Face recognition failure handling method and device based on agent

Abstract

The application relates to the technical field of Internet, and discloses a face recognition fault processing method and device based on an intelligent agent, the method comprising the following steps: in response to a face recognition authentication request initiated by a driver terminal, determining available authentication channels based on a preset authentication strategy; controlling the driver terminal to perform corresponding face recognition authentication through the available authentication channels; if the authentication fails, triggering the intelligent agent to analyze system operation state indexes and authentication failure information to determine a fault type, and triggering a corresponding strategy of target channel degradation or temporary exemption of face recognition based on the fault type to complete fault processing; the system operation state indexes comprise at least one of a query rate per second, a response time, an authentication failure rate and failure rate change data. Through the technical scheme, the problems of low reliability and weak fault tolerance of the face recognition authentication for the online car-hailing driver in the related art are solved, and the reliability and fault tolerance of the face recognition authentication for the online car-hailing driver are improved.

Description

Technical Field

[0001] This application relates to the field of Internet technology, specifically to a method and apparatus for handling facial recognition faults based on intelligent agents. Background Technology

[0002] With the rapid development of the ride-hailing industry, the security and reliability of driver identity authentication have become crucial to ensuring passenger safety. Facial recognition, as a core authentication method, is widely used in the pre-departure identity verification process for ride-hailing drivers. Related technologies include optimizations for driver facial recognition, such as triggering facial recognition under multiple conditions to ensure identity consistency, or improving the authentication pass rate through priority ranking of multiple authentication nodes.

[0003] However, significant shortcomings remain in practical applications: On the one hand, related technologies primarily focus on "identity consistency verification" or "improving authentication pass rates in normal scenarios," lacking adaptability to facial recognition system failures, authentication channel anomalies, or high-concurrency scenarios. For example, when a single authentication channel service becomes unavailable, the failure rate of all channels surges, or regional network failures cause continuous authentication failures, the system can only provide a binary result of "allowed to drive" or "prohibited from driving," failing to ensure core business continuity through flexible degradation mechanisms. This results in a large number of compliant drivers being unable to drive immediately due to system anomalies, severely impacting operational efficiency and driver experience. On the other hand, related technologies rely on fixed rules to trigger authentication or node switching, lacking intelligent decision-making capabilities. They cannot analyze the system's operating status in real time to accurately pinpoint the type of failure and do not consider issues such as compatibility between failed channels and special channels, further reducing authentication efficiency and system fault tolerance.

[0004] In other words, the facial recognition authentication for ride-hailing drivers in the relevant technologies suffers from low reliability and weak fault tolerance. Summary of the Invention

[0005] This application provides a face recognition fault handling method and device based on intelligent agents to solve the problems of low reliability and weak fault tolerance in face recognition authentication for ride-hailing drivers in related technologies.

[0006] In a first aspect, this application provides a face recognition fault handling method based on intelligent agents, the method comprising: In response to a facial recognition authentication request initiated by the driver, the available authentication channels are determined based on a preset authentication strategy; The driver's terminal controls the corresponding facial recognition authentication to be performed through the available authentication channel; If authentication fails, the intelligent agent is triggered to analyze the system operation status indicators and authentication failure information to determine the fault type. Based on the fault type, the agent triggers the corresponding strategy of target channel downgrading or face recognition temporary exemption to complete the fault handling. The system operation status indicators include at least one of the following: query rate per second, response time, authentication failure rate, and failure rate change.

[0007] In one optional implementation, the available authentication channels are determined based on a preset authentication policy, including: The initial authentication method and candidate authentication channels are determined based on the preset authentication strategy. The preset authentication strategy includes scene adaptation rules, time period matching rules and special population adaptation rules. The initial authentication method includes live face authentication and static face authentication. The candidate authentication channels are at least two platform channels that provide face comparison services. Obtain the driver's authentication channel failure records within a preset time period, remove the failed channels corresponding to the failure records from the candidate authentication channels, and obtain the available authentication channels.

[0008] In one optional implementation, obtaining the driver's authentication channel failure records within a preset time period includes: All driver failure records are stored and retrieved through a distributed cache. All failure records include the channel identifier, failure time, and failure return code. Retain failure records whose failure time is close to the preset duration from all failure records, and use them as authentication channel failure records.

[0009] In one optional implementation, determining the initial authentication method and candidate authentication channels based on a preset authentication strategy includes: Based on the driver's identity information in the facial recognition authentication request, determine whether the driver is initiating authentication for the first time; If the driver is initiating authentication for the first time, the initial authentication method will be determined as live face authentication; otherwise, the initial authentication method will be determined as either live face authentication or static face authentication. Based on the current business scenario and time period, candidate authentication channels are selected from all preset authentication channels according to the time period allocation rules; If the driver's identity information indicates that the driver belongs to a special group, a dedicated authentication channel will be determined from the candidate authentication channels according to the special group adaptation rules, and the dedicated authentication channel will be used as a candidate authentication channel.

[0010] In one optional implementation, the agent is triggered to analyze system operating status indicators and authentication failure information to determine the type of failure, including: If the failure rate of a single authentication channel exceeds the first threshold and the change in failure rate exceeds the second threshold, and / or the failure return code in the authentication failure information is a service unavailable identifier, the failure type is determined to be a channel-level failure. If the failure rate of the full authentication channel exceeds the third threshold, or if the failure rate of driver face recognition authentication in a specific area is abnormally higher than the preset benchmark value, the fault type is determined to be a global fault.

[0011] In one optional implementation, a corresponding strategy for triggering target channel degradation or temporary exemption for face recognition is implemented based on the fault type to complete the fault handling, including: If the fault type is a channel-level fault, disable the authentication channel corresponding to the channel-level fault, select available channels sequentially from the list of authentication channels supported by the driver according to the preset polling strategy, and control the driver to re-execute face recognition authentication through the currently selected channel until authentication is successful or all available channels fail to attempt. If the fault type is a global fault, determine whether the driver meets the preset exemption conditions. If the driver meets the preset exemption conditions, temporarily exempt the driver from facial recognition authentication and allow the driver to drive. Otherwise, return an authentication failure message and record the reason for the failure. The preset exemption conditions include at least one of the following: the driver is not logging in for the first time, there are no abnormalities in the historical authentication records within the preset period, there are no security violation records within the preset period, and the account-bound device has not been changed.

[0012] Secondly, this application provides a face recognition fault handling device based on intelligent agents, the device comprising: The channel determination module is used to respond to the facial recognition authentication request initiated by the driver and determine the available authentication channels based on the preset authentication strategy. The face authentication module is used to control the driver's terminal to perform the corresponding face recognition authentication through the available authentication channel; The fault handling module is used to trigger the intelligent agent to analyze the system operation status indicators and authentication failure information if authentication fails, in order to determine the fault type, and trigger the corresponding strategy of target channel downgrading or face recognition temporary exemption based on the fault type to complete the fault handling. The system operation status indicators include at least one of the following: query rate per second, response time, authentication failure rate, and failure rate change data.

[0013] Thirdly, this application provides an electronic device, including: a memory and a processor, which are communicatively connected to each other. The memory stores computer instructions, and the processor executes the computer instructions to perform the agent-based face recognition fault handling method described in the first aspect or any corresponding embodiment.

