TWM658166U - Service System - Google Patents

Abstract

A service system includes a processing unit and a storage unit electrically connected to the processing unit. The storage unit stores a face image verification model, which is configured to identify whether the face presented by the image belongs to a living body. The processing unit is used to: input a first face image part in an image data to be verified into the face image verification model to obtain a recognition result; determine whether the recognition result is a positive recognition result or a negative judgment result; if the recognition result is determined to be the positive recognition result, grant a service request and execute a service program corresponding to the service request; if the recognition result is determined to be the negative judgment result, reject the service request and output a text message data message contained in the negative judgment result.

Description

Service System

The present invention relates to a service system, and in particular to a service system suitable for providing online services to users through the Internet.

With the rapid development and popularization of Internet technology, the Internet has become an indispensable part of people's daily life. From financial transactions, online shopping to social media interactions, the Internet provides users with unlimited convenience and opportunities.

However, these conveniences also lead to challenges in personal data security and privacy protection. For example, on some online services that require users to register before using them, personal data may be used by malicious persons to register.

From the perspective of Internet service providers, the existing technology allows users to register with information such as email addresses, mobile phone numbers, and even ID numbers. In essence, it is impossible to confirm whether there is a situation where someone has registered in someone else's name. The main reason is that the above personal information is very easy to leak out and be used by someone with ulterior motives. Therefore, how to improve the above problem has become the issue to be discussed in this case.

In order to prevent someone from using other people's personal information to register fraudulently, the purpose of this invention is to provide a service system that can verify facial images.

The novel service system includes a processing unit and a storage unit electrically connected to the processing unit. The storage unit stores a face image verification model pre-trained by machine learning technology, wherein the face image verification model is implemented by a visual language model and is configured to identify whether the face presented by the input image belongs to a living body, and when the recognition result is negative, generate and output an explanatory text message presented in natural language and related to the recognition result. The processing unit is used to: after obtaining an image data to be verified corresponding to a service request, input a first face image portion included in the image data to be verified into the face image verification model to obtain a recognition result generated and output by the face image verification model based on the first face image portion; determine whether the recognition result is a positive recognition result indicating successful verification or a negative recognition result indicating successful verification. a negative judgment result indicating a verification failure and including a text message data; if the recognition result is determined to be the positive recognition result, the service request is granted and a service procedure corresponding to the service request is executed; if the recognition result is determined to be the negative judgment result, the service request is rejected and the text message data message included in the negative judgment result is output.

In some embodiments of the novel service system, the processing unit is adapted to be electrically connected to a client electronic device. The processing unit receives the image data to be verified from the client electronic device, and, when the text message data includes a prompting instruction text message indicating a suggested adjustment of the shooting method, the processing unit outputs the text message data in a manner including transmitting the prompting instruction text message to the client electronic device so that the prompting instruction text message can be displayed by the client electronic device.

In some implementations of the novel service system, the storage unit further stores a warning database. When the text message data includes a warning description text message indicating a specific type of image counterfeiting, the processing unit outputs the text message data in a manner that includes generating a warning verification record including the warning description text message and a warning feature data, and providing the warning verification record to the storage unit so that the warning verification record is added to the warning database, wherein the warning feature data is generated by the processing unit based on the image data to be verified.

In some implementations of the novel service system, the processing unit is adapted to be electrically connected to a service-end electronic device. The storage unit also stores a facial feature matching rule suitable for application to facial image recognition. The processing unit, when determining that the recognition result is the affirmative recognition result, grants the service request and executes the service program in a manner including: when determining that the recognition result is the affirmative recognition result, determining, according to the facial feature matching rule, whether the first facial image portion matches a second facial image portion included in the image data to be verified, wherein the second facial image portion presents a facial image on a certificate; when determining that the first facial image portion matches the second facial image portion, the first facial image portion and the second facial image portion are matched. One of the facial image parts is used as an image data to be checked, and it is determined whether there is any warning verification record matching the image data to be checked in the warning database; when it is determined that there is no warning verification record matching the image data to be checked in the warning database, the service request is approved and the service program is executed; when it is determined that there is a warning verification record matching the image data to be checked in the warning database, the service request is not approved, and a manual review notification corresponding to the service request is generated and transmitted to the service-end electronic device.

