, as shown in fig. 3, in the step S1042, based on the obtained image fusion feature vector, the identifying the authenticity of the target certificate is performed to obtain a model output result, which specifically includes:

s10421, inputting the image fusion characteristic vectors at a plurality of target moments in a preset time period to a pre-trained time sequence neural network model;

specifically, after an image fusion feature vector is obtained by using a convolutional neural network model and based on multi-frame certificate image information, the image fusion feature vector is used as input data of a time sequence neural network model.

S10422, performing time sequence response on the image fusion characteristic vectors at a plurality of target moments by using the time sequence neural network model to obtain a comprehensive time sequence characteristic vector;

the time sequence neural network model can be an LTSM model, specifically, image fusion characteristic vectors at a plurality of target moments obtained by using a convolutional neural network model are sequentially input into the time sequence neural network model, and the output of the time sequence neural network model is determined as a comprehensive time sequence characteristic vector; then inputting the comprehensive time sequence characteristic vector to a preset full connection layer;

specifically, the integrated time series feature vector is determined according to a plurality of time series hidden vectors obtained by extracting sequence features of the image fusion feature vector at a plurality of target times by using a time series neural network model, for example, any one selected from the obtained plurality of time series hidden vectors may be determined as the integrated time series feature vector, or the integrated time series feature vector may be obtained by performing a weighted operation on the obtained plurality of time series hidden vectors.

S10423, performing full-connection operation on the obtained comprehensive time sequence characteristic vector by using a preset full-connection layer to obtain a model output result;

specifically, the comprehensive time sequence feature vector obtained by using the time sequence neural network model is input into a preset full connection layer, full connection operation is carried out on the comprehensive time sequence feature vector by using the full connection layer, the output of the full connection network is 2-dimensional output (namely the position of 0 and the position of 1), namely a certificate two-classification recognition result, wherein the probability value corresponding to the position of 0 represents a second probability that the target certificate is a forged certificate; correspondingly, the probability value corresponding to the position of 1 represents the first probability that the target certificate is the real certificate.

Further, in the process of identifying the authenticity of the certificate by using a pre-trained deep learning network model, corresponding processing is mainly carried out on multi-frame certificate image information by adopting an image fusion mode to obtain an image fusion characteristic vector, the image fusion characteristic vector is used for representing the information change condition of the certificate anti-counterfeiting mark in the multi-frame certificate image information, and the authenticity of the target certificate is identified based on the image fusion characteristic vector; considering that the larger the posture difference in the multi-frame certificate image information is, the more obvious the attribute information change of the anti-counterfeit mark in the real certificate is, therefore, in order to further improve the identification accuracy of the certificate authenticity, for the acquisition process of the multi-frame certificate image for certificate authenticity identification, based on this, as shown in fig. 4, the above S102 acquires the multi-frame certificate image information of the target certificate, which specifically includes:

s1021, extracting a plurality of video key frames from the video stream information of the target certificate, wherein the video stream information comprises: certificate image frames of the target certificate under a plurality of postures;

specifically, after a client detects a certificate image acquisition request, a camera device is used for acquiring video stream information of a target certificate, and then a plurality of video key frames are extracted from the video stream information, wherein prompt information for representing shaking of the certificate can be displayed for a user in order to ensure posture difference between certificate image frames in the video stream information, namely, the camera device is used for acquiring the video stream information of the target certificate when the target certificate is in a shaking state.

S1022, aiming at each video key frame, determining the corresponding certificate attitude information of the video key frame by using a preset certificate attitude identification algorithm;

specifically, after a plurality of video key frames are extracted, the plurality of video key frames are used as input data of a preset certificate gesture recognition algorithm, the preset certificate gesture recognition algorithm is utilized to perform image preprocessing on the plurality of video key frames, gesture recognition is performed on the plurality of preprocessed video key frames, and certificate gesture information corresponding to each video key frame is obtained; wherein the image preprocessing comprises: and at least one of certificate corner positioning, image alignment and size adjustment.

S1023, selecting a preset number of video key frames according to the certificate attitude information corresponding to each video key frame, wherein the certificate attitude difference among the preset number of video key frames is greater than a preset attitude difference threshold value;

specifically, after certificate posture information corresponding to each video key frame is determined, comparing the certificate posture information, and selecting a plurality of video key frames meeting preset posture difference requirements, for example, selecting N video key frames with large posture difference from M video key frames, wherein M is larger than N;

in the process of selecting N video key frames, a plurality of video key frames may be classified according to a gesture type based on certificate gesture information corresponding to each video key frame to obtain a plurality of video key frame sets, where the gesture type may include: classifying the postures of the front, the left slant, the upper slant and the like, and then respectively selecting a certain number of video key frames from the video key frame sets corresponding to different posture types to obtain N video key frames with large posture difference.