[0014] Fourthly, this application provides a computer-readable storage medium storing computer instructions for causing a computer to execute the agent-based face recognition fault handling method described in the first aspect or any corresponding embodiment.

[0015] Fifthly, this application provides a computer program product, including computer instructions for causing a computer to execute the agent-based face recognition fault handling method described in the first aspect or any corresponding embodiment.

[0016] The face recognition fault handling method based on intelligent agents proposed in this application achieves the following beneficial technical effects compared with the prior art: In response to a facial recognition authentication request initiated by the driver, the system determines available authentication channels based on a preset authentication strategy. A failed channel filtering mechanism removes recently failed channels, and combined with time-based allocation and special population adaptation rules, precise selection of authentication channels is achieved, effectively avoiding invalid attempts and improving authentication efficiency. The system controls the driver to perform facial recognition authentication through available channels. When all candidate channels are unavailable, a polling switching mechanism is triggered to repeatedly try channels built into the client, maximizing the utilization of available resources and ensuring timely processing of authentication requests. If authentication fails, an intelligent agent is triggered to analyze system operating status indicators and authentication failure information to determine the fault type. Based on the fault type, a corresponding strategy of target channel downgrading or temporary facial recognition exemption is triggered to complete the fault handling. System operating status indicators include at least one of the following: query rate per second, response time, authentication failure rate, and change in failure rate. This is achieved through an artificial intelligence agent (AI). The agent monitors multi-dimensional indicators in real time and makes intelligent decisions, achieving second-level fault response and precise degradation. In the event of a global failure, it temporarily exempts compliant drivers to ensure their continued operation, and automatically switches to an available channel when a channel fails. This balances system security and business continuity, while significantly reducing operational costs by replacing manual intervention. It also improves the reliability and fault tolerance of facial recognition authentication for ride-hailing drivers. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the specific embodiments of this application or the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0018] Figure 1 This is a schematic diagram illustrating an application scenario according to an embodiment of this application; Figure 2 This is a flowchart of a face recognition fault handling method based on an intelligent agent according to an embodiment of this application; Figure 3This is a schematic diagram of a normal authentication process and an AI Agent downgrade process according to an embodiment of this application; Figure 4 This is a structural block diagram of a face recognition fault handling device based on an intelligent agent according to an embodiment of this application; Figure 5 This is a schematic diagram of the hardware structure of an electronic device according to an embodiment of this application. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0020] It is understood that before using the technical solutions disclosed in the various embodiments of this application, users should be informed of the types, scope of use, and usage scenarios of the personal information involved in this application in an appropriate manner in accordance with relevant laws and regulations, and user authorization should be obtained.

[0021] The terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this application, "multiple" means two or more, unless otherwise explicitly specified.

[0022] As one optional application scenario in the embodiments of this application, such as Figure 1 As shown, the agent-based facial recognition fault handling system may include at least one terminal device and at least one server. Figure 1 The system is illustrated in the example, which includes a computer 101, a mobile terminal 102, and a server 103, and the terminal devices such as the computer 101 and the mobile terminal 102 are connected to the server 103 through a network 110.

[0023] Specifically, the terminal device can be a smartphone, tablet, laptop, PDA, desktop computer, game console, smart TV, smart wearable device, in-vehicle terminal, VR (Virtual Reality) device, AR (Augmented Reality) device, etc. Server 103 can be a standalone physical server, a server cluster, a distributed system, or a cloud server providing cloud services. Network 110 can be a wired or wireless network, examples of which include, but are not limited to, the Internet, corporate intranet, local area network, wide area network, mobile communication network, and combinations thereof.

[0024] It should be noted that, Figure 1 This is merely an example of an application scenario and does not limit the scope of protection of this application.

[0025] The embodiments of this application will be described below with reference to the accompanying drawings. It should be understood that the pages shown in the drawings are merely examples, and various page designs are possible in practice. The various graphic elements on the page may have different arrangements and different visual representations, one or more elements may be omitted or replaced, and one or more other elements may also be present; no limitations are imposed on the embodiments of this application. Furthermore, the embodiments are primarily described below with reference to terminal device 110. It should be understood that the actions described relative to terminal device 110 can be performed by application 101 on terminal device 110, or can be performed by application 101 in conjunction with its server (e.g., server 120).

[0026] This application provides a face recognition fault handling method based on an intelligent agent. By using AI Agent intelligent decision-making and dynamic degradation mechanism, it solves the problem of drivers being unable to drive due to system failure, channel abnormality or continuous authentication failure, thereby improving the reliability and fault tolerance of face recognition authentication.

[0027] According to an embodiment of this application, an embodiment of a face recognition fault handling method based on intelligent agents is provided. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0028] This embodiment provides a face recognition fault handling method based on intelligent agents, which can be used in the aforementioned server. Figure 2 This is a flowchart of a face recognition fault handling method based on an intelligent agent according to an embodiment of this application, such as... Figure 2 As shown, the process includes the following steps: Step S201: In response to the facial recognition authentication request initiated by the driver, determine the available authentication channels based on the preset authentication strategy.

[0029] Specifically, the driver's end refers to the terminal device used by the driver, such as a smartphone or tablet, on which the ride-hailing driver application runs. A facial recognition authentication request is a data request triggered when a driver initiates a trip through the driver's application to verify their identity. This request carries the driver's identification and the information required for authentication. Pre-configured authentication strategies refer to a set of pre-configured rules for selecting authentication methods and channels, including first-time / non-first-time authentication rules, time-based allocation rules (e.g., prioritizing Tencent Cloud during morning rush hour), and special population adaptation rules (e.g., specifying Tencent Cloud for drivers from Hong Kong, Macau, and Taiwan). Available authentication channels refer to the authentication service channels that, after initial screening and failure record filtering, can ultimately be used for this facial recognition, such as Alibaba Cloud Real Person Authentication, Tencent Cloud Smart Eye, and Megvii Cloud platform channels.

[0030] Based on the authentication request initiated by the driver, and combined with preset authentication strategies (such as requiring a live ID for the first authentication, allowing live or static IDs for subsequent authentications; filtering candidate channels based on the current time period and matching rules; and matching dedicated channels for specific groups), a preliminary list of candidate authentication channels is determined. Then, by reading the driver's recent failed channel records from a distributed cache (such as Redis), recently failed channels are removed from the candidate list, thus obtaining the available authentication channels. This approach achieves accurate pre-selection of authentication channels, avoids repeated attempts at failed channels, improves authentication efficiency, and simultaneously considers business scenario adaptability and compatibility with specific groups.

[0031] In one possible implementation, when the driver's application (APP) initiates a dispatch request, the server first determines whether it is the first time authenticating based on the driver's identity information: if it is the first time, then mandatory liveness authentication is enforced; if it is not the first time, either liveness or static authentication is allowed. The server then retrieves the channel allocation strategy from the configuration center based on the current time period (e.g., morning rush hour) (e.g., prioritizing Tencent Cloud during morning rush hour, and Alibaba Cloud and Megvii Cloud in a 3:7 ratio during off-peak hours), filtering out a candidate channel list. Next, the server queries Redis for the driver's failed channel records within the last 5 minutes. If a channel from that record exists among the candidate channels, it is removed, and the remaining channels are considered available authentication channels.

[0032] Step S202: Control the driver to perform the corresponding face recognition authentication through the available authentication channel.

