Multi-protocol intelligent recognition method and system based on multi-manufacturer face recognition algorithm process scheduling

Abstract

The invention provides a multi-protocol intelligent recognition method and system based on multi-manufacturer face recognition algorithm process scheduling, and the method comprises the steps: obtaining a recognition request submitted by a calling party, and the recognition request comprises a photo and algorithm preference information; selecting an optimal face recognition algorithm from multi-manufacturer face recognition algorithms according to the algorithm preference information of the calling party; judging whether the optimal face recognition algorithm has a dynamic library conflict with a face recognition algorithm being executed in the core sub-process or not, if not, scheduling the core sub-process, executing the calling of an SDK (Software Development Kit) of the optimal face recognition algorithm, and performing face recognition on the photo; if the dynamic library conflict exists, scheduling an independent sub-process, and executing the calling of an SDK (Software Development Kit) of an optimal face recognition algorithm; and obtaining a face recognition result returned by the core sub-process or the independent sub-process, responding to a calling party request in a standard format, and recording a calling log, a performance index and a recognition result at the same time.

Description

Technical Field

[0001] This invention relates to the field of data synchronization technology, and more specifically, to an intelligent recognition method and system based on process scheduling of multi-vendor face recognition algorithms. Background Technology

[0002] With the widespread application of facial recognition technology in security, finance, transportation, and other fields, numerous facial recognition algorithm SDKs from various vendors have emerged in the market. These algorithms differ in feature extraction accuracy, performance optimization, and facial quality detection standards. In practical applications, it is often necessary to call different vendors' or different versions of algorithms depending on the scenario. Referring to the invention patent application with application number 202411727138.1, a method for deploying algorithm services that integrates multiple facial recognition algorithms and provides a unified interface is provided. This method matches the algorithm requirements in the request data with the interface descriptions in all interface documents to determine the target algorithm API interface. Based on the request format of the target algorithm API interface, the image to be recognized in the request data is converted to the correct format before the target algorithm API interface is called for processing. However, due to symbol conflicts at the dynamic library level between different vendors' and different versions of algorithm SDKs, they cannot coexist in the same process, thus limiting the system's flexibility and scalability.

[0003] Traditional solutions often rely on fixed algorithms or single-vendor SDKs, making it difficult to support multiple versions coexisting or achieve unified scheduling and high-concurrency distributed deployment. Therefore, there is an urgent need for a face recognition system that can dynamically schedule multi-process algorithm services and support multi-protocol access.

[0004] In order to solve the above problems, people have been seeking an ideal technological solution. Summary of the Invention

[0005] Therefore, it is necessary to provide an intelligent recognition system based on multi-vendor algorithm process scheduling to address the aforementioned technical problems, aiming to solve the problems of dynamic library conflicts and inflexible algorithm scheduling in traditional face recognition algorithm programs.

[0006] To achieve the above objectives, the first aspect of the present invention provides an intelligent recognition method based on multi-vendor face recognition algorithm process scheduling, comprising the following steps: Obtain the recognition request submitted by the caller, the recognition request including the photo and algorithm preference information; The optimal face recognition algorithm is selected from multiple vendors' face recognition algorithms based on the caller's algorithm preference information; Determine if there is a dynamic library conflict between the optimal face recognition algorithm and the face recognition algorithm being executed in the core subprocess. If there is no dynamic library conflict, schedule the core subprocess to execute the SDK of the optimal face recognition algorithm and perform face recognition on the photo; if there is a dynamic library conflict, schedule an independent subprocess to execute the SDK of the optimal face recognition algorithm. Retrieve the face recognition results returned by the core subprocess or independent subprocess, respond to the caller's request in a standard format, and record the call log, performance metrics, and recognition results.

[0007] This system achieves an optimal balance between resource efficiency and service stability by perfectly combining intelligent scheduling and conflict isolation, ensuring recognition accuracy. Specifically, based on algorithm preference information provided by the caller, the system can dynamically select the optimal solution from multiple vendor algorithms to ensure the most suitable recognition effect for different business scenarios. Through a dynamic library conflict detection mechanism, it intelligently distinguishes between core subprocesses and independent subprocesses, fundamentally solving the symbol conflict problem of multiple vendor algorithm SDKs in the same process. When there is no conflict, the core subprocess is reused to reduce process creation overhead; when there is a conflict, independent subprocesses are activated to achieve resource isolation and parallel processing, maximizing system throughput. In addition, regardless of the algorithm and process scheduling method used internally, a unified standard format response is provided externally, simplifying the integration complexity for callers and improving the user experience.

[0008] In one specific embodiment, the algorithm preference information includes at least one of the following: business scenario type, accuracy requirement, response time requirement, and cost constraint; The optimal face recognition algorithm is selected based on the caller's algorithm preference information, including: Based on multi-dimensional algorithm preference information, a weighted scoring model is used to calculate the comprehensive fit score of face recognition algorithms from multiple vendors. The algorithms are then sorted from high to low according to their comprehensive fit scores, and the optimal face recognition algorithm is selected to perform the face recognition task.