S1024, determining the selected video key frames with the preset number as multi-frame certificate image information of the target certificate;

specifically, after N frames of certificate image information with large posture difference are selected based on the certificate posture information, the N frames of certificate image information are input into a certificate authenticity identification model trained in advance together, so that the certificate authenticity identification model identifies the authenticity of the target certificate based on the N frames of certificate image information in an image fusion processing mode.

It should be noted that, in the case that the server identifies the authenticity of the target certificate, after the client acquires the video stream information of the target certificate by using the camera device, the client may select the multi-frame certificate image information with the posture difference meeting the preset requirement by using the steps S1021 to S1024, and specifically, the client uploads the multi-frame certificate image information to the server, so that the server identifies the authenticity of the target certificate based on the multi-frame certificate image information by using the steps S104 to S106;

correspondingly, the server may also select multi-frame certificate image information with the posture difference meeting the preset requirement by using the steps S1021 to S1024, specifically, the client uploads the acquired video stream information of the target certificate to the server, so that the server selects multi-frame certificate image information with the posture difference meeting the preset requirement by using the steps S1021 to S1024 based on the video stream information; and based on the multi-frame certificate image information, the authenticity of the target certificate is identified by the steps S104 to S106.

As shown in fig. 5, before acquiring the multi-frame certificate image information of the target certificate in S102, the training process for the certificate authenticity identification model further includes:

s108, sample video streams corresponding to a plurality of sample certificates are obtained, wherein each sample video stream comprises: certificate image frames of the sample certificate under a plurality of postures;

wherein at least one sample video stream can be captured for each sample document, which can be captured with the camera device while the sample document is in a shaken state.

S110, extracting a plurality of sample key frames from a sample video stream of each sample certificate;

specifically, each sample certificate corresponds to at least one sample video stream, key frame extraction is performed on each sample video stream, a plurality of sample key frames corresponding to each sample video stream are obtained, that is, each sample video stream corresponds to a plurality of sample key frames, and then a plurality of model training samples corresponding to the sample video stream are determined based on the plurality of sample key frames.

S112, determining a sample certificate image set based on a plurality of sample key frames corresponding to each sample certificate, wherein the sample certificate image set comprises: a plurality of model training samples, each model training sample comprising: multiple frames of sample certificate images;

specifically, at least one group of sample certificate images with preset number is selected based on a plurality of sample key frames corresponding to each sample video stream; determining each group of the certificate images of the samples with the preset number as a model training sample; determining the combination of a plurality of model training samples corresponding to each sample certificate as a sample certificate image set;

wherein each model training sample may include: and in the training process of the certificate authenticity identification model, the selected N frames of sample certificate images with the posture difference meeting the preset difference requirement are taken as a whole, and the N frames of sample certificate images are subjected to fusion processing to perform optimization training on the parameters of the preset deep learning network model.

And S114, performing fusion training on the multi-frame sample certificate images by using a deep learning method and based on the plurality of model training samples to obtain a certificate authenticity identification model.

Wherein, the sample certificate of choosing includes: two certificates, a true certificate and a counterfeit certificate; deep learning is carried out on the corresponding relation between the change information of the certificate anti-counterfeiting mark in the multi-frame sample certificate image in each model training sample and the authenticity of the certificate by utilizing a deep learning method and adopting an image fusion mode, so that model parameters in a preset deep learning network model are optimized and trained to obtain a certificate authenticity identification model;

specifically, for the process of training a sample based on multiple models to obtain a certificate authenticity identification model, reference may be made to the above three image fusion modes, where the fusion mode used in the use process of the certificate authenticity identification model is the same as the fusion mode used in the training process of the certificate authenticity identification model; for example, in the training process of the certificate authenticity identification model, the convolutional layer in the convolutional neural network model is used as the network layer for image fusion, in the process of identifying the authenticity of the target certificate by using the certificate authenticity identification model, the convolutional layer in the convolutional neural network model is also used as the network layer for image fusion, that is, the second fusion mode is adopted, and the image fusion characteristic vector is obtained by performing image fusion processing on the image information of a plurality of frames of certificates by using the pre-trained deep learning network model.