[0033] Specifically, facial recognition authentication refers to the process of verifying identity using facial biometrics. Depending on the scenario, it can be divided into live face authentication (requiring drivers to blink, shake their heads, etc. to prevent photo / video forgery) and static face authentication (comparing only a single frame of facial image). The specific implementation depends on the software development kit (SDK) or application programming interface (API) capabilities provided by the selected authentication channel.

[0034] After identifying an available authentication channel, the system sends an authentication command to the driver's terminal. The driver's terminal, based on the specified authentication method (live or static) and channel (such as Alibaba Cloud, Tencent Cloud, etc.), calls the corresponding channel's SDK or interface to collect the driver's facial image or video stream, performs liveness detection and facial feature extraction, and sends the feature data or image to the corresponding authentication service platform for a 1:1 comparison. Finally, the driver receives the authentication result (success or failure). This step completes the actual execution of facial recognition authentication, providing foundational data for subsequent failure handling or successful authentication.

[0035] In one possible implementation, if the available authentication channel is Tencent Cloud and the authentication method is liveness detection, the driver's app will load the Tencent Cloud real-person authentication SDK, guiding the driver to complete specified actions such as blinking and opening their mouth. Real-time video streaming will be captured for liveness detection. Once the detection is successful, facial features will be extracted and uploaded to the Tencent Cloud server for comparison. The server will return the comparison result to the driver's app, simultaneously recording the authentication channel identifier, return code, and authentication result. If authentication is successful, the process ends, and the driver is allowed to drive; if authentication fails, the subsequent AI Agent downgrade analysis process will be triggered.

[0036] Step S203: If authentication fails, the intelligent agent is triggered to analyze the system operation status indicators and authentication failure information to determine the fault type, and trigger the corresponding strategy of target channel downgrading or face recognition temporary exemption based on the fault type to complete the fault handling. The system operation status indicators include at least one of the following: query rate per second, response time, authentication failure rate, and failure rate change data.

[0037] Specifically, an intelligent agent refers to an automated program module that autonomously analyzes and makes decisions based on real-time monitoring data, capable of sensing system anomalies and dynamically triggering response strategies, such as the AI ​​Agent in this solution. System operation status indicators refer to a multi-dimensional data set used to assess the health of the authentication system, including queries per second (QPS), response time (RT), authentication failure rate, and changes in failure rate. Authentication failure information refers to a detailed record of the authentication failure, including at least the failed channel identifier and failure return code, such as specific error codes like "service unavailable" or "liveness detection failed." Fault type refers to the fault classification derived from indicator analysis, including channel-level faults (single channel anomalies) and global faults (all channels or regions experiencing anomalies). Target channel degradation refers to the response measure of disabling the faulty channel and switching to other available channels for retrying according to a preset strategy when a specific channel fault is detected. Temporary exemption for facial recognition refers to an emergency handling method that temporarily exempts drivers who meet safety conditions from facial recognition authentication during a global fault, allowing them to drive directly. Query per second (SPS) refers to the number of authentication requests processed by the system per unit of time, used to measure system load and traffic pressure. Response time refers to the time elapsed from initiating an authentication request to receiving the authentication result, used to assess the service quality and latency of the channel. Authentication failure rate refers to the proportion of authentication failures to total requests per unit of time, used to measure the health and availability of the channel. Failure rate change data refers to the magnitude of change in the authentication failure rate compared to the same period last year (year-on-year) or the previous consecutive period (month-on-month), used to determine the degree of exacerbation of abnormal trends, such as a 50% year-on-year increase or a 30% month-on-month increase.

[0038] When facial recognition authentication fails, the system triggers an AI agent to analyze the fault. The agent collects online monitoring metrics (QPS, RT, authentication failure rate, year-on-year / month-on-month failure rate data) and information about the current authentication failure (channel identifier, return code) in real time, and determines the fault type based on preset rules. If it is determined to be a channel-level fault (such as a sudden increase in single-channel failure rate or frequent occurrence of specific return codes), a target channel degradation strategy is triggered, disabling the faulty channel and switching to another available channel for re-authentication. If it is determined to be a global fault (such as the overall channel failure rate exceeding a threshold or regional anomalies), a temporary facial recognition exemption strategy is triggered, temporarily exempting drivers who meet the exemption conditions (such as not first-time logins or normal historical behavior) from authentication and allowing them to drive. Through this intelligent decision-making and dual-path degradation, fault response and automated processing are achieved within seconds, maximizing driver efficiency while ensuring safety.

[0039] In one possible implementation, the AI ​​Agent continuously monitors real-time metrics for each authentication channel. Suppose the Tencent Cloud channel failure rate suddenly spikes to 35%, increasing by 60% compared to the previous 5 minutes, and the "503 Service Unavailable" error code appears frequently in the authentication failure messages. The agent then determines this as a channel-level failure and immediately generates a downgrade instruction: disabling the Tencent Cloud channel and switching subsequent authentication requests to Alibaba Cloud or Megvii Cloud channels in a round-robin order for retry. If monitoring shows that the failure rate for all channels exceeds 15%, and the overall failure rate for drivers in a certain city abnormally rises to 25%, the agent determines this as a global failure. In this case, the agent verifies the exemption conditions for the requesting driver: if the driver is not logging in for the first time, has no security violations in the past 7 days, and the device has not been changed, the agent generates a temporary exemption instruction, allowing the driver to start driving directly, and records the exemption log; if the driver does not meet the conditions, the agent returns an authentication failure and records the reason for subsequent manual verification.

[0040] The face recognition fault handling method based on intelligent agents provided in this embodiment responds to face recognition authentication requests initiated by the driver's end. Based on a preset authentication strategy, it determines available authentication channels, eliminates recently failed channels through a failure channel filtering mechanism, and combines time period allocation and special population adaptation rules to achieve accurate screening of authentication channels, effectively avoiding invalid attempts and improving authentication efficiency. It controls the driver's end to perform corresponding face recognition authentication through available authentication channels. When all candidate channels are unavailable, a polling switching mechanism is triggered to try channels built into the client in a loop, maximizing the utilization of available resources and ensuring timely processing of authentication requests. If authentication fails, the intelligent agent is triggered to analyze system operating status indicators and authentication failure information to determine the fault type, and triggers corresponding strategies such as target channel downgrading or temporary face recognition exemption based on the fault type to complete fault handling. System operating status indicators include at least one of the following: query rate per second, response time, authentication failure rate, and change in failure rate. This is achieved through AI... The agent monitors multi-dimensional indicators in real time and makes intelligent decisions, achieving second-level fault response and precise degradation. In the event of a global failure, it temporarily exempts compliant drivers to ensure their continued operation. In the event of a channel failure, it automatically switches to an available channel, balancing system security and business continuity. At the same time, it significantly reduces operational costs by replacing manual intervention. It also improves the reliability and fault tolerance of facial recognition authentication for ride-hailing drivers.

[0041] In some optional implementations, step S201 above, which determines the available authentication channel based on a preset authentication policy, includes the following steps: Step a1: Determine the initial authentication method and candidate authentication channels based on the preset authentication strategy. The preset authentication strategy includes scene adaptation rules, time period matching rules and special population adaptation rules. The initial authentication method includes live face authentication and static face authentication. The candidate authentication channels are at least two platform channels that provide face comparison services.

