Signature verification method and device, computer device and storage medium

Abstract

The application relates to a signature verification method and device, computer equipment and a storage medium, and relates to the technical field of computers. The method can be used in the field of financial technology or other related fields. The method comprises the following steps: acquiring a video image photographed in a remote signature process of a user performing a financial service, determining a target image in which the user is in a signature state from the video image, acquiring a hand region image of the user from the target image, and acquiring an identity verification result of the user according to the hand region image; in the case that the identity verification result represents that the identity verification is passed, acquiring a user signature image corresponding to the remote signature process performed by the user; and performing signature verification on the user signature image to obtain a verification result of the user signature image. The method can avoid the identity of a signature user being stolen by other users to perform remote signature, and can ensure that the user signature image after remote signature can pass the verification and meet the requirements of the financial service, so that the intelligence of remote signature is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of computer, in particular to a signature verification method and device, computer equipment, storage medium and computer program product. BACKGROUND

[0002] With the development of computer technology, a technology of using signature verification to handle financial business appears. In the process of handling financial business, the user usually needs to sign some files required for performing financial business. For example, for resource transfer business, in order to ensure the safety of the resource transfer process, the user needs to determine the amount of resources to be transferred, and then the user needs to sign the resources to be transferred to ensure the smooth execution of the resource transfer business. At the same time, in order to facilitate the user to handle financial business, some financial business can also be handled through remote handling, so a remote signature technology appears. The user can complete remote signature in the financial business program, which can ensure the smooth execution of the financial business transaction process.

[0003] The existing remote signature technology usually further executes the financial business after the user completes the signature. However, remote signature is different from on-site signature. Other users may steal the identity of the signature user to perform remote signature. In addition, the signature picture obtained after remote signature may not meet the requirements of the financial business in the absence of guidance. Therefore, the existing remote signature technology is not intelligent enough. SUMMARY

[0004] Therefore, it is necessary to provide a signature verification method, device, computer equipment, computer readable storage medium and computer program product to solve the above technical problems.

[0005] In a first aspect, the present application provides a signature verification method, which comprises:

[0006] obtaining a video image photographed in a remote signature process of a user performing a financial business, and determining a target image in which the user is in a signature state from the video image;

[0007] obtaining a hand region image of the user from the target image, and obtaining an identity verification result of the user according to the hand region image;

[0008] in a case where the identity verification result represents that the identity verification is passed, obtaining a user signature image corresponding to the remote signature process performed by the user;

[0009] performing signature verification on the user signature image to obtain a verification result of the user signature image.

[0010] In one of the embodiments, the video image comprises a plurality of video image frames; the target image is composed of a plurality of target image frames in which the user is in a signature state; the determining the target image in which the user is in a signature state from the video image comprises: inputting the plurality of video image frames into a pre-trained video classification model, obtaining image features corresponding to each of the video image frames and motion features corresponding to each of the video image frames through the video classification model; fusing the image features and the motion features to obtain image classification features corresponding to each of the video image frames; obtaining image classification identifiers of each of the video image frames according to the image classification features, and determining the target image frames from the video image frames based on the image classification identifiers; the image classification identifier represents whether the video image frame is taken in a signature state of the user; and the target image frames are determined from the video image frames based on the image classification identifiers.

[0011] In one of the embodiments, the video classification model comprises a first branch and a second branch; the inputting the plurality of video image frames into a pre-trained video classification model, obtaining image features corresponding to each of the video image frames and motion features corresponding to each of the video image frames through the video classification model comprises: obtaining a current video image frame and a neighboring video image frame corresponding to the current video image frame; the neighboring video image frame is adjacent in time to the current video image frame; inputting the current video image frame into the first branch of the video classification model to obtain the image features of the current video image frame through the first branch; inputting the current video image frame and the neighboring video image frame into the second branch of the video classification model to obtain motion information corresponding to the current video image frame through the second branch, and to obtain the motion features of the current video image frame according to the motion information.

[0012] In one of the embodiments, the obtaining the hand region image of the user from the target image comprises: inputting each of the target image frames into a pre-trained image region segmentation model to obtain a plurality of segmentation region images contained in each of the target image frames and region types of each of the segmentation region images through the image region segmentation model; and taking the segmentation region image whose region type represents a hand region type as the hand region image.

[0013] In one of the embodiments, the obtaining the identity verification result of the user according to the hand region image comprises: inputting the hand region image into a pre-trained first identity verification model and a pre-trained second identity verification model respectively, obtaining a first identity confidence of the user through the first identity verification model, and obtaining a second identity confidence of the user through the second identity verification model; determining a user identity confidence of the user according to the first identity confidence and the second identity confidence; and obtaining the identity verification result of the user according to the user identity confidence.

[0014] In one of the embodiments, the number of the hand region images is multiple, and each of the hand region images corresponds to a target image frame of the target image; the inputting the hand region image into the pre-trained first identity verification model and the pre-trained second identity verification model respectively, the obtaining the first identity confidence of the user through the first identity verification model, and the obtaining the second identity confidence of the user through the second identity verification model comprise: obtaining a current hand region image and a neighboring hand region image corresponding to the current hand region image; the target image frame corresponding to the neighboring hand region image is adjacent to the target image frame corresponding to the current hand region image; inputting the current hand region image into the first identity verification model to obtain a handwriting posture feature of the current hand region image through the first identity verification model, and obtaining the first identity confidence according to the handwriting posture feature and a pre-stored reference posture feature of the user; the reference posture feature is obtained based on a posture feature in a historical signature process of the user; inputting the current hand region image and the neighboring hand region image into the second identity verification model to obtain a handwriting action feature of the current hand region image through the second identity verification model, and obtaining the second identity confidence according to the handwriting action feature and a pre-stored reference action feature of the user; the reference action feature is obtained based on an action feature in the historical signature process of the user.

[0015] In one of the embodiments, the signature verification on the user signature image to obtain the verification result of the user signature image comprises: inputting the user signature image into a pre-trained signature verification model, obtaining a plurality of character block images contained in the user signature image through the signature verification model; obtaining character block features of each of the character block images, and obtaining character block classification identifiers corresponding to each of the character block images according to the character block features; the character block classification identifiers represent whether the character block images can be used to identify character block contents; in the case that all of the character block classification identifiers represent that the character block images can be used to identify character block contents, it is determined that the verification result of the user signature image is verification passed; in the case that any one of the character block classification identifiers represents that the character block image cannot be used to identify character block contents, it is determined that the verification result of the user signature image is verification failed.