Wherein, in order to further improve the recognition accuracy of the certificate true and false recognition model that the training obtained, when choosing for use model training sample, consider the gesture difference between multiframe sample certificate image in every model training sample, from a plurality of sample key frames that every sample video stream corresponds, choose the multiframe sample certificate image that the gesture difference is big as a model training sample, obtain the sample certificate image set that is used for discerning model training by a plurality of model training samples again, based on this, above-mentioned S112, based on a plurality of sample key frames that each sample certificate corresponds, confirm sample certificate image set, specifically include:

determining certificate posture information corresponding to a plurality of sample key frames of each sample certificate by using a preset certificate posture recognition algorithm aiming at each sample certificate;

specifically, after a plurality of sample key frames are extracted from each sample video stream, the plurality of sample key frames are used as input data of a preset certificate gesture recognition algorithm, the preset certificate gesture recognition algorithm is utilized to perform image preprocessing on the plurality of sample key frames, and gesture recognition is performed on the plurality of preprocessed sample key frames to obtain certificate gesture information corresponding to each sample key frame; wherein the image preprocessing comprises: and at least one of certificate corner positioning, image alignment and size adjustment.

Selecting a plurality of groups of preset number of sample certificate images according to the certificate attitude information corresponding to each sample key frame, wherein the certificate attitude difference between each group of preset number of sample certificate images is greater than a preset attitude difference threshold value;

selecting a plurality of sample key frames with the posture difference meeting the preset posture difference requirement from a plurality of sample key frames corresponding to each sample video stream based on the certificate posture information of each sample key frame, and taking the plurality of sample key frames as a group of sample certificate images with the preset number;

for example, selecting N sample key frames with large attitude difference from M sample key frames corresponding to each sample video stream as a group of preset number of sample certificate images, wherein M is larger than N; selecting N different sample key frames with large attitude difference for multiple times from M sample key frames corresponding to each sample video stream to obtain a plurality of model training samples corresponding to the sample video stream, wherein each model training sample comprises: and the posture difference meets the N frames of sample certificate images with the preset difference requirement.

Specifically, in the process of selecting N sample key frames, a plurality of sample key frames may be classified according to a posture type based on certificate posture information corresponding to each sample key frame to obtain a plurality of sample key frame sets, where the posture type may include: classifying the postures of the front face, the left slant, the upper slant and the like, then respectively selecting a certain number of sample key frames from sample key frame sets corresponding to different posture types to obtain N sample key frames with large posture difference, and taking the N sample key frames as a group of sample certificate images with preset number.

Determining a sample certificate image set according to a plurality of groups of preset number of sample certificate images corresponding to each sample certificate;

the method comprises the steps that a plurality of groups of preset number of sample certificate images can be obtained through one sample video stream of each sample certificate, each group of preset number of sample certificate images is determined to be a model training sample, and therefore a sample certificate image set consisting of a plurality of model training samples is formed; the more the number of the sample certificates is, or the more the number of the sample video streams of each sample certificate is, the more the obtained model training samples are.

In a specific embodiment, as shown in fig. 6, a schematic diagram of a specific implementation principle of certificate authenticity identification is provided, specifically:

A. aiming at the process of training to obtain a certificate authenticity identification model based on a sample certificate image set, the method specifically comprises the following steps:

a1, acquiring sample video streams corresponding to a plurality of sample certificates respectively, wherein each sample video stream is acquired by a camera device when the sample certificate is in a shaking state, and the sample certificate comprises: true sample certificates and counterfeit sample certificates;

it should be noted that, for convenience of description, in fig. 6, each sample certificate corresponds to one sample video stream as an example, and in a specific implementation, multiple sample video streams may also be acquired for each sample certificate;

a2, extracting a plurality of sample key frames from a sample video stream corresponding to each sample certificate;

a3, aiming at each sample certificate, determining certificate posture information corresponding to a plurality of extracted sample key frames by using a preset certificate posture recognition algorithm;

a4, aiming at each sample certificate, selecting a plurality of model training samples according to certificate posture information corresponding to a plurality of extracted sample key frames respectively to obtain a plurality of model training samples containing N sample key frames, wherein the certificate posture difference between the models of the N sample key frames in each model training sample is larger than a preset posture difference threshold value, and the corresponding relation between the N sample key frames and the certificate authenticity is known;

for example, in fig. 6, for a sample video stream 1 corresponding to a sample certificate 1, m model training samples selected are: the method comprises the following steps of 1, obtaining a model training sample 11, 12 and …, wherein each model training sample comprises N sample key frames; aiming at a sample video stream n corresponding to a sample certificate n, m model training samples selected are respectively as follows: model training sample N1, model training samples N2 and … model training samples nm, wherein each model training sample comprises N sample key frames;

a5, performing fusion training on N sample key frames in each model training sample by using a deep learning method and based on the obtained multiple model training samples to obtain a certificate authenticity identification model, wherein the certificate authenticity identification model can comprise: the system comprises a CNN model and an LSTM model, wherein the last network layer of the LSTM model is a preset full connection layer;