[0042] Specifically, the initial authentication method refers to the type of facial recognition to be used in this authentication, determined according to the first / non-first authentication rules. This includes liveness detection (requiring action commands) and static facial recognition (single-frame image comparison only). Candidate authentication channels refer to a list of multiple third-party or self-developed facial recognition service platforms initially selected based on rules such as scenario and time period allocation, such as Tencent Cloud Smart Eye, Alibaba Cloud Real Person Authentication, and Megvii Cloud. Scenario adaptation rules refer to the strategy rules for configuring differentiated authentication requirements based on different business scenarios (such as vehicle dispatch authentication and order acceptance authentication). Time period allocation rules refer to the rules for dynamically adjusting the traffic allocation ratio of each authentication channel according to time periods (such as morning peak and off-peak periods). For example, during morning peak, Tencent Cloud accounts for 100%, while during off-peak periods, Alibaba Cloud:Megvii Cloud = 3:7. Special population adaptation rules refer to the rules for configuring dedicated authentication channels for specific groups (such as drivers from Hong Kong, Macau, and Taiwan, and elderly drivers) to avoid compatibility issues. Liveness detection requires drivers to perform specified actions such as blinking, shaking their head, and opening their mouth, using infrared or depth cameras to detect liveness features and prevent photo or video forgery. Static face detection only captures a single frame of a face image for feature comparison and does not include a liveness detection step; it is typically used for low-risk or non-first-time authentication scenarios.

[0043] The system determines whether it's a first-time or subsequent authentication based on the driver's identity information, and then selects the initial authentication method (forced liveness detection for first-time users, optional for subsequent users). Combining the current business scenario and time period, it filters a list of candidate channels from all preset channels according to time period matching rules. If the driver belongs to a special group, a dedicated channel is determined from the candidate channels according to special group adaptation rules, and this dedicated channel is then used as a candidate channel. Through this multi-dimensional strategy combination, the system achieves accurate initial screening of authentication methods and channels, laying the foundation for subsequent failure filtering and authentication execution.

[0044] Step a2: Obtain the driver's authentication channel failure records within a preset time period, remove the failed channels corresponding to the failure records from the candidate authentication channels, and obtain the available authentication channels.

[0045] Specifically, authentication channel failure records refer to the recent authentication failure history data of drivers stored in a distributed cache (such as Redis). Each record includes at least the channel identifier, failure time, and failure return code. A failed channel refers to a channel through which the driver's authentication failed within a preset time period (e.g., the last 5 minutes). For example, if a driver failed 3 times within 5 minutes on a Tencent Cloud channel, then Tencent Cloud is marked as a failed channel. Available authentication channels refer to the list of channels remaining after removing failed channels from the candidate authentication channels. If all candidate channels are removed, the available authentication channels are empty.

[0046] The system reads all authentication failure records for the driver within a preset time period (e.g., the last 5 minutes) from the distributed cache, extracts the channel identifiers to create a list of failed channels, and then compares the candidate authentication channels determined in step a1 with this list. Channels already in the failed channel list are removed, and the remaining channels are the available authentication channels. This failed channel filtering mechanism effectively avoids repeatedly trying recently failed channels, improving authentication efficiency and user experience.

[0047] In one possible implementation, driver Zhang San initiates an authentication request. Step a1 determines the candidate channels as [Tencent Cloud, Alibaba Cloud, Megvii Cloud]. The system queries Redis for Zhang San's failure records in the last 5 minutes and finds two records: {"channel":"Tencent Cloud", "time":"10:32:15", "code":"503"}, {"channel":"Tencent Cloud", "time":"10:34:22", "code":"503"}. Therefore, the list of failed channels is [Tencent Cloud]. After removing Tencent Cloud from the candidate channels, the available authentication channels are [Alibaba Cloud, Megvii Cloud].

[0048] Through this implementation method, the system can proactively avoid recently failed channels before authentication is performed, reduce invalid authentication attempts, reduce waiting time caused by repeated failures, alleviate request pressure on faulty channels, and improve overall authentication efficiency and system stability.

[0049] In some optional implementations, obtaining the driver's authentication channel failure records within a preset time period in step a2 above includes the following steps: Step b1 involves storing and retrieving all driver failure records through a distributed cache. All failure records include the channel identifier, failure time, and failure return code.

[0050] Specifically, a distributed cache refers to a high-speed key-value storage system deployed in memory for real-time reading and writing of authentication failure records, characterized by low latency and high throughput, such as Redis. All failure records refer to the data set written to the cache each time the driver's authentication failed historically. Each record contains at least a channel identifier, a failure timestamp, and a failure return code. The channel identifier is an identifier used to uniquely identify the authentication service provider, such as "tencent_cloud", "aliyun_cloud", or "megvii_cloud". The failure time refers to the specific moment the authentication failure occurred, typically a Unix timestamp or a time string accurate to milliseconds. The failure return code refers to the error code or descriptive information returned by the authentication service channel, used to identify the reason for the failure, such as "503 Service Unavailable", "400 Bad Request", "Liveness detection failed", or "Face comparison failed".

[0051] After each failed face recognition authentication, the system encapsulates the failure information (including channel identifier, failure time, and failure return code) into a record and writes it to a distributed cache (such as Redis). The driver's unique identifier is used as the key, and all failure records are stored as a list or ordered set. When a failure record needs to be retrieved, the system reads all historical failure records for that driver from the cache using the same key, providing a data foundation for subsequent time-based filtering. This mechanism achieves real-time persistence and fast access to failure records, providing a reliable data source for failure channel filtering.

[0052] Step b2: Select failure records from all failure records whose failure time is within a preset time period as authentication channel failure records.

[0053] Specifically, from all the failure records obtained in step b1, the system filters out all records whose failure time is after "current time minus preset duration" based on the current time and a preset duration threshold (e.g., 5 minutes). These records are then used as "authentication channel failure records" available for this authentication. By filtering by time, only recent failures are retained, avoiding the influence of older failure records on the current channel selection, while also reducing the amount of data that needs to be processed and improving filtering efficiency.

[0054] In one possible implementation, the preset duration is 5 minutes. The system reads a complete list of driver Zhang San's failure records from Redis, containing the following records: {"channel":"Tencent Cloud", "time":"2024-05-20 10:32:15", "code":"503"} {"channel":"Tencent Cloud", "time":"2024-05-20 10:34:22", "code":"503"} {"channel":"Alibaba Cloud", "time":"2024-05-20 09:15:03", "code":"400"} {"channel":"Megvii Cloud", "time":"2024-05-19 18:20:47", "code":"Liveness detection failed"} The current time is 2024-05-20 10:36:00. With a preset duration of 5 minutes, records failing after 10:31:00 (i.e., the first two Tencent Cloud records) will be retained as authentication channel failure records. Records from Alibaba Cloud and Megvii Cloud are filtered out because they exceed the 5-minute timeframe and will not be used for this filtering.

[0055] Through this implementation method, the system can accurately obtain information on drivers' recent failed channels, ensuring the timeliness and accuracy of filtering decisions, while reducing the amount of cached data and computational overhead, and improving overall processing efficiency.

[0056] In some optional implementations, step a1 above, which determines the initial authentication method and candidate authentication channels based on a preset authentication strategy, includes the following steps: Step c1: Based on the driver's identity information in the facial recognition authentication request, determine whether the driver is initiating authentication for the first time.