In some implementations of the novel service system, the storage unit further stores a certificate verification rule suitable for being applied to certificate image recognition and indicating a certificate eligibility condition. The processing unit is also used to determine whether the image data to be verified presents a certificate that meets the certificate eligibility condition according to the certificate verification rule after obtaining the image data to be verified, and the processing unit only approves the service request and executes the service procedure when it is determined that the image data to be verified presents a certificate that meets the certificate eligibility condition and the recognition result is the positive recognition result.

The effect of the present invention is that after the processing unit obtains the image data to be verified corresponding to the service request, it can use the face image verification model to determine whether the face in the first face image part is alive, and if the face image verification model outputs a positive recognition result, the service request is granted, otherwise the service request is rejected. In this way, the service system can determine whether to grant a service request by performing liveness detection on the face image, so it helps to prevent malicious people from using other people's data to send service requests.

Before the present invention is described in detail, it should be noted that, unless otherwise specifically defined, the term "electrically connected" in this patent specification is used to describe the "coupled" relationship between computer hardware (e.g., electronic systems, equipment, devices, units, components), and generally refers to "wired electrical connections" achieved by physically connecting multiple computer hardware through conductors/semiconductor materials, and "radio connections" that achieve wireless data transmission using wireless communication technology (e.g., but not limited to wireless networks, Bluetooth, and electromagnetic induction, etc.). On the other hand, unless otherwise specifically defined, the "electrical connection" described in this patent specification also generally refers to a "direct electrical connection" achieved by directly coupling multiple computer hardware to each other, and an "indirect electrical connection" achieved by indirectly coupling multiple computer hardware through other computer hardware.

The "unit" described in this patent specification represents computer hardware rather than software. For example, a "processing unit" is used to represent computer hardware with data processing capabilities. On the other hand, the "unit" described in this patent specification can refer to a single computer hardware with a specific function, or a group of computer hardware with similar functions. For example, a "processing unit" can refer to a single processor with data processing capabilities, but it can also refer to a collection of a group of processors.

Referring to FIG. 1 , an embodiment of the novel service system 1 belongs to a service institution. In the application of the embodiment, the service institution is, for example, a financial service institution (e.g., a commercial bank), and the service system 1 is suitable for use with multiple client electronic devices 10 (only one of which is shown in FIG. 1 ) and a service-side electronic device 20. For ease of understanding, the following description only uses the client electronic device 10 shown in FIG. 1 to illustrate the embodiment .

The client electronic device 10 belongs to a user, and can be a smart phone or a tablet computer, or a laptop or a desktop computer with a camera. On the other hand, the service electronic device 20 is, for example, a desktop computer, and is used by the business personnel working in the financial service institution to operate.

In this embodiment, the service system 1 is implemented as a server device as a whole, and the service system 1 includes a processing unit 11 and a storage unit 12 electrically connected to the processing unit 11, and the processing unit 11 is suitable for being electrically connected to the client electronic device 10 and the server electronic device 20 through a network, thereby communicating with the client electronic device 10 and the server electronic device 20.

More specifically, in this embodiment, the processing unit 11 is a processor implemented by an integrated circuit and having data operation and instruction transmission and reception functions, and the storage unit 12 is a data storage device for storing digital data (such as a hard disk, or other types of computer-readable recording media). However, in similar embodiments, the processing unit 11 can also be a circuit component including a processor and a circuit board, and the storage unit 12 can also be a collection of multiple storage devices of the same or different types. Furthermore, in other embodiments, the service system 1 may also be implemented as a plurality of mutually electrically connected server devices (equivalent to a server device set), in which case the processing unit 11 may be implemented as a set of a plurality of processors/circuit components respectively possessed by the server devices, and the storage unit 12 may be implemented as a set of a plurality of storage devices respectively possessed by the server devices. Based on the above, the actual implementation of the service system 1 in terms of computer hardware is not limited to this embodiment.