B. To the certificate true and false recognition model that utilizes training to based on the multiframe certificate image information of target certificate, carry out the process of discerning to the target certificate true and false, specifically include:

b1, acquiring multi-frame certificate image information of the target certificate, wherein the target certificate comprises: at least one certificate anti-counterfeiting mark, the multiframe certificate image information;

specifically, after detecting a certificate image acquisition request, a client acquires a certificate video stream of a target certificate by using a camera device, and selects N frames of certificate image information with a certificate attitude difference larger than a preset attitude difference threshold value from the certificate video stream;

b2, inputting the selected N frames of certificate image information into a pre-trained CNN model, and carrying out image fusion processing on the multi-frame certificate image information by using the CNN model to obtain an image fusion characteristic vector, wherein the image fusion characteristic vector comprises: the characteristic data is used for representing the change information of the at least one certificate anti-counterfeiting mark;

the input of the CNN model is w h (3N), wherein w and h represent the size of the certificate image, 3 represents the number of channels of the image RGB, and N represents the number of the certificate images.

B3, sequentially inputting the image fusion characteristic vectors at a plurality of target moments in a preset time period to a pre-trained LSTM model, and performing time sequence characteristic extraction on the image fusion characteristic vectors by using the LSTM model to obtain a comprehensive time sequence characteristic vector;

b4, performing full-connection operation on the obtained comprehensive time sequence characteristic vector by using a preset full-connection layer to obtain a two-classification recognition result; wherein, the two classification recognition results comprise: a first probability that the target certificate is a real certificate and/or a second probability that the target certificate is a fake certificate;

b5, judging whether the first probability is larger than a first preset threshold value, if so, determining that the target certificate is a real certificate, and if not, determining that the target certificate is a fake certificate.

It should be noted that the processes from a1 to a5 are preferably executed by a server, and the processes from B1 to B5 may be executed by a client, especially an information processing module in the client, and may also be executed by the server.

In one or more embodiments of the present specification, a certificate authenticity identification method includes, first, obtaining multi-frame certificate image information of a target certificate, where the target certificate includes: at least one certificate security feature; then, identifying the authenticity of the target certificate by using a pre-trained certificate authenticity identification model based on the certificate anti-counterfeiting mark in the acquired multi-frame certificate image information to obtain a model output result; and finally, determining the authenticity identification result aiming at the target certificate according to the obtained model output result. In one or more embodiments of the present specification, after obtaining information of multiple frames of certificate images of a target certificate, the multi-frame certificate images of the target certificate are fused and identified by using a pre-trained certificate authenticity identification model in a multi-frame image fusion manner, so as to obtain an authenticity identification result for the target certificate, which can accurately identify information change conditions of a certificate anti-counterfeiting mark in the multi-frame certificate images based on the multi-frame certificate images, thereby improving accuracy of certificate authenticity identification, so as to intercept a situation that the target certificate is a counterfeit certificate in time, and further improve accuracy of subsequent service response.

Corresponding to the certificate authenticity identification method described in fig. 1 to 6, based on the same technical concept, one or more embodiments of the present specification further provide a certificate authenticity identification apparatus, and fig. 7 is a schematic diagram of a first module composition of the certificate authenticity identification apparatus provided in one or more embodiments of the present specification, the apparatus is configured to perform the certificate authenticity identification method described in fig. 1 to 6, and as shown in fig. 7, the apparatus includes:

a certificate image acquisition module 701, which acquires multi-frame certificate image information of a target certificate, wherein the target certificate includes: at least one certificate security feature;

the certificate authenticity identification module 702 is used for identifying the authenticity of the target certificate based on the certificate anti-counterfeiting mark in the multi-frame certificate image information by utilizing a pre-trained certificate authenticity identification model to obtain a model output result; the certificate authenticity identification model is obtained by performing multi-frame image fusion training on a preset sample certificate image set by utilizing a deep learning method;

and the authenticity result determining module 703 is configured to determine an authenticity identification result for the target certificate according to the model output result.

Optionally, the certificate authenticity identification model comprises: a deep learning network model;

the certificate authenticity identification module 702, which:

and carrying out image fusion processing on the multi-frame certificate image information by using the deep learning network model to obtain an image fusion characteristic vector, wherein the image fusion characteristic vector comprises: the characteristic data is used for representing the change information of the at least one certificate anti-counterfeiting mark;

and identifying the authenticity of the target certificate based on the image fusion feature vector to obtain a model output result.