[0057] Specifically, driver identity information refers to data used to uniquely identify a driver, including driver ID, mobile phone number, and identification number, which can be obtained by parsing from the authentication request. "First-time authentication" means that the driver account has never completed facial recognition authentication before, i.e., there are no successful authentication records for this driver in the system.

[0058] The system parses the driver's identity information from the authentication request and queries the database or cache for the driver's historical authentication records. If no successful authentication records are found, it is determined that the authentication is being initiated for the first time; otherwise, it is not the first time. This determination provides a basis for selecting subsequent authentication methods, ensuring that the first authentication must use the highest security level of liveness detection.

[0059] Step c2: If the driver is initiating authentication for the first time, the initial authentication method will be determined as live face authentication; otherwise, the initial authentication method will be determined as either live face authentication or static face authentication.

[0060] Specifically, the initial authentication method refers to the specific authentication type to be used in this facial recognition authentication, including live face authentication (which requires blinking, head shaking, and other actions) and static face authentication (single-frame image comparison only).

[0061] Based on the judgment result of step c1, if the driver is authenticating for the first time, the initial authentication method will be forcibly set to live face authentication to ensure the security of identity verification. If it is not the first time authenticating, the system can flexibly choose live face authentication or static authentication according to business needs, risk level, or other strategies. For example, static authentication will be given priority in low-risk periods or scenarios to improve efficiency. Through differentiated configuration, security and user experience are balanced.

[0062] Step c3: Based on the current business scenario and time period, select candidate authentication channels from all preset authentication channels according to the time period allocation rules.

[0063] Specifically, the current business scenario refers to the specific business context in which the driver initiates authentication, such as vehicle departure authentication, order acceptance authentication, night operation authentication, etc. Different scenarios may correspond to different security requirements.

[0064] The system obtains the current time (e.g., 8:30 AM) and the current business scenario (e.g., vehicle dispatch authentication), searches for pre-configured time-slot allocation rules (e.g., Tencent Cloud accounts for 100% during peak hours 7:00-9:00 AM, and Alibaba Cloud and Megvii Cloud are allocated in a 3:7 ratio during off-peak hours 9:00-5:00 PM), and filters out a list of channels that meet the criteria from all preset authentication channels (e.g., Tencent Cloud, Alibaba Cloud, Megvii Cloud, Baidu Cloud, etc.) as candidate authentication channels. Through dynamic allocation, load balancing and resource optimization of channel traffic are achieved.

[0065] Step c4: If the driver's identity information indicates that the driver belongs to a special group, determine the exclusive authentication channel from the candidate authentication channels according to the special group adaptation rules, and use the exclusive authentication channel as a candidate authentication channel.

[0066] Specifically, "special groups" refers to driver groups that require special authentication processing due to reasons such as document type, region, or device compatibility, such as drivers from Hong Kong, Macao, and Taiwan, foreign drivers, and drivers using specific mobile phone models. "Dedicated authentication channels" refers to authentication service channels pre-designated for special groups and offering the best compatibility. For example, drivers from Hong Kong, Macao, and Taiwan are required to use the Tencent Cloud channel to support the recognition of Mainland Travel Permits for Hong Kong and Macao Residents.

[0067] The system checks the driver's identity information for special population markers (such as "Mainland Travel Permit for Hong Kong and Macao Residents"). If a special population is identified, a dedicated channel is determined from the candidate authentication channels according to the special population adaptation rules (such as "Hong Kong, Macao and Taiwan drivers → Tencent Cloud", "Foreign drivers → Alibaba Cloud"). (If the dedicated channel is not in the candidate list, it may be forcibly added to or replace the entire candidate list.) This dedicated channel is then used as the final candidate authentication channel. This adaptation mechanism effectively solves the compatibility problem for special populations and improves the authentication pass rate.

[0068] In one possible implementation, driver Wang Wu (driver ID: 888888) initiates a vehicle dispatch authentication request. The system checks historical records and finds that Wang Wu successfully authenticated yesterday, indicating this is not his first time. Therefore, the initial authentication method can be either live or static (the system selects static authentication based on the current low-risk period). The current time is 3 PM (off-peak), the business scenario is a regular dispatch, and the time allocation rule is "off-peak Alibaba Cloud: Megvii Cloud = 3:7". The system filters candidate channels as [Alibaba Cloud, Megvii Cloud] (randomly selected proportionally). Further checking the driver's identity information reveals that Wang Wu's document type is a "Mainland Travel Permit for Hong Kong and Macao Residents," belonging to a special group. The special group adaptation rule stipulates that "drivers from Hong Kong, Macao, and Taiwan should prioritize using the Tencent Cloud channel." Therefore, the system forcibly changes the candidate channel to [Tencent Cloud] and uses Tencent Cloud as the candidate channel for this authentication.

[0069] Through this implementation method, the system can accurately determine candidate authentication channels by comprehensively considering the initial authentication determination, time period allocation, and adaptation to special groups. This not only ensures the security of the initial authentication but also enables dynamic scheduling of channel load. At the same time, it solves the compatibility problem of special groups, significantly improving authentication efficiency and success rate.

[0070] In some optional implementations, the triggering agent in step S203 above analyzes the system operating status indicators and authentication failure information to determine the fault type, including: Step d1: If the failure rate of a single authentication channel exceeds the first threshold and the change in failure rate exceeds the second threshold, and / or the failure return code in the authentication failure information is a service unavailable identifier, the fault type is determined to be a channel-level fault.

[0071] Specifically, the first threshold refers to the failure rate threshold used to determine whether a single channel has failed, for example, 30%. Exceeding this value indicates that the channel may be abnormal. The second threshold refers to the threshold used to determine the abnormal increase in the failure rate, for example, 50%, indicating that the failure rate has increased significantly compared to the same period in history or the previous period. The change in failure rate refers to the change in the current failure rate compared to the baseline period (such as the previous hour or the same period yesterday), usually expressed as a percentage. Channel-level failure refers to a failure type in which a single authentication service channel is abnormal, but other channels can still provide normal service. For example, Tencent Cloud channel service is unavailable, but Alibaba Cloud channel is normal.

[0072] The AI ​​Agent monitors the failure rate of each authentication channel in real time. When it detects that the failure rate of a certain channel exceeds a preset first threshold (e.g., 30%), and the change in the failure rate of that channel (year-on-year or month-on-month) exceeds a second threshold (e.g., a 50% increase), it indicates that the channel not only has a high failure rate but is also showing a rapidly deteriorating trend. Alternatively, when the authentication failure information returns a clear "service unavailable" identifier (e.g., 503, service busy, system error, etc.), the AI ​​Agent directly determines that the channel has experienced a channel-level failure. By combining multiple indicators, channel-level anomalies can be accurately identified, avoiding misjudgments caused by occasional fluctuations.

[0073] Step d2: If the failure rate of the full authentication channel exceeds the third threshold, or the failure rate of driver face recognition authentication in a specific area is abnormally higher than the preset benchmark value, the fault type is determined to be a global fault.

[0074] Specifically, the third threshold refers to the critical value of the average failure rate across all channels used to determine a global failure, such as 15%. Exceeding this value indicates that the system as a whole may experience large-scale anomalies. The full authentication channel refers to the collection of all currently available authentication service channels, including all platform channels such as Tencent Cloud, Alibaba Cloud, and Megvii Cloud. The preset baseline value refers to the reference value for the face recognition authentication failure rate under normal conditions in a specific region (such as a city), for example, the average failure rate during off-peak hours in that city is 5%. A global failure refers to a type of failure that involves all or most authentication channels, or a large number of users in a specific region who are unable to complete authentication, such as core data center network outages or regional DNS hijacking.