[0016] In one of the embodiments, after it is determined that the verification result of the user signature image is verification failed, the method further comprises: displaying prompt information for prompting the user to re-perform the remote signature process.

[0017] In one of the embodiments, the obtaining of the video image photographed in the remote signature process of the user performing the financial business comprises: in response to a remote signature request triggered by the user in the process of performing the financial business, displaying a remote signature handwriting interface for the user to input the user signature image, and setting an image acquisition device to an open state; in response to an operation of the user closing the remote signature handwriting interface, closing the open state of the image acquisition device, and taking an image photographed by the image acquisition device in the open state as the video image photographed in the remote signature process of the user performing the financial business.

[0018] In a second aspect, the application further provides a signature verification device, the device comprising:

[0019] a target image acquisition module, configured to obtain a video image photographed in a remote signature process of a user performing a financial business, and determine a target image in which the user is in a signature state from the video image;

[0020] a user identity verification module, configured to obtain a hand region image of the user from the target image, and obtain an identity verification result of the user according to the hand region image;

[0021] a signature image acquisition module, configured to obtain a user signature image corresponding to the remote signature process of the user in the case that the identity verification result represents that the identity verification is passed;

[0022] a user signature verification module, configured to perform signature verification on the user signature image to obtain a verification result of the user signature image.

[0023] In a third aspect, the present application also provides a computer device. The computer device comprises a memory and a processor, the memory stores a computer program, and the processor implements the following steps when executing the computer program:

[0024] obtaining a video image photographed in a remote signature process of a user performing a financial service, determining a target image in which the user is in a signature state from the video image;

[0025] obtaining a hand region image of the user from the target image, and obtaining an identity verification result of the user according to the hand region image;

[0026] in a case where the identity verification result represents that the identity verification is passed, obtaining a user signature image corresponding to the remote signature process performed by the user;

[0027] performing signature verification on the user signature image to obtain a verification result of the user signature image.

[0028] In a fourth aspect, the present application also provides a computer readable storage medium. The computer readable storage medium stores a computer program, and the computer program is executed by a processor to implement the following steps:

[0029] obtaining a video image photographed in a remote signature process of a user performing a financial service, determining a target image in which the user is in a signature state from the video image;

[0030] obtaining a hand region image of the user from the target image, and obtaining an identity verification result of the user according to the hand region image;

[0031] in a case where the identity verification result represents that the identity verification is passed, obtaining a user signature image corresponding to the remote signature process performed by the user;

[0032] performing signature verification on the user signature image to obtain a verification result of the user signature image.

[0033] In a fifth aspect, the present application also provides a computer program product. The computer program product comprises a computer program, and the computer program is executed by a processor to implement the following steps:

[0034] obtaining a video image photographed in a remote signature process of a user performing a financial service, determining a target image in which the user is in a signature state from the video image;

[0035] obtaining a hand region image of the user from the target image, and obtaining an identity verification result of the user according to the hand region image;

[0036] In a case where the identity verification result indicates that the identity verification is passed, a user signature image corresponding to the remote signature process performed by the user is acquired;

[0037] The user signature image is subjected to signature verification, and a verification result of the user signature image is obtained.

[0038] The signature verification method, device, computer device, storage medium and computer program product described above acquire a video image captured in a remote signature process performed by a user for a financial service, determine a target image in which the user is in a signature state from the video image, acquire a hand region image of the user from the target image, and obtain an identity verification result of the user according to the hand region image. In a case where the identity verification result indicates that the identity verification is passed, a user signature image corresponding to the remote signature process performed by the user is acquired. The user signature image is subjected to signature verification, and a verification result of the user signature image is obtained. The present application identifies a target image in which the user is in a signature state from a video image captured in a remote signature process performed by a user for a financial service, and obtains an identity verification result of the user by using a hand region image contained in the target image. Therefore, only after the identity verification is passed, a user signature image of remote signature is acquired, and the user signature image is further subjected to verification to obtain a verification result, so that other users can be prevented from stealing the identity of a signature user to perform remote signature, and the user signature image after remote signature can be guaranteed to pass the verification and meet the requirements of a financial service. Therefore, the intelligence of remote signature can be improved. BRIEF DESCRIPTION OF DRAWINGS

[0039] Figure 1 A flowchart of a signature verification method in an embodiment is shown;

[0040] Figure 2 A flowchart of determining a target image to be captured in an embodiment is shown;

[0041] Figure 3 A flowchart of acquiring image features and motion features in an embodiment is shown;

[0042] Figure 4 A flowchart of obtaining an identity verification result of a user in an embodiment is shown;

[0043] Figure 5 A flowchart of obtaining a first identity credibility and a second identity credibility in an embodiment is shown;

[0044] Figure 6 A flowchart of obtaining a verification result of a user signature image in an embodiment is shown;

[0045] Figure 7 A block diagram of a signature verification device in an embodiment is shown;

[0046] Figure 8 Figure 1 is a schematic diagram of the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0047] In order to make the purposes, technical solutions and advantages of the present application clearer, the present application will be further described in detail below with reference to the drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and not used to limit the present application.

[0048] In one embodiment, as shown in Figure 1 a signature verification method is provided, and the present embodiment takes the method applied to a terminal as an example. It should be understood that the method can also be applied to a server, and can also be applied to a system including a terminal and a server, and is implemented through the interaction of the terminal and the server. In the present embodiment, the method includes the following steps:

[0049] In step S101, a video image captured in a remote signature process of a user performing a financial service is obtained, and a target image in which the user is in a signature state is determined from the video image.

[0050] The video image refers to a video image captured in a remote signature process of a user performing a financial service, for example, a video image captured in a process from when the user opens a handwritten page for remote signature to when the user submits a signature image and closes the handwritten page. The video image can include three process video images, i.e., a video image of a process from when the handwritten page is opened to when the user starts to sign, a video image of a process in which the user signs, i.e., a video image in which the user is in a signature state, and a video image of a process in which the signature is completed, the signature image is submitted, and the handwritten page is closed. The target image refers to the video image in which the user signs, i.e., the video image in which the user is in a signature state.

[0051] Specifically, in a remote signature process of a user performing a financial service, the remote signature process of the user can be captured to obtain a video image. Then, the video image in which the user is signing, i.e., the video image in which the user is in a signature state, can be determined as a target image from the video image.

[0052] In step S102, a hand region image of the user is obtained from the target image, and an identity verification result of the user is obtained according to the hand region image.

[0053] The hand region image refers to an image containing a hand region of a user in a target image, and the identity verification result refers to a user identity verification result of the user. In this embodiment, the terminal can verify the identity of the user according to the hand region image of the user, that is, the image of the hand region of the user in the determined target image is obtained as the hand region image of the user, and then the identity of the user is verified according to the hand region image to obtain the identity verification result of the user.