[0009] By leveraging multi-dimensional preference information such as business scenario type, accuracy requirements, response time, and cost constraints, we achieve precise matching between algorithm selection and business needs, ensuring a high degree of consistency between technical solutions and business objectives. A weighted scoring model transforms subjective preferences into objective values, eliminating arbitrariness in selection, avoiding biases in technology selection, promoting the healthy development of the algorithm ecosystem, and ensuring that the optimal technical solution is always chosen.

[0010] In one specific embodiment, determining whether the optimal face recognition algorithm conflicts with the face recognition algorithm being executed in the core subprocess via dynamic library includes: Pre-obtain the dynamic library conflict relationships between face recognition algorithms from multiple vendors and construct an algorithm conflict matrix; After selecting the optimal face recognition algorithm, query the dynamic library conflict relationship between the optimal similarity algorithm in the algorithm conflict matrix and the face recognition algorithm being executed in the core subprocess.

[0011] The library conflict detection mechanism, based on a pre-built algorithm conflict matrix, achieves a fundamental shift from passive response to proactive prevention by pre-abstracting complex dynamic library dependencies into a fast-queryable conflict relationship graph. This system can accurately identify potential conflicts through efficient matrix queries before algorithm scheduling, completely avoiding the performance overhead of runtime dynamic detection while ensuring extreme stability in multi-vendor algorithm environments. This design not only minimizes conflict detection response time but also greatly improves system maintainability through centralized configuration management, allowing new algorithms to be integrated with only an updated conflict matrix without modifying the core architecture.

[0012] To achieve the above objectives, a second aspect of the present invention provides a multi-protocol intelligent recognition system based on multi-vendor face recognition algorithm process scheduling, comprising: A unified API gateway is configured to expose a unified interface to receive and obtain recognition requests submitted by callers, the recognition requests including photos and algorithm preference information; The main control process is configured to perform algorithm task distribution, child process management, and caller request forwarding based on algorithm preference information. The core subprocess is configured to run multiple face recognition algorithm SDKs that do not conflict with dynamic libraries; An independent algorithm subprocess is configured to load and run a face recognition algorithm that has dynamic library conflicts. The communication framework is configured to enable remote procedure calls between processes via the RPCX framework. The algorithm feature pool is configured to store registered facial feature data locally for feature comparison and similar person retrieval during the execution of the facial recognition algorithm.

[0013] To achieve the above objectives, a third aspect of the present invention provides a computer device, including a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other through the communication bus. Memory, used to store computer programs; The processor, when executing a program stored in memory, implements the intelligent recognition method as described in the first aspect.

[0014] To achieve the above objectives, a fourth aspect of the present invention provides a computer program product, including a computer program that, when executed by a processor, implements the intelligent recognition method as described in the first aspect.

[0015] The beneficial effects of this invention are as follows: This invention constructs a highly flexible and stable face recognition system through an innovative process scheduling and multi-algorithm subprocess combination architecture. The system adopts a modular design, encapsulating face recognition algorithms such as face feature extraction, quality verification, and most similar person identification based on a local feature pool into independent algorithm SDKs. A unified master control process enables the collaborative work of these processes.

[0016] To address the long-standing technical challenge of loading different versions of algorithm dynamic libraries in parallel within the same process, this invention proposes an isolated execution scheme based on independent algorithm subprocesses. Each algorithm SDK runs in its own dedicated subprocess environment, fundamentally eliminating issues such as dynamic library symbol conflicts, resource contention, and memory pollution. A lightweight RPCX framework is used to construct an efficient inter-process communication mechanism, ensuring that each algorithm subprocess maintains complete operational independence while achieving low-latency data exchange and task collaboration.

[0017] This system architecture not only solves the compatibility problem of multiple algorithms coexisting, but also achieves optimal utilization of computing resources through dynamic resource allocation and load balancing mechanisms. When a new algorithm SDK is added, the system can quickly create a corresponding independent subprocess, completing hot algorithm updates without restarting the service. Simultaneously, based on a pre-built algorithm conflict matrix, the system can intelligently avoid potential library conflict risks, ensuring stable service performance even in complex multi-algorithm environments. This design provides a solid technical foundation for the large-scale commercial application of facial recognition technology and has broad practical value in fields such as security, finance, and the Internet of Things. Attached Figure Description

[0018] Figure 1 This is a flowchart of process scheduling and identification.

[0019] Figure 2 System architecture diagram. Detailed Implementation

[0020] The technical solution of the present invention will be further described in detail below through specific embodiments.