[0075] The AI ​​Agent calculates the average failure rate across all authentication channels. If this value exceeds a preset third threshold (e.g., 15%), it indicates an overall system anomaly. Alternatively, it monitors the authentication failure rate of drivers within a specific region (e.g., a city). If the failure rate in that region is significantly higher than the historical baseline for that region (e.g., soaring from 5% to 25%), it indicates a potential regional fault. Meeting either condition triggers a global fault assessment by the AI ​​Agent. This dual-dimensional monitoring of all channels and regions comprehensively covers different fault scenarios, ensuring timely triggering of appropriate degradation strategies.

[0076] In one possible implementation, the AI ​​Agent monitors the metrics of each channel in real time. At 10:05 AM, it was detected that the failure rate of the Tencent Cloud channel surged from 5% to 35%, a 600% increase compared to the previous 5 minutes, and the "503 Service Unavailable" error accounted for over 90% of the failure return codes, meeting the conditions of step d1 and being determined as a channel-level failure. Meanwhile, the failure rates of the Alibaba Cloud and Megvii Cloud channels remained at normal levels of 3% and 4%, respectively. The average failure rate for all channels was (35%+3%+4%) / 3=14%, which did not exceed the third threshold of 15%, and the failure rates in each region were normal, therefore, no global failure was triggered.

[0077] In another scenario, a network outage occurred in the core data center of a certain city, causing all authentication channel requests to time out. The AI ​​Agent detected that the average failure rate across all channels surged to 85%, far exceeding the third threshold of 15%, thus satisfying the first condition of step d2; at the same time, the driver authentication failure rate in the city surged from the usual 4% to 80%, far exceeding the preset benchmark value, thus satisfying the second condition, and the AI ​​Agent determined it to be a global failure.

[0078] Through this implementation method, the AI ​​Agent can accurately distinguish between channel-level faults and global faults, providing an accurate basis for subsequent differentiated degradation strategies. This avoids erroneously triggering global exemptions during channel-level faults (which may pose security risks) and also avoids blindly switching channels during global faults (which wastes resources and fails to solve the problem). This achieves accurate identification and classification of fault types.

[0079] In some optional implementations, in conjunction with steps d1 and d2 above, the corresponding strategy in step S203 of triggering target channel degradation or temporary exemption for face recognition based on fault type completes the fault handling, including: Step e1: If the fault type is a channel-level fault, disable the authentication channel corresponding to the channel-level fault, select available channels sequentially from the list of authentication channels supported by the driver according to the preset polling strategy, and control the driver to re-execute face recognition authentication through the currently selected channel until authentication is successful or all available channels fail to attempt.

[0080] Specifically, the preset polling strategy refers to the pre-defined channel switching order and retry rules, such as trying channels in a fixed order of Channel 1 → Channel 2 → Channel 3 → ... → Channel n, or dynamically adjusting the priority based on the historical success rate of the channels. The authentication channel list refers to the collection of all face recognition authentication channels supported by the driver's application, such as Tencent Cloud, Alibaba Cloud, Megvii Cloud, Baidu Cloud, etc. This list does not depend on the server-side configuration and serves as a pool of alternative resources for polling switching.

[0081] When the AI ​​Agent determines a channel-level failure, the system immediately disables the faulty authentication channel and marks it as unavailable. Simultaneously, it obtains a complete list of authentication channels supported by the driver's device and selects available channels sequentially according to a preset polling strategy (e.g., fixed-order polling). After each channel is selected, the driver re-initiates facial recognition authentication through that channel and waits for the authentication result. If authentication is successful, the process ends; if authentication fails and the return code indicates that the channel also has a problem, the system continues to select the next channel to try until authentication succeeds or all available channels in the list fail. This mechanism proactively avoids the failure point when a single channel fails, utilizes multi-channel resources to automatically recover authentication requests, and improves the final success rate of a single authentication attempt.

[0082] Step e2: If the fault type is a global fault, determine whether the driver meets the preset exemption conditions. If the driver meets the preset exemption conditions, temporarily exempt the driver from face recognition authentication and allow the driver to drive. Otherwise, return an authentication failure message and record the reason for failure. The preset exemption conditions include at least one of the following: the driver is not logging in for the first time, there are no abnormalities in the historical authentication records within the preset period, there are no security violation records within the preset period, and the account-bound device has not been changed.

[0083] Specifically, the preset exemption conditions refer to a set of pre-set security access rules used to determine whether a driver can be temporarily exempted from facial recognition authentication in the event of a global failure. The purpose is to grant exemption permissions only to low-risk compliant drivers to prevent security vulnerabilities.

[0084] When the AI ​​Agent determines a global failure, the system initiates an exemption condition verification process. First, it extracts the driver's historical behavior data and verifies each preset exemption condition: whether it's a previous login (proving it's not a newly registered user), whether there are any anomalies in the historical authentication records within a preset period (e.g., 7 days) (e.g., no frequent failures, no logins from unusual locations), whether there are any safety violations within the preset period (e.g., no complaints, no dangerous driving), and whether the device bound to the account has not changed (device fingerprint matches historical records). If the driver meets all or at least one of the core exemption conditions, the system generates a temporary exemption instruction, allowing the driver to drive directly without facial recognition authentication, and records the exemption in the log for future reference. If the driver does not meet the exemption conditions, an authentication failure message is returned, and the reason for the failure is recorded for subsequent manual review or risk control processing. Through this mechanism, in the event of a global system failure, the driving rights of high-credit drivers are protected while maintaining a safety baseline, avoiding safety risks caused by large-scale exemptions.

[0085] In one possible implementation, after the AI ​​Agent determines that a channel-level failure has occurred in the Tencent Cloud channel, it proceeds to step e1. The system obtains the list of authentication channels supported by the driver's client, which is [Tencent Cloud, Alibaba Cloud, Megvii Cloud]. It immediately disables Tencent Cloud, leaving [Alibaba Cloud, Megvii Cloud] as the remaining available channels. According to a preset polling strategy (fixed order: Alibaba Cloud first, Megvii Cloud second), the system first selects the Alibaba Cloud channel, controls the driver's client to re-initiate authentication, and if authentication is successful, the process ends.

[0086] In another scenario, the AI ​​Agent determines a global failure (all channels unavailable). The system verifies the exemption conditions for the currently requesting driver, Li Si: Li Si is not a first-time login (registered for 3 months), has no abnormal authentication records in the past 7 days (successful authentication every day), has no security violation records in the past 7 days, and the device used for this login is consistent with the historically bound device, meeting all exemption conditions. The system generates a temporary exemption instruction, allowing Li Si to drive directly. Simultaneously, another driver, Zhao Wu, also initiates a request, but Zhao Wu is a first-time login (newly registered driver), not meeting the "not a first-time login" condition. The system returns an authentication failure message and records "global failure and failure to meet exemption conditions" as the reason for the failure.

[0087] Through this implementation method, the system can accurately adopt differentiated processing strategies based on the type of fault: in the event of a channel-level fault, it ensures the success rate of authentication by polling and switching; in the event of a global fault, it protects the driving rights of compliant drivers by screening for exemption conditions. This achieves refined management and optimal resource allocation in fault scenarios, taking into account system stability, business continuity and security.