[0054] In step S103, if the identity verification result indicates that the identity verification is passed, a user signature image corresponding to the remote signature process performed by the user is obtained.

[0055] The user signature image is a signature image provided by the user after completing the remote signature. If the identity verification result indicates that the identity verification is passed, that is, the terminal confirms that the user currently performing the remote signature is not stolen by other users, the user signature image provided by the user after completing the remote signature process can be further obtained.

[0056] In step S104, the signature of the user signature image is verified to obtain a verification result of the user signature image.

[0057] The signature verification is used to verify the standardization of the user signature image. Since the user signature image is obtained by remote manual signature of the user, the user signature process may be arbitrary, and the obtained user signature image may be difficult to identify, so as to not meet the processing requirements of the financial business. Therefore, after obtaining the user signature image, the terminal can further verify the user signature image to determine whether the user signature image can meet the processing requirements of the financial business. Only if the signature verification is passed, that is, the user signature image can meet the processing requirements of the financial business, the terminal will further use the user signature image to perform the financial business. If the signature verification is not passed, the user needs to perform remote signature again to obtain a new user signature image to perform the financial business.

[0058] In the signature verification method, a video image captured in a remote signature process of a user performing a financial service is obtained, a target image in which the user is in a signature state is determined from the video image, a hand region image of the user is obtained from the target image, and an identity verification result of the user is obtained according to the hand region image; in a case where the identity verification result indicates that the identity verification is passed, a user signature image corresponding to the remote signature process of the user is obtained; and the user signature image is subjected to signature verification to obtain a verification result of the user signature image. The target image in which the user is in the signature state is identified from the video image captured in the remote signature process of the user performing the financial service, the identity verification of the user is obtained by using the hand region image contained in the target image, and thus, after the identity verification is passed, the user signature image of the remote signature is obtained, and the user signature image is further subjected to verification to obtain the verification result, so that the intelligent property of the remote signature can be improved, the identity of the signature user can be prevented from being stolen by other users to perform the remote signature, and the user signature image after the remote signature can be ensured to pass the verification and meet the requirements of the financial service.

[0059] In one embodiment, the video image includes a plurality of video image frames; the target image is composed of a plurality of target image frames in which the user is in the signature state; as shown in Figure 2

[0060] In step S201, the plurality of video image frames are input into the pre-trained video classification model, and image features corresponding to each video image frame and motion features corresponding to each video image frame are obtained by the video classification model.

[0061] In this embodiment, the video image can be composed of a plurality of video frames, i.e., a plurality of video image frames, and the target image can also be composed of a plurality of image frames, i.e., target image frames, in which the user is in the signature state. The video classification model is a model for distinguishing the categories of video image frames, which can be used to identify whether the user in the signature state is captured in the video image frame, the image features can be features for representing the image content contained in each video image frame, and the motion features can be motion features related to the image content that changes in the plurality of video image frames. After the terminal obtains each video image frame captured in the remote signature process of the user performing the financial service, the video classification model can be used to extract the image features corresponding to each image frame and the motion features corresponding to each image frame.

[0062] In step S202, the image features and the motion features are fused to obtain image classification features corresponding to each video image frame.

[0063] ​The image classification feature is finally used for video image frame classification. In this embodiment, after the terminal obtains the image feature and the motion feature corresponding to each video image frame, the terminal can fuse the image feature and the motion feature, so as to obtain the image classification feature corresponding to each video image frame for video image frame classification.

[0064] In step S203, the image classification identifier of each video image frame is obtained according to the image classification feature, and the target image frame is determined from the video image frame based on the image classification identifier. The image classification identifier indicates whether the user is in a signature state when the video image frame is captured.

[0065] In step S204, the target image frame is determined from the video image frame based on the image classification identifier.

[0066] The image classification identifier is a classification identifier used to indicate whether the user is in a signature state when the video image frame is captured. For example, the image classification identifier 1 can indicate that the user is in a signature state when the video image frame is captured, and the image classification identifier 0 can indicate that the user is not in a signature state when the video image frame is captured. In this embodiment, after the terminal obtains the image classification feature of each video image frame, the terminal can further classify each video image frame based on the image classification feature, determine whether the user is in a signature state when the video image frame is captured, set the video image frame to the image classification identifier 1 if the user is in a signature state when the video image frame is captured, and set the video image frame to the image classification identifier 0 if the user is not in a signature state when the video image frame is captured. Then, the terminal can further extract the video image frame with the image classification identifier 1 as the target image frame.

[0067] In this embodiment, the terminal can extract the image feature and the motion feature of each video image frame through the video classification model, classify the video image frame based on the image feature and the motion feature, and determine the target image frame according to the classification result. By using the image feature and the motion feature of the video image frame for image classification at the same time, the accuracy of the image classification result can be improved.

[0068] Further, the video classification model includes a first branch and a second branch. As shown in FIG. 2, step S201 can further include: Figure 3

[0069] In step S301, the current video image frame and the adjacent video image frame corresponding to the current video image frame are obtained. The adjacent video image frame is adjacent to the current video image frame in time.

[0070] ​The current video image frame refers to any one of the plurality of video image frames, and the adjacent video image frame refers to the image frame adjacent in shooting time to the current video image frame. In the embodiment, the plurality of video image frames can correspond to different shooting times respectively, so that the plurality of video image frames can be sorted according to the shooting times, and the adjacent video image frame can refer to the video image frame adjacent in sorting to the current video image frame, that is, the video image frame adjacent in shooting time to the current video image frame.

[0071] In step S302, the current video image frame is input into the first branch of the video classification model, and the image feature of the current video image frame is obtained through the first branch.

[0072] The first branch refers to the network branch of the video classification model for extracting image features. In the embodiment, the video classification model can be implemented by a dual-stream convolutional neural network, which can include two branches, one for processing two-dimensional image information and the other for processing three-dimensional motion information. The first branch can be the branch for processing two-dimensional image information. For the first branch, the terminal can input the current video image frame into the first branch of the model, and then extract the image feature of the current video image frame through the first branch of the model, so as to obtain the image feature of the current video image frame through the first branch.

[0073] In step S303, the current video image frame and the adjacent video image frame are input into the second branch of the video classification model, the motion information corresponding to the current video image frame is obtained through the second branch, and the motion feature of the current video image frame is obtained according to the motion information.