Optionally, the deep learning network model includes: a convolutional neural network model;

the certificate authenticity identification module 702, which:

carrying out image fusion processing on the multi-frame certificate image information by utilizing an input layer in the convolutional neural network model to obtain certificate image fusion data;

carrying out convolution processing on the certificate image fusion data by utilizing a convolution layer in the convolution neural network model to obtain convolution characteristics of the fusion data;

and pooling the convolution features of the fusion data by using a pooling layer in the convolution neural network model to obtain an image fusion feature vector.

the certificate authenticity identification module 702, which:

inputting the multi-frame certificate image information to a convolutional layer by utilizing an input layer in the convolutional neural network model;

carrying out convolution processing and image fusion processing on the multi-frame certificate image information by using the convolution layer to obtain certificate image fusion data;

and performing pooling processing on the certificate image fusion data by using a pooling layer in the convolutional neural network model to obtain an image fusion characteristic vector.

the certificate authenticity identification module 702, which:

respectively carrying out convolution processing on the multiple frames of certificate image information by using the convolution layer to obtain the certificate image convolution characteristics of each frame of certificate image information;

pooling the certificate image convolution characteristics by using a pooling layer in the convolution neural network model to obtain certificate image pooling characteristics of each frame of the certificate image information;

and performing fusion processing on the plurality of certificate image pooling features by using a full connection layer in the convolutional neural network model to obtain an image fusion feature vector.

Optionally, the certificate authenticity identification model further comprises: a time-series neural network model;

the certificate authenticity identification module 702, which:

inputting the image fusion characteristic vectors at a plurality of target moments in a preset time period into the time sequence neural network model;

performing time sequence response on the image fusion characteristic vectors at a plurality of target moments by using the time sequence neural network model to obtain a comprehensive time sequence characteristic vector;

and carrying out full-connection operation on the comprehensive time sequence characteristic vector by utilizing a preset full-connection layer to obtain a model output result.

Optionally, the credential image acquisition module 701 is to:

extracting a plurality of video key frames from video stream information of a target certificate, wherein the video stream information comprises: certificate image frames of the target certificate under a plurality of postures;

determining certificate gesture information corresponding to each video key frame by using a preset certificate gesture recognition algorithm aiming at each video key frame;

selecting a preset number of video key frames according to the certificate attitude information of each video key frame, wherein the certificate attitude difference between the preset number of video key frames is greater than a preset attitude difference threshold value;

and determining the selected video key frames of the preset number as multi-frame certificate image information of the target certificate.

Optionally, as shown in fig. 8, the apparatus further includes: a credential recognition model training module 703 that:

acquiring sample video streams corresponding to a plurality of sample certificates respectively, wherein each sample video stream comprises: certificate image frames of the sample certificate under a plurality of postures;

for each sample document, extracting a plurality of sample key frames from the sample video stream of the sample document;

determining a sample certificate image set based on the plurality of sample key frames corresponding to each sample certificate, wherein the sample certificate image set comprises: a plurality of model training samples, each of the model training samples comprising: multiple frames of sample certificate images;

and performing fusion training on the multi-frame sample certificate images by using a deep learning method and based on the plurality of model training samples to obtain a certificate authenticity identification model.

Optionally, the certificate recognition model training module 703 is to:

determining certificate gesture information corresponding to the plurality of sample key frames of each sample certificate by using a preset certificate gesture recognition algorithm;

selecting a plurality of groups of preset number of sample certificate images according to the certificate attitude information of each sample key frame, wherein the certificate attitude difference between each group of preset number of sample certificate images is greater than a preset attitude difference threshold value;

and determining a sample certificate image set according to the multiple groups of preset number of sample certificate images corresponding to the sample certificates.

In one or more embodiments of the present specification, a certificate authenticity identification apparatus first obtains multi-frame certificate image information of a target certificate, where the target certificate includes: at least one certificate security feature; then, identifying the authenticity of the target certificate by using a pre-trained certificate authenticity identification model based on the certificate anti-counterfeiting mark in the acquired multi-frame certificate image information to obtain a model output result; and finally, determining the authenticity identification result aiming at the target certificate according to the obtained model output result. In one or more embodiments of the present specification, after obtaining information of multiple frames of certificate images of a target certificate, the multi-frame certificate images of the target certificate are fused and identified by using a pre-trained certificate authenticity identification model in a multi-frame image fusion manner, so as to obtain an authenticity identification result for the target certificate, which can accurately identify information change conditions of a certificate anti-counterfeiting mark in the multi-frame certificate images based on the multi-frame certificate images, thereby improving accuracy of certificate authenticity identification, so as to intercept a situation that the target certificate is a counterfeit certificate in time, and further improve accuracy of subsequent service response.

It should be noted that the embodiment of the certificate authenticity identification device in this specification and the embodiment of the certificate authenticity identification method in this specification are based on the same inventive concept, so that the specific implementation of this embodiment may refer to the implementation of the corresponding certificate authenticity identification method, and repeated details are not repeated.