In this embodiment, as shown in FIG. 1 , the storage unit 12 stores a face image verification model M, a document verification rule R2, a face feature matching rule R1, a registration database DB1, and a warning database DB2.

The face image verification model M is pre-trained by combining liveness detection technology and natural language processing technology and using machine learning (or deep learning) technology. More specifically, the face image verification model M is a vision-language model, which includes a convolutional neural network part and a transformer part, wherein the convolutional neural network part is used to process and analyze the image input to the face image verification model M, and the transformer part is used to generate natural language text messages used to describe the image according to the analysis results of the convolutional neural network part on the image. Furthermore, the visual language model may adopt a model with multi-modal capabilities such as GPT-4V, Gemini, CLIP or LLaVA, but is not limited thereto.

In this embodiment, the facial image verification model M is configured to: identify whether the face presented in the image is a living body based on the image input into the facial image verification model M itself, and when the recognition result is negative (that is, when it is recognized as non-living and/or the recognition confidence score is insufficient), generate and output an explanatory text message presented in natural language and related to the recognition result. To be more specific, based on liveness detection technology, the facial image verification model M can be used to identify whether the face presented in the image is a face of a living human, a flat face mask, a three-dimensional face mask, a virtual face generated by deepfake technology, or the result of reshooting a facial image (such as a facial photo or video). Therefore, the aforementioned "identifying whether the face presented in the input image is a living person" is equivalent to judging whether the image itself is a "photograph directly taken of a real living human" based on the face presented in the image.

Furthermore, when the facial image verification model M determines that the face in the image is not alive, the explanatory text message generated by the facial image verification model M is used to indicate a specific type of image forgery that it has identified, such as "image copying", "3D mask" or "deep fake", but not limited to this. On the other hand, when the imaging quality of the image itself is poor, resulting in an insufficient confidence score in the recognition result of the facial image verification model M, the explanatory text message generated and output by the facial image verification model M will indicate at least one recommended adjustment method for shooting, such as "move to a place with sufficient light" or "keep the lens stable", but not limited to this.

In addition, during the training phase of the face image verification model M, the data used by the face image verification model M for machine learning includes, for example, a group of visual language training data, and the group of visual language training data includes a group of first-category visual language training data and a group of second-category visual language training data.

Specifically, for each first-category visual language training data, the first-category visual language training data includes a first-category training image portion that has undergone image preprocessing, and a first-category training text portion corresponding to the first-category training image portion. The first-category training image portion is a photo showing a human face, and the first-category training text portion describes in natural language whether the human face in the first-category training image portion is a face of a living human, a flat face mask, a three-dimensional face mask, a face generated by deep fake technology, or a result of reshooting a facial image. In other words, the first-category training text portion is used to annotate the actual attributes of the first-category training image portion.

On the other hand, for each second-category visual language training data, the second-category visual language training data includes a second-category training image portion that has undergone image preprocessing, and a second-category training text portion corresponding to the second-category training image portion. The second-category training image portion is a photo showing a face, and the second-category training text portion describes the image quality characteristics of the second-category training image in natural language, as well as the applicable recommended adjustment shooting method, such as "the image is clear, no adjustment is required", "the face is not bright enough, it is recommended to move to a place with sufficient light and reshoot", or "the face is blurred, it is recommended to keep the lens stable when shooting".

By using the group of first-category visual language training data for machine learning, the face image verification model M is able to learn the relationship between the first-category training image parts and the corresponding first-category training text parts, thereby establishing a mapping relationship between image features and natural language context in vector space, so as to have the ability to perform liveness detection on faces in images and describe the detection results in natural language. On the other hand, by using the group of second-category visual language training data for machine learning, the face image verification model M is able to learn the relationship between the second-category training image parts and the corresponding second-category training text parts, thereby establishing a mapping relationship between image features and natural language meanings in vector space, and thus having the ability to evaluate image quality and generate recommendations for adjusting the shooting method.

It should be understood that there are many existing solutions for the technologies such as liveness detection, natural language processing, and visual language models described in this embodiment. In addition, there are many references on the training methods of visual language models, such as " Alec et al., " Learning Transferable Visual Models From Natural Language Supervision " (2021) " or " Andrej Karpathy, Li Fei-Fei, Department of Computer Science, Stanford University, "Deep Visual-Semantic Alignments for Generating Image Descriptions" (2014) " and other papers. Therefore, this embodiment will not go into too much detail about the training method of the face image verification model M.

The facial feature matching rule R1 is used for the processing unit 11 to perform facial image recognition on the image, and can be used for the processing unit 11 to recognize a face from an image, and to compare whether the facial features of two faces in the same image or different images match each other, thereby determining whether the two faces belong to the same person. Furthermore, the facial feature matching rule R1 can be implemented, for example, by a recognition algorithm based on facial feature points (feature-based recognition algorithms), or by a recognition algorithm using a neural network (recognition algorithms using neural network). Since the facial feature matching rule R1 can be achieved using many different existing technologies, and its technical details are not the technical focus of this patent, it will not be described in detail here.

The document verification rule R2 indicates a document eligibility condition corresponding to one or more specific documents (such as but not limited to identity cards and/or health insurance cards), and can be used for the processing unit 11 to perform document image recognition on the image to identify the document from the image and verify whether it meets the format specifications of the specific document. More specifically, in one embodiment, the document verification rule R2 is a software module that can be loaded and executed by the processing unit 11, which includes recognition conditions and verification conditions set according to the appearance characteristics of the specific document (such as the layout format of the document), and is implemented in a rule-based manner. Alternatively, in another embodiment, the document verification rule R2 may also be included in a document image recognition model that can be run by the processing unit 11, and the document image recognition model may be, for example, pre-trained by machine learning with images of specific documents. Since the document verification rule R2 can be achieved using many different existing technologies, and its technical details are not the technical focus of this patent, it will not be described in detail here.

The registration database DB1 is established for an online service provided by the financial service institution. Specifically, in the application of this embodiment, the online service is, for example, an online registration service for a user to register a new account (such as but not limited to a financial account). Furthermore, the registration database DB1 is used to store the registration information of the successfully registered user.

The warning database DB2 is also established for the online service provided by the financial service institution. However, unlike the registration database DB1, the warning database DB2 is used to store the record of the user's unsuccessful registration due to verification failure. More specifically, in this embodiment, the warning database DB2 is used to store one or more warning verification records, and each warning verification record is a record of a user who has used the online registration service but failed to register successfully due to verification failure.

With reference to FIG. 2 , the following uses the client electronic device 10 and the server electronic device 20 as examples to exemplarily explain how the service system 1 of this embodiment implements a facial image verification method.

First, in step S1, when the processing unit 11 receives a service request from the client electronic device 10, the processing unit 11 generates an image request corresponding to the service request, and transmits the image request to the client electronic device 10 in response to the service request.

Specifically, in this embodiment, the service request is generated by the client electronic device 10 according to a manual operation of a user, and is a registration service request corresponding to the online registration service. On the other hand, the image request, for example, includes an image verification prompt message, and the image verification prompt message, for example, is used to prompt the user to use the client electronic device 10 to take a picture of his own face (i.e., a selfie), and to take a picture of a specific certificate held by the user (e.g., the user's ID card), and the shooting result is returned to the processing unit 11.

After the processing unit 11 transmits the image request to the client electronic device 10, the process proceeds to step S2.

In step S2, when the processing unit 11 receives a shooting result from the client electronic device 10 and corresponding to the image request, the processing unit 11 uses the shooting result as image data to be verified corresponding to the service request, and obtains a first facial image portion showing a human face and a document image portion showing an ID from the image data to be verified.

More specifically, in this embodiment, the image data to be verified includes, for example, a first image file having the first facial image portion, and a second image file having the ID image portion, so the processing unit 11 extracts the first facial image portion from the first image file, and extracts the ID image portion from the second image file. It should be noted that, since both the first facial image portion and the ID image portion are portions extracted from an image, the first facial image portion and the ID image portion are also each an image file.