[0074] The second branch refers to the network branch of the video classification model for extracting motion features, that is, the branch for processing three-dimensional motion information in the dual-stream convolutional neural network. For the second branch, since the motion change of the current video image frame needs to be identified, in addition to inputting the current video image frame into the second branch, the adjacent video image frame adjacent in shooting time to the current video image frame, that is, the last video image frame before the current video image frame in shooting time and the next video image frame after the current video image frame in shooting time, are also input into the second branch, so that the motion feature corresponding to the current video image frame is obtained through the second branch. That is, the motion feature of each video image frame can be obtained through the above-mentioned manner.

[0075] In this embodiment, the terminal can obtain the image feature and the motion feature corresponding to each video image frame through the first branch and the second branch of the video classification model. The image feature can be obtained by inputting the current video image frame into the first branch of the video classification model, and the motion feature can be obtained by inputting the current video image frame and the adjacent video image frame into the second branch of the video classification model. In this way, the image feature and the motion feature of each video image frame can be accurately obtained.

[0076] In addition, step S102 can further include: inputting each target image frame into a pre-trained image region segmentation model, obtaining a plurality of segmentation region images contained in each target image frame and a region type of each segmentation region image through the image region segmentation model; and taking the segmentation region image with the region type representing a hand region type as a hand region image.

[0077] The image region segmentation model refers to a neural network model used for image region segmentation processing of an image. The model can identify image features of different image regions in the image, thereby dividing the image region into different region types. The segmentation region image refers to an image of each segmentation region obtained after image region segmentation. In this embodiment, after obtaining the target image frame, the terminal can input the target image frame into a pre-trained image region segmentation model, for example, a PolarMask model. The model introduces polar coordinate representation into example segmentation, uses angle and distance as coordinates for positioning, represents the example segmentation problem as an optimization problem of instance center classification and dense distance regression in polar coordinates, and further increases the bottom-up path enhancement based on PFN, which can better utilize bottom layer information. Therefore, the PolarMask model can be used to segment the image region of each target image frame and determine the region type of each segmentation region image, so that the segmentation region image with the region type representing a hand region type is taken as a hand region image.

[0078] In this embodiment, after the terminal segments the target image frame through the pre-trained image region segmentation model and obtains a plurality of segmentation region images contained in each target image frame, the segmentation region image with the region type representing a hand region type is taken as a hand region image, thereby realizing segmentation and acquisition of the hand region image and further improving the accuracy of hand region image acquisition.

[0079] In one embodiment, as shown in Figure 4 step S102 can further include:

[0080] Step S401, input the hand region image into the pre-trained first identity authentication model and the pre-trained second identity authentication model respectively, obtain the first identity confidence of the user through the first identity authentication model, and obtain the second identity confidence of the user through the second identity authentication model.

[0081] The first identity authentication model and the second identity authentication model are neural network models for user identity authentication using hand region images in different ways, for example, the first identity authentication model can be a model for verifying the identity of a user using the user's handwriting posture. Since different users have different handwriting postures, the first identity authentication model can identify the user's handwriting posture through the hand region image, thereby performing identity authentication using the handwriting posture to obtain the confidence of the user's identity, i.e., the first identity confidence. The second identity authentication model can be a model for verifying the identity of a user using the user's handwriting action. Since different users have different handwriting action habits, the second identity authentication model can identify the user's handwriting action through the hand region image, thereby performing identity authentication using the handwriting action to obtain the confidence of the user's identity, i.e., the second identity confidence.

[0082] Step S402, determine the user identity confidence of the user according to the first identity confidence and the second identity confidence;

[0083] Step S403, obtain the identity authentication result of the user according to the user identity confidence.

[0084] The user identity confidence is the final identity confidence for verifying the identity of the user. The terminal can determine the identity authentication result of the user based on the identity confidence, for example, if the identity confidence is greater than a certain set confidence threshold, it is determined that the user identity authentication is passed, and if the identity confidence does not satisfy the confidence threshold, it is determined that the identity authentication is not passed. Specifically, after the terminal obtains the first identity confidence based on the first identity authentication model and the second identity confidence based on the second identity authentication model in step S401, the terminal can also determine the user identity confidence according to the first identity confidence and the second identity confidence, for example, by weighting the first identity confidence and the second identity confidence to obtain the user identity confidence, thereby obtaining the identity authentication result of the user using the user identity confidence.

[0085] In this embodiment, the terminal can also obtain the first identity confidence and the second identity confidence of the user through the first identity authentication model and the second identity authentication model respectively, thereby obtaining the identity authentication result of the user using the first identity confidence and the second identity confidence. The above method can improve the accuracy of user identity authentication.

[0086] Further, the number of hand region images is multiple, and each of the hand region images corresponds to a plurality of target image frames of the composed target image. Figure 5 As shown in FIG. 4, step S401 can further include:

[0087] In step S501, a current hand region image and a neighboring hand region image corresponding to the current hand region image are obtained. The shooting time of a target image frame corresponding to the neighboring hand region image is adjacent to that of the target image frame corresponding to the current hand region image.

[0088] In this embodiment, the number of hand region images can also be multiple, and each of the hand region images corresponds to a target image frame of the composed target image. The terminal can extract a hand region image from each of the recognized target image frames, thereby obtaining a plurality of hand region images. The current hand region image refers to any one of the plurality of hand region images, and the neighboring hand region image refers to a hand region image extracted from an image frame adjacent in shooting time to the target image frame corresponding to the current hand region image. In this embodiment, the plurality of target image frames can correspond to different shooting times. After the terminal determines the current hand region image, it can further determine a target video frame from which the current hand region image is extracted and a video frame adjacent in shooting time to the target video frame, and extract a hand region image from the video frame as the neighboring hand region image.

[0089] In step S502, the current hand region image is input into the first identity authentication model, the handwriting posture feature of the current hand region image is obtained through the first identity authentication model, and the first identity confidence of the user is obtained according to the handwriting posture feature and the pre-stored reference posture feature of the user. The reference posture feature is obtained based on the posture feature in the historical signature process of the user.

[0090] The reference posture feature can be the standard posture feature of the user determined based on the posture feature in the historical signature process of the user. The terminal can record the posture feature of each signature process of the user in each signature process, and form the reference posture feature of the user according to the posture feature. After the current hand region image is obtained, the current hand region image is input into the first identity authentication model, the posture feature of the current hand region image is obtained through the first identity authentication model, i.e., the handwriting posture feature is obtained through the first identity authentication model, so as to obtain the first identity confidence of the user by using the handwriting posture feature and the pre-stored reference posture feature of the user.