Further, corresponding to the methods shown in fig. 1 to 6, based on the same technical concept, one or more embodiments of the present specification further provide a certificate authenticity identification apparatus for performing the certificate authenticity identification method, as shown in fig. 9.

The certificate authenticity identification device can generate large differences due to different configurations or performances, and can comprise one or more processors 901 and a memory 902, wherein one or more stored applications or data can be stored in the memory 902. Memory 902 may be, among other things, transient storage or persistent storage. The application stored in the memory 902 may include one or more modules (not shown), each of which may include a series of computer-executable instructions for a certificate authenticity identification device. Still further, the processor 901 may be configured to communicate with the memory 902 to execute a series of computer-executable instructions in the memory 902 on the certificate authenticity identification device. The certificate authenticity identification apparatus may also include one or more power supplies 903, one or more wired or wireless network interfaces 904, one or more input output interfaces 905, one or more keyboards 906, and the like.

In one embodiment, the certificate authenticity identification apparatus comprises a memory, and one or more programs, wherein the one or more programs are stored in the memory, and the one or more programs may comprise one or more modules, and each module may comprise a series of computer-executable instructions for the certificate authenticity identification apparatus, and the one or more programs configured to be executed by one or more processors comprise computer-executable instructions for:

acquiring multi-frame certificate image information of a target certificate, wherein the target certificate comprises: at least one certificate security feature;

identifying the authenticity of the target certificate by using a pre-trained certificate authenticity identification model based on the certificate anti-counterfeiting mark in the multi-frame certificate image information to obtain a model output result; the certificate authenticity identification model is obtained by performing multi-frame image fusion training on a preset sample certificate image set by utilizing a deep learning method;

and determining the authenticity identification result aiming at the target certificate according to the model output result.

Optionally, the computer executable instructions, when executed, the certificate authenticity identification model comprises: a deep learning network model;

utilize the certificate true and false recognition model that trains in advance, based on among the multiframe certificate image information certificate false proof mark is right the true and false of target certificate is discerned, obtains model output result, includes:

Optionally, the computer executable instructions, when executed, the deep learning network model comprises: a convolutional neural network model;

the image fusion processing is carried out on the multi-frame certificate image information by utilizing the deep learning network model to obtain an image fusion characteristic vector, and the method comprises the following steps:

Optionally, the computer executable instructions, when executed, further comprise: a time-series neural network model;

the identifying the authenticity of the target certificate based on the image fusion feature vector to obtain a model output result comprises the following steps:

Optionally, the computer executable instructions, when executed, obtain a plurality of frames of document image information of a target document, comprising:

Optionally, when executed, the computer executable instructions are trained to:

Optionally, when executed, the computer-executable instructions, when executed, determine a sample document image set based on the plurality of sample key frames corresponding to each of the sample documents, including:

Further, based on the same technical concept, corresponding to the methods shown in fig. 1 to fig. 6, one or more embodiments of the present specification further provide a storage medium for storing computer-executable instructions, where in a specific embodiment, the storage medium may be a usb disk, an optical disk, a hard disk, and the like, and the storage medium stores computer-executable instructions that, when executed by a processor, implement the following processes:

Optionally, the storage medium stores computer executable instructions that, when executed by the processor, cause the certificate authenticity identification model to comprise: a deep learning network model;

Optionally, the storage medium stores computer-executable instructions that, when executed by the processor, the deep learning network model comprises: a convolutional neural network model;

Optionally, the storage medium stores computer executable instructions that, when executed by the processor, further comprise: a time-series neural network model;

Optionally, the storage medium stores computer executable instructions that, when executed by the processor, obtain a plurality of frames of document image information of a target document, including:

Optionally, the storage medium stores computer executable instructions that, when executed by the processor, the certificate authenticity identification model is trained by:

Optionally, the storage medium stores computer-executable instructions that, when executed by the processor, determine a sample document image set based on the plurality of sample keyframes corresponding to each of the sample documents, including:

The storage medium in one or more embodiments of the present description stores computer-executable instructions that, when executed by a processor, first acquire a plurality of frames of document image information of a target document, the target document including: at least one certificate security feature; then, identifying the authenticity of the target certificate by using a pre-trained certificate authenticity identification model based on the certificate anti-counterfeiting mark in the acquired multi-frame certificate image information to obtain a model output result; and finally, determining the authenticity identification result aiming at the target certificate according to the obtained model output result. In one or more embodiments of the present specification, after obtaining information of multiple frames of certificate images of a target certificate, the multi-frame certificate images of the target certificate are fused and identified by using a pre-trained certificate authenticity identification model in a multi-frame image fusion manner, so as to obtain an authenticity identification result for the target certificate, which can accurately identify information change conditions of a certificate anti-counterfeiting mark in the multi-frame certificate images based on the multi-frame certificate images, thereby improving accuracy of certificate authenticity identification, so as to intercept a situation that the target certificate is a counterfeit certificate in time, and further improve accuracy of subsequent service response.