[0021] Example 1 This embodiment provides an intelligent recognition method based on multi-vendor face recognition algorithm process scheduling, including the following steps: Obtain the recognition request submitted by the caller, the recognition request including the photo and algorithm preference information; In specific implementation, the algorithm preference information includes at least one of the following: business scenario type, accuracy requirement, response time requirement, and cost constraint.

[0022] Specifically, the business scenarios include facial recognition payment, identity verification, remote account opening, and security monitoring. Different business scenarios have different requirements for the accuracy and response time of facial recognition algorithms. For example, in financial payment scenarios, priority is given to ensuring transaction security and real-time response; in identity verification scenarios, the pursuit of ultimate accuracy is sought; in remote account opening scenarios, user experience and compliance requirements are balanced; and in security monitoring scenarios, the efficiency of large-scale retrieval is optimized.

[0023] The identification request includes not only photos and algorithm preference information, but also request headers, path parameters, query parameters, and request body.

[0024] The optimal face recognition algorithm is selected from multiple vendors' face recognition algorithms based on the caller's algorithm preference information.

[0025] In one embodiment, selecting the optimal face recognition algorithm based on the caller's algorithm preference information includes: Based on multi-dimensional algorithm preference information, a weighted scoring model is used to calculate the comprehensive fit score of face recognition algorithms from multiple vendors. The algorithms are then sorted from high to low according to their comprehensive fit scores, and the optimal face recognition algorithm is selected to perform the face recognition task.

[0026] This method employs an intelligent weight allocation mechanism to dynamically adjust the algorithm evaluation dimensions based on the characteristics of the business scenario. For facial recognition payment scenarios, low-latency algorithms are prioritized while ensuring necessary accuracy; for identity verification scenarios, high-precision algorithms are emphasized, with a slight relaxation of response time requirements; and in security monitoring scenarios, priority is given to algorithm recall and adaptability to complex environments. By establishing a precise mapping between scenario features and algorithm capabilities, the optimal technical solution is ensured for each business scenario.

[0027] It's important to note that face recognition algorithms typically include face detection, facial feature extraction, quality detection, and feature comparison. The role of quality detection is crucial; it involves self-evaluating the reliability and confidence of the extracted features, either during or after feature extraction. The goal is to assign a quality score or confidence level to the extracted feature vector. This score directly reflects the algorithm's confidence in the extracted features. If the quality score falls below a certain internal threshold, it may directly determine that the recognition has failed or refuse to output the feature vector.

[0028] The system determines whether there is a dynamic library conflict between the optimal face recognition algorithm and the face recognition algorithm being executed in the core subprocess. If there is no dynamic library conflict, the core subprocess is scheduled to execute the SDK of the optimal face recognition algorithm to perform face recognition on the photo. If there is a dynamic library conflict, an independent subprocess is scheduled to execute the SDK of the optimal face recognition algorithm.

[0029] Specifically, determining whether the optimal face recognition algorithm conflicts with the face recognition algorithm being executed in the core subprocess involves checking for dynamic library conflicts, including: Pre-obtain the dynamic library conflict relationships between face recognition algorithms from multiple vendors and construct an algorithm conflict matrix; After selecting the optimal face recognition algorithm, query the dynamic library conflict relationship between the optimal similarity algorithm in the algorithm conflict matrix and the face recognition algorithm being executed in the core subprocess.

[0030] Based on a pre-built algorithm conflict matrix, the system can make intelligent process scheduling decisions. Once the optimal algorithm is selected, the system immediately queries the conflict matrix. If a library conflict is found between the algorithm and the algorithm running in the core subprocess, the algorithm is automatically scheduled to run in an independent subprocess, ensuring that each algorithm runs stably in a completely isolated environment.

[0031] It is understandable that during the registration process of a face recognition algorithm, the main control process proactively acquires and records dynamic library conflict information for each algorithm SDK through a combination of static analysis and dynamic testing, establishing a complete algorithm conflict matrix. This proactive conflict detection mechanism allows the system to anticipate potential library conflict risks before algorithm scheduling, laying the foundation for accurate process isolation.

[0032] Retrieve the face recognition results returned by the core subprocess or independent subprocess, respond to the caller's request in a standard format, and record the call log, performance metrics, and recognition results.

[0033] In the above method, only a unified interface is provided to the outside world, so that users no longer need to pay attention to the specific interface, connection method and configuration of the face algorithm. The platform caller can access the corresponding face algorithm by calling the unified external interface to realize the photo processing function.

[0034] Example 2 The difference between this embodiment and Embodiment 1 is that: Before selecting the optimal face recognition algorithm from multiple vendors' face recognition algorithms based on the caller's algorithm preference information, the quality detection subprocess is called to execute the image quality detection algorithm to perform a preliminary quality assessment of the input image. The assessment indicators include at least one of sharpness, brightness, and face size. If the initial quality assessment meets the feature extraction quality threshold, the optimal face recognition algorithm will be selected from multiple vendors' face recognition algorithms based on the caller's algorithm preference information; otherwise, the face recognition process will end and the result of poor image quality will be returned to the caller.

[0035] It's understandable that the image quality detection algorithm here differs from the pre-set image quality detection in face recognition algorithms. It's positioned at the very beginning of the process, using a unified and configurable standard to quickly screen input images for quality. This rejects images with extremely poor quality (such as severely blurry, lacking faces, or having excessively small faces), preventing them from entering the subsequent computationally intensive feature extraction and comparison stages. This saves significant computing resources and improves the overall system throughput.

[0036] Furthermore, when the initial quality assessment meets the feature extraction quality threshold, and the optimal face recognition algorithm is selected from multiple vendors' face recognition algorithms based on the caller's algorithm preference information, the optimal face recognition algorithm can also be selected from multiple vendors' face recognition algorithms by combining the initial quality assessment results.

[0037] The initial quality assessment identifies specific quality defects in the photos (such as severe side profile, low light, or occlusion). Based on these defects and the caller's algorithm preference information, the system intelligently selects the feature extraction algorithm best suited to handle these problems. This assigns different quality challenges to the most capable algorithm to solve them, thereby maximizing the overall system's recognition rate.

[0038] Example 3 This embodiment provides a multi-protocol intelligent recognition system based on multi-vendor face recognition algorithm process scheduling, such as... Figure 2 As shown, it includes: A unified API gateway is configured to expose a unified interface to receive and obtain recognition requests submitted by callers, the recognition requests including photos and algorithm preference information; The main control process is configured to perform algorithm task distribution, child process management, and caller request forwarding based on algorithm preference information. The core subprocess is configured to run multiple face recognition algorithm SDKs that do not conflict with dynamic libraries; An independent algorithm subprocess is configured to load and run a face recognition algorithm that has dynamic library conflicts. The communication framework is configured to enable remote procedure calls between processes via the RPCX framework. The algorithm feature pool is configured to store registered facial feature data locally for feature comparison and similar person retrieval during the execution of the facial recognition algorithm.

[0039] It is understood that the main control process is also configured with: The conflict detection and isolation module is used to detect library conflicts between algorithm SDKs and ensure the parallel execution of conflicting algorithms through process-level isolation; The unified scheduling and management module is used to dynamically select and execute algorithm services based on business needs and system status.

[0040] Specifically, the system supports hot-swapping of new algorithm SDKs. To achieve this functionality, the main control process also includes a plugin management module, which continuously monitors a preset plugin directory. When a new face recognition algorithm SDK is added, the algorithm dynamic library and description file are placed in the plugin directory. The plugin management module of the main control process detects a change in the specified directory, extracts key metadata from the description file, and registers it in the algorithm service registry. When a face recognition algorithm SDK is removed, the corresponding registration information is removed from the algorithm service registry.

[0041] Furthermore, the main control process also monitors the running status of the child process in real time through RPCX, terminates the abnormal child process when it runs abnormally, and then starts a new child process according to the configuration information of the abnormal child process.

[0042] In practical applications, a main control process and multiple sub-processes can be deployed on a single-node server. The main control process dynamically selects the algorithm sub-process to execute the recognition task based on the caller's algorithm preferences. This mode is simple to deploy and suitable for small to medium-scale recognition task scenarios.

[0043] Furthermore, multiple node servers can be configured, each running an independent algorithm service instance. These multiple node servers form a distributed cluster via NATS. One node service serves as the master scheduling node, responsible for receiving requests and intelligently distributing them, assigning tasks to different node servers based on load. This clustered distributed deployment mode enables high-concurrency recognition and is suitable for scenarios such as smart parks, security monitoring, and airport access control.

[0044] It should be understood that although the steps in the flowcharts of the above embodiments are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the above embodiments may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0045] Example 5 This embodiment provides a computer device, including a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other through the communication bus; Memory, used to store computer programs; When the processor executes a program stored in the memory, it implements the intelligent recognition method as described in any one of Examples 1-2.

[0046] Example 6 Based on the above embodiments, this embodiment provides a computer program product, including a computer program that, when executed by a processor, implements the intelligent recognition method as described in any one of embodiments 1-2.

[0047] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them; although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications can still be made to the specific implementation of the present invention or equivalent substitutions can be made to some technical features without departing from the spirit of the technical solutions of the present invention, and all such modifications and substitutions should be covered within the scope of the technical solutions claimed in the present invention.

Claims

1. An intelligent recognition method based on multi-vendor face recognition algorithm process scheduling, characterized in that, Includes the following steps: Obtain the recognition request submitted by the caller, the recognition request including the photo and algorithm preference information; The optimal face recognition algorithm is selected from multiple vendors' face recognition algorithms based on the caller's algorithm preference information; Determine if there is a dynamic library