[0091] In step S503, the current hand region image and the adjacent hand region image are input into the second identity authentication model, the handwriting action feature of the current hand region image is obtained through the second identity authentication model, and the second identity credibility is obtained according to the handwriting action feature and the reference action feature of the user stored in advance. The reference action feature is obtained based on the action features in the historical signature process of the user.

[0092] Similarly, the reference action feature can be the standard action feature of the user determined based on the action features in the historical signature process of the user. The terminal can record the action features of each signature process and form the reference action feature of the user according to the action features in each signature process. Since the action feature depends on the motion process of the hand region image of the user, the terminal needs to input the adjacent hand region image into the second identity authentication model in addition to the current hand region image, so as to obtain the action feature of the current hand region image through the second identity authentication model, that is, to obtain the handwriting action feature through the second identity authentication model, and to obtain the second identity credibility of the user by using the handwriting action feature and the reference action feature of the user stored in advance.

[0093] In the embodiment, the terminal can also pre-store the reference posture feature obtained based on the posture features in the historical signature process of the user and the reference action feature obtained based on the action features in the historical signature process, so as to determine the first identity credibility by using the handwriting posture feature obtained by the first identity authentication model and the reference posture feature, and to determine the second identity credibility by using the handwriting action feature obtained by the second identity authentication model and the reference action feature. Since the reference posture feature and the reference action feature are obtained through the historical signature process of the user, the accuracy of the obtained first identity credibility and the second identity credibility can be improved.

[0094] In one embodiment, as shown in FIG. 1 1, Figure 6 the step S104 can further include:

[0095] In step S601, the user signature image is input into the pre-trained signature verification model, and a plurality of character block images contained in the user signature image are obtained through the signature verification model.

[0096] The signature verification model refers to a pre-trained neural network model used for verifying whether a user signature image meets the available standard, i.e., whether it meets the financial service standard. Generally, a remote signature needs to be able to identify the word blocks included in the signature image, and can be used for processing of the financial service. Therefore, the signature verification model can be a model used for identifying whether the word blocks included in the user signature image can be used to identify the content of the word blocks, and the word block image is an image of each word block presented in the user signature image. The terminal can input the user signature image into the pre-trained signature verification model, so as to extract each word block image included in the user signature image by the signature verification model.

[0097] In step S602, the word block features of each word block image are obtained, and the word block classification identifiers respectively corresponding to each word block image are obtained according to the word block features. The word block classification identifier represents whether the word block image can be used to identify the content of the word block.

[0098] The word block feature refers to the feature corresponding to the word block image, and the word block classification identifier is identifier information used for representing whether the word block can be used to identify the content of the word block. For example, if a word block can be used to identify the content of the word block, the word block classification identifier corresponding to the word block image can be classification identifier 1. In addition, if a word block cannot be used to identify the content of the word block, the word block classification identifier corresponding to the word block image can be classification identifier 0. In this embodiment, the terminal can obtain the word block feature corresponding to each word block image, and can obtain the word block classification identifier respectively corresponding to each word block image according to the word block feature.

[0099] In step S603, in the case that all the word block classification identifiers represent that the word block image can be used to identify the content of the word block, it is determined that the verification result of the user signature image is verification passed.

[0100] In step S604, in the case that any one of the word block classification identifiers represents that the word block image cannot be used to identify the content of the word block, it is determined that the verification result of the user signature image is verification failed.

[0101] After that, if all the word block classification identifiers represent that the word block image can be used to identify the content of the word block, i.e., all the word block images can be used to identify the content of the word block, it is indicated that the user signature image is available, and can meet the financial service standard. Therefore, the terminal can determine that the verification result of the user signature image is verification passed. If one of the word block classification identifiers represents that the word block image cannot be used to identify the content of the word block, i.e., there is a word block image that cannot be used to identify the content of the word block, it is indicated that the user signature image is unavailable, and cannot meet the financial service standard. Therefore, the terminal can determine that the verification result of the user signature image is verification failed.

[0102] In this embodiment, the terminal can use the pre-trained signature verification model to divide the user signature image into multiple character block images, so as to determine whether each character block image can be used to identify the character block content, and use the above identification result to complete the verification of the user signature image. In this way, the user signature image that passes the verification can be suitable for financial business, and the efficiency of financial business can be further improved.

[0103] In addition, after step S604, the method can further include: displaying prompt information for prompting the user to re-perform the remote signature process.

[0104] If the user signature image fails the verification, i.e., the user signature image cannot be used to handle the financial business, in order to ensure the smooth handling of the financial business, the user needs to re-perform the remote signature operation process. At this time, the terminal can display prompt information prompting the user to re-perform the remote signature process, so as to inform the user to re-perform the remote signature operation.

[0105] In this embodiment, after determining that the user signature image fails the verification, the terminal can also display prompt information prompting the user to re-perform the remote signature process to handle the financial business, so as to further improve the efficiency of the financial business.

[0106] In one embodiment, step S101 can further include: in response to a remote signature request triggered by the user during the execution of the financial business process, displaying a remote signature handwriting interface for the user to input a user signature image, and setting an image acquisition device to an open state; in response to an operation of the user closing the remote signature handwriting interface, closing the open state of the image acquisition device, and taking an image of the image acquisition device in the open state as a video image taken during the remote signature process of the user performing the financial business.

[0107] In this embodiment, the video image can be taken by an image acquisition device, which can be installed on the terminal, such as a camera installed on a mobile terminal. The remote signature handwriting interface is a display interface on the terminal for the user to perform handwriting signature operation, and the remote signature request is a request triggered by the user to perform remote signature operation, such as triggering by the user clicking a control for handwriting signature in a financial application. Specifically, when the user triggers the remote signature request, the terminal can display a remote signature handwriting interface for the user to input handwriting signature, i.e., input a user signature image, and also start the image acquisition device to start collecting video images.

[0108] After that, if the user completes the handwritten signature and closes the remote signature handwritten interface, for example, by clicking the handwritten signature submission control to perform the operation of closing the remote signature handwritten interface, at this time the terminal can close the opening state of the image acquisition device, and the image captured by the image acquisition device in the opening state is taken as the video image captured in the remote signature process of the user performing the financial business. Since the image acquisition device is opened when the remote signature handwritten interface starts to display, and is closed when the remote signature handwritten interface is closed, the video image is the image captured when the remote signature handwritten interface is displayed.

[0109] In this embodiment, the terminal can only capture video images when the remote signature handwritten interface is opened, so as to reduce the captured video images as much as possible under the premise of ensuring the integrity of the signature process, thereby improving the target image screening efficiency.

[0110] In one embodiment, a signature verification method applied to remote signature is also provided. The method can acquire the hand image and signature motion trajectory of the signature user, and determine whether the identity of the user performing remote signature is stolen according to the above information. If not, the signature image of the user is further acquired, and the signature image is further processed to determine whether the signature image passes the verification. The method can specifically include the following steps:

[0111] Step 1: Acquire the video frame of the user in the signature state when performing remote signature.

[0112] When the user opens the handwritten interface of the remote signature through the APP provided by the mobile terminal, the camera of the mobile terminal can be opened at the same time, and the video frame of the user from opening the handwritten interface to closing the signature interface after completing the signature is captured through the camera. Then, the above video frame can be input into the pre-trained key frame extraction model, and the signature video frame of the user in the signature state is extracted from the above video frame through the key frame extraction model.

[0113] The video classification network structure used by the method is a double-flow convolutional neural network. Two data flows process two-dimensional image information and three-dimensional motion information respectively. The two-dimensional image information processing part takes the image of the current state frame as input, and the three-dimensional motion information processing part accepts the optical flow information in the adjacent time as input. After the above processing is completed, the features extracted from the motion information can be fused with the features obtained through the image information, so as to determine the part where the action occurs in the similar image scene, and improve the accuracy of the final classification result of the image information flow branch.

[0114] Step 2: Perform hand region segmentation on the signature video frame to obtain the hand region image in the signature video frame.

[0115] Input the signature video frame obtained in step 1 into the trained hand region recognition model, and obtain the hand region image in the signature video frame through the hand region recognition model.

[0116] There are feature information such as the geometric shape of the hand, the fingerprint of the finger, and the pen holding posture in the signature hand image, which is helpful for identity recognition. In this paper, the signature hand in the signature video frame is segmented using the exemplar segmentation method, and PolarMask is selected as the basic network. The PolarMask model introduces polar coordinate representation into exemplar segmentation, uses angle and distance as coordinates to locate, represents the exemplar segmentation problem as an optimization problem of instance center classification and dense distance regression in polar coordinates, and further increases the bottom-up path enhancement based on PFN, which can better utilize the bottom layer information.

[0117] Step 3: Identity recognition verification using hand image.

[0118] Input the hand region image obtained in step 2 into the trained action detection model, extract the action features and hand image features corresponding to the hand region image through the action detection model, and use the action features and hand image features to verify the identity of the signature user to determine the identity verification result of the signature user.

[0119] Because the pen holding posture of each person is different when writing, a subnetwork for processing two-dimensional information based on attention mechanism is used to identify the identity of the writer. The nomination module in the network is responsible for mining the unique part of the writer's writing posture, and the evaluation module sorts the reliability of the region extracted by the above nomination module. In addition to the hand posture information, the unique hand writing action in the signature process can also assist in identity recognition. In this paper, two-dimensional convolution operation and three-dimensional convolution operation are used to determine the corresponding video frame. Based on two-dimensional convolution operation, two-dimensional features on the image are extracted, and then three-dimensional convolution operation is used to extract the difference in each time period to further obtain the motion features with space-time information.

[0120] Step 4: If the identity verification result is verified, obtain the signature image and perform detection verification on the signature image.

[0121] After the segmented signature image is preliminarily processed by the method of graying and binarization, the signature image is segmented into word block pictures by the projection method, and then the word block pictures are normalized to unify the size and format of the images. In view of the handwriting identification, the skip connection is introduced into the improved ResNet-50 network to further deepen the network while avoiding the occurrence of gradient explosion or gradient disappearance. The local features in the signature image are extracted through the perception field in the network, and the parameters in the network are trained through the error back propagation algorithm. The more discriminative and robust handwriting convolution features are extracted through the above-mentioned manner to perform signature verification.

[0122] Step 5: If the signature image detection fails, prompt the user to sign again.

[0123] If the detected signature image fails the detection, a prompt information is displayed on the APP end to prompt the user to perform the signature operation again.

[0124] The signature verification method applied to remote signature provided in the embodiment collects the hand image and signature motion trajectory and the like of a signature user, and judges whether the identity of the user performing remote signature is stolen according to the above information. If not, the signature image of the user is further acquired, and the signature image is further processed to judge whether the signature image passes the verification. Compared with the remote signature method provided in the prior art, the present application can obtain the hand image and signature motion trajectory and the like by acquiring the video frame of the signature state, so as to judge the user identity to avoid the identity from being stolen. Meanwhile, the signature image can be further detected and verified, so that the signature can pass the detection and verification, and the intelligence of remote signature verification is further improved.

[0125] It should be understood that, although each step in the flowchart involved in each embodiment as described above is displayed in sequence according to the arrow, these steps are not necessarily executed in sequence according to the arrow. Unless otherwise specified herein, the execution of these steps is not strictly limited in sequence, and these steps can be executed in other sequences. Moreover, at least part of the steps in the flowchart involved in each embodiment as described above can include multiple steps or stages, which are not necessarily executed at the same time, but can be executed at different times, and the execution sequence of these steps or stages is not necessarily sequential, but can be executed in rotation or alternation with at least part of other steps or steps or stages in other steps.

[0126] Based on the same inventive concept, the embodiments of the present application also provide a signature verification device for implementing the signature verification method described above. The implementation scheme of the device for solving the problem is similar to the implementation scheme described in the above method, so the specific limitations in one or more signature verification device embodiments provided below can refer to the limitations of the signature verification method in the foregoing, which will not be described here again.

[0127] In one embodiment, as shown in Figure 7 A signature verification device is provided, comprising: a target image acquisition module 701, a user identity verification module 702, a signature image acquisition module 703, and a user signature verification module 704, wherein:

[0128] The target image acquisition module 701 is configured to acquire a video image captured during a remote signature process of a user performing a financial service, and determine a target image in which the user is in a signature state from the video image.

[0129] The user identity verification module 702 is configured to acquire a hand region image of the user from the target image, and obtain an identity verification result of the user according to the hand region image.

[0130] The signature image acquisition module 703 is configured to acquire a user signature image corresponding to the remote signature process of the user in the case where the identity verification result represents that the identity verification is passed.

[0131] The user signature verification module 704 is configured to perform signature verification on the user signature image to obtain a verification result of the user signature image.

[0132] In one embodiment, the video image comprises a plurality of video image frames; the target image is composed of a plurality of target image frames in which the user is in the signature state; the target image acquisition module 701 is further configured to input the plurality of video image frames into a pre-trained video classification model, acquire image features and motion features corresponding to each video image frame through the video classification model, fuse the image features and the motion features to obtain image classification features corresponding to each video image frame, acquire image classification identifiers of each video image frame according to the image classification features, and determine target image frames from the video image frames based on the image classification identifiers; the image classification identifier represents whether the video image frame captures the user in the signature state; and the target image frames are determined from the video image frames based on the image classification identifiers.

[0133] In one embodiment, the video classification model comprises a first branch and a second branch; the target image acquisition module 701 is further configured to acquire a current video image frame and a neighboring video image frame corresponding to the current video image frame; the neighboring video image frame is adjacent in time to the current video image frame; the current video image frame is input into the first branch of the video classification model, and image features of the current video image frame are obtained through the first branch; the current video image frame and the neighboring video image frame are input into the second branch of the video classification model, motion information corresponding to the current video image frame is obtained through the second branch, and motion features of the current video image frame are obtained according to the motion information.

[0134] In one embodiment, the user identity authentication module 702 is further configured to input each target image frame into a pre-trained image region segmentation model, obtain a plurality of segmentation region images contained in each target image frame and region types of each segmentation region image through the image region segmentation model, and take the segmentation region images with the region type of a hand region type as hand region images.

[0135] In one embodiment, the user identity authentication module 702 is further configured to input the hand region images into a pre-trained first identity authentication model and a pre-trained second identity authentication model respectively, obtain a first identity confidence of the user through the first identity authentication model, and obtain a second identity confidence of the user through the second identity authentication model; determine a user identity confidence of the user according to the first identity confidence and the second identity confidence; and obtain an identity authentication result of the user according to the user identity confidence.

[0136] In one embodiment, the number of hand region images is a plurality, which respectively correspond to a plurality of target image frames constituting a target image; the user identity authentication module 702 is further configured to acquire a current hand region image and a neighboring hand region image corresponding to the current hand region image; the neighboring hand region image corresponds to a target image frame adjacent in time to a target image frame corresponding to the current hand region image; the current hand region image is input into the first identity authentication model, handwriting posture features of the current hand region image are obtained through the first identity authentication model, and the first identity confidence is obtained according to the handwriting posture features and pre-stored reference posture features of the user; the reference posture features are obtained based on posture features in a historical signature process of the user; the current hand region image and the neighboring hand region image are input into the second identity authentication model, handwriting action features of the current hand region image are obtained through the second identity authentication model, and the second identity confidence is obtained according to the handwriting action features and pre-stored reference action features of the user; the reference action features are obtained based on action features in the historical signature process of the user.

[0137] In an embodiment, the user signature verification module 704 is further configured to input the user signature image into a pre-trained signature verification model, obtain a plurality of character block images contained in the user signature image by using the signature verification model, obtain character block features of each character block image, and obtain character block classification identifiers corresponding to each character block image according to the character block features. The character block classification identifiers represent whether the character block image can be used to identify the character block content. In a case where all the character block classification identifiers represent that the character block image can be used to identify the character block content, it is determined that the verification result of the user signature image is verification passed. In a case where any one of the character block classification identifiers represents that the character block image cannot be used to identify the character block content, it is determined that the verification result of the user signature image is verification failed.

[0138] In an embodiment, the user signature verification module 704 is further configured to display prompt information for prompting the user to re-perform the remote signature process.

[0139] In an embodiment, the target image acquisition module 701 is further configured to, in response to a remote signature request triggered by the user in the process of performing the financial service, display a remote signature handwriting interface for the user to input a user signature image, and set an image acquisition device to an open state; in response to an operation of the user closing the remote signature handwriting interface, close the open state of the image acquisition device, and take an image captured by the image acquisition device in the open state as a video image captured in a remote signature process of the user performing the financial service.

[0140] The above modules in the signature verification apparatus can be all or partially implemented by software, hardware, or a combination thereof. The above modules can be embedded in or independent of a processor in a computer device in a hardware form, or stored in a memory in the computer device in a software form, so as to be called and executed by a processor to perform operations corresponding to the above modules.

[0141] In an embodiment, a computer device is provided, which can be a terminal. An internal structure diagram of the computer device can be as shown in FIG. 8. Figure 8As shown in the figure. The computer device includes a processor, a memory, a communication interface, a display screen and an input device connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operating system and the computer program in the non-volatile storage medium to run. The communication interface of the computer device is used for wired or wireless communication with external terminals. Wireless communication can be achieved through WIFI, mobile cellular network, NFC (near field communication) or other technologies. The computer program is executed by the processor to implement a signature verification method. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen. The input device of the computer device can be a touch layer overlaid on the display screen, or a key, trackball or touchpad arranged on the shell of the computer device, or an external keyboard, touchpad or mouse, etc.

[0142] Those skilled in the art can understand that, Figure 8 The structure shown in the figure is only a block diagram of part of the structure related to the scheme of the present application, and does not constitute a limitation on the computer device to which the scheme of the present application is applied. A specific computer device can include more or fewer components than those shown in the figure, or combine certain components, or have a different component arrangement.

[0143] In one embodiment, a computer device is also provided, including a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the steps in the above method embodiments.

[0144] In one embodiment, a computer readable storage medium is provided, which stores a computer program, and the computer program is executed by a processor to implement the steps in the above method embodiments.

[0145] In one embodiment, a computer program product is provided, including a computer program, and the computer program is executed by a processor to implement the steps in the above method embodiments.

[0146] It should be noted that the user information (including but not limited to user equipment information, user personal information, etc.) and data (including but not limited to data for analysis, stored data, displayed data, etc.) involved in the present application are all information and data authorized by the user or authorized by all parties.

[0147] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a non-volatile computer readable storage medium, and when the computer program is executed, the processes of the above-mentioned embodiments of the methods can be included. Any reference to memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (Read-Only Memory, ROM), magnetic tape, floppy disk, flash memory, optical storage, high-density embedded non-volatile memory, resistive memory (ReRAM), magnetoresistive random access memory (Magnetoresistive Random Access Memory, MRAM), ferroelectric memory (Ferroelectric Random Access Memory, FRAM), phase change memory (Phase Change Memory, PCM), graphene memory, etc. Volatile memory can include random access memory (Random Access Memory, RAM) or external cache memory, etc. As an illustration but not limitation, RAM can be in various forms, such as static random access memory (Static Random Access Memory, SRAM) or dynamic random access memory (Dynamic Random Access Memory, DRAM), etc. The database involved in the embodiments provided in the present application can include at least one of a relational database and a non-relational database. The non-relational database can include a distributed database based on a block chain, etc., without being limited thereto. The processor involved in the embodiments provided in the present application can be a general-purpose processor, a central processing unit, a graphics processing unit, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., without being limited thereto.

[0148] Any combination of the technical features of the above embodiments can be made. In order to make the description simple, all possible combinations of the technical features in the above embodiments are not described, however, as long as the combination of the technical features does not exist contradictory, it should be considered as the scope of the present application.

[0149] The above embodiments only express several implementation manners of the present application, and the description is more specific and detailed, but it should not be understood as a limitation on the scope of the patent of the present application. It should be pointed out that for ordinary skilled in the art, without departing from the concept of the present application, a number of modifications and improvements can be made, which are within the scope of protection of the present application. Therefore, the protection scope of the present application should be subject to the appended claims.

Claims

1. A signature verification method characterized by, The method comprises: acquiring a video image photographed in a remote signature process of a user performing a financial service, determining a target image in which the user is in a signature state from the video image; acquiring a hand region image of the user from the target image, and acquiring an identity verification result of the user according to the hand region image; in a case where the identity verification result represents that the identity verification is passed, acquiring a user signature image corresponding to the remote signature process performed by the user; performing signature verification on the user signature image, inputting a pre-trained signature verification model, and acquiring a plurality of character block images contained in the user signature image through the signature verification model; acquiring character block features of each of the character block images, and acquiring character block classification identifiers corresponding to each of the character block images according to the character block features; the character block classification identifiers represent whether the character block images can be used to identify character block content; in a case where each of the character block classification identifiers represents that the character block image can be used to identify the character block content, determining that a verification result of the user signature image is verification passed; in a case where any one of the character block classification identifiers represents that the character block image cannot be used to identify the character block content, determining that the verification result of the user signature image is verification failed.

2. The method of claim 1, wherein, The video image comprises a plurality of video image frames; the target image is composed of a plurality of target image frames in which the user is in a signature state; The method comprises: inputting the plurality of video image frames into a pre-trained video classification model, acquiring image features and motion features corresponding to each of the video image frames through the video classification model; fusing the image features and the motion features to obtain image classification features corresponding to each of the video image frames; acquiring image classification identifiers of each of the video image frames according to the image classification features, and determining the target image frames from the video image frames based on the image classification identifiers; the image classification identifiers represent whether the video image frames photograph the user in a signature state; determining the target image frames from each of the video image frames based on the image classification identifiers.

3. The method of claim 2, wherein, The video classification model comprises a first branch and a second branch; The method comprises: acquiring a current video image frame and a neighboring video image frame corresponding to the current video image frame; the neighboring video image frame is adjacent in time to the current video image frame; inputting the current video image frame into the first branch of the video classification model, and acquiring image features of the current video image frame through the first branch; input the current video image frame and the adjacent video image frame into a second branch of a video classification model, acquire motion information corresponding to the current video image frame through the second branch, and obtain motion features of the current video image frame according to the motion information.

4. The method of claim 2, wherein, The method further includes: inputting each target image frame into a pre-trained image region segmentation model to obtain a plurality of segmentation region images contained in each target image frame and a region type of each segmentation region image through the image region segmentation model; segmentation region images with the region type being a hand region type are taken as the hand region images.

5. The method of claim 1, wherein, The method further includes: inputting the hand region images into a pre-trained first identity verification model and a pre-trained second identity verification model respectively, acquiring a first identity credibility of the user through the first identity verification model, and acquiring a second identity credibility of the user through the second identity verification model; determining a user identity credibility of the user according to the first identity credibility and the second identity credibility; acquiring an identity verification result of the user according to the user identity credibility.

6. The method of claim 5, wherein, The number of the hand region images is a plurality, and each hand region image corresponds to a different target image frame constituting the target image. The method further includes: acquiring a current hand region image and an adjacent hand region image corresponding to the current hand region image; a target image frame corresponding to the adjacent hand region image is adjacent to a target image frame corresponding to the current hand region image in shooting time; inputting the current hand region image into the first identity verification model to acquire a handwriting posture feature of the current hand region image through the first identity verification model, and acquiring the first identity credibility according to the handwriting posture feature and a reference posture feature of the user pre-stored; the reference posture feature is acquired based on posture features in a historical signature process of the user; inputting the current hand region image and the adjacent hand region image into the second identity verification model to acquire a handwriting action feature of the current hand region image through the second identity verification model, and acquiring the second identity credibility according to the handwriting action feature and a reference action feature of the user pre-stored; the reference action feature is acquired based on action features in the historical signature process of the user.

7. The method of claim 1, wherein, After determining that the verification result of the user signature image is not passed, the method further includes: displaying prompt information for prompting the user to perform the remote signature process again.

8. The method according to any one of claims 1 to 7, characterized in that, The method further includes: In response to a remote signature request triggered by the user in the process of performing the financial business, a remote signature handwriting interface for the user to input the user signature image is displayed, and the image acquisition device is set to an open state; In response to an operation of the user closing the remote signature handwriting interface, the open state of the image acquisition device is closed, and an image captured by the image acquisition device in the open state is taken as a video image captured in a remote signature process of the user performing the financial business.

9. A signature verification apparatus characterized by comprising: The apparatus comprises: a target image acquisition module configured to acquire a video image captured in a remote signature process of a user performing a financial business, and determine a target image in which the user is in a signature state from the video image; a user identity verification module configured to acquire a hand region image of the user from the target image, and acquire an identity verification result of the user according to the hand region image; a signature image acquisition module configured to acquire a user signature image corresponding to the remote signature process of the user in a case where the identity verification result represents that the identity verification is passed; a user signature verification module configured to perform signature verification on the user signature image, input a pre-trained signature verification model, and acquire a plurality of character block images contained in the user signature image through the signature verification model; acquire character block features of each of the character block images, and acquire a character block classification identifier corresponding to each of the character block images according to the character block features; the character block classification identifier represents whether the character block image can be used to identify character block content; in a case where each of the character block classification identifiers represents that the character block image can be used to identify the character block content, determine that a verification result of the user signature image is verification passed; in a case where any one of the character block classification identifiers represents that the character block image cannot be used to identify the character block content, determine that the verification result of the user signature image is verification failed. 10.A computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the computer device is configured to perform the method according to any one of claims 1-9. The processor executes the computer program to implement the steps of the method of any one of claims 1 to 8.

11. A computer readable storage medium having stored thereon a computer program, characterized in that The computer program is executed by the processor to implement the steps of the method of any one of claims 1 to 8.

12. A computer program product comprising a computer program, characterized in that, The computer program is executed by the processor to implement the steps of the method of any one of claims 1 to 8.

Images