[0088] Figure 3 This is a schematic diagram illustrating a normal authentication process and an AI Agent downgrade process according to an embodiment of this application. For example... Figure 3 As shown, the process consists of two parts: a normal authentication phase and an intelligent degradation phase.

[0089] During the normal authentication phase, after the driver initiates a facial recognition authentication request, the process first executes step 1: obtaining the authentication method, determining whether to use live face authentication or static face authentication based on whether it is the driver's first authentication. Next, step 2: obtaining the authentication channel, filtering candidate authentication channels based on the multi-platform facial recognition strategy configured in the backend (combining scenario, time period, and matching rules), and adapting dedicated channels for special groups (such as drivers from Hong Kong, Macau, and Taiwan). Then, step 4: filtering failed channels, retrieving the driver's failure records from Redis within the last 5 minutes and removing failed channels from the candidate channels (e.g., Tencent Cloud, Alibaba Cloud, Megvii Cloud, etc.). Then, step 5: determining if there are any available channels. If an available channel exists, proceed directly to step 6 for facial recognition; if all candidate channels are unavailable, obtain the complete list of all channels supported by the driver's built-in system (channel 1, channel 2… channel n), select channels sequentially according to a polling strategy, and perform facial recognition until authentication is successful or all channels have been tried. If authentication is successful, the process ends, and the driver is allowed to drive; if authentication fails, the AI ​​Agent degradation process is triggered.

[0090] In the AI ​​Agent degradation process, the AI ​​Agent collects online monitoring metrics (including QPS, response time, failure rate, and year-on-year / month-on-month data) in real time, combines them with authentication failure information (such as failure return codes) for intelligent analysis, determines the fault type (channel-level fault or global fault), and generates corresponding degradation instructions. Degradation instructions include two strategies: face recognition exemption degradation (temporarily exempting drivers from authentication for global faults who meet the exemption conditions) and designated channel degradation (disabling the faulty channel and switching to another available channel for channel-level faults). The degradation instructions are fed back to the authentication channel acquisition or failed channel filtering steps to achieve dynamic adjustment and ensure that subsequent authentication requests can be executed according to the new strategy.

[0091] This embodiment also provides a face recognition fault handling device based on an intelligent agent. This device is used to implement the above embodiments and preferred embodiments, and details already described will not be repeated. As used below, the term "module" can be a combination of software and / or hardware that implements a predetermined function. Although the device described in the following embodiments is preferably implemented in software, hardware implementation, or a combination of software and hardware, is also possible and contemplated.

[0092] This embodiment provides a face recognition fault handling device based on intelligent agents, such as... Figure 4 As shown, it includes: The channel determination module 401 is used to determine the available authentication channels in response to the face recognition authentication request initiated by the driver's terminal, based on a preset authentication strategy. The face authentication module 402 is used to control the driver to perform the corresponding face recognition authentication through the available authentication channel; The fault handling module 403 is used to trigger the intelligent agent to analyze the system operation status indicators and authentication failure information if authentication fails, in order to determine the fault type, and trigger the corresponding strategy of target channel downgrading or face recognition temporary exemption based on the fault type to complete the fault handling. The system operation status indicators include at least one of the following: query rate per second, response time, authentication failure rate, and failure rate change data.

[0093] In some alternative implementations, the channel determination module 401 includes: The determining unit is used to determine the initial authentication method and candidate authentication channels based on the preset authentication strategy. The preset authentication strategy includes scene adaptation rules, time period matching rules and special population adaptation rules. The initial authentication method includes live face authentication and static face authentication. The candidate authentication channels are at least two platform channels that provide face comparison services. The acquisition unit is used to acquire the driver's authentication channel failure records within a preset time period, remove the failed channels corresponding to the failure records from the candidate authentication channels, and obtain the available authentication channels.

[0094] In some optional implementations, the acquisition unit includes: The record reading subunit is used to store and read all failure records of the driver through a distributed cache. All failure records include the channel identifier, failure time and failure return code. The failure record determination sub-unit is used to retain failure records whose failure time is within a preset duration from all failure records, as authentication channel failure records.

[0095] In some alternative implementations, the determining unit includes: The judgment subunit is used to determine whether the driver is initiating authentication for the first time based on the driver's identity information in the face recognition authentication request; The initial authentication processing subunit is used to determine the initial authentication method as live face authentication if the driver is initiating authentication for the first time; otherwise, it determines the initial authentication method as either live face authentication or static face authentication. The candidate authentication channel determination sub-unit is used to select candidate authentication channels from all preset authentication channels according to the time period allocation rules, based on the current business scenario and time period. The dedicated authentication channel determination subunit is used to determine the dedicated authentication channel from the candidate authentication channels according to the special group adaptation rules if the driver's identity information indicates that the driver belongs to a special group, and to use the dedicated authentication channel as a candidate authentication channel.

[0096] In some alternative implementations, the fault handling module 403 includes: The channel-level fault determination unit is used to determine the fault type as a channel-level fault if the failure rate of a single authentication channel exceeds a first threshold and the change in failure rate exceeds a second threshold, and / or the failure return code in the authentication failure information is a service unavailable identifier. The global fault determination unit is used to determine the fault type as a global fault if the failure rate of the full authentication channel exceeds the third threshold, or if the failure rate of driver face recognition authentication in a specific area is abnormally higher than the preset benchmark value.

[0097] In some alternative implementations, the fault handling module 403 includes: The channel-level fault handling unit is used to disable the authentication channel corresponding to the channel-level fault if the fault type is a channel-level fault, select an available channel from the list of authentication channels supported by the driver terminal in sequence according to a preset polling strategy, and control the driver terminal to re-execute face recognition authentication through the currently selected channel until the authentication is successful or all available channels fail to attempt. The global fault handling unit is used to determine whether the driver meets the preset exemption conditions if the fault type is a global fault. If the driver meets the preset exemption conditions, the driver is temporarily exempted from facial recognition authentication and allowed to drive. Otherwise, the authentication failure prompt is returned and the reason for failure is recorded. The preset exemption conditions include at least one of the following: the driver is not logging in for the first time, there are no abnormalities in the historical authentication records within the preset period, there are no security violation records within the preset period, and the account-bound device has not been changed.

[0098] The agent-based face recognition fault handling device provided in this application can execute the agent-based face recognition fault handling method provided in any embodiment of this application, and has the corresponding functional modules and beneficial effects for executing the method. Further functional descriptions of the above modules and units are the same as those in the corresponding embodiments described above, and will not be repeated here.

[0099] Figure 5 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application.

[0100] The following is a detailed reference. Figure 5 The diagram illustrates a structural schematic suitable for implementing the electronic device described in the embodiments of this application. The electronic device may include a processor (e.g., a central processing unit, graphics processor, etc.) 501, which can perform various appropriate actions and processes according to a program stored in read-only memory (ROM) 502 or a program loaded from memory 508 into random access memory (RAM) 503. The RAM 503 also stores various programs and data required for the operation of the electronic device. The processor 501, ROM 502, and RAM 503 are interconnected via a bus 504. An input / output (I / O) interface 505 is also connected to the bus 504.