It should be noted that the embodiment of the storage medium in this specification and the embodiment of the certificate authenticity identification method in this specification are based on the same inventive concept, and therefore, for specific implementation of this embodiment, reference may be made to the implementation of the corresponding certificate authenticity identification method, and repeated details are not repeated.

The foregoing description has been directed to specific embodiments of this disclosure. Other embodiments are within the scope of the following claims. In some cases, the actions or steps recited in the claims may be performed in a different order than in the embodiments and still achieve desirable results. In addition, the processes depicted in the accompanying figures do not necessarily require the particular order shown, or sequential order, to achieve desirable results. In some embodiments, multitasking and parallel processing may also be possible or may be advantageous.

In the 90 s of the 20 th century, improvements in a technology could clearly distinguish between improvements in hardware (e.g., improvements in circuit structures such as diodes, transistors, switches, etc.) and improvements in software (improvements in process flow). However, as technology advances, many of today's process flow improvements have been seen as direct improvements in hardware circuit architecture. Designers almost always obtain the corresponding hardware circuit structure by programming an improved method flow into the hardware circuit. Thus, it cannot be said that an improvement in the process flow cannot be realized by hardware physical modules. For example, a Programmable Logic Device (PLD), such as a Field Programmable Gate Array (FPGA), is an integrated circuit whose Logic functions are determined by programming the Device by a user. A digital system is "integrated" on a PLD by the designer's own programming without requiring the chip manufacturer to design and fabricate application-specific integrated circuit chips. Furthermore, nowadays, instead of manually making an integrated Circuit chip, such Programming is often implemented by "logic compiler" software, which is similar to a software compiler used in program development and writing, but the original code before compiling is also written by a specific Programming Language, which is called Hardware Description Language (HDL), and HDL is not only one but many, such as abel (advanced Boolean Expression Language), ahdl (alternate Language Description Language), traffic, pl (core unified Programming Language), Cal, jhdware Description Language, langua, mylar, pams, Hardware (Hardware Description Language), langva, Lola, HDL, palmware, Hardware (Hardware Description Language), VHDL (Hardware Description Language), and the like, which are currently used in the most popular languages. It will also be apparent to those skilled in the art that hardware circuitry that implements the logical method flows can be readily obtained by merely slightly programming the method flows into an integrated circuit using the hardware description languages described above.

The controller may be implemented in any suitable manner, for example, the controller may take the form of, for example, a microprocessor or processor and a computer-readable medium storing computer-readable program code (e.g., software or firmware) executable by the (micro) processor, logic gates, switches, an Application Specific Integrated Circuit (ASIC), a programmable logic controller, and an embedded microcontroller, examples of which include, but are not limited to, the following microcontrollers: ARC 625D, Atmel AT91SAM, Microchip PIC18F26K20, and Silicone Labs C8051F320, the memory controller may also be implemented as part of the control logic for the memory. Those skilled in the art will also appreciate that, in addition to implementing the controller as pure computer readable program code, the same functionality can be implemented by logically programming method steps such that the controller is in the form of logic gates, switches, application specific integrated circuits, programmable logic controllers, embedded microcontrollers and the like. Such a controller may thus be considered a hardware component, and the means included therein for performing the various functions may also be considered as a structure within the hardware component. Or even means for performing the functions may be regarded as being both a software module for performing the method and a structure within a hardware component.

The systems, devices, modules or units illustrated in the above embodiments may be implemented by a computer chip or an entity, or by a product with certain functions. One typical implementation device is a computer. In particular, the computer may be, for example, a personal computer, a laptop computer, a cellular telephone, a camera phone, a smartphone, a personal digital assistant, a media player, a navigation device, an email device, a game console, a tablet computer, a wearable device, or a combination of any of these devices.

For convenience of description, the above devices are described as being divided into various units by function, and are described separately. Of course, the functionality of the various elements may be implemented in the same one or more software and/or hardware implementations of one or more of the present descriptions.

As will be appreciated by one skilled in the art, one or more embodiments of the present description may be provided as a method, system, or computer program product. Accordingly, one or more of the present description may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, one or more of the present description may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied in the medium.

One or more of the present specification has been described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to one or more embodiments of the specification. It will be understood that each flow and/or block of the flow diagrams and/or block diagrams, and combinations of flows and/or blocks in the flow diagrams and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.

These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.

These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.