It is further explained that, in normal application, the first image file and the second image file are both photos taken by the client electronic device 10, and the first image file is a self-photographed photo of the user using the client electronic device 10, and the second image file is a photo of the specific ID card taken by the user using the client electronic device 10. Furthermore, the ID card image portion has a second face image portion showing another face, wherein the face in the second face image portion is a face image formed on the specific ID card, that is, a headshot on the specific ID card.

However, in "abnormal" application situations, the first image file may also be a photo of a face mask, a photo of a facial image, or a virtual face image generated using deep fake technology. On the other hand, the ID card presented by the second image file may not be the user's own ID card, but the ID card of a person whose ID card is being used.

After the processing unit 11 obtains the first facial image portion and the ID image portion (including the second facial image portion in the ID image portion), as shown in FIG. 2 , the process proceeds to step S3 and step S6 together.

In step S3, the processing unit 11 inputs the first facial image portion into the facial image verification model M for it to recognize the first facial image portion, and thereby obtains a recognition result generated and outputted according to the first facial image portion by the facial image verification model M. Then, the process proceeds to step S4.

In step S4, the processing unit 11 determines whether the recognition result output by the face image verification model M is a positive recognition result indicating successful verification or a negative recognition result indicating failed verification and including a text message data.

Specifically, in one embodiment, the recognition result has, for example, an indicator character data (which may be, for example, text, code, or symbol), and the processing unit 11 determines whether the recognition result is a positive recognition result or a negative judgment result based on the indicator character data. Alternatively, in another embodiment, the processing unit 11 may also determine whether the recognition result is a positive recognition result or a negative judgment result based on whether the recognition result includes the text message data.

It should be noted that if the recognition result is a positive recognition result, it means that the facial image verification model M determines that the face in the first facial image part is indeed the face of a living human being. If the recognition result is a negative recognition result, it means that the facial image verification model M determines that the facial part in the first facial image is "not" the face of a living human being, or it means that the confidence score of the facial image verification model M in the recognition is insufficient.

If the determination result of the processing unit 11 in step S4 is a negative identification result, the process proceeds to step S5. If the determination result of the processing unit 11 is a positive identification result, the process proceeds to step S8.

In step S5 following step S4, once the processing unit 11 determines that the recognition result is the negative recognition result, the processing unit 11 determines that the verification of the image data to be verified has failed. In this case, the processing unit 11 rejects the service request and outputs the text message data message included in the negative judgment result.

More specifically, in an exemplary case of the present embodiment, the text message data of the negative recognition result includes a prompt explanation text message and a warning explanation text message. The prompt explanation text message indicates a recommended adjustment shooting method determined by the face image verification model M in natural language, and the warning explanation text message indicates a specific image forgery type determined by the face image verification model M in natural language. For example, the prompt explanation text message may be, for example, "Please reshoot in a well-lit place" or "Please reshoot and reduce lens shaking", and the warning explanation text message may be, for example, "face mask" or "deep fake", but is not limited to this.

Furthermore, in this exemplary case, the processing unit 11 outputs the text message data by executing a first output process and a second output process. The first output process is to transmit the prompt instruction text message to the client electronic device 10 so that the prompt instruction text message can be displayed by the client electronic device 10, thereby prompting the user to appropriately adjust the shooting method to reshoot and re-provide new image data to be verified. On the other hand, the second output process is to generate a warning verification record including the warning instruction text message and a warning feature data, and provide the warning verification record to the storage unit 12 so that the warning verification record is added to the warning database DB2. The warning feature data is generated by the processing unit 11 according to the image data to be verified. More specifically, the warning feature data at least includes the first face image portion, and optionally also includes the ID image portion. In addition, optionally, the second output processing may further include transmitting the warning verification record to the service end electronic device 20 for output, thereby notifying relevant business personnel to perform appropriate processing (e.g., notifying relevant investigation units).

In other exemplary cases of this embodiment, according to the result determined by the face image verification model M, the text message data may also include only one of the prompt description text message and the warning description text message. If the text message data only includes the prompt description text message, the processing unit 11 outputs the text message data by only performing the first output process. On the other hand, if the text message data only includes the warning description text message, the processing unit 11 outputs the text message data by only performing the second output process.

It is worth noting that, since the face image verification model M is implemented as a visual language model, it can output a prompt explanation text message and/or a warning explanation text message in the form of natural language when the recognition result is non-live, so that the user operating the client electronic device 10 and/or the business personnel of the financial service institution can understand the reason why the face image verification model M determines that it is non-live. In this way, compared with the method of directly outputting confidence scores or only outputting two opposite results (i.e., positive results or negative results), this embodiment can provide richer feedback on the live detection results, which is beneficial for the service institution to further analyze user behavior or further optimize the overall verification mechanism of the service system 1.

After the processing unit 11 determines that the verification has failed based on the negative recognition result and outputs the text message data, the process of this embodiment ends.

In step S6 following step S2, the processing unit 11 determines, based on the certificate verification rule R2, whether the certificate image portion extracted from the image data to be verified presents a certificate that meets the certificate eligibility conditions, that is, whether the certificate presented in the certificate image portion meets the certificate eligibility conditions.

In one embodiment, the certificate eligibility condition, for example, means that the certificate presented by the image portion of the certificate belongs to the specific certificate. Alternatively, in another embodiment, the certificate eligibility condition may also mean that the certificate presented by the image portion of the certificate has a face image and/or specific identification information (such as ID number, date of birth, etc.). The actual form of the certificate eligibility condition can be set and adjusted according to actual needs, so this embodiment does not impose any special restrictions on the certificate eligibility condition.

If the determination result of the processing unit 11 in step S6 is no, the process proceeds to step S7. If the determination result of the processing unit 11 is yes, the process proceeds to step S8.

In step S7 following step S6, once the processing unit 11 determines that the image data to be verified does not present a certificate that meets the certificate qualification conditions, the processing unit 11 determines that the verification of the image data to be verified has failed. In this case, the processing unit 11 rejects the service request, and generates and transmits a certificate verification failure notification to the client electronic device 10, so that the client electronic device 10 displays the certificate verification failure notification for the user to browse. Optionally, the processing unit 11 also generates a warning verification record based on the judgment result that the certificate qualification conditions are not met, and adds the warning verification record to the warning database DB2. At this point, the process of this embodiment ends.

In step S8 following step S4 and step S6, once the processing unit 11 determines in step S4 that the recognition result is the positive recognition result, and also determines in step S6 that the ID image portion in the image data to be verified presents an ID that meets the ID qualification condition, the processing unit 11 determines, based on the facial feature matching rule R1, whether the facial features of the face in the first facial image portion match the facial features of the face in the second facial image portion in the ID image portion. It should be noted that the "match" mentioned here means that the face in the first facial image portion and the face in the second facial image portion are substantially the same person's face.

If the determination result of the processing unit 11 in step S8 is no, the process proceeds to step S9. If the determination result of the processing unit 11 is yes, the process proceeds to step S10.

In step S9 following step S8, once the processing unit 11 determines that the first face image portion does not match the second face image portion (i.e., determines that the face in the first face image portion and the face in the second face image portion are not the same person), the processing unit 11 determines that the verification of the image data to be verified has failed. In this case, the processing unit 11 rejects the service request, and generates and transmits a facial matching failure notification to the client electronic device 10, so that the client electronic device 10 displays the facial matching failure notification for the user to browse. Optionally, the processing unit 11 also generates a warning verification record according to the judgment result that the first face image part does not match the second face image part, and adds the warning verification record to the warning database DB2. At this point, the process of this embodiment ends.

In step S10 following step S8, once the processing unit 11 determines that the first face image portion matches the second face image portion (i.e., determines that the face in the first face image portion and the face in the second face image portion are indeed the same person), the processing unit 11 uses one of the first face image portion and the second face image portion as image data to be checked, and determines whether there is any warning verification record matching the image data to be checked in the warning database DB2. In this embodiment, the processing unit 11 uses the first face image portion as the image data to be checked, but is not limited thereto.

If the determination result of the processing unit 11 in step S10 is yes, the process proceeds to step S11. If the determination result of the processing unit 11 is no, the process proceeds to step S12.

In step S11 following step S10, once the processing unit 11 determines that there is any warning verification record matching the image data to be checked in the warning database DB2, it means that the user may have tried to use the online registration service but failed, or the user's identity may have been impersonated. In this case, the processing unit 11 does not approve the service request, and generates and transmits a manual review notification corresponding to the service request to the service-end electronic device 20 to remind the business personnel that there is a service request that needs to be manually reviewed for subsequent processing.

In step S12 following step S10, once the processing unit 11 determines that there is no warning verification record matching the image data to be checked in the warning database DB2, the processing unit 11 determines the verification process for the image data to be verified and approves the service request, and executes a service program corresponding to the service request. In this embodiment, the service program is a registration service program for implementing the online registration service, and the processing unit 11 executes the service program in a manner that includes, for example, adding the registration data provided by the user (such as the data presented in the image portion of the ID card) to the registration database DB1. At this point, the process of this embodiment ends.

In summary, in the present embodiment, the processing unit 11 will approve the service request and execute the service procedure only when it determines that the recognition result output by the facial image verification model M is the affirmative recognition result, that the ID card image portion of the image data to be verified presents an ID card that meets the ID card eligibility conditions, and that the first facial image portion matches the second facial image portion.

The above is an example of how the service system 1 of this embodiment implements the facial image verification method.

It is to be noted that in another embodiment, the image data to be verified may also be a single image file, and in normal application, the image file is a selfie of the user holding the specific ID. In other words, in the other embodiment, the first face image portion and the ID image portion may also be from the same photo, and this is not limited to the present embodiment.

It should be noted that steps S1 to S12 of the present embodiment and the flowchart of FIG2 are only used to illustrate one of the practicable modes of the present novel facial image verification method. It should be understood that even if steps S1 to S12 are merged, split or adjusted in sequence, if the process after merging, splitting or adjusting in sequence achieves substantially the same effect as the present embodiment in substantially the same manner, it still belongs to the practicable mode of the present novel facial image verification method. Therefore, steps S1 to S12 of the present embodiment and the flowchart of FIG2 are not used to limit the practicable scope of the present novel.

In summary, by implementing the facial image verification method, the processing unit 11 of the service system 1 will require the user to provide a selfie and a photo of the ID for verification, and after the processing unit 11 obtains the image data to be verified including the first facial image portion and the second facial image portion, the processing unit 11 can use the facial image verification model M to determine whether the face in the first facial image portion is a living body, use the ID verification rule R2 to determine whether the ID presented in the image data to be verified meets the ID qualification conditions, and determine whether the facial features in the first facial image portion match the facial features in the second facial image portion, and only when the above judgment results are all affirmative, will the service request be approved and the service program be executed. In particular, when the facial image verification model M determines that the facial part in the first facial image is "not" alive or the recognition confidence score is insufficient, the facial image verification model M can output an explanatory text message in natural language for the user or business personnel to refer to, so as to clarify the reason why the liveness verification failed. In this way, this embodiment can not only confirm that the selfie provided by the user is not a fake non-live photo, but also confirm that the user is indeed requesting services with his own information, and can provide rich semantic feedback when a non-live photo is detected. In this way, this embodiment provides a more secure and reliable verification mechanism compared to the existing technology, and can effectively prevent people with bad intentions from arbitrarily using other people's information to request services, so it can indeed achieve the purpose of this novel technology.

However, the above is only an example of the implementation of the present invention, and it cannot be used to limit the scope of the implementation of the present invention. All simple equivalent changes and modifications made according to the scope of the patent application of the present invention and the content of the patent specification are still within the scope of the present patent.

1: Service system 11: Processing unit 12: Storage unit M: Face image verification model R1: Facial feature matching rules R2: Document verification rules DB1: Registration database DB2: Alert database 10: Client electronic device 20: Server electronic device S1~S12: Steps

Other features and functions of the present invention will be clearly presented in the implementation method of the reference figures, wherein: FIG. 1 is a block diagram, which exemplarily represents an implementation example of the present service system, and a client electronic device and a server electronic device suitable for use with the implementation example; and FIG. 2 is a flow chart, which exemplarily illustrates how the implementation example implements a facial image verification method.

1: Service system

11: Processing unit

12: Storage unit

M: Face image verification model

R1: Facial feature matching rules

R2: Certificate verification rules

DB1: Register database

DB2: Alert database

10: Client electronic devices

20: Server-side electronic devices

Claims (5)

A service system, comprising: a processing unit; and a storage unit, electrically connected to the processing unit, and storing a face image verification model pre-trained using machine learning technology, wherein the face image verification model is implemented as a visual language model and is configured to identify whether the face presented by the input image belongs to a living body, and when the recognition result is negative, generate and output an explanatory text message presented in natural language and related to the recognition result; wherein the processing unit is used to: After obtaining an image data to be verified corresponding to a service request, a first facial image portion included in the image data to be verified is input into the facial image verification model to obtain a recognition result generated and output by the facial image verification model based on the first facial image portion; Determine whether the recognition result is a positive recognition result indicating successful verification or a negative judgment result indicating failed verification and including a text message data; If the recognition result is determined to be the positive recognition result, the service request is granted, and a service procedure corresponding to the service request is executed; and If the identification result is determined to be a negative judgment result, the service request is rejected, and the text message data message included in the negative judgment result is output. In the service system as described in claim 1, the processing unit is adapted to be electrically connected to a client electronic device; wherein the processing unit receives the image data to be verified from the client electronic device, and, when the text message data includes a prompt instruction text message indicating a recommended adjustment of the shooting method, the processing unit outputs the text message data in a manner that includes transmitting the prompt instruction text message to the client electronic device so that the prompt instruction text message can be displayed by the client electronic device. A service system as described in claim 1, wherein: The storage unit further stores a warning database; and When the text message data includes a warning description text message indicating a specific type of image counterfeiting, the processing unit outputs the text message data in a manner that includes generating a warning verification record including the warning description text message and a warning feature data, and providing the warning verification record to the storage unit so that the warning verification record is added to the warning database, wherein the warning feature data is generated by the processing unit based on the image data to be verified. As described in claim 3, the processing unit is suitable for being electrically connected to a service-end electronic device; wherein: The storage unit also stores a facial feature matching rule suitable for being applied to facial image recognition; and The processing unit grants the service request and executes the service program in the case where the recognition result is determined to be the positive recognition result, including: In the case where the recognition result is determined to be the positive recognition result, judging whether the first facial image portion matches a second facial image portion included in the image data to be verified according to the facial feature matching rule, wherein the second facial image portion presents a facial image on a certificate; When it is determined that the first facial image portion matches the second facial image portion, one of the first facial image portion and the second facial image portion is used as an image data to be checked, and it is determined whether there is any warning verification record matching the image data to be checked in the warning database; When it is determined that there is no warning verification record matching the image data to be checked in the warning database, the service request is approved and the service procedure is executed; and When it is determined that there is a warning verification record matching the image data to be checked in the warning database, the service request is not approved, and a manual review notification corresponding to the service request is generated and transmitted to the service end electronic device. A service system as described in claim 1, wherein: The storage unit also stores a certificate verification rule suitable for being applied to certificate image recognition and indicating a certificate eligibility condition; and The processing unit is also used to determine whether the image data to be verified presents a certificate that meets the certificate eligibility condition according to the certificate verification rule after obtaining the image data to be verified, and the processing unit only approves the service request and executes the service procedure when it is determined that the image data to be verified presents a certificate that meets the certificate eligibility condition and also determines that the recognition result is the positive recognition result.

Images