[0101] Typically, the following devices can be connected to I / O interface 505: input devices 506 including, for example, touchscreens, touchpads, keyboards, mice, cameras, microphones, accelerometers, gyroscopes, etc.; output devices 507 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; memory devices 508 including, for example, magnetic tapes, hard disks, etc.; and communication devices 509. Communication device 509 allows electronic devices to communicate wirelessly or wiredly with other devices to exchange data. Although Figure 5 Electronic devices with various devices are shown, but it should be understood that it is not required to implement or have all of the devices shown, and more or fewer devices may be implemented or have instead.

[0102] Specifically, according to embodiments of this application, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of this application include a computer program product comprising a computer program carried on a non-transitory computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication device 509, or installed from a memory 508, or installed from a ROM 502. When the computer program is executed by the processor 501, it performs the functions defined in the agent-based face recognition fault handling method of this application.

[0103] Figure 5 The electronic device shown is merely an example and should not impose any limitation on the functionality and scope of use of the embodiments of this application.

[0104] This application also provides a computer-readable storage medium. The methods described above according to this application can be implemented in hardware or firmware, or implemented as computer code that can be recorded on a storage medium, or implemented as computer code downloaded via a network and originally stored on a remote storage medium or a non-transitory machine-readable storage medium and then stored on a local storage medium. Thus, the methods described herein can be processed by software stored on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. The storage medium can be a magnetic disk, optical disk, read-only memory, random access memory, flash memory, hard disk, or solid-state drive, etc.; further, the storage medium can also include combinations of the above types of memory. It is understood that computers, processors, microprocessor controllers, or programmable hardware include storage components capable of storing or receiving software or computer code. When the software or computer code is accessed and executed by the computer, processor, or hardware, the agent-based face recognition fault handling method shown in the above embodiments is implemented.

[0105] A portion of this application can be applied as a computer program product, such as computer program instructions, which, when executed by a computer, can invoke or provide the methods and / or technical solutions according to this application through the operation of the computer. Those skilled in the art will understand that the forms in which computer program instructions exist in a computer-readable medium include, but are not limited to, source files, executable files, installation package files, etc. Correspondingly, the ways in which computer program instructions are executed by a computer include, but are not limited to: the computer directly executing the instructions, or the computer compiling the instructions and then executing the corresponding compiled program, or the computer reading and executing the instructions, or the computer reading and installing the instructions and then executing the corresponding installed program. Here, the computer-readable medium can be any available computer-readable storage medium or communication medium accessible to a computer.

[0106] Although embodiments of this application have been described in conjunction with the accompanying drawings, those skilled in the art can make various modifications and variations without departing from the spirit and scope of this application, and all such modifications and variations fall within the scope defined by the appended claims.

Claims

1. A face recognition fault handling method based on intelligent agents, characterized in that, The method includes: In response to a facial recognition authentication request initiated by the driver, the available authentication channels are determined based on a preset authentication strategy; The driver terminal is controlled to perform the corresponding facial recognition authentication through the available authentication channel; If authentication fails, the intelligent agent is triggered to analyze the system operation status indicators and authentication failure information to determine the fault type, and based on the fault type, trigger the corresponding strategy of target channel downgrading or face recognition temporary exemption to complete the fault handling. The system operation status indicators include at least one of the following: query rate per second, response time, authentication failure rate, and failure rate change data.

2. The method according to claim 1, characterized in that, The process of determining available authentication channels based on a preset authentication strategy includes: The initial authentication method and candidate authentication channels are determined based on a preset authentication strategy. The preset authentication strategy includes scene adaptation rules, time period matching rules, and special population adaptation rules. The initial authentication method includes live face authentication and static face authentication. The candidate authentication channels are at least two platform channels that provide face comparison services. Obtain the authentication channel failure records of the driver within a preset time period, remove the failed channels corresponding to the failure records from the candidate authentication channels, and obtain the available authentication channels.

3. The method according to claim 2, characterized in that, The step of obtaining the driver's authentication channel failure records within a preset time period includes: All failure records of the driver are stored and read through a distributed cache. The complete failure records include the channel identifier, failure time, and failure return code. The failure records whose failure time is within a preset time period are retained from all the failure records and used as the authentication channel failure records.

4. The method according to claim 2, characterized in that, The process of determining the initial authentication method and candidate authentication channels based on a preset authentication strategy includes: Based on the driver's identity information in the facial recognition authentication request, determine whether the driver is initiating authentication for the first time; If the driver is initiating authentication for the first time, the initial authentication method will be determined as live face authentication; otherwise, the initial authentication method will be determined as either live face authentication or static face authentication. Based on the current business scenario and time period, the candidate authentication channels are selected from all preset authentication channels according to the time period allocation rules; If the driver's identity information indicates that the driver belongs to a special group, a dedicated authentication channel is determined from the candidate authentication channels according to the special group adaptation rules, and the dedicated authentication channel is used as the candidate authentication channel.

5. The method according to claim 1, characterized in that, The triggering agent analyzes system operating status indicators and authentication failure information to determine the fault type, including: If the failure rate of a single authentication channel exceeds the first threshold and the change in failure rate increases beyond the second threshold, and / or the failure return code in the authentication failure information is a service unavailable identifier, the fault type is determined to be a channel-level fault. If the failure rate of the full authentication channel exceeds the third threshold, or if the failure rate of driver face recognition authentication in a specific area is abnormally higher than the preset benchmark value, the fault type is determined to be a global fault.

6. The method according to claim 5, characterized in that, The corresponding strategy for triggering target channel downgrade or temporary exemption for face recognition based on the fault type completes the fault handling, including: If the fault type is the channel-level fault, disable the authentication channel corresponding to the channel-level fault, select an available channel from the list of authentication channels supported by the driver terminal in sequence according to the preset polling strategy, and control the driver terminal to re-execute face recognition authentication through the currently selected channel until authentication is successful or all available channels fail to attempt. If the fault type is the global fault, determine whether the driver meets the preset exemption conditions. If the driver meets the preset exemption conditions, temporarily exempt the driver from face recognition authentication and allow the driver to drive. Otherwise, return an authentication failure prompt and record the reason for failure. The preset exemption conditions include at least one of the following: the driver is not logging in for the first time, there are no abnormalities in the historical authentication records within the preset period, there are no security violation records within the preset period, and the account-bound device has not been changed.

7. A face recognition fault handling device based on intelligent agents, characterized in that, The device includes: The channel determination module is used to respond to the facial recognition authentication request initiated by the driver and determine the available authentication channels based on the preset authentication strategy. The face authentication module is used to control the driver's terminal to perform corresponding face recognition authentication through the available authentication channel; The fault handling module is used to trigger the intelligent agent to analyze the system operation status indicators and authentication failure information if authentication fails, in order to determine the fault type, and trigger the corresponding strategy of target channel downgrading or face recognition temporary exemption based on the fault type to complete the fault handling. The system operation status indicators include at least one of the following: query rate per second, response time, authentication failure rate, and failure rate change data.

8. An electronic device, characterized in that, include: The system includes a memory and a processor, which are interconnected. The memory stores computer instructions, and the processor executes the computer instructions to perform the face recognition fault handling method based on any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions for causing the computer to execute the agent-based face recognition fault handling method according to any one of claims 1 to 6.

10. A computer program product, characterized in that, It includes computer instructions for causing a computer to execute the agent-based face recognition fault handling method according to any one of claims 1 to 6.

Images