In a typical configuration, a computing device includes one or more processors (CPUs), input/output interfaces, network interfaces, and memory.

The memory may include forms of volatile memory in a computer readable medium, Random Access Memory (RAM) and/or non-volatile memory, such as Read Only Memory (ROM) or flash memory (flash RAM). Memory is an example of a computer-readable medium.

Computer-readable media, including both non-transitory and non-transitory, removable and non-removable media, may implement information storage by any method or technology. The information may be computer readable instructions, data structures, modules of a program, or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), Static Random Access Memory (SRAM), Dynamic Random Access Memory (DRAM), other types of Random Access Memory (RAM), Read Only Memory (ROM), Electrically Erasable Programmable Read Only Memory (EEPROM), flash memory or other memory technology, compact disc read only memory (CD-ROM), Digital Versatile Discs (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices, or any other non-transmission medium that can be used to store information that can be accessed by a computing device. As defined herein, a computer readable medium does not include a transitory computer readable medium such as a modulated data signal and a carrier wave.

It should also be noted that the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising an … …" does not exclude the presence of other like elements in a process, method, article, or apparatus that comprises the element.

One or more of the present specification can be described in the general context of computer-executable instructions, such as program modules, being executed by a computer. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform particular tasks or implement particular abstract data types. One or more of the present specification can also be practiced in distributed computing environments where tasks are performed by remote processing devices that are linked through a communications network. In a distributed computing environment, program modules may be located in both local and remote computer storage media including memory storage devices.

The embodiments in the present specification are described in a progressive manner, and the same and similar parts among the embodiments are referred to each other, and each embodiment focuses on the differences from the other embodiments. In particular, for the system embodiment, since it is substantially similar to the method embodiment, the description is simple, and for the relevant points, reference may be made to the partial description of the method embodiment.

The above description is merely illustrative of one or more embodiments of the present disclosure and is not intended to limit one or more embodiments of the present disclosure. Various modifications and alterations to one or more of the present descriptions will be apparent to those skilled in the art. Any modification, equivalent replacement, improvement or the like made within the spirit and principle of one or more of the present specification should be included in the scope of one or more claims of the present specification.

Claims

1. A method of authenticating a document, comprising:

2. The method of claim 1, wherein the certificate authenticity identification model comprises: a deep learning network model;

3. The method of claim 2, wherein the deep learning network model comprises: a convolutional neural network model;

4. The method of claim 2, wherein the deep learning network model comprises: a convolutional neural network model;

5. The method of claim 2, wherein the deep learning network model comprises: a convolutional neural network model;

6. The method of claim 2, wherein the document authenticity identification model further comprises: a time-series neural network model;

7. The method of claim 1, wherein the acquiring multiple frames of document image information of a target document comprises:

8. The method of claim 1, wherein the certificate authenticity identification model is trained by:

9. The method of claim 8, wherein said determining a sample document image set based on the plurality of sample keyframes corresponding to each of the sample documents comprises:

10. A document authenticity identification device comprising:

the certificate image acquisition module acquires multi-frame certificate image information of a target certificate, wherein the target certificate comprises: at least one certificate security feature;

the certificate authenticity identification module is used for identifying the authenticity of the target certificate based on the certificate anti-counterfeiting mark in the multi-frame certificate image information by utilizing a pre-trained certificate authenticity identification model to obtain a model output result; the certificate authenticity identification model is obtained by performing multi-frame image fusion training on a preset sample certificate image set by utilizing a deep learning method;

and the authenticity result determining module is used for determining an authenticity identification result aiming at the target certificate according to the model output result.

11. The apparatus of claim 10, wherein the document authenticity identification model comprises: a deep learning network model;

the certificate true and false identification module comprises:

12. The apparatus of claim 11, wherein the deep learning network model comprises: a convolutional neural network model;

the certificate true and false identification module comprises:

13. The apparatus of claim 11, wherein the deep learning network model comprises: a convolutional neural network model;

the certificate true and false identification module comprises:

14. The apparatus of claim 11, wherein the deep learning network model comprises: a convolutional neural network model;

the certificate true and false identification module comprises:

15. The apparatus of claim 11, wherein the document authenticity identification model further comprises: a time-series neural network model;

the certificate true and false identification module comprises:

16. The apparatus of claim 10, wherein the credential image acquisition module:

17. The apparatus of claim 10, wherein the apparatus further comprises: certificate recognition model training module, it:

18. The apparatus of claim 17, wherein the credential recognition model training module is to:

19. A document authenticity identification device comprising:

a processor; and

a memory arranged to store computer executable instructions that, when executed, cause the processor to:

20. A storage medium storing computer-executable instructions that, when executed by a processor, implement a method of: