WO2020238552A1 - 基于微表情识别的审批指令生成方法、装置和计算机设备 - Google Patents

基于微表情识别的审批指令生成方法、装置和计算机设备 Download PDF

Info

Publication number
WO2020238552A1
WO2020238552A1 PCT/CN2020/088005 CN2020088005W WO2020238552A1 WO 2020238552 A1 WO2020238552 A1 WO 2020238552A1 CN 2020088005 W CN2020088005 W CN 2020088005W WO 2020238552 A1 WO2020238552 A1 WO 2020238552A1
Authority
WO
WIPO (PCT)
Prior art keywords
micro
image
terminal
expression recognition
expression
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Ceased
Application number
PCT/CN2020/088005
Other languages
English (en)
French (fr)
Inventor
龚春燕
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
OneConnect Smart Technology Co Ltd
Original Assignee
OneConnect Smart Technology Co Ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by OneConnect Smart Technology Co Ltd filed Critical OneConnect Smart Technology Co Ltd
Publication of WO2020238552A1 publication Critical patent/WO2020238552A1/zh
Anticipated expiration legal-status Critical
Ceased legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F3/00—Input arrangements for transferring data to be processed into a form capable of being handled by the computer; Output arrangements for transferring data from processing unit to output unit, e.g. interface arrangements
    • G06F3/01—Input arrangements or combined input and output arrangements for interaction between user and computer
    • G06F3/011—Arrangements for interaction with the human body, e.g. for user immersion in virtual reality
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/10—Office automation; Time management
    • G06Q10/103—Workflow collaboration or project management
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
    • G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
    • G06V40/16—Human faces, e.g. facial parts, sketches or expressions
    • G06V40/161—Detection; Localisation; Normalisation
    • G06V40/166—Detection; Localisation; Normalisation using acquisition arrangements
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
    • G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
    • G06V40/16—Human faces, e.g. facial parts, sketches or expressions
    • G06V40/168—Feature extraction; Face representation
    • G06V40/171—Local features and components; Facial parts ; Occluding parts, e.g. glasses; Geometrical relationships
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
    • G06V40/40—Spoof detection, e.g. liveness detection
    • G06V40/45—Detection of the body part being alive
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2203/00—Indexing scheme relating to G06F3/00 - G06F3/048
    • G06F2203/01—Indexing scheme relating to G06F3/01
    • G06F2203/011—Emotion or mood input determined on the basis of sensed human body parameters such as pulse, heart rate or beat, temperature of skin, facial expressions, iris, voice pitch, brain activity patterns

Definitions

  • This application relates to the field of artificial intelligence, and in particular to a method, device, computer equipment and storage medium for generating approval instructions based on micro-expression recognition.
  • the main purpose of this application is to provide a method, device, computer equipment and storage medium for generating approval instructions based on micro-expression recognition, aiming to accurately and quickly identify risky behaviors.
  • this application proposes a method for generating approval instructions based on micro-expression recognition, including the following steps:
  • multiple whole-body images of the user corresponding to the terminal are collected through the terminal's camera, and the multiple whole-body images are used to perform in vivo verification of the user corresponding to the terminal according to preset living verification rules And after the living body verification is passed, at least one of the multiple whole-body images is selected as the initial image of the user corresponding to the terminal, wherein the initial image includes at least the face of the user corresponding to the terminal;
  • the facial image is input into a preset micro-expression recognition model based on neural network model training to perform calculations to obtain a micro-expression recognition result, wherein the micro-expression recognition model is based on the facial image and is related to the person Trained from sample data composed of micro-expression categories associated with face images;
  • micro expression recognition result does not belong to the preset malicious micro expression list, it is determined that the approval application is normal, and an approval instruction is generated according to the approval type.
  • This application provides a device for generating approval instructions based on micro-expression recognition, including:
  • the approval application receiving unit is configured to receive an approval application sent by the terminal, where the approval application specifies an approval type
  • the initial image acquisition unit is configured to collect multiple whole body images of the user corresponding to the terminal through the camera of the terminal according to the approval application, and use the multiple whole body images to compare the The user corresponding to the terminal performs in vivo verification; and after the in vivo verification is passed, at least one of the multiple whole-body images is selected as the initial image of the user corresponding to the terminal, wherein the initial image includes at least the corresponding terminal The user's face;
  • a facial image acquisition unit configured to perform region recognition processing on the initial image to identify a facial region, and use an image within the facial region as a facial image
  • the micro-expression recognition unit is used to input the facial image into a preset micro-expression recognition model based on neural network model training for calculation, thereby obtaining a micro-expression recognition result, wherein the micro-expression recognition model is based on a human face Image, and sample data composed of micro-expression categories associated with the face image through training;
  • the malicious micro-expression judging unit is used to judge whether the micro-expression recognition result belongs to a preset malicious micro-expression list
  • the approval instruction generating unit is configured to determine that the approval application is normal if the micro-expression recognition result does not belong to the preset malicious micro-expression list, and generate an approval instruction according to the type of approval.
  • the present application provides a computer device including a memory and a processor, the memory stores a computer program, and the processor implements the steps of any one of the above methods when the computer program is executed.
  • the present application provides a computer-readable storage medium on which a computer program is stored, and when the computer program is executed by a processor, the steps of any of the above methods are implemented.
  • the method, device, computer equipment, and storage medium for generating approval instructions based on micro-expression recognition of the present application realize accurate and quick approval instruction generation.
  • FIG. 1 is a schematic flowchart of an approval instruction generation method based on micro-expression recognition according to an embodiment of the application
  • FIG. 2 is a schematic block diagram of the structure of an approval instruction generation apparatus based on micro-expression recognition according to an embodiment of the application;
  • FIG. 3 is a schematic block diagram of the structure of a computer device according to an embodiment of the application.
  • an embodiment of the present application provides a method for generating approval instructions based on micro-expression recognition, including the following steps:
  • the approval application collect multiple whole-body images of the user corresponding to the terminal through the camera of the terminal, and use the multiple whole-body images to perform the operation on the user corresponding to the terminal according to preset living body verification rules. In vivo verification; and after the in vivo verification is passed, at least one of the multiple whole-body images is selected as the initial image of the user corresponding to the terminal, wherein the initial image includes at least the face of the user corresponding to the terminal;
  • micro-expression recognition result does not belong to the preset malicious micro-expression list, determine that the approval application is normal, and generate an approval instruction according to the type of approval.
  • the approval application sent by the terminal is received.
  • Traditional methods can only use written data and other materials to determine whether it is malicious, and cannot accurately and quickly screen out malicious.
  • This application uses micro-expression recognition technology to accurately and quickly determine whether it is malicious, so as to realize effective generation of approval instructions.
  • the approval application refers to an application sent by a person for approval of the intended item.
  • the approval application may also include approval materials, such as the borrower's ID number, name, income status, mortgage status and credit status.
  • the approval can be any approval, such as loan approval.
  • step S2 according to the approval application, multiple whole-body images of the user corresponding to the terminal are collected through the camera of the terminal, and according to preset living body verification rules, the multiple whole-body images are used to compare the The user corresponding to the terminal performs in vivo verification; and after the in vivo verification is passed, at least one of the multiple whole-body images is selected as the initial image of the user corresponding to the terminal, wherein the initial image includes at least the corresponding terminal The user's face.
  • the method of collecting the initial image of the user corresponding to the terminal through the camera of the terminal is, for example, collecting the video data of the user corresponding to the terminal in a video mode through the camera of the terminal, and from the video data
  • the initial image is collected according to the preset image selection rules; or, the terminal’s camera is used to collect multiple images of the user corresponding to the terminal in a camera mode, and one or more of the multiple images As the initial image.
  • the living body verification method includes, for example, according to the approval application, through the microphone of the terminal, voice prompts the user corresponding to the terminal to make a specified physical action, and collect the whole body of the user through the camera of the terminal An image, wherein the whole body image includes the face of the user corresponding to the terminal; it is determined whether the whole body image includes the specified limb movement; if the whole body image includes the specified limb movement, it is determined that the living body verification is passed.
  • the initial image is subjected to area recognition processing to recognize the face area, and the image within the face area is taken as the face image. Since the initial image can include a large area, this application only needs to perform micro-expression recognition through facial areas, so facial areas are recognized through area recognition processing, thereby reducing the amount of data calculation and reducing the computational burden.
  • the method of performing area recognition processing on the initial image to identify the facial area can be any feasible method, for example: dividing the initial image into multiple areas, comparing the multiple areas with preset facial areas, The area whose similarity does not exceed the threshold is regarded as the face area; or, the initial image is divided into multiple areas, and the image data of each area is compared with the preset eye image data to obtain the image data of each area and the eye image data The area where the difference does not exceed the preset value is recorded as the eye area, and the image data of each area is compared with the preset mouth image data to obtain the difference between the image data of each area and the mouth image data.
  • the area whose difference does not exceed the preset value is recorded as the mouth area, and the standard facial image is called, and the center of the eye area in the standard facial image is reduced or enlarged to make the center of the eye area in the standard facial image and the eye area in the initial image And the center of the mouth area in the standard facial image coincides with the center of the mouth area in the initial image; and the initial image and the standard facial image after the scaling down or scaling up operation
  • the overlapping area is recorded as a face area, and the image within the range of the face area is taken as a face image.
  • the facial image is input into a preset micro-expression recognition model based on neural network model training to perform calculations to obtain a micro-expression recognition result, wherein the micro-expression recognition model is based on a human face
  • the image and the sample data composed of the micro-expression categories associated with the face image are trained.
  • the neural network model can be any model, such as VGG16 model, VGG19 model, VGG-F model, ResNet152 model, ResNet50 model, DPN131 model, IXception model, AlexNet model, DenseNet model, etc.
  • the DPN model is preferred.
  • DPN Dual Path Network
  • the above DPN, ResNeXt, and DenseNet are existing network structures, so I won’t repeat them here. Accordingly, the facial image is calculated by the micro-expression recognition model trained based on the neural network model to obtain the micro-expression recognition result.
  • the micro-expression recognition results include: happy, angry, scared, surprised, etc., and can also be subdivided into happy smiling, angry, trembling, and shocked.
  • step S5 it is determined whether the micro-expression recognition result belongs to a preset malicious micro-expression list.
  • the preset malicious micro-expression list records specific malicious micro-expressions, such as fake smiles, secret joys, etc. If the micro-expression recognition result belongs to a preset malicious micro-expression list, such as a fake smile, it indicates that it is likely to be a malicious application and should not be passed.
  • step S6 if the micro-expression recognition result does not belong to the preset malicious micro-expression list, it is determined that the approval application is normal, and an approval instruction is generated.
  • the preset malicious micro-expression list is used to distinguish whether the micro-expression is malicious or not. If the micro-expression recognition result does not belong to the preset malicious micro-expression list, it indicates that the expression is normal and belongs to a normal application, and normal approval procedures should be performed, and an approval instruction is generated accordingly. Wherein, the approval instruction is used to instruct the approval application to be approved according to a preset approval procedure.
  • multiple whole-body images of the user corresponding to the terminal are collected through the camera of the terminal, and according to preset living body verification rules, the multiple whole-body images
  • the user corresponding to the terminal performs in vivo verification; and after the in vivo verification is passed, at least one of the multiple whole-body images is selected as the initial image of the user corresponding to the terminal, wherein the initial image includes at least the terminal corresponding
  • the step S2 of the user's face includes:
  • voice prompts the user corresponding to the terminal to make a specified physical action through the microphone of the terminal, and collect multiple whole body images of the user corresponding to the terminal through the camera of the terminal. Wherein at least one of the multiple whole-body images includes the face of the user corresponding to the terminal;
  • this embodiment adopts a living body identification method to ensure that the user is a living user.
  • the terminal’s microphone is used to voice prompt the user corresponding to the terminal to make a specified physical action
  • the terminal’s camera is used to collect the full body image of the user, wherein the whole body The image includes the face of the user corresponding to the terminal; it is determined whether the whole body image includes the specified limb movement; if the whole body image includes the specified limb movement, the whole body image is used as the initial image.
  • the designated limb movement includes multiple types, and if the whole body image includes the designated limb movement, determining that the living body verification is passed includes: collecting multiple types of the designated limb movement in the multiple whole body images And mark it as a living body verification image; determine whether the order of occurrence of the specified limb actions in the living body verification image is the same as the preset sequence; if they are the same, judge that the living body verification passes.
  • the step S3 of performing region recognition processing on the initial image to recognize a facial region, and using an image within the facial region as a facial image includes:
  • S301 Divide the initial image into multiple regions, and compare the image data of each region with preset eye image data to obtain the difference between the image data of each region and the eye image data, and the difference does not exceed the preset value. Set the value area as the eye area;
  • the area recognition processing on the initial image is implemented to recognize the face area, and the image within the face area is used as the face image.
  • the eye image data is standard image data that can be used to identify eye features (for example, the data of the human eye image area collected in advance)
  • the mouth image data is the standard image data that can be used to identify mouth features (such as the data collected in advance).
  • the data of the human mouth image area), the image data is, for example, image pixels (three primary colors, etc.).
  • the specific method for comparing image data can be any conventional comparison method, which will not be repeated here.
  • the eye area is larger than a single divided area, then multiple continuous areas whose difference does not exceed a preset value are taken as the eye area; similarly, multiple continuous areas whose difference does not exceed the preset value For the nose area. Since the five sense organs in a human face are distributed according to a certain geometric structure ratio, if the eye area and the mouth area are determined, the approximate facial contour can be obtained.
  • the standard facial image is called, and the center of the eye area in the standard facial image is coincident with the center of the eye area in the initial image through an equal-scale reduction or enlargement operation, and the standard facial image
  • the center of the mouth area coincides with the center of the mouth area in the initial image, and then the area in the initial image that overlaps the standard facial image after the scaling down or enlargement operation is recorded as the face area, and
  • the image within the range of the face area is used as a face image.
  • the said facial image is input into a preset micro-expression recognition model based on neural network model training for calculation, thereby obtaining a micro-expression recognition result, wherein the micro-expression recognition model is based on human Before step S4 trained by the sample data composed of the face image and the micro-expression categories associated with the face image, it includes:
  • S401 Obtain a specified number of sample data, and divide the sample data into a training set and a test set; wherein the sample data includes a face image and a micro-expression category associated with the face image;
  • S402 Input the sample data of the training set into a preset neural network model for training to obtain an initial micro-expression recognition model, where the stochastic gradient descent method is used in the training process;
  • the micro-expression recognition model is set.
  • This embodiment is based on a neural network model to train a micro expression recognition model.
  • the neural network model can be VGG16 model, VGG19 model, VGG-F model, ResNet152 model, ResNet50 model, DPN131 model, IXception model, AlexNet model, DenseNet model, etc., preferably DPN model.
  • the stochastic gradient descent method is to randomly sample some training data to replace the entire training set. If the sample size is large (for example, hundreds of thousands), then only tens of thousands or thousands of samples may be used and iterated When the optimal solution is reached, the training speed can be improved. Further, the training process can also use the reverse conduction law to update the parameters of each layer of the neural network model.
  • the reverse conduction law is based on the gradient descent method.
  • the input-output relationship of the BP network is essentially a mapping relationship: the function of a BP neural network with n inputs and m outputs is from n-dimensional Euclidean A continuous mapping from space to a finite field in m-dimensional Euclidean space. This mapping is highly non-linear and facilitates the update of the parameters of each layer of the neural network model.
  • the sample data of the test set is then used to verify the initial micro-expression recognition model, and if the verification passes, the initial micro-expression recognition model is recorded as the micro-expression recognition model.
  • the said facial image is input into a preset micro-expression recognition model based on neural network model training for calculation, thereby obtaining a micro-expression recognition result, wherein the micro-expression recognition model is based on human Before step S4 trained by the sample data composed of the face image and the micro-expression categories associated with the face image, it includes:
  • sample data of the test set to verify the initial micro-expression recognition model, where the sample data includes a face image and a micro-expression category associated with the face image;
  • transfer learning is used to quickly obtain a micro-expression recognition model. If you have a trained neural network model, you can eliminate the training step, and directly get the initial micro-expression recognition model. That is, the weight parameters of each layer are initialized to the initial weight parameters of each layer of the initial micro-expression recognition model to obtain the initial micro-expression recognition model, thereby omitting the training step. In order to prevent the initial micro-expression recognition model from being unsuitable, the initial micro-expression recognition model is also verified with sample data of the test set, where the sample data includes a face image and a micro-expression category associated with the face image If the verification is passed, the initial micro-expression recognition model is recorded as the micro-expression recognition model. So as to ensure that the finally obtained micro-expression recognition model is correct and usable.
  • the method includes:
  • micro-expression recognition result belongs to a preset malicious micro-expression list, determine that the approval application is abnormal, and return to the terminal a message that the approval is not passed.
  • the preset malicious micro-expression list is a criterion for judging whether the user's micro-expression is malicious, so as to determine whether the loan is a malicious loan. If the micro-expression recognition result belongs to the preset malicious micro-expression list, it indicates that the user has a high possibility of maliciousness, so it is determined that the micro-expression is malicious and the approval is not passed information is returned to the terminal. Further, after the information that the approval is not passed is returned to the terminal, the user corresponding to the terminal is recorded as a suspicious user, so as to be used as a reference when the user makes a next loan application.
  • the method for generating approval instructions is applied to an approval terminal, which is a blockchain node in a pre-built blockchain network, and the judgment of whether the micro-expression recognition result belongs to a preset After step S5 of the list of malicious micro expressions, it includes:
  • micro-expression recognition result belongs to the preset malicious micro-expression list, in the pre-built blockchain network, obtain multiple audit areas selected according to the preset consensus mechanism of the blockchain network Block chain node;
  • S53 Send the designated data record application to the final audit blockchain node, and receive the audit result of the final audit blockchain node, and determine whether the audit result meets the preset blockchain recording conditions, where
  • the designated data record application is accompanied by the user corresponding to the approval application and the micro-expression recognition result;
  • the consensus mechanism adopted in this embodiment can be any consensus mechanism, and the share authorization certification mechanism is preferred.
  • the share authorization proof mechanism means that all blockchain nodes vote for multiple entrusted blockchain nodes, so that multiple entrusted blockchain nodes will judge whether to record data in the blockchain on behalf of all blockchain nodes. , So as to avoid the shortcomings of slow operation of the blockchain when there are too many blockchain nodes.
  • this embodiment also uses the The node sends connection confirmation information, and records the audit blockchain node that replies to the connection confirmation information as the final audit blockchain node method to determine the audit blockchain nodes that can be audited, and record it as the final audit blockchain
  • the node sends the micro-expression recognition result to multiple final audit blockchain nodes, thereby reducing unnecessary network overhead and ensuring the accuracy of subsequent judgments (excluding nodes that cannot participate in the audit).
  • the preset blockchain recording condition is, for example, that the audit result is that the number of final audited blockchain nodes passed is greater than a preset threshold.
  • the method for generating approval instructions based on micro-expression recognition of the present application receives an approval application sent by a terminal; performs in vivo verification on the user corresponding to the terminal; and after the in vivo verification is passed, selects at least one of the multiple full-body images As the initial image of the user corresponding to the terminal; perform region recognition processing on the initial image to recognize a facial image; input the facial image into a preset micro-expression recognition model based on neural network model training for calculation ; If the micro-expression recognition result does not belong to the preset malicious micro-expression list, it is determined that the approval application is normal, and an approval instruction is generated. Thereby realizing accurate and quick generation of approval instructions.
  • an embodiment of the present application provides an approval instruction generation device based on micro-expression recognition, including:
  • the approval application receiving unit 10 is configured to receive an approval application sent by the terminal;
  • the initial image acquisition unit 20 is configured to collect multiple whole-body images of the user corresponding to the terminal through the camera of the terminal according to the approval application, and use the multiple whole-body images to compare all the images according to preset living body verification rules.
  • the user corresponding to the terminal performs in vivo verification; and after the in vivo verification is passed, at least one of the multiple whole-body images is selected as the initial image of the user corresponding to the terminal, wherein the initial image includes at least the terminal corresponding The user's face;
  • the facial image acquisition unit 30 is configured to perform region recognition processing on the initial image to identify a facial region, and use an image within the facial region as a facial image;
  • the micro-expression recognition unit 40 is configured to input the facial image into a preset micro-expression recognition model based on neural network model training for calculation, thereby obtaining a micro-expression recognition result, wherein the micro-expression recognition model is based on a human A face image, and sample data composed of micro-expression categories associated with the face image through training;
  • the malicious micro-expression judging unit 50 is configured to determine whether the micro-expression recognition result belongs to a preset malicious micro-expression list
  • the approval instruction generating unit 60 is configured to determine that the approval application is normal if the micro-expression recognition result does not belong to the preset malicious micro-expression list, and generate an approval instruction.
  • the approval application sent by the terminal is received.
  • Traditional methods can only use written data and other materials to determine whether it is malicious, and cannot accurately and quickly screen out malicious.
  • This application uses micro-expression recognition technology to accurately and quickly determine whether it is malicious, so as to realize effective generation of approval instructions.
  • the approval application refers to an application sent by a person for approval of the intended item. Further, the approval application may also include approval materials, such as the borrower's ID number, name, income status, mortgage status and credit status.
  • multiple whole-body images of the user corresponding to the terminal are collected through the camera of the terminal, and according to preset living body verification rules, the multiple whole-body images are used to compare the The user corresponding to the terminal performs in vivo verification; and after the in vivo verification is passed, at least one of the multiple whole-body images is selected as the initial image of the user corresponding to the terminal, wherein the initial image includes at least the corresponding terminal The user's face.
  • the method of collecting the initial image of the user corresponding to the terminal through the camera of the terminal is, for example, collecting the video data of the user corresponding to the terminal in a video mode through the camera of the terminal, and from the video data
  • the initial image is collected according to the preset image selection rules; or, the terminal’s camera is used to collect multiple images of the user corresponding to the terminal in a camera mode, and one or more of the multiple images As the initial image.
  • the living body verification method includes, for example, according to the approval application, through the microphone of the terminal, voice prompts the user corresponding to the terminal to make a specified physical action, and collect the whole body of the user through the camera of the terminal An image, wherein the whole body image includes the face of the user corresponding to the terminal; it is determined whether the whole body image includes the specified limb movement; if the whole body image includes the specified limb movement, it is determined that the living body verification is passed.
  • the initial image is subjected to area recognition processing to recognize the face area, and the image within the face area is taken as the face image. Since the initial image can include a large area, this application only needs to perform micro-expression recognition through facial areas, so facial areas are recognized through area recognition processing, thereby reducing the amount of data calculation and reducing the computational burden.
  • the method of performing area recognition processing on the initial image to identify the facial area can be any feasible method, for example: dividing the initial image into multiple areas, comparing the multiple areas with preset facial areas, The area whose similarity does not exceed the threshold is regarded as the face area; or, the initial image is divided into multiple areas, and the image data of each area is compared with the preset eye image data to obtain the image data of each area and the eye image data The area where the difference does not exceed the preset value is recorded as the eye area, and the image data of each area is compared with the preset mouth image data to obtain the difference between the image data of each area and the mouth image data.
  • the area whose difference does not exceed the preset value is recorded as the mouth area, and the standard facial image is called, and the center of the eye area in the standard facial image is reduced or enlarged to make the center of the eye area in the standard facial image and the eye area in the initial image And the center of the mouth area in the standard facial image coincides with the center of the mouth area in the initial image; and the initial image and the standard facial image after the scaling down or scaling up operation
  • the overlapping area is recorded as a face area, and the image within the range of the face area is taken as a face image.
  • the facial image is input into a preset micro-expression recognition model based on neural network model training to perform calculations to obtain a micro-expression recognition result, wherein the micro-expression recognition model is based on a human face
  • the image and the sample data composed of the micro-expression categories associated with the face image are trained.
  • the neural network model can be any model, such as VGG16 model, VGG19 model, VGG-F model, ResNet152 model, ResNet50 model, DPN131 model, IXception model, AlexNet model, DenseNet model, etc.
  • the DPN model is preferred.
  • DPN Deep Path Network
  • ResNeXt ResNeXt
  • DenseNet DenseNet
  • the above DPN, ResNeXt, and DenseNet are existing network structures, so I won’t repeat them here. Accordingly, the facial image is calculated by the micro-expression recognition model trained based on the neural network model to obtain the micro-expression recognition result.
  • the micro-expression recognition results include: happy, angry, scared, surprised, etc., and can also be subdivided into happy smiling, angry, trembling, and shocked.
  • the micro-expression recognition result belongs to a preset malicious micro-expression list.
  • the preset malicious micro-expression list records specific malicious micro-expressions, such as fake smiles, secret joys, etc. If the micro-expression recognition result belongs to a preset malicious micro-expression list, such as a fake smile, it indicates that it is likely to be a malicious application and should not be passed.
  • the micro-expression recognition result does not belong to the preset malicious micro-expression list, it is determined that the approval application is normal, and an approval instruction is generated.
  • the preset malicious micro-expression list is used to distinguish whether the micro-expression is malicious or not. If the micro-expression recognition result does not belong to the preset malicious micro-expression list, it indicates that the expression is normal and belongs to a normal application, and normal approval procedures should be performed, and an approval instruction is generated accordingly. Wherein, the approval instruction is used to instruct the approval application to be approved according to a preset approval procedure.
  • the initial image acquisition unit 20 includes:
  • the whole body image collection subunit is used to prompt the user corresponding to the terminal to make a specified physical action through the microphone of the terminal according to the approval application, and to collect the user corresponding to the terminal through the camera of the terminal Multiple whole-body images in, wherein at least one of the multiple whole-body images includes the face of the user corresponding to the terminal;
  • the designated limb movement judgment subunit is used to judge whether the whole body image includes the designated limb movement
  • the living body verification pass judgment subunit is used to judge that the living body verification is passed if the whole body image includes the specified limb movement;
  • the initial image acquisition subunit is configured to select at least one of the multiple whole-body images as the initial image of the user corresponding to the terminal, wherein the initial image includes at least the face of the user corresponding to the terminal.
  • this embodiment adopts a living body identification method to ensure that the user is a living user.
  • the terminal’s microphone is used to voice prompt the user corresponding to the terminal to make a specified physical action
  • the terminal’s camera is used to collect the full body image of the user, wherein the whole body The image includes the face of the user corresponding to the terminal; it is determined whether the whole body image includes the specified limb movement; if the whole body image includes the specified limb movement, the whole body image is used as the initial image.
  • the designated limb movement includes multiple types, and if the whole body image includes the designated limb movement, determining that the living body verification is passed includes: collecting multiple types of the designated limb movement in the multiple whole body images And mark it as a living body verification image; determine whether the order of occurrence of the specified limb actions in the living body verification image is the same as the preset sequence; if they are the same, judge that the living body verification passes.
  • the facial image acquisition unit 30 includes:
  • the eye area acquisition subunit is used to divide the initial image into multiple areas, compare the image data of each area with preset eye image data, and obtain the difference between the image data of each area and the eye image data, Mark the area where the difference does not exceed the preset value as the eye area;
  • the mouth area acquisition subunit is used to compare the image data of each area with the preset mouth image data to obtain the difference between the image data of each area and the mouth image data, and record the area where the difference does not exceed the preset value Is the mouth area;
  • the coincidence processing subunit is used to call the standard facial image, and make the center of the eye area in the standard facial image coincide with the center of the eye area in the initial image through an equal-scale reduction or enlargement operation, and the standard The center of the mouth area in the facial image coincides with the center of the mouth area in the initial image;
  • the facial image labeling subunit is used to mark the area in the initial image that overlaps the standard facial image after the equal-scale reduction or enlargement operation as a facial region, and use the image within the facial region as a facial image .
  • the area recognition processing on the initial image is implemented to recognize the face area, and the image within the face area is used as the face image.
  • the eye image data is standard image data that can be used to identify eye features
  • the mouth image data is standard image data that can be used to identify mouth features.
  • the image data is, for example, image pixels (three primary colors, etc.).
  • the specific method for comparing image data can be any conventional comparison method, which will not be repeated here. Further, if the eye area is larger than a single divided area, then multiple continuous areas whose difference does not exceed a preset value are regarded as the eye area; similarly, multiple consecutive areas whose difference does not exceed a preset value For the nose area.
  • the standard facial image is called, and the center of the eye area in the standard facial image is coincident with the center of the eye area in the initial image through an equal-scale reduction or enlargement operation, and the standard facial image
  • the center of the mouth area coincides with the center of the mouth area in the initial image, and then the area in the initial image that overlaps the standard facial image after the scaling down or enlargement operation is recorded as the face area, and The image within the range of the face area is used as a face image.
  • the device includes:
  • the sample data acquisition unit is configured to acquire a specified number of sample data, and divide the sample data into a training set and a test set; wherein the sample data includes a face image and a micro-expression category associated with the face image;
  • the training unit is used to input the sample data of the training set into the preset neural network model for training to obtain the initial micro-expression recognition model, where the stochastic gradient descent method is used in the training process;
  • a verification unit for verifying the initial micro-expression recognition model by using the sample data of the test set
  • the micro-expression recognition model marking unit is configured to record the initial micro-expression recognition model as the micro-expression recognition model if the verification is passed.
  • the micro-expression recognition model is set.
  • This embodiment is based on a neural network model to train a micro expression recognition model.
  • the neural network model can be VGG16 model, VGG19 model, VGG-F model, ResNet152 model, ResNet50 model, DPN131 model, IXception model, AlexNet model, DenseNet model, etc., preferably DPN model.
  • the stochastic gradient descent method is to randomly sample some training data to replace the entire training set. If the sample size is large (for example, hundreds of thousands), then only tens of thousands or thousands of samples may be used and iterated When the optimal solution is reached, the training speed can be improved. Further, the training process can also use the reverse conduction law to update the parameters of each layer of the neural network model.
  • the reverse conduction law is based on the gradient descent method.
  • the input-output relationship of the BP network is essentially a mapping relationship: the function of a BP neural network with n inputs and m outputs is from n-dimensional Euclidean A continuous mapping from space to a finite field in m-dimensional Euclidean space. This mapping is highly non-linear and facilitates the update of the parameters of each layer of the neural network model.
  • the sample data of the test set is then used to verify the initial micro-expression recognition model, and if the verification passes, the initial micro-expression recognition model is recorded as the micro-expression recognition model.
  • the device includes:
  • the weight parameter acquisition unit is used to acquire the weight parameters of each layer of the existing neural network model that has been trained
  • An initialization unit configured to initialize the weight parameters of each layer to the initial weight parameters of each layer of the initial micro-expression recognition model
  • the test verification unit is configured to verify the initial micro-expression recognition model using sample data of a test set, wherein the sample data includes a face image and a micro-expression category associated with the face image;
  • the micro-expression recognition model acquisition unit is configured to record the initial micro-expression recognition model as the micro-expression recognition model if the verification is passed.
  • transfer learning is used to quickly obtain a micro-expression recognition model. If you have a trained neural network model, you can eliminate the training step, and directly get the initial micro-expression recognition model. That is, the weight parameters of each layer are initialized to the initial weight parameters of each layer of the initial micro-expression recognition model to obtain the initial micro-expression recognition model, thereby omitting the training step. In order to prevent the initial micro-expression recognition model from being unsuitable, the initial micro-expression recognition model is also verified with sample data of the test set, where the sample data includes a face image and a micro-expression category associated with the face image If the verification is passed, the initial micro-expression recognition model is recorded as the micro-expression recognition model. So as to ensure that the finally obtained micro-expression recognition model is correct and usable.
  • the device includes:
  • the abnormality determination unit is configured to determine that the approval application is abnormal if the micro-expression recognition result belongs to the preset malicious micro-expression list, and return information that the approval is not passed to the terminal.
  • the preset malicious micro-expression list is a criterion for judging whether the user's micro-expression is malicious, so as to determine whether the loan is a malicious loan. If the micro-expression recognition result belongs to the preset malicious micro-expression list, it indicates that the user has a high possibility of maliciousness, so it is determined that the micro-expression is malicious and the approval is not passed information is returned to the terminal. Further, after the information that the approval is not passed is returned to the terminal, the user corresponding to the terminal is recorded as a suspicious user, so that it can be used as a reference when the user makes a next loan application.
  • the approval instruction generation method is applied to an approval terminal, the approval terminal is a blockchain node in a pre-built blockchain network, and the device includes:
  • the auditing block chain node obtaining unit is used to obtain a preset consensus mechanism based on the block chain network in the pre-built block chain network if the micro expression recognition result belongs to the preset malicious micro expression list Multiple audit blockchain nodes selected by voting;
  • the confirmation information sending unit is configured to send connection confirmation information to all the audit blockchain nodes, and record the audit blockchain node that replies to the connection confirmation information as the final audit blockchain node;
  • the record application sending unit is used to send a designated data record application to the final audit blockchain node, and receive the audit result of the final audit blockchain node, and determine whether the audit result meets a preset block Chain record condition, wherein the designated data record application is accompanied by the user corresponding to the approval application and the micro-expression recognition result;
  • the recording unit is configured to record the user corresponding to the approval application and the micro-expression recognition result in the blockchain network if the review result meets the preset blockchain recording condition.
  • the consensus mechanism adopted in this embodiment can be any consensus mechanism, and the share authorization certification mechanism is preferred.
  • the share authorization proof mechanism means that all blockchain nodes vote for multiple entrusted blockchain nodes, so that multiple entrusted blockchain nodes will judge whether to record data in the blockchain on behalf of all blockchain nodes. , So as to avoid the shortcomings of slow operation of the blockchain when there are too many blockchain nodes.
  • this embodiment also uses the The node sends connection confirmation information, and records the audit blockchain node that replies to the connection confirmation information as the final audit blockchain node method to determine the audit blockchain nodes that can be audited, and record it as the final audit blockchain
  • the node sends the micro-expression recognition result to multiple final audit blockchain nodes, thereby reducing unnecessary network overhead and ensuring the accuracy of subsequent judgments (excluding nodes that cannot participate in the audit).
  • the preset blockchain recording condition is, for example, that the audit result is that the number of final audited blockchain nodes passed is greater than a preset threshold.
  • the micro-expression recognition-based approval instruction generation device of the present application receives an approval application sent by a terminal; performs in vivo verification of the user corresponding to the terminal; and after the in vivo verification is passed, selects at least one of the multiple whole body images As the initial image of the user corresponding to the terminal; perform region recognition processing on the initial image to recognize a facial image; input the facial image into a preset micro-expression recognition model based on neural network model training for calculation ; If the micro-expression recognition result does not belong to the preset malicious micro-expression list, it is determined that the approval application is normal, and an approval instruction is generated. Thereby realizing accurate and quick generation of approval instructions.
  • an embodiment of the present invention also provides a computer device.
  • the computer device may be a server, and its internal structure may be as shown in the figure.
  • the computer equipment includes a processor, a memory, a network interface and a database connected through a system bus. Among them, the computer designed processor is used to provide calculation and control capabilities.
  • the memory of the computer device includes a non-volatile storage medium and an internal memory.
  • the non-volatile storage medium stores an operating system, a computer program, and a database.
  • the memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium.
  • the database of the computer equipment is used to store data used in the method of generating approval instructions based on micro-expression recognition.
  • the network interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, an approval instruction generation method based on micro-expression recognition is realized.
  • the processor executes the above micro-expression recognition-based approval instruction generation method, including the following steps: receiving an approval application sent by a terminal; according to the approval application, collecting multiple whole body images of the user corresponding to the terminal through the terminal's camera According to the preset living body verification rules, the user corresponding to the terminal is verified in vivo using the multiple whole-body images; and after the living body verification is passed, at least one of the multiple whole-body images is selected as the terminal
  • the initial image of the corresponding user where the initial image includes at least the face of the user corresponding to the terminal; performing area recognition processing on the initial image to identify the face area, and use the image within the face area as the face Image; input the facial image into a preset micro-expression recognition model based on neural network model training to perform operations to obtain a micro-expression recognition result, wherein the micro-expression recognition model is based on the face image, and the It is obtained by training sample data composed of micro-expression categories associated with the face image; determining whether the micro-expression recognition result belongs to the preset malicious micro
  • multiple whole-body images of the user corresponding to the terminal are collected through the camera of the terminal, and according to preset living body verification rules, the multiple whole-body images
  • the user corresponding to the terminal performs in vivo verification; and after the in vivo verification is passed, at least one of the multiple whole-body images is selected as the initial image of the user corresponding to the terminal, wherein the initial image includes at least the terminal corresponding
  • the step of the user’s face includes: according to the approval application, through the microphone of the terminal, voice prompting the user corresponding to the terminal to make a specified physical action, and collecting the corresponding physical action through the camera of the terminal Multiple whole-body images of the user, wherein at least one of the multiple whole-body images includes the face of the user corresponding to the terminal; judging whether the whole-body image includes the specified limb movement; if the whole-body image includes the If a physical action is specified, it is determined that the living body verification is passed; at least one of the multiple whole-body images is selected as the
  • the step of performing region recognition processing on the initial image to identify a facial region, and using an image within the facial region as a facial image includes: dividing the initial image into multiple Area, compare the image data of each area with the preset eye image data to obtain the difference between the image data of each area and the eye image data, and record the area where the difference does not exceed the preset value as the eye area; The image data of each area is compared with the preset mouth image data, and the difference between the image data of each area and the mouth image data is obtained.
  • the area whose difference does not exceed the preset value is recorded as the mouth area; the standard facial image is called, and Through an equal-scale reduction or enlargement operation, the center of the eye region in the standard facial image coincides with the center of the eye region in the initial image, and the center of the mouth region in the standard facial image coincides with the initial image
  • the center of the mouth area in the overlapped; the area in the initial image that overlaps the standard facial image after the equal-scale reduction or enlargement operation is recorded as the facial area, and the image within the facial area is taken as the facial image .
  • the said facial image is input into a preset micro-expression recognition model based on neural network model training for calculation, thereby obtaining a micro-expression recognition result, wherein the micro-expression recognition model is based on human
  • the step includes: obtaining a specified number of sample data, and dividing the sample data into a training set and a test set; wherein, the The sample data includes the face image and the micro-expression category associated with the face image; the sample data of the training set is input into the preset neural network model for training, and the initial micro-expression recognition model is obtained.
  • the training process Using the stochastic gradient descent method; using the sample data of the test set to verify the initial micro expression recognition model; if the verification passes, the initial micro expression recognition model is recorded as the micro expression recognition model.
  • the said facial image is input into a preset micro-expression recognition model based on neural network model training for calculation, thereby obtaining a micro-expression recognition result, wherein the micro-expression recognition model is based on human
  • the step includes: obtaining the weight parameters of each layer of the existing neural network model that has been trained; The parameters are initialized to the initial weight parameters of each layer of the initial micro-expression recognition model; the sample data of the test set is used to verify the initial micro-expression recognition model, wherein the sample data includes a face image and the associated face image Micro-expression category; if the verification is passed, the initial micro-expression recognition model is recorded as the micro-expression recognition model.
  • the method includes: if the micro-expression recognition result belongs to the preset malicious micro-expression list, determining the The approval application is abnormal, and information that the approval is not passed is returned to the terminal.
  • the method for generating approval instructions is applied to an approval terminal, which is a blockchain node in a pre-built blockchain network, and the judgment of whether the micro-expression recognition result belongs to a preset
  • the step of the list of malicious micro-expressions it includes: if the micro-expression recognition result belongs to the preset list of malicious micro-expressions, obtaining a preset consensus based on the blockchain network in a pre-built blockchain network Multiple audit blockchain nodes selected by the mechanism; send connection confirmation information to all the audit blockchain nodes, and record the audit blockchain node that replies to the connection confirmation information as the final audit blockchain node; Send the designated data record application to the final audit blockchain node, and receive the audit result of the final audit blockchain node, and determine whether the audit result meets the preset blockchain recording conditions, wherein The designated data record application is accompanied by the user corresponding to the approval application and the micro-expression recognition result; if the review result meets the preset blockchain recording conditions, the user corresponding to the approval application and the micro-expression The recognition result is recorded in the blockchain
  • the computer device of this application receives the approval application sent by the terminal; performs in vivo verification on the user corresponding to the terminal; and after the in vivo verification is passed, selects at least one of the multiple full-body images as the user corresponding to the terminal
  • the initial image perform region recognition processing on the initial image to recognize a facial image; input the facial image into a preset micro-expression recognition model based on neural network model training for calculation; if the micro-expression recognition result is not If it belongs to the preset malicious micro-expression list, it is determined that the approval application is normal, and an approval instruction is generated. Thereby realizing accurate and quick generation of approval instructions.
  • An embodiment of the application also provides a computer-readable storage medium on which a computer program is stored.
  • a method for generating an approval instruction based on micro-expression recognition is implemented, including the following steps: receiving an approval application sent by a terminal
  • multiple whole-body images of the user corresponding to the terminal are collected through the terminal's camera, and according to preset living body verification rules, the multiple whole-body images are used to perform the living body of the user corresponding to the terminal Verification; and after the in vivo verification is passed, at least one of the multiple full-body images is selected as the initial image of the user corresponding to the terminal, wherein the initial image includes at least the face of the user corresponding to the terminal;
  • the initial image is subjected to region recognition processing to identify the facial region, and the image within the facial region is used as a facial image; the facial image is input into a preset micro-expression recognition model based on neural network model training.
  • micro-expression recognition model is trained based on the face image and the sample data composed of the micro-expression categories associated with the face image; it is determined whether the micro-expression recognition result is It belongs to the preset malicious micro-expression list; if the micro-expression recognition result does not belong to the preset malicious micro-expression list, it is determined that the approval application is normal, and an approval instruction is generated.
  • multiple whole-body images of the user corresponding to the terminal are collected through the camera of the terminal, and according to preset living body verification rules, the multiple whole-body images
  • the user corresponding to the terminal performs in vivo verification; and after the in vivo verification is passed, at least one of the multiple whole-body images is selected as the initial image of the user corresponding to the terminal, wherein the initial image includes at least the terminal corresponding
  • the step of the user’s face includes: according to the approval application, through the microphone of the terminal, voice prompting the user corresponding to the terminal to make a specified physical action, and collecting the corresponding physical action through the camera of the terminal Multiple whole-body images of the user, wherein at least one of the multiple whole-body images includes the face of the user corresponding to the terminal; judging whether the whole-body image includes the specified limb movement; if the whole-body image includes the If a physical action is specified, it is determined that the living body verification is passed; at least one of the multiple whole-body images is selected as the
  • the step of performing region recognition processing on the initial image to identify a facial region, and using an image within the facial region as a facial image includes: dividing the initial image into multiple Area, compare the image data of each area with the preset eye image data to obtain the difference between the image data of each area and the eye image data, and record the area where the difference does not exceed the preset value as the eye area; The image data of each area is compared with the preset mouth image data, and the difference between the image data of each area and the mouth image data is obtained.
  • the area whose difference does not exceed the preset value is recorded as the mouth area; the standard facial image is called, and Through an equal-scale reduction or enlargement operation, the center of the eye region in the standard facial image coincides with the center of the eye region in the initial image, and the center of the mouth region in the standard facial image coincides with the initial image
  • the center of the mouth area in the overlapped; the area in the initial image that overlaps the standard facial image after the equal-scale reduction or enlargement operation is recorded as the facial area, and the image within the facial area is taken as the facial image .
  • the said facial image is input into a preset micro-expression recognition model based on neural network model training for calculation, thereby obtaining a micro-expression recognition result, wherein the micro-expression recognition model is based on human
  • the step includes: obtaining a specified number of sample data, and dividing the sample data into a training set and a test set; wherein, the The sample data includes the face image and the micro-expression category associated with the face image; the sample data of the training set is input into the preset neural network model for training, and the initial micro-expression recognition model is obtained.
  • the training process Using the stochastic gradient descent method; using the sample data of the test set to verify the initial micro expression recognition model; if the verification passes, the initial micro expression recognition model is recorded as the micro expression recognition model.
  • the said facial image is input into a preset micro-expression recognition model based on neural network model training for calculation, thereby obtaining a micro-expression recognition result, wherein the micro-expression recognition model is based on human
  • the step includes: obtaining the weight parameters of each layer of the existing neural network model that has been trained; The parameters are initialized to the initial weight parameters of each layer of the initial micro-expression recognition model; the sample data of the test set is used to verify the initial micro-expression recognition model, wherein the sample data includes a face image and the associated face image Micro expression category; if the verification is passed, the initial micro expression recognition model is recorded as the micro expression recognition model.
  • the method includes: if the micro-expression recognition result belongs to the preset malicious micro-expression list, determining the The approval application is abnormal, and information that the approval is not passed is returned to the terminal.
  • the method for generating approval instructions is applied to an approval terminal, which is a blockchain node in a pre-built blockchain network, and the judgment of whether the micro-expression recognition result belongs to a preset
  • the step of the list of malicious micro-expressions it includes: if the micro-expression recognition result belongs to the preset list of malicious micro-expressions, obtaining a preset consensus based on the blockchain network in a pre-built blockchain network Multiple audit blockchain nodes selected by the mechanism; send connection confirmation information to all the audit blockchain nodes, and record the audit blockchain node that replies to the connection confirmation information as the final audit blockchain node; Send the designated data record application to the final audit blockchain node, and receive the audit result of the final audit blockchain node, and determine whether the audit result meets the preset blockchain recording conditions, wherein The designated data record application is accompanied by the user corresponding to the approval application and the micro-expression recognition result; if the review result meets the preset blockchain recording conditions, the user corresponding to the approval application and the micro-expression The recognition result is recorded in the blockchain
  • the computer-readable storage medium of the present application receives an approval application sent by a terminal; performs in vivo verification on the user corresponding to the terminal; and after the in vivo verification is passed, selects at least one of the multiple full-body images as the terminal
  • the corresponding initial image of the user perform region recognition processing on the initial image to recognize a facial image; input the facial image into a preset micro-expression recognition model based on neural network model training for calculation; If the recognition result does not belong to the preset malicious micro-expression list, it is determined that the approval application is normal, and an approval instruction is generated. Thereby realizing accurate and quick generation of approval instructions.
  • the computer-readable storage medium may be non-volatile or volatile.
  • Non-volatile memory may include read only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory.
  • Volatile memory may include random access memory (RAM) or external cache memory.
  • RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual-rate data rate SDRAM (SSRSDRAM), enhanced SDRAM (ESDRAM), synchronous Link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Human Computer Interaction (AREA)
  • Health & Medical Sciences (AREA)
  • Oral & Maxillofacial Surgery (AREA)
  • Business, Economics & Management (AREA)
  • Human Resources & Organizations (AREA)
  • Multimedia (AREA)
  • General Health & Medical Sciences (AREA)
  • Strategic Management (AREA)
  • General Engineering & Computer Science (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Data Mining & Analysis (AREA)
  • Economics (AREA)
  • Marketing (AREA)
  • Operations Research (AREA)
  • Quality & Reliability (AREA)
  • Tourism & Hospitality (AREA)
  • General Business, Economics & Management (AREA)
  • Image Analysis (AREA)

Abstract

本申请涉及人工智能技术领域，揭示了一种基于微表情识别的审批指令生成方法、装置、计算机设备和存储介质，所述方法包括：接收终端发送的审批申请；通过所述终端的摄像头采集所述终端对应的用户的多幅全身图像，根据预设的活体验证规则进行活体验证；并在活体验证通过之后，从所述多幅全身图像中选取至少一张作为所述终端对应的用户的初始图像；对所述初始图像进行区域识别处理从而识别出面部图像；将所述面部图像输入到预设的基于神经网络模型训练完成的微表情识别模型中进行运算，从而得到微表情识别结果；若所述微表情识别结果不属于预设的恶意微表情列表，则判定所述审批申请正常，并生成审批指令。从而实现了准确快捷地鉴别风险行为。

Description

基于微表情识别的审批指令生成方法、装置和计算机设备
本申请要求于2019年5月30日提交中国专利局、申请号为201910463916.3，发明名称为“基于微表情识别的审批指令生成方法、装置和计算机设备”的中国专利申请的优先权，其全部内容通过引用结合在本申请中。
技术领域
本申请涉及人工智能领域，特别是涉及到一种基于微表情识别的审批指令生成方法、装置、计算机设备和存储介质。
背景技术
传统方法对于风险行为的审批，只能通过书面数据等材料的判断，来确定风险行为是否属于恶意行为，无法准确快捷筛选出恶意行为。然而，发明人意识到，这种审核书面数据的方式来进行，容易被恶意申请人钻空子，例如通过伪造材料等方式欺骗审核人员。并且这种审核书面数据的方式，需要对书面材料进行核实，从而需要进行大量保密数据的查询，涉及到权限等方面的要求，门槛高流程复杂。因此传统方法在防止风险行为方面缺乏准确快捷的风险控制方案。
技术问题
本申请的主要目的为提供一种基于微表情识别的审批指令生成方法、装置、计算机设备和存储介质，旨在准确快捷地鉴别风险行为。
技术解决方案
为了实现上述发明目的，本申请提出一种基于微表情识别的审批指令生成方法，包括以下步骤：
接收终端发送的审批申请，其中所述审批申请指定了审批的类型；
根据所述审批申请，通过所述终端的摄像头采集所述终端对应的用户的多幅全身图像，根据预设的活体验证规则，利用所述多幅全身图像对所述终端对应的用户进行活体验证；并在活体验证通过之后，从所述多幅全身图像中选取至少一张作为所述终端对应的用户的初始图像，其中所述初始图像至少包括所述终端对应的用户的面部；
对所述初始图像进行区域识别处理从而识别出面部区域，并将所述面部区域范围内的图像作为面部图像；
将所述面部图像输入到预设的基于神经网络模型训练完成的微表情识别模型中进行运算，从而得到微表情识别结果，其中，所述微表情识别模型基于人脸图像，以及与所述人脸图像关联的微表情类别组成的样本数据训练而成；
判断所述微表情识别结果是否属于预设的恶意微表情列表；
若所述微表情识别结果不属于预设的恶意微表情列表，则判定所述审批申请正常，并根据所述审批的类型生成审批指令。
本申请提供一种基于微表情识别的审批指令生成装置，包括：
审批申请接收单元，用于接收终端发送的审批申请，其中所述审批申请指定了审批的类型；
初始图像获取单元，用于根据所述审批申请，通过所述终端的摄像头采集所述终端对应的用户的多幅全身图像，根据预设的活体验证规则，利用所述多幅全身图像对所述终端对应的用户进行活体验证；并在活体验证通过之后，从所述多幅全身图像中选取至少一张作为所述终端对应的用户的初始图像，其中所述初始图像至少包括所述终端对应的用户的面部；
面部图像获取单元，用于对所述初始图像进行区域识别处理从而识别出面部区域，并将所述面部区域范围内的图像作为面部图像；
微表情识别单元，用于将所述面部图像输入到预设的基于神经网络模型训练完成的微表情识别模型中进行运算，从而得到微表情识别结果，其中，所述微表情识别模型基于人脸图像，以及与所述人脸图像关联的微表情类别组成的样本数据训练而成；
恶意微表情判断单元，用于判断所述微表情识别结果是否属于预设的恶意微表情列表；
审批指令生成单元，用于若所述微表情识别结果不属于预设的恶意微表情列表，则判定所述审批申请正常，并根据所述审批的类型生成审批指令。
本申请提供一种计算机设备，包括存储器和处理器，所述存储器存储有计算机程序，所述处理器执行所述计算机程序时实现上述任一项所述方法的步骤。
本申请提供一种计算机可读存储介质，其上存储有计算机程序，所述计算机程序被处理器执行时实现上述任一项所述的方法的步骤。
有益效果
本申请的基于微表情识别的审批指令生成方法、装置、计算机设备和存储介质，实现了准确快捷的审批指令生成。
附图说明
图1为本申请一实施例的基于微表情识别的审批指令生成方法的流程示意图；
图2为本申请一实施例的基于微表情识别的审批指令生成装置的结构示意框图；
图3为本申请一实施例的计算机设备的结构示意框图。
本发明的最佳实施方式
参照图1，本申请实施例提供一种基于微表情识别的审批指令生成方法，包括以下步骤：
S1、接收终端发送的审批申请，其中所述审批申请指定了审批的类型；
S2、根据所述审批申请，通过所述终端的摄像头采集所述终端对应的用户的多幅全身图像，根据预设的活体验证规则，利用所述多幅全身图像对所述终端对应的用户进行活体验证；并在活体验证通过之后，从所述多幅全身图像中选取至少一张作为所述终端对应的用户的初始图像，其中所述初始图像至少包括所述终端对应的用户的面部；
S3、对所述初始图像进行区域识别处理从而识别出面部区域，并将所述面部区域范围内的图像作为面部图像；
S4、将所述面部图像输入到预设的基于神经网络模型训练完成的微表情识别模型中进行运算，从而得到微表情识别结果，其中，所述微表情识别模型基于人脸图像，以及与所述人脸图像关联的微表情类别组成的样本数据训练而成；
S5、判断所述微表情识别结果是否属于预设的恶意微表情列表；
S6、若所述微表情识别结果不属于预设的恶意微表情列表，则判定所述审批申请正常，并根据所述审批的类型生成审批指令。
如上述步骤S1所述，接收终端发送的审批申请。传统方法对于的审批，只能通过书面数据等材料的判断，来确定是否属于恶意，无法准确快捷筛选出恶意。本申请利用微表情识别技术，以准确快捷地判断出是否属于恶意，以实现有效的审批指令生成。其中，所述审批申请指人发送的对意图事项进行审批的申请。进一步地，所述审批申请还可以包括审批材料，例如借款人的身份证号、姓名、收入情况、抵押物状况和信用情况等。其中，所述审批可以为任意审批，例如为贷款审批等。
如上述步骤S2所述，根据所述审批申请，通过所述终端的摄像头采集所述终端对应的用户的多幅全身图像，根据预设的活体验证规则，利用所述多幅全身图像对所述终端对应的用户进行活体验证；并在活体验证通过之后，从所述多幅全身图像中选取至少一张作为所述终端对应的用户的初始图像，其中所述初始图像至少包括所述终端对应的用户的面部。其中，通过所述终端的摄像头，采集所述终端对应的用户的初始图像的方式例如为：通过所述终端的摄像头，以视频模式采集所述终端对应的用户的视频数据，从所述视频数据中根据预设的图像选择规则采集初始图像；或者，通过所述终端的摄像头，以照相模式采集所述终端对应的用户的多张图像，并以所述多张图像中的一张或者多张作为所述初始图像。所述活体验证的方式例如包括：根据所述审批申请，通过所述终端的麦克风，语音提示所述终端对应的用户做出指定肢体动作，并通过所述终端的摄像头以采集所述用户的全身图像，其中所述全身图像包括所述终端对应的用户的面部；判断所述全身图像是否包括所述指定肢体动作；若所述全身图像包括所述指定肢体动作，则判定活体验证通过。
如上述步骤S3所述，对所述初始图像进行区域识别处理从而识别出面部区域，并将所述面部区域范围内的图像作为面部图像。由于初始图像可以包括较大的区域，而本申请通过面部区域进行微表情识别即可，因此通过区域识别处理从而识别出面部区域，从而减少数据计算量，减轻计算负担。其中对所述初始图像进行区域识别处理从而识别出面部区域的方式可为任意可行方式，例如：将所述初始图像划分为多个区域，将多个区域与预设的面部区域进行对比，将相似程度不超过阈值的区域作为面部区域；或者，所述初始图像划分为多个区域，将每个区域的图像数据与预设的眼睛图像数据进行对比，得到每个区域图像数据与眼睛图像数据的差值，将差值不超过预设数值的区域记为眼睛区域，将每个区域的图像数据与预设的嘴巴图像数据进行比较，得到每个区域图像数据与嘴巴图像数据的差值，将差值不超过预设数值的区域记为嘴巴区域，调用标准面部图像，并通过等比例缩小或者放大操作，使所述标准面部图像中的眼睛区域的中心与所述初始图像中的眼睛区域的中心重合，以及所述标准面部图像中的嘴巴区域的中心与所述初始图像中的嘴巴区域的中心重合；将所述初始图像中与经过所述等比例缩小或者放大操作后的标准面部图像重叠的区域记为面部区域，并将所述面部区域范围内的图像作为面部图像。
如上述步骤S4所述，将所述面部图像输入到预设的基于神经网络模型训练完成的微表情识别模型 中进行运算，从而得到微表情识别结果，其中，所述微表情识别模型基于人脸图像，以及与所述人脸图像关联的微表情类别组成的样本数据训练而成。其中神经网络模型可以为任意模型，例如VGG16模型、VGG19模型、VGG-F模型、ResNet152模型、ResNet50模型、DPN131模型、IXception模型、AlexNet模型和DenseNet模型等，优选DPN模型。DPN(Dual Path Network)是神经网络结构，在ResNeXt的基础上引入了DenseNet的核心内容，使得模型对特征的利用更加充分。上述DPN、ResNeXt和DenseNet是现有的网络结构，在此不在赘述。据此，通过基于神经网络模型训练完成的微表情识别模型对所述面部图像进行运算，从而得到微表情识别结果。其中所述微表情识别结果包括：开心、生气、害怕、惊讶等类型，还可细分为喜笑颜开、怒火中烧、胆颤心惊、大吃一惊等类型。
如上述步骤S5所述，判断所述微表情识别结果是否属于预设的恶意微表情列表。预设的恶意微表情列表记录了具体的恶意微表情，例如假意微笑、心存窃喜等。若所述微表情识别结果属于预设的恶意微表情列表，例如为假意微笑，则表明很可能属于恶意申请，不应通过。
如上述步骤S6所述，若所述微表情识别结果不属于预设的恶意微表情列表，则判定所述审批申请正常，并生成审批指令。预设的恶意微表情列表是用于区分微表情是否恶意的判断标准。若所述微表情识别结果不属于预设的恶意微表情列表，表明表情正常，属于正常申请，应该进行正常的审批程序，据此生成审批指令。其中，所述审批指令用于指示对所述审批申请按照预设的审批程序进行审批。
在一个实施方式中，所述根据所述审批申请，通过所述终端的摄像头采集所述终端对应的用户的多幅全身图像，根据预设的活体验证规则，利用所述多幅全身图像对所述终端对应的用户进行活体验证；并在活体验证通过之后，从所述多幅全身图像中选取至少一张作为所述终端对应的用户的初始图像，其中所述初始图像至少包括所述终端对应的用户的面部的步骤S2，包括：
S201、根据所述审批申请，通过所述终端的麦克风，语音提示所述终端对应的用户做出指定肢体动作，并通过所述终端的摄像头以采集所述终端对应的用户的多幅全身图像，其中所述多幅全身图像中的至少一张包括所述终端对应的用户的面部；
S202、判断所述全身图像是否包括所述指定肢体动作；
S203、若所述全身图像包括所述指定肢体动作，则判定活体验证通过；
S204、从所述多幅全身图像中选取至少一张作为所述终端对应的用户的初始图像，其中所述初始图像至少包括所述终端对应的用户的面部。
如上所述,实现了先进行活体识别，再采集所述终端对应的用户的初始图像。为了防止静态图像、伪造视频等欺骗摄像头以进行欺诈贷款的行为，本实施方式采用活体识别的方式以保证用户为活体用户。具体地，根据所述审批申请，通过所述终端的麦克风，语音提示所述终端对应的用户做出指定肢体动作，并通过所述终端的摄像头以采集所述用户的全身图像，其中所述全身图像包括所述终端对应的用户的面部；判断所述全身图像是否包括所述指定肢体动作；若所述全身图像包括所述指定肢体动作，则将所述全身图像作为所述初始图像。即，若用户为活体用户，那么将根据语音提示做出指定肢体动作，因此所述全身图像包括所述指定肢体动作可以表明用户为活体用户。因此此时的全身图像是可用的，据此将所述全身图像作为所述初始图像。进一步地，所述指定肢体动作包括多种，所述若所述全身图像包括所述指定肢体动作，则判定活体验证通过，包括：采集所述多幅全身图像中出现多种所述指定肢体动作的图像，并标记为活体验证图像；判断所述活体验证图像中的所述指定肢体动作出现的顺序是否与预设顺序相同；若相同，则判定活体验证通过。
在一个实施方式中，所述对所述初始图像进行区域识别处理从而识别出面部区域，并将所述面部区域范围内的图像作为面部图像的步骤S3，包括：
S301、将所述初始图像划分为多个区域，将每个区域的图像数据与预设的眼睛图像数据进行对比，得到每个区域图像数据与眼睛图像数据的差值，将差值不超过预设数值的区域记为眼睛区域；
S302、将每个区域的图像数据与预设的嘴巴图像数据进行比较，得到每个区域图像数据与嘴巴图像数据的差值，将差值不超过预设数值的区域记为嘴巴区域；
S303、调用标准面部图像，并通过等比例缩小或者放大操作，使所述标准面部图像中的眼睛区域的中心与所述初始图像中的眼睛区域的中心重合，以及所述标准面部图像中的嘴巴区域的中心与所述初始图像中的嘴巴区域的中心重合；
S304、将所述初始图像中与经过所述等比例缩小或者放大操作后的标准面部图像重叠的区域记为面部区域，并将所述面部区域范围内的图像作为面部图像。
如上所述，实现了对所述初始图像进行区域识别处理从而识别出面部区域，并将所述面部区域范围内的图像作为面部图像。其中，眼睛图像数据为标准的可用于标识眼睛特征的图像数据(例如预先采集得到的人的眼睛图像区域的数据)，嘴巴图像数据为标准的可用于标识嘴巴特征的图像数据(例如预先采集得到的人的嘴巴图像区域的数据)，所述图像数据例如为图像像素(三原色等)等。而具体比对图像数据的方法可采用任意的传统比较方式，在此不赘述。进一步地，若所述眼睛区域大于划分的单个区域， 则以差值不超过预设数值的多个连续的区域为眼睛区域；同理，以差值不超过预设数值的多个连续的区域为鼻子区域。由于人的面部中的五官是按一定的几何结构比例分布的，若确定眼睛区域与嘴巴区域，即可获知大致的面部轮廓。据此，调用标准面部图像，并通过等比例缩小或者放大操作，使所述标准面部图像中的眼睛区域的中心与所述初始图像中的眼睛区域的中心重合，同时使所述标准面部图像中的嘴巴区域的中心与所述初始图像中的嘴巴区域的中心重合，再将所述初始图像中与经过所述等比例缩小或者放大操作后的标准面部图像重叠的区域记为面部区域，并将所述面部区域范围内的图像作为面部图像。
在一个实施方式中，所述将所述面部图像输入到预设的基于神经网络模型训练完成的微表情识别模型中进行运算，从而得到微表情识别结果，其中，所述微表情识别模型基于人脸图像，以及与所述人脸图像关联的微表情类别组成的样本数据训练而成的步骤S4之前，包括：
S401、获取指定数量的样本数据，并将样本数据分成训练集和测试集；其中，所述样本数据包括人脸图像，以及与所述人脸图像关联的微表情类别；
S402、将训练集的样本数据输入到预设的神经网络模型中进行训练，得到初始微表情识别模型，其中，训练的过程中采用随机梯度下降法；
S403、利用测试集的样本数据验证所述初始微表情识别模型；
S404、若验证通过，则将所述初始微表情识别模型记为所述微表情识别模型。
如上所述，实现了设置微表情识别模型。本实施方式基于神经网络模型以训练出微表情识别模型。其中神经网络模型可为VGG16模型、VGG19模型、VGG-F模型、ResNet152模型、ResNet50模型、DPN131模型、IXception模型、AlexNet模型和DenseNet模型等，优选DPN模型。其中，随机梯度下降法就是随机取样一些训练数据，替代整个训练集，如果样本量很大的情况(例如几十万)，那么可能只用其中几万条或者几千条的样本，就已经迭代到最优解了，可以提高训练速度。进一步地，训练的过程还可以采用反向传导法则更新所述神经网络模型各层的参数。其中反向传导法则(BP)建立在梯度下降法的基础上，BP网络的输入输出关系实质上是一种映射关系：一个n输入m输出的BP神经网络所完成的功能是从n维欧氏空间向m维欧氏空间中一有限域的连续映射，这一映射具有高度非线性，有利于神经网络模型各层的参数的更新。从而获得初始微表情识别模型。再利用测试集的样本数据验证所述初始微表情识别模型，若验证通过，则将所述初始微表情识别模型记为所述微表情识别模型。
在一个实施方式中，所述将所述面部图像输入到预设的基于神经网络模型训练完成的微表情识别模型中进行运算，从而得到微表情识别结果，其中，所述微表情识别模型基于人脸图像，以及与所述人脸图像关联的微表情类别组成的样本数据训练而成的步骤S4之前，包括：
S411、获取已经训练完成的已有神经网络模型的各层权重参数；
S412、将所述各层权重参数初始化为初始微表情识别模型的各层初始权重参数；
S413、利用测试集的样本数据验证所述初始微表情识别模型，其中，所述样本数据包括人脸图像，以及与所述人脸图像关联的微表情类别；
S414、若验证通过，则将所述初始微表情识别模型记为所述微表情识别模型。
如上所述，实现了采用迁移学习，快速获得微表情识别模型。若具有已经训练好的神经网络模型，即可以免去训练的步骤，从而直接得到初始微表情识别模型。即将所述各层权重参数初始化为初始微表情识别模型的各层初始权重参数，以获得初始微表情识别模型，从而省去了训练步骤。为了防止初始微表情识别模型并不适用，还利用测试集的样本数据验证所述初始微表情识别模型，其中，所述样本数据包括人脸图像，以及与所述人脸图像关联的微表情类别，若验证通过，则将所述初始微表情识别模型记为所述微表情识别模型。从而保证最终获得的微表情识别模型正确可用。
在一个实施方式中，所述判断所述微表情识别结果是否属于预设的恶意微表情列表的步骤S5之后，包括：
S7、若所述微表情识别结果属于预设的恶意微表情列表，则判定所述审批申请异常，并向所述终端返回审批不通过的信息。
如上所述，实现了向所述终端返回审批不通过的信息。如前述，预设的恶意微表情列表是判断所述用户的微表情是否恶意的标准，从而判断所述贷款是否为恶意贷款。若所述微表情识别结果属于预设的恶意微表情列表中，表明用户存在恶意的可能性很高，从而则判定所述为恶意，并向所述终端返回审批不通过的信息。进一步地，在向所述终端返回审批不通过的信息之后，将所述终端对应的用户记为嫌疑用户，以便在以后所述用户进行下一次贷款申请时作为参考。
在一个实施方式中，所述审批指令生成方法应用于审批终端，所述审批终端是预先构建的区块链网络中的一个区块链节点，所述判断所述微表情识别结果是否属于预设的恶意微表情列表的步骤S5之后，包括：
S51、若所述微表情识别结果属于预设的恶意微表情列表，则在预先搭建的区块链网络中，获取根据所述区块链网络的预设共识机制投选出的多个审核区块链节点；
S52、向所有所述审核区块链节点发送连接确认信息，并将回复所述连接确认信息的审核区块链节点记为最终审核区块链节点；
S53、将指定数据记录申请发送给所述最终审核区块链节点，并接收所述最终审核区块链节点的审核结果，并判断所述审核结果是否满足预设的区块链记录条件，其中所述指定数据记录申请附带有所述审批申请对应的用户和所述微表情识别结果；
S54、若所述审核结果满足预设的区块链记录条件，则将所述审批申请对应的用户和所述微表情识别结果记录入所述区块链网络中。
如上所述，实现了将所述审批申请对应的用户和所述微表情识别结果记录入所述区块链网络中。本实施方式采用的共识机制可为任意共识机制，优选股份授权证明机制。其中股份授权证明机制是指，所有区块链节点投选出多个委托区块链节点，从而多个委托区块链节点代表所有区块链节点对是否将数据记录入区块链做出判断，从而避免当区块链节点数量过多时区块链运作缓慢的缺点。并且，为了防止部分审核区块链节点不能参与审核(例如部分审核区块链节点未开机，或者通讯故障等原因)而导致最终决定不准确，本实施方式还采用向所有所述审核区块链节点发送连接确认信息，并将回复所述连接确认信息的审核区块链节点记为最终审核区块链节点的方式，以确定能够审核的审核区块链节点，并记为最终审核区块链节点，将所述微表情识别结果发送给多个所述最终审核区块链节点，从而既减少了不必要的网络开销，又能保证后续判断的准确性(排除了不能参与审核的节点)。其中预设的区块链记录条件例如为所述审核结果为通过的最终审核区块链节点数量大于预设阈值。
本申请的基于微表情识别的审批指令生成方法，接收终端发送的审批申请；对所述终端对应的用户进行活体验证；并在活体验证通过之后，从所述多幅全身图像中选取至少一张作为所述终端对应的用户的初始图像；对所述初始图像进行区域识别处理从而识别出面部图像；将所述面部图像输入到预设的基于神经网络模型训练完成的微表情识别模型中进行运算；若微表情识别结果不属于预设的恶意微表情列表，则判定所述审批申请正常，并生成审批指令。从而实现了准确快捷的审批指令生成。
参照图2，本申请实施例提供一种基于微表情识别的审批指令生成装置，包括：
审批申请接收单元10，用于接收终端发送的审批申请；
初始图像获取单元20，用于根据所述审批申请，通过所述终端的摄像头采集所述终端对应的用户的多幅全身图像，根据预设的活体验证规则，利用所述多幅全身图像对所述终端对应的用户进行活体验证；并在活体验证通过之后，从所述多幅全身图像中选取至少一张作为所述终端对应的用户的初始图像，其中所述初始图像至少包括所述终端对应的用户的面部；
面部图像获取单元30，用于对所述初始图像进行区域识别处理从而识别出面部区域，并将所述面部区域范围内的图像作为面部图像；
微表情识别单元40，用于将所述面部图像输入到预设的基于神经网络模型训练完成的微表情识别模型中进行运算，从而得到微表情识别结果，其中，所述微表情识别模型基于人脸图像，以及与所述人脸图像关联的微表情类别组成的样本数据训练而成；
恶意微表情判断单元50，用于判断所述微表情识别结果是否属于预设的恶意微表情列表；
审批指令生成单元60，用于若所述微表情识别结果不属于预设的恶意微表情列表，则判定所述审批申请正常，并生成审批指令。
如上述单元10所述，接收终端发送的审批申请。传统方法对于的审批，只能通过书面数据等材料的判断，来确定是否属于恶意，无法准确快捷筛选出恶意。本申请利用微表情识别技术，以准确快捷地判断出是否属于恶意，以实现有效的审批指令生成。其中，所述审批申请指人发送的对意图事项进行审批的申请。进一步地，所述审批申请还可以包括审批材料，例如借款人的身份证号、姓名、收入情况、抵押物状况和信用情况等。
如上述单元20所述，根据所述审批申请，通过所述终端的摄像头采集所述终端对应的用户的多幅全身图像，根据预设的活体验证规则，利用所述多幅全身图像对所述终端对应的用户进行活体验证；并在活体验证通过之后，从所述多幅全身图像中选取至少一张作为所述终端对应的用户的初始图像，其中所述初始图像至少包括所述终端对应的用户的面部。其中，通过所述终端的摄像头，采集所述终端对应的用户的初始图像的方式例如为：通过所述终端的摄像头，以视频模式采集所述终端对应的用户的视频数据，从所述视频数据中根据预设的图像选择规则采集初始图像；或者，通过所述终端的摄像头，以照相模式采集所述终端对应的用户的多张图像，并以所述多张图像中的一张或者多张作为所述初始图像。所述活体验证的方式例如包括：根据所述审批申请，通过所述终端的麦克风，语音提示所述终端对应的用户做出指定肢体动作，并通过所述终端的摄像头以采集所述用户的全身图像，其中所述全身图像包括所述终端对应的用户的面部；判断所述全身图像是否包括所述指定肢体动作；若所述全身图像包括所述指定肢体动作，则判定活体验证通过。
如上述单元30所述，对所述初始图像进行区域识别处理从而识别出面部区域，并将所述面部区域 范围内的图像作为面部图像。由于初始图像可以包括较大的区域，而本申请通过面部区域进行微表情识别即可，因此通过区域识别处理从而识别出面部区域，从而减少数据计算量，减轻计算负担。其中对所述初始图像进行区域识别处理从而识别出面部区域的方式可为任意可行方式，例如：将所述初始图像划分为多个区域，将多个区域与预设的面部区域进行对比，将相似程度不超过阈值的区域作为面部区域；或者，所述初始图像划分为多个区域，将每个区域的图像数据与预设的眼睛图像数据进行对比，得到每个区域图像数据与眼睛图像数据的差值，将差值不超过预设数值的区域记为眼睛区域，将每个区域的图像数据与预设的嘴巴图像数据进行比较，得到每个区域图像数据与嘴巴图像数据的差值，将差值不超过预设数值的区域记为嘴巴区域，调用标准面部图像，并通过等比例缩小或者放大操作，使所述标准面部图像中的眼睛区域的中心与所述初始图像中的眼睛区域的中心重合，以及所述标准面部图像中的嘴巴区域的中心与所述初始图像中的嘴巴区域的中心重合；将所述初始图像中与经过所述等比例缩小或者放大操作后的标准面部图像重叠的区域记为面部区域，并将所述面部区域范围内的图像作为面部图像。
如上述单元40所述，将所述面部图像输入到预设的基于神经网络模型训练完成的微表情识别模型中进行运算，从而得到微表情识别结果，其中，所述微表情识别模型基于人脸图像，以及与所述人脸图像关联的微表情类别组成的样本数据训练而成。其中神经网络模型可以为任意模型，例如VGG16模型、VGG19模型、VGG-F模型、ResNet152模型、ResNet50模型、DPN131模型、IXception模型、AlexNet模型和DenseNet模型等，优选DPN模型。DPN(Dual Path Network)是神经网络结构，在ResNeXt的基础上引入了DenseNet的核心内容，使得模型对特征的利用更加充分。上述DPN、ResNeXt和DenseNet是现有的网络结构，在此不在赘述。据此，通过基于神经网络模型训练完成的微表情识别模型对所述面部图像进行运算，从而得到微表情识别结果。其中所述微表情识别结果包括：开心、生气、害怕、惊讶等类型，还可细分为喜笑颜开、怒火中烧、胆颤心惊、大吃一惊等类型。
如上述单元50所述，判断所述微表情识别结果是否属于预设的恶意微表情列表。预设的恶意微表情列表记录了具体的恶意微表情，例如假意微笑、心存窃喜等。若所述微表情识别结果属于预设的恶意微表情列表，例如为假意微笑，则表明很可能属于恶意申请，不应通过。
如上述单元60所述，若所述微表情识别结果不属于预设的恶意微表情列表，则判定所述审批申请正常，并生成审批指令。预设的恶意微表情列表是用于区分微表情是否恶意的判断标准。若所述微表情识别结果不属于预设的恶意微表情列表，表明表情正常，属于正常申请，应该进行正常的审批程序，据此生成审批指令。其中，所述审批指令用于指示对所述审批申请按照预设的审批程序进行审批。
在一个实施方式中，所述初始图像获取单元20，包括：
全身图像采集子单元，用于根据所述审批申请，通过所述终端的麦克风，语音提示所述终端对应的用户做出指定肢体动作，并通过所述终端的摄像头以采集所述终端对应的用户的多幅全身图像，其中所述多幅全身图像中的至少一张包括所述终端对应的用户的面部；
指定肢体动作判断子单元，用于判断所述全身图像是否包括所述指定肢体动作；
活体验证通过判定子单元，用于若所述全身图像包括所述指定肢体动作，则判定活体验证通过；
初始图像获取子单元，用于从所述多幅全身图像中选取至少一张作为所述终端对应的用户的初始图像，其中所述初始图像至少包括所述终端对应的用户的面部。
如上所述,实现了先进行活体识别，再采集所述终端对应的用户的初始图像。为了防止静态图像、伪造视频等欺骗摄像头以进行欺诈贷款的行为，本实施方式采用活体识别的方式以保证用户为活体用户。具体地，根据所述审批申请，通过所述终端的麦克风，语音提示所述终端对应的用户做出指定肢体动作，并通过所述终端的摄像头以采集所述用户的全身图像，其中所述全身图像包括所述终端对应的用户的面部；判断所述全身图像是否包括所述指定肢体动作；若所述全身图像包括所述指定肢体动作，则将所述全身图像作为所述初始图像。即，若用户为活体用户，那么将根据语音提示做出指定肢体动作，因此所述全身图像包括所述指定肢体动作可以表明用户为活体用户。因此此时的全身图像是可用的，据此将所述全身图像作为所述初始图像。进一步地，所述指定肢体动作包括多种，所述若所述全身图像包括所述指定肢体动作，则判定活体验证通过，包括：采集所述多幅全身图像中出现多种所述指定肢体动作的图像，并标记为活体验证图像；判断所述活体验证图像中的所述指定肢体动作出现的顺序是否与预设顺序相同；若相同，则判定活体验证通过。
在一个实施方式中，所述面部图像获取单元30，包括：
眼睛区域获取子单元，用于将所述初始图像划分为多个区域，将每个区域的图像数据与预设的眼睛图像数据进行对比，得到每个区域图像数据与眼睛图像数据的差值，将差值不超过预设数值的区域记为眼睛区域；
嘴巴区域获取子单元，用于将每个区域的图像数据与预设的嘴巴图像数据进行比较，得到每个区域图像数据与嘴巴图像数据的差值，将差值不超过预设数值的区域记为嘴巴区域；
重合处理子单元，用于调用标准面部图像，并通过等比例缩小或者放大操作，使所述标准面部图像 中的眼睛区域的中心与所述初始图像中的眼睛区域的中心重合，以及所述标准面部图像中的嘴巴区域的中心与所述初始图像中的嘴巴区域的中心重合；
面部图像标记子单元，用于将所述初始图像中与经过所述等比例缩小或者放大操作后的标准面部图像重叠的区域记为面部区域，并将所述面部区域范围内的图像作为面部图像。
如上所述，实现了对所述初始图像进行区域识别处理从而识别出面部区域，并将所述面部区域范围内的图像作为面部图像。其中，眼睛图像数据为标准的可用于标识眼睛特征的图像数据，嘴巴图像数据为标准的可用于标识嘴巴特征的图像数据，所述图像数据例如为图像像素(三原色等)等。而具体比对图像数据的方法可采用任意的传统比较方式，在此不赘述。进一步地，若所述眼睛区域大于划分的单个区域，则以差值不超过预设数值的多个连续的区域为眼睛区域；同理，以差值不超过预设数值的多个连续的区域为鼻子区域。由于人的面部中的五官是按一定的几何结构比例分布的，若确定眼睛区域与嘴巴区域，即可获知大致的面部轮廓。据此，调用标准面部图像，并通过等比例缩小或者放大操作，使所述标准面部图像中的眼睛区域的中心与所述初始图像中的眼睛区域的中心重合，同时使所述标准面部图像中的嘴巴区域的中心与所述初始图像中的嘴巴区域的中心重合，再将所述初始图像中与经过所述等比例缩小或者放大操作后的标准面部图像重叠的区域记为面部区域，并将所述面部区域范围内的图像作为面部图像。
在一个实施方式中，所述装置，包括：
样本数据获取单元，用于获取指定数量的样本数据，并将样本数据分成训练集和测试集；其中，所述样本数据包括人脸图像，以及与所述人脸图像关联的微表情类别；
训练单元，用于将训练集的样本数据输入到预设的神经网络模型中进行训练，得到初始微表情识别模型，其中，训练的过程中采用随机梯度下降法；
验证单元，用于利用测试集的样本数据验证所述初始微表情识别模型；
微表情识别模型标记单元，用于若验证通过，则将所述初始微表情识别模型记为所述微表情识别模型。
如上所述，实现了设置微表情识别模型。本实施方式基于神经网络模型以训练出微表情识别模型。其中神经网络模型可为VGG16模型、VGG19模型、VGG-F模型、ResNet152模型、ResNet50模型、DPN131模型、IXception模型、AlexNet模型和DenseNet模型等，优选DPN模型。其中，随机梯度下降法就是随机取样一些训练数据，替代整个训练集，如果样本量很大的情况(例如几十万)，那么可能只用其中几万条或者几千条的样本，就已经迭代到最优解了，可以提高训练速度。进一步地，训练的过程还可以采用反向传导法则更新所述神经网络模型各层的参数。其中反向传导法则(BP)建立在梯度下降法的基础上，BP网络的输入输出关系实质上是一种映射关系：一个n输入m输出的BP神经网络所完成的功能是从n维欧氏空间向m维欧氏空间中一有限域的连续映射，这一映射具有高度非线性，有利于神经网络模型各层的参数的更新。从而获得初始微表情识别模型。再利用测试集的样本数据验证所述初始微表情识别模型，若验证通过，则将所述初始微表情识别模型记为所述微表情识别模型。
在一个实施方式中，所述装置，包括：
权重参数获取单元，用于获取已经训练完成的已有神经网络模型的各层权重参数；
初始化单元，用于将所述各层权重参数初始化为初始微表情识别模型的各层初始权重参数；
测试验证单元，用于利用测试集的样本数据验证所述初始微表情识别模型，其中，所述样本数据包括人脸图像，以及与所述人脸图像关联的微表情类别；
微表情识别模型获取单元，用于若验证通过，则将所述初始微表情识别模型记为所述微表情识别模型。
如上所述，实现了采用迁移学习，快速获得微表情识别模型。若具有已经训练好的神经网络模型，即可以免去训练的步骤，从而直接得到初始微表情识别模型。即将所述各层权重参数初始化为初始微表情识别模型的各层初始权重参数，以获得初始微表情识别模型，从而省去了训练步骤。为了防止初始微表情识别模型并不适用，还利用测试集的样本数据验证所述初始微表情识别模型，其中，所述样本数据包括人脸图像，以及与所述人脸图像关联的微表情类别，若验证通过，则将所述初始微表情识别模型记为所述微表情识别模型。从而保证最终获得的微表情识别模型正确可用。
在一个实施方式中，所述装置，包括：
异常判定单元，用于若所述微表情识别结果属于预设的恶意微表情列表，则判定所述审批申请异常，并向所述终端返回审批不通过的信息。
如上所述，实现了向所述终端返回审批不通过的信息。如前述，预设的恶意微表情列表是判断所述用户的微表情是否恶意的标准，从而判断所述贷款是否为恶意贷款。若所述微表情识别结果属于预设的恶意微表情列表中，表明用户存在恶意的可能性很高，从而则判定所述为恶意，并向所述终端返回审批不通过的信息。进一步地，在向所述终端返回审批不通过的信息之后，将所述终端对应的用户记为嫌疑 用户，以便在以后所述用户进行下一次贷款申请时作为参考。
在一个实施方式中，所述审批指令生成方法应用于审批终端，所述审批终端是预先构建的区块链网络中的一个区块链节点，所述装置，包括：
审核区块链节点获取单元，用于若所述微表情识别结果属于预设的恶意微表情列表，则在预先搭建的区块链网络中，获取根据所述区块链网络的预设共识机制投选出的多个审核区块链节点；
确认信息发送单元，用于向所有所述审核区块链节点发送连接确认信息，并将回复所述连接确认信息的审核区块链节点记为最终审核区块链节点；
记录申请发送单元，用于将指定数据记录申请发送给所述最终审核区块链节点，并接收所述最终审核区块链节点的审核结果，并判断所述审核结果是否满足预设的区块链记录条件，其中所述指定数据记录申请附带有所述审批申请对应的用户和所述微表情识别结果；
记录单元，用于若所述审核结果满足预设的区块链记录条件，则将所述审批申请对应的用户和所述微表情识别结果记录入所述区块链网络中。
如上所述，实现了将所述审批申请对应的用户和所述微表情识别结果记录入所述区块链网络中。本实施方式采用的共识机制可为任意共识机制，优选股份授权证明机制。其中股份授权证明机制是指，所有区块链节点投选出多个委托区块链节点，从而多个委托区块链节点代表所有区块链节点对是否将数据记录入区块链做出判断，从而避免当区块链节点数量过多时区块链运作缓慢的缺点。并且，为了防止部分审核区块链节点不能参与审核(例如部分审核区块链节点未开机，或者通讯故障等原因)而导致最终决定不准确，本实施方式还采用向所有所述审核区块链节点发送连接确认信息，并将回复所述连接确认信息的审核区块链节点记为最终审核区块链节点的方式，以确定能够审核的审核区块链节点，并记为最终审核区块链节点，将所述微表情识别结果发送给多个所述最终审核区块链节点，从而既减少了不必要的网络开销，又能保证后续判断的准确性(排除了不能参与审核的节点)。其中预设的区块链记录条件例如为所述审核结果为通过的最终审核区块链节点数量大于预设阈值。
本申请的基于微表情识别的审批指令生成装置，接收终端发送的审批申请；对所述终端对应的用户进行活体验证；并在活体验证通过之后，从所述多幅全身图像中选取至少一张作为所述终端对应的用户的初始图像；对所述初始图像进行区域识别处理从而识别出面部图像；将所述面部图像输入到预设的基于神经网络模型训练完成的微表情识别模型中进行运算；若微表情识别结果不属于预设的恶意微表情列表，则判定所述审批申请正常，并生成审批指令。从而实现了准确快捷的审批指令生成。
参照图3，本发明实施例中还提供一种计算机设备，该计算机设备可以是服务器，其内部结构可以如图所示。该计算机设备包括通过系统总线连接的处理器、存储器、网络接口和数据库。其中，该计算机设计的处理器用于提供计算和控制能力。该计算机设备的存储器包括非易失性存储介质、内存储器。该非易失性存储介质存储有操作系统、计算机程序和数据库。该内存器为非易失性存储介质中的操作系统和计算机程序的运行提供环境。该计算机设备的数据库用于存储基于微表情识别的审批指令生成方法所用数据。该计算机设备的网络接口用于与外部的终端通过网络连接通信。该计算机程序被处理器执行时以实现一种基于微表情识别的审批指令生成方法。
上述处理器执行上述基于微表情识别的审批指令生成方法，包括以下步骤：接收终端发送的审批申请；根据所述审批申请，通过所述终端的摄像头采集所述终端对应的用户的多幅全身图像，根据预设的活体验证规则，利用所述多幅全身图像对所述终端对应的用户进行活体验证；并在活体验证通过之后，从所述多幅全身图像中选取至少一张作为所述终端对应的用户的初始图像，其中所述初始图像至少包括所述终端对应的用户的面部；对所述初始图像进行区域识别处理从而识别出面部区域，并将所述面部区域范围内的图像作为面部图像；将所述面部图像输入到预设的基于神经网络模型训练完成的微表情识别模型中进行运算，从而得到微表情识别结果，其中，所述微表情识别模型基于人脸图像，以及与所述人脸图像关联的微表情类别组成的样本数据训练而成；判断所述微表情识别结果是否属于预设的恶意微表情列表；若所述微表情识别结果不属于预设的恶意微表情列表，则判定所述审批申请正常，并生成审批指令。
在一个实施方式中，所述根据所述审批申请，通过所述终端的摄像头采集所述终端对应的用户的多幅全身图像，根据预设的活体验证规则，利用所述多幅全身图像对所述终端对应的用户进行活体验证；并在活体验证通过之后，从所述多幅全身图像中选取至少一张作为所述终端对应的用户的初始图像，其中所述初始图像至少包括所述终端对应的用户的面部的步骤，包括：根据所述审批申请，通过所述终端的麦克风，语音提示所述终端对应的用户做出指定肢体动作，并通过所述终端的摄像头以采集所述终端对应的用户的多幅全身图像，其中所述多幅全身图像中的至少一张包括所述终端对应的用户的面部；判断所述全身图像是否包括所述指定肢体动作；若所述全身图像包括所述指定肢体动作，则判定活体验证通过；从所述多幅全身图像中选取至少一张作为所述终端对应的用户的初始图像，其中所述初始图像至少包括所述终端对应的用户的面部。
在一个实施方式中，所述对所述初始图像进行区域识别处理从而识别出面部区域，并将所述面部区域范围内的图像作为面部图像的步骤，包括：将所述初始图像划分为多个区域，将每个区域的图像数据与预设的眼睛图像数据进行对比，得到每个区域图像数据与眼睛图像数据的差值，将差值不超过预设数值的区域记为眼睛区域；将每个区域的图像数据与预设的嘴巴图像数据进行比较，得到每个区域图像数据与嘴巴图像数据的差值，将差值不超过预设数值的区域记为嘴巴区域；调用标准面部图像，并通过等比例缩小或者放大操作，使所述标准面部图像中的眼睛区域的中心与所述初始图像中的眼睛区域的中心重合，以及所述标准面部图像中的嘴巴区域的中心与所述初始图像中的嘴巴区域的中心重合；将所述初始图像中与经过所述等比例缩小或者放大操作后的标准面部图像重叠的区域记为面部区域，并将所述面部区域范围内的图像作为面部图像。
在一个实施方式中，所述将所述面部图像输入到预设的基于神经网络模型训练完成的微表情识别模型中进行运算，从而得到微表情识别结果，其中，所述微表情识别模型基于人脸图像，以及与所述人脸图像关联的微表情类别组成的样本数据训练而成的步骤之前，包括：获取指定数量的样本数据，并将样本数据分成训练集和测试集；其中，所述样本数据包括人脸图像，以及与所述人脸图像关联的微表情类别；将训练集的样本数据输入到预设的神经网络模型中进行训练，得到初始微表情识别模型，其中，训练的过程中采用随机梯度下降法；利用测试集的样本数据验证所述初始微表情识别模型；若验证通过，则将所述初始微表情识别模型记为所述微表情识别模型。
在一个实施方式中，所述将所述面部图像输入到预设的基于神经网络模型训练完成的微表情识别模型中进行运算，从而得到微表情识别结果，其中，所述微表情识别模型基于人脸图像，以及与所述人脸图像关联的微表情类别组成的样本数据训练而成的步骤之前，包括：获取已经训练完成的已有神经网络模型的各层权重参数；将所述各层权重参数初始化为初始微表情识别模型的各层初始权重参数；利用测试集的样本数据验证所述初始微表情识别模型，其中，所述样本数据包括人脸图像，以及与所述人脸图像关联的微表情类别；若验证通过，则将所述初始微表情识别模型记为所述微表情识别模型。
在一个实施方式中，所述判断所述微表情识别结果是否属于预设的恶意微表情列表的步骤之后，包括：若所述微表情识别结果属于预设的恶意微表情列表，则判定所述审批申请异常，并向所述终端返回审批不通过的信息。
在一个实施方式中，所述审批指令生成方法应用于审批终端，所述审批终端是预先构建的区块链网络中的一个区块链节点，所述判断所述微表情识别结果是否属于预设的恶意微表情列表的步骤之后，包括：若所述微表情识别结果属于预设的恶意微表情列表，则在预先搭建的区块链网络中，获取根据所述区块链网络的预设共识机制投选出的多个审核区块链节点；向所有所述审核区块链节点发送连接确认信息，并将回复所述连接确认信息的审核区块链节点记为最终审核区块链节点；将指定数据记录申请发送给所述最终审核区块链节点，并接收所述最终审核区块链节点的审核结果，并判断所述审核结果是否满足预设的区块链记录条件，其中所述指定数据记录申请附带有所述审批申请对应的用户和所述微表情识别结果；若所述审核结果满足预设的区块链记录条件，则将所述审批申请对应的用户和所述微表情识别结果记录入所述区块链网络中。
本领域技术人员可以理解，图中示出的结构，仅仅是与本申请方案相关的部分结构的框图，并不构成对本申请方案所应用于其上的计算机设备的限定。
本申请的计算机设备，接收终端发送的审批申请；对所述终端对应的用户进行活体验证；并在活体验证通过之后，从所述多幅全身图像中选取至少一张作为所述终端对应的用户的初始图像；对所述初始图像进行区域识别处理从而识别出面部图像；将所述面部图像输入到预设的基于神经网络模型训练完成的微表情识别模型中进行运算；若微表情识别结果不属于预设的恶意微表情列表，则判定所述审批申请正常，并生成审批指令。从而实现了准确快捷的审批指令生成。
本申请一实施例还提供一种计算机可读存储介质，其上存储有计算机程序，计算机程序被处理器执行时实现基于微表情识别的审批指令生成方法，包括以下步骤：接收终端发送的审批申请；根据所述审批申请，通过所述终端的摄像头采集所述终端对应的用户的多幅全身图像，根据预设的活体验证规则，利用所述多幅全身图像对所述终端对应的用户进行活体验证；并在活体验证通过之后，从所述多幅全身图像中选取至少一张作为所述终端对应的用户的初始图像，其中所述初始图像至少包括所述终端对应的用户的面部；对所述初始图像进行区域识别处理从而识别出面部区域，并将所述面部区域范围内的图像作为面部图像；将所述面部图像输入到预设的基于神经网络模型训练完成的微表情识别模型中进行运算，从而得到微表情识别结果，其中，所述微表情识别模型基于人脸图像，以及与所述人脸图像关联的微表情类别组成的样本数据训练而成；判断所述微表情识别结果是否属于预设的恶意微表情列表；若所述微表情识别结果不属于预设的恶意微表情列表，则判定所述审批申请正常，并生成审批指令。
在一个实施方式中，所述根据所述审批申请，通过所述终端的摄像头采集所述终端对应的用户的多幅全身图像，根据预设的活体验证规则，利用所述多幅全身图像对所述终端对应的用户进行活体验证； 并在活体验证通过之后，从所述多幅全身图像中选取至少一张作为所述终端对应的用户的初始图像，其中所述初始图像至少包括所述终端对应的用户的面部的步骤，包括：根据所述审批申请，通过所述终端的麦克风，语音提示所述终端对应的用户做出指定肢体动作，并通过所述终端的摄像头以采集所述终端对应的用户的多幅全身图像，其中所述多幅全身图像中的至少一张包括所述终端对应的用户的面部；判断所述全身图像是否包括所述指定肢体动作；若所述全身图像包括所述指定肢体动作，则判定活体验证通过；从所述多幅全身图像中选取至少一张作为所述终端对应的用户的初始图像，其中所述初始图像至少包括所述终端对应的用户的面部。
在一个实施方式中，所述对所述初始图像进行区域识别处理从而识别出面部区域，并将所述面部区域范围内的图像作为面部图像的步骤，包括：将所述初始图像划分为多个区域，将每个区域的图像数据与预设的眼睛图像数据进行对比，得到每个区域图像数据与眼睛图像数据的差值，将差值不超过预设数值的区域记为眼睛区域；将每个区域的图像数据与预设的嘴巴图像数据进行比较，得到每个区域图像数据与嘴巴图像数据的差值，将差值不超过预设数值的区域记为嘴巴区域；调用标准面部图像，并通过等比例缩小或者放大操作，使所述标准面部图像中的眼睛区域的中心与所述初始图像中的眼睛区域的中心重合，以及所述标准面部图像中的嘴巴区域的中心与所述初始图像中的嘴巴区域的中心重合；将所述初始图像中与经过所述等比例缩小或者放大操作后的标准面部图像重叠的区域记为面部区域，并将所述面部区域范围内的图像作为面部图像。
在一个实施方式中，所述将所述面部图像输入到预设的基于神经网络模型训练完成的微表情识别模型中进行运算，从而得到微表情识别结果，其中，所述微表情识别模型基于人脸图像，以及与所述人脸图像关联的微表情类别组成的样本数据训练而成的步骤之前，包括：获取指定数量的样本数据，并将样本数据分成训练集和测试集；其中，所述样本数据包括人脸图像，以及与所述人脸图像关联的微表情类别；将训练集的样本数据输入到预设的神经网络模型中进行训练，得到初始微表情识别模型，其中，训练的过程中采用随机梯度下降法；利用测试集的样本数据验证所述初始微表情识别模型；若验证通过，则将所述初始微表情识别模型记为所述微表情识别模型。
在一个实施方式中，所述将所述面部图像输入到预设的基于神经网络模型训练完成的微表情识别模型中进行运算，从而得到微表情识别结果，其中，所述微表情识别模型基于人脸图像，以及与所述人脸图像关联的微表情类别组成的样本数据训练而成的步骤之前，包括：获取已经训练完成的已有神经网络模型的各层权重参数；将所述各层权重参数初始化为初始微表情识别模型的各层初始权重参数；利用测试集的样本数据验证所述初始微表情识别模型，其中，所述样本数据包括人脸图像，以及与所述人脸图像关联的微表情类别；若验证通过，则将所述初始微表情识别模型记为所述微表情识别模型。
在一个实施方式中，所述判断所述微表情识别结果是否属于预设的恶意微表情列表的步骤之后，包括：若所述微表情识别结果属于预设的恶意微表情列表，则判定所述审批申请异常，并向所述终端返回审批不通过的信息。
在一个实施方式中，所述审批指令生成方法应用于审批终端，所述审批终端是预先构建的区块链网络中的一个区块链节点，所述判断所述微表情识别结果是否属于预设的恶意微表情列表的步骤之后，包括：若所述微表情识别结果属于预设的恶意微表情列表，则在预先搭建的区块链网络中，获取根据所述区块链网络的预设共识机制投选出的多个审核区块链节点；向所有所述审核区块链节点发送连接确认信息，并将回复所述连接确认信息的审核区块链节点记为最终审核区块链节点；将指定数据记录申请发送给所述最终审核区块链节点，并接收所述最终审核区块链节点的审核结果，并判断所述审核结果是否满足预设的区块链记录条件，其中所述指定数据记录申请附带有所述审批申请对应的用户和所述微表情识别结果；若所述审核结果满足预设的区块链记录条件，则将所述审批申请对应的用户和所述微表情识别结果记录入所述区块链网络中。
本申请的计算机可读存储介质，接收终端发送的审批申请；对所述终端对应的用户进行活体验证；并在活体验证通过之后，从所述多幅全身图像中选取至少一张作为所述终端对应的用户的初始图像；对所述初始图像进行区域识别处理从而识别出面部图像；将所述面部图像输入到预设的基于神经网络模型训练完成的微表情识别模型中进行运算；若微表情识别结果不属于预设的恶意微表情列表，则判定所述审批申请正常，并生成审批指令。从而实现了准确快捷的审批指令生成。该计算机可读存储介质可以是非易失性的，也可以是易失性的。
本领域普通技术人员可以理解实现上述实施例方法中的全部或部分流程，是可以通过计算机程序来指令相关的硬件来完成，所述的计算机程序可存储于一非易失性计算机可读取存储介质中，该计算机程序在执行时，可包括如上述各方法的实施例的流程。其中，本申请所提供的和实施例中所使用的对存储器、存储、数据库或其它介质的任何引用，均可包括非易失性和/或易失性存储器。非易失性存储器可以包括只读存储器(ROM)、可编程ROM(PROM)、电可编程ROM(EPROM)、电可擦除可编程ROM(EEPROM)或闪存。易失性存储器可包括随机存取存储器(RAM)或者外部高速缓冲存储器。作为说明而非局限， RAM以多种形式可得，诸如静态RAM(SRAM)、动态RAM(DRAM)、同步DRAM(SDRAM)、双速据率SDRAM(SSRSDRAM)、增强型SDRAM(ESDRAM)、同步链路(Synchlink)DRAM(SLDRAM)、存储器总线(Rambus)直接RAM(RDRAM)、直接存储器总线动态RAM(DRDRAM)、以及存储器总线动态RAM(RDRAM)等。

Claims (20)

  1. 一种基于微表情识别的审批指令生成方法，其中，包括：
    接收终端发送的审批申请，其中所述审批申请指定了审批的类型；
    根据所述审批申请，通过所述终端的摄像头采集所述终端对应的用户的多幅全身图像，根据预设的活体验证规则，利用所述多幅全身图像对所述终端对应的用户进行活体验证；并在活体验证通过之后，从所述多幅全身图像中选取至少一张作为所述终端对应的用户的初始图像，其中所述初始图像至少包括所述终端对应的用户的面部；
    对所述初始图像进行区域识别处理从而识别出面部区域，并将所述面部区域范围内的图像作为面部图像；
    将所述面部图像输入到预设的基于神经网络模型训练完成的微表情识别模型中进行运算，从而得到微表情识别结果，其中，所述微表情识别模型基于人脸图像，以及与所述人脸图像关联的微表情类别组成的样本数据训练而成；
    判断所述微表情识别结果是否属于预设的恶意微表情列表；
    若所述微表情识别结果不属于预设的恶意微表情列表，则判定所述审批申请正常，并根据所述审批的类型生成审批指令。
  2. 根据权利要求1所述的基于微表情识别的审批指令生成方法，其中，所述根据所述审批申请，通过所述终端的摄像头采集所述终端对应的用户的多幅全身图像，根据预设的活体验证规则，利用所述多幅全身图像对所述终端对应的用户进行活体验证；并在活体验证通过之后，从所述多幅全身图像中选取至少一张作为所述终端对应的用户的初始图像，其中所述初始图像至少包括所述终端对应的用户的面部的步骤，包括：
    根据所述审批申请，通过所述终端的麦克风语音提示所述终端对应的用户做出指定肢体动作，并通过所述终端的摄像头以采集所述终端对应的用户的多幅全身图像，其中所述多幅全身图像中的至少一张包括所述终端对应的用户的面部；
    判断所述全身图像是否包括所述指定肢体动作；
    若所述全身图像包括所述指定肢体动作，则判定活体验证通过；
    从所述多幅全身图像中选取至少一张作为所述终端对应的用户的初始图像，其中所述初始图像至少包括所述终端对应的用户的面部。
  3. 根据权利要求1所述的基于微表情识别的审批指令生成方法，其中，所述对所述初始图像进行区域识别处理从而识别出面部区域，并将所述面部区域范围内的图像作为面部图像的步骤，包括：
    将所述初始图像划分为多个区域，将每个区域的图像数据与预设的眼睛图像数据进行对比，得到每个区域图像数据与眼睛图像数据的差值，将差值不超过预设数值的区域记为眼睛区域；
    将每个区域的图像数据与预设的嘴巴图像数据进行比较，得到每个区域图像数据与嘴巴图像数据的差值，将差值不超过预设数值的区域记为嘴巴区域；
    调用标准面部图像，并通过等比例缩小或者放大操作，使所述标准面部图像中的眼睛区域的中心与所述初始图像中的眼睛区域的中心重合，以及所述标准面部图像中的嘴巴区域的中心与所述初始图像中的嘴巴区域的中心重合；
    将所述初始图像中与经过所述等比例缩小或者放大操作后的标准面部图像重叠的区域记为面部区域，并将所述面部区域范围内的图像作为面部图像。
  4. 根据权利要求1所述的基于微表情识别的审批指令生成方法，其中，所述将所述面部图像输入到预设的基于神经网络模型训练完成的微表情识别模型中进行运算，从而得到微表情识别结果，其中，所述微表情识别模型基于人脸图像，以及与所述人脸图像关联的微表情类别组成的样本数据训练而成的步骤之前，包括：
    获取指定数量的样本数据，并将样本数据分成训练集和测试集；其中，所述样本数据包括人脸图像，以及与所述人脸图像关联的微表情类别；
    将训练集的样本数据输入到预设的神经网络模型中进行训练，得到初始微表情识别模型，其中，训练的过程中采用随机梯度下降法；
    利用测试集的样本数据验证所述初始微表情识别模型；
    若验证通过，则将所述初始微表情识别模型记为所述微表情识别模型。
  5. 根据权利要求1所述的基于微表情识别的审批指令生成方法，其中，所述将所述面部图像输入到预设的基于神经网络模型训练完成的微表情识别模型中进行运算，从而得到微表情识别结果，其中，所述微表情识别模型基于人脸图像，以及与所述人脸图像关联的微表情类别组成的样本数据训练而成的 步骤之前，包括：
    获取已经训练完成的已有神经网络模型的各层权重参数；
    将所述各层权重参数初始化为初始微表情识别模型的各层初始权重参数；
    利用测试集的样本数据验证所述初始微表情识别模型，其中，所述样本数据包括人脸图像，以及与所述人脸图像关联的微表情类别；
    若验证通过，则将所述初始微表情识别模型记为所述微表情识别模型。
  6. 根据权利要求1所述的基于微表情识别的审批指令生成方法，其中，所述判断所述微表情识别结果是否属于预设的恶意微表情列表的步骤之后，包括：
    若所述微表情识别结果属于预设的恶意微表情列表，则判定所述审批申请异常，并向所述终端返回审批不通过的信息。
  7. 根据权利要求1所述的基于微表情识别的审批指令生成方法，其中，所述审批指令生成方法应用于审批终端，所述审批终端是预先构建的区块链网络中的一个区块链节点，所述判断所述微表情识别结果是否属于预设的恶意微表情列表的步骤之后，包括：
    若所述微表情识别结果属于预设的恶意微表情列表，则在预先搭建的区块链网络中，获取根据所述区块链网络的预设共识机制投选出的多个审核区块链节点；
    向所有所述审核区块链节点发送连接确认信息，并将回复所述连接确认信息的审核区块链节点记为最终审核区块链节点；
    将指定数据记录申请发送给所述最终审核区块链节点，并接收所述最终审核区块链节点的审核结果，并判断所述审核结果是否满足预设的区块链记录条件，其中所述指定数据记录申请附带有所述审批申请对应的用户和所述微表情识别结果；
    若所述审核结果满足预设的区块链记录条件，则将所述审批申请对应的用户和所述微表情识别结果记录入所述区块链网络中。
  8. 一种基于微表情识别的审批指令生成装置，其中，包括：
    审批申请接收单元，用于接收终端发送的审批申请，其中所述审批申请指定了审批的类型；
    初始图像获取单元，用于根据所述审批申请，通过所述终端的摄像头采集所述终端对应的用户的多幅全身图像，根据预设的活体验证规则，利用所述多幅全身图像对所述终端对应的用户进行活体验证；并在活体验证通过之后，从所述多幅全身图像中选取至少一张作为所述终端对应的用户的初始图像，其中所述初始图像至少包括所述终端对应的用户的面部；
    面部图像获取单元，用于对所述初始图像进行区域识别处理从而识别出面部区域，并将所述面部区域范围内的图像作为面部图像；
    微表情识别单元，用于将所述面部图像输入到预设的基于神经网络模型训练完成的微表情识别模型中进行运算，从而得到微表情识别结果，其中，所述微表情识别模型基于人脸图像，以及与所述人脸图像关联的微表情类别组成的样本数据训练而成；
    恶意微表情判断单元，用于判断所述微表情识别结果是否属于预设的恶意微表情列表；
    审批指令生成单元，用于若所述微表情识别结果不属于预设的恶意微表情列表，则判定所述审批申请正常，并根据所述审批的类型生成审批指令。
  9. 一种计算机设备，包括存储器和处理器，所述存储器存储有计算机程序，其中，所述处理器执行所述计算机程序时实现如下步骤：
    接收终端发送的审批申请，其中所述审批申请指定了审批的类型；
    根据所述审批申请，通过所述终端的摄像头采集所述终端对应的用户的多幅全身图像，根据预设的活体验证规则，利用所述多幅全身图像对所述终端对应的用户进行活体验证；并在活体验证通过之后，从所述多幅全身图像中选取至少一张作为所述终端对应的用户的初始图像，其中所述初始图像至少包括所述终端对应的用户的面部；
    对所述初始图像进行区域识别处理从而识别出面部区域，并将所述面部区域范围内的图像作为面部图像；
    将所述面部图像输入到预设的基于神经网络模型训练完成的微表情识别模型中进行运算，从而得到微表情识别结果，其中，所述微表情识别模型基于人脸图像，以及与所述人脸图像关联的微表情类别组成的样本数据训练而成；
    判断所述微表情识别结果是否属于预设的恶意微表情列表；
    若所述微表情识别结果不属于预设的恶意微表情列表，则判定所述审批申请正常，并根据所述审批的类型生成审批指令。
  10. 如权利要求9所述的计算机设备，其中，所述处理器执行所述计算机程序时实现所述根据所述审批申请，通过所述终端的摄像头采集所述终端对应的用户的多幅全身图像，根据预设的活体验证规则， 利用所述多幅全身图像对所述终端对应的用户进行活体验证；并在活体验证通过之后，从所述多幅全身图像中选取至少一张作为所述终端对应的用户的初始图像，其中所述初始图像至少包括所述终端对应的用户的面部的步骤，包括：
    根据所述审批申请，通过所述终端的麦克风语音提示所述终端对应的用户做出指定肢体动作，并通过所述终端的摄像头以采集所述终端对应的用户的多幅全身图像，其中所述多幅全身图像中的至少一张包括所述终端对应的用户的面部；
    判断所述全身图像是否包括所述指定肢体动作；
    若所述全身图像包括所述指定肢体动作，则判定活体验证通过；
    从所述多幅全身图像中选取至少一张作为所述终端对应的用户的初始图像，其中所述初始图像至少包括所述终端对应的用户的面部。
  11. 如权利要求9所述的计算机设备，其中，所述处理器执行所述计算机程序时实现所述对所述初始图像进行区域识别处理从而识别出面部区域，并将所述面部区域范围内的图像作为面部图像的步骤，包括：
    将所述初始图像划分为多个区域，将每个区域的图像数据与预设的眼睛图像数据进行对比，得到每个区域图像数据与眼睛图像数据的差值，将差值不超过预设数值的区域记为眼睛区域；
    将每个区域的图像数据与预设的嘴巴图像数据进行比较，得到每个区域图像数据与嘴巴图像数据的差值，将差值不超过预设数值的区域记为嘴巴区域；
    调用标准面部图像，并通过等比例缩小或者放大操作，使所述标准面部图像中的眼睛区域的中心与所述初始图像中的眼睛区域的中心重合，以及所述标准面部图像中的嘴巴区域的中心与所述初始图像中的嘴巴区域的中心重合；
    将所述初始图像中与经过所述等比例缩小或者放大操作后的标准面部图像重叠的区域记为面部区域，并将所述面部区域范围内的图像作为面部图像。
  12. 如权利要求9所述的计算机设备，其中，所述处理器执行所述计算机程序时实现所述将所述面部图像输入到预设的基于神经网络模型训练完成的微表情识别模型中进行运算，从而得到微表情识别结果的步骤之前，还用于实现如下步骤：
    获取指定数量的样本数据，并将样本数据分成训练集和测试集；其中，所述样本数据包括人脸图像，以及与所述人脸图像关联的微表情类别；
    将训练集的样本数据输入到预设的神经网络模型中进行训练，得到初始微表情识别模型，其中，训练的过程中采用随机梯度下降法；
    利用测试集的样本数据验证所述初始微表情识别模型；
    若验证通过，则将所述初始微表情识别模型记为所述微表情识别模型。
  13. 如权利要求9所述的计算机设备，其中，所述处理器执行所述计算机程序时实现所述将所述面部图像输入到预设的基于神经网络模型训练完成的微表情识别模型中进行运算，从而得到微表情识别结果的步骤之前，还用于执行如下步骤：
    获取已经训练完成的已有神经网络模型的各层权重参数；
    将所述各层权重参数初始化为初始微表情识别模型的各层初始权重参数；
    利用测试集的样本数据验证所述初始微表情识别模型，其中，所述样本数据包括人脸图像，以及与所述人脸图像关联的微表情类别；
    若验证通过，则将所述初始微表情识别模型记为所述微表情识别模型。
  14. 如权利要求9所述的计算机设备，其中，所述处理器执行所述计算机程序时实现所述判断所述微表情识别结果是否属于预设的恶意微表情列表的步骤之后，还用于实现如下步骤：
    若所述微表情识别结果属于预设的恶意微表情列表，则判定所述审批申请异常，并向所述终端返回审批不通过的信息。
  15. 一种计算机可读存储介质，其上存储有计算机程序，其中，所述计算机程序被处理器执行时实现如下步骤：
    接收终端发送的审批申请，其中所述审批申请指定了审批的类型；
    根据所述审批申请，通过所述终端的摄像头采集所述终端对应的用户的多幅全身图像，根据预设的活体验证规则，利用所述多幅全身图像对所述终端对应的用户进行活体验证；并在活体验证通过之后，从所述多幅全身图像中选取至少一张作为所述终端对应的用户的初始图像，其中所述初始图像至少包括所述终端对应的用户的面部；
    对所述初始图像进行区域识别处理从而识别出面部区域，并将所述面部区域范围内的图像作为面部图像；
    将所述面部图像输入到预设的基于神经网络模型训练完成的微表情识别模型中进行运算，从而得到 微表情识别结果，其中，所述微表情识别模型基于人脸图像，以及与所述人脸图像关联的微表情类别组成的样本数据训练而成；
    判断所述微表情识别结果是否属于预设的恶意微表情列表；
    若所述微表情识别结果不属于预设的恶意微表情列表，则判定所述审批申请正常，并根据所述审批的类型生成审批指令。
  16. 如权利要求15所述的存储介质，其中，所述计算机程序被处理器执行时实现所述根据所述审批申请，通过所述终端的摄像头采集所述终端对应的用户的多幅全身图像，根据预设的活体验证规则，利用所述多幅全身图像对所述终端对应的用户进行活体验证；并在活体验证通过之后，从所述多幅全身图像中选取至少一张作为所述终端对应的用户的初始图像，其中所述初始图像至少包括所述终端对应的用户的面部的的步骤，包括：
    根据所述审批申请，通过所述终端的麦克风语音提示所述终端对应的用户做出指定肢体动作，并通过所述终端的摄像头以采集所述终端对应的用户的多幅全身图像，其中所述多幅全身图像中的至少一张包括所述终端对应的用户的面部；
    判断所述全身图像是否包括所述指定肢体动作；
    若所述全身图像包括所述指定肢体动作，则判定活体验证通过；
    从所述多幅全身图像中选取至少一张作为所述终端对应的用户的初始图像，其中所述初始图像至少包括所述终端对应的用户的面部。
  17. 如权利要求15所述的存储介质，其中，所述计算机程序被处理器执行时实现所述对所述初始图像进行区域识别处理从而识别出面部区域，并将所述面部区域范围内的图像作为面部图像的步骤，包括：
    将所述初始图像划分为多个区域，将每个区域的图像数据与预设的眼睛图像数据进行对比，得到每个区域图像数据与眼睛图像数据的差值，将差值不超过预设数值的区域记为眼睛区域；
    将每个区域的图像数据与预设的嘴巴图像数据进行比较，得到每个区域图像数据与嘴巴图像数据的差值，将差值不超过预设数值的区域记为嘴巴区域；
    调用标准面部图像，并通过等比例缩小或者放大操作，使所述标准面部图像中的眼睛区域的中心与所述初始图像中的眼睛区域的中心重合，以及所述标准面部图像中的嘴巴区域的中心与所述初始图像中的嘴巴区域的中心重合；
    将所述初始图像中与经过所述等比例缩小或者放大操作后的标准面部图像重叠的区域记为面部区域，并将所述面部区域范围内的图像作为面部图像。
  18. 如权利要求15所述的存储介质，其中，所述计算机程序被处理器执行时实现所述将所述面部图像输入到预设的基于神经网络模型训练完成的微表情识别模型中进行运算，从而得到微表情识别结果的步骤之前，包括还用于实现如下步骤：
    获取指定数量的样本数据，并将样本数据分成训练集和测试集；其中，所述样本数据包括人脸图像，以及与所述人脸图像关联的微表情类别；
    将训练集的样本数据输入到预设的神经网络模型中进行训练，得到初始微表情识别模型，其中，训练的过程中采用随机梯度下降法；
    利用测试集的样本数据验证所述初始微表情识别模型；
    若验证通过，则将所述初始微表情识别模型记为所述微表情识别模型。
  19. 如权利要求15所述的存储介质，其中，所述计算机程序被处理器执行时实现所述将所述面部图像输入到预设的基于神经网络模型训练完成的微表情识别模型中进行运算，从而得到微表情识别结果的步骤之前，还用于实现如下步骤：
    获取已经训练完成的已有神经网络模型的各层权重参数；
    将所述各层权重参数初始化为初始微表情识别模型的各层初始权重参数；
    利用测试集的样本数据验证所述初始微表情识别模型，其中，所述样本数据包括人脸图像，以及与所述人脸图像关联的微表情类别；
    若验证通过，则将所述初始微表情识别模型记为所述微表情识别模型。
  20. 如权利要求15所述的存储介质，其中，所述计算机程序被处理器执行时实现所述判断所述微表情识别结果是否属于预设的恶意微表情列表的步骤之后，还用于实现如下步骤：
    若所述微表情识别结果属于预设的恶意微表情列表，则判定所述审批申请异常，并向所述终端返回审批不通过的信息。
PCT/CN2020/088005 2019-05-30 2020-04-30 基于微表情识别的审批指令生成方法、装置和计算机设备 Ceased WO2020238552A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201910463916.3 2019-05-30
CN201910463916.3A CN110333778A (zh) 2019-05-30 2019-05-30 基于微表情识别的审批指令生成方法、装置和计算机设备

Publications (1)

Publication Number Publication Date
WO2020238552A1 true WO2020238552A1 (zh) 2020-12-03

Family

ID=68140693

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2020/088005 Ceased WO2020238552A1 (zh) 2019-05-30 2020-04-30 基于微表情识别的审批指令生成方法、装置和计算机设备

Country Status (2)

Country Link
CN (1) CN110333778A (zh)
WO (1) WO2020238552A1 (zh)

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113254994A (zh) * 2021-05-27 2021-08-13 平安普惠企业管理有限公司 数据库访问方法、装置、存储介质和计算机设备
CN113329137A (zh) * 2021-05-31 2021-08-31 口碑(上海)信息技术有限公司 图片传输方法、装置、计算机设备及计算机可读存储介质
CN113837208A (zh) * 2021-10-18 2021-12-24 北京远鉴信息技术有限公司 一种异常图像的确定方法、装置、电子设备及存储介质

Families Citing this family (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110333778A (zh) * 2019-05-30 2019-10-15 深圳壹账通智能科技有限公司 基于微表情识别的审批指令生成方法、装置和计算机设备
CN111600922B (zh) * 2020-03-25 2023-08-15 平安银行股份有限公司 基于调节比例的请求发送方法、装置和计算机设备
CN112800997B (zh) * 2020-04-10 2024-01-05 支付宝(杭州)信息技术有限公司 一种活体检测方法、装置及设备
CN111768190A (zh) * 2020-05-29 2020-10-13 平安银行股份有限公司 自助设备的安全使用方法、装置和计算机设备
CN114297426B (zh) * 2021-12-31 2024-12-17 上海浦东发展银行股份有限公司 企业异常贷款行为识别方法、装置、设备及存储介质

Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US9552510B2 (en) * 2015-03-18 2017-01-24 Adobe Systems Incorporated Facial expression capture for character animation
CN109165608A (zh) * 2018-08-30 2019-01-08 深圳壹账通智能科技有限公司 多视角微表情识别方法、装置、存储介质和计算机设备
CN109325422A (zh) * 2018-08-28 2019-02-12 深圳壹账通智能科技有限公司 表情识别方法、装置、终端及计算机可读存储介质
CN109447155A (zh) * 2018-10-29 2019-03-08 国家电网有限公司 一种人脸表情识别模型训练方法、装置及设备
CN109472690A (zh) * 2018-10-25 2019-03-15 深圳壹账通智能科技有限公司 基于区块链的贷款审批方法、装置、存储介质及电子设备
CN109697665A (zh) * 2018-12-15 2019-04-30 深圳壹账通智能科技有限公司 基于人工智能的贷款审核方法、装置、设备及介质
CN110333778A (zh) * 2019-05-30 2019-10-15 深圳壹账通智能科技有限公司 基于微表情识别的审批指令生成方法、装置和计算机设备

Family Cites Families (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107273845B (zh) * 2017-06-12 2020-10-02 大连海事大学 一种基于置信区域和多特征加权融合的人脸表情识别方法
CN107491757A (zh) * 2017-08-18 2017-12-19 上海二三四五金融科技有限公司 一种基于活体特征的防欺诈系统及控制方法

Patent Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US9552510B2 (en) * 2015-03-18 2017-01-24 Adobe Systems Incorporated Facial expression capture for character animation
CN109325422A (zh) * 2018-08-28 2019-02-12 深圳壹账通智能科技有限公司 表情识别方法、装置、终端及计算机可读存储介质
CN109165608A (zh) * 2018-08-30 2019-01-08 深圳壹账通智能科技有限公司 多视角微表情识别方法、装置、存储介质和计算机设备
CN109472690A (zh) * 2018-10-25 2019-03-15 深圳壹账通智能科技有限公司 基于区块链的贷款审批方法、装置、存储介质及电子设备
CN109447155A (zh) * 2018-10-29 2019-03-08 国家电网有限公司 一种人脸表情识别模型训练方法、装置及设备
CN109697665A (zh) * 2018-12-15 2019-04-30 深圳壹账通智能科技有限公司 基于人工智能的贷款审核方法、装置、设备及介质
CN110333778A (zh) * 2019-05-30 2019-10-15 深圳壹账通智能科技有限公司 基于微表情识别的审批指令生成方法、装置和计算机设备

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113254994A (zh) * 2021-05-27 2021-08-13 平安普惠企业管理有限公司 数据库访问方法、装置、存储介质和计算机设备
CN113329137A (zh) * 2021-05-31 2021-08-31 口碑(上海)信息技术有限公司 图片传输方法、装置、计算机设备及计算机可读存储介质
CN113837208A (zh) * 2021-10-18 2021-12-24 北京远鉴信息技术有限公司 一种异常图像的确定方法、装置、电子设备及存储介质
CN113837208B (zh) * 2021-10-18 2024-01-23 北京远鉴信息技术有限公司 一种异常图像的确定方法、装置、电子设备及存储介质

Also Published As

Publication number Publication date
CN110333778A (zh) 2019-10-15

Similar Documents

Publication Publication Date Title
WO2020238552A1 (zh) 基于微表情识别的审批指令生成方法、装置和计算机设备
CN112528259B (zh) 身份验证方法、装置、计算机设备和存储介质
US20230034040A1 (en) Face liveness detection method, system, and apparatus, computer device, and storage medium
CN212846789U (zh) 电子设备
CN113435362B (zh) 异常行为检测方法、装置、计算机设备及存储介质
CN109767321A (zh) 问答过程优化方法、装置、计算机设备和存储介质
US11269983B2 (en) Thermally enriched multi-modal and multi-channel biometric authentication
US20190147218A1 (en) User specific classifiers for biometric liveness detection
WO2021217899A1 (zh) 显示信息的加密方法、装置、设备及存储介质
CN111666973B (zh) 车辆损伤图片处理方法、装置、计算机设备及存储介质
CN118521936A (zh) 基于aigc检测算法识别ai换脸风险的方法及系统
WO2022267181A1 (zh) 代码清单的生成方法、装置、计算机设备和存储介质
CN112967216B (zh) 人脸图像关键点的检测方法、装置、设备以及存储介质
Mahmood et al. A Novel Approach to Liveness Detection: Real-time Eye Blink, Head Movement, and Lip Movement Analysis
CN113657155A (zh) 一种行为检测方法、装置、计算机设备和存储介质
CN117240607B (zh) 一种基于安全计算机的安全认证方法
CN113158920B (zh) 特定动作识别模型的训练方法、装置以及计算机设备
CN113792683B (zh) 文本识别模型的训练方法、装置、设备以及存储介质
CN114511891B (zh) 人脸识别处理方法、装置、终端及可读存储介质
EP4371120A1 (en) Device, system and method for verified self-diagnosis
Irwansyah et al. Attendance System Using Two Factor Authentication Based on Secure App with Flutter
Haluška et al. Facial liveness detection for biometric user identification
CN120602128B (zh) 多模态投毒攻击防御方法和系统
CN114596615B (zh) 基于对抗学习的人脸活体检测方法、装置、设备及介质
US20260030335A1 (en) Data processing

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 20814417

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 20814417

Country of ref document: EP

Kind code of ref document: A1

32PN Ep: public notification in the ep bulletin as address of the adressee cannot be established

Free format text: NOTING OF LOSS OF RIGHTS PURSUANT TO RULE 112(1) EPC (EPO FORM 1205A DATED 18.03.2022)

122 Ep: pct application non-entry in european phase

Ref document number: 20814417

Country of ref document: EP

Kind code of ref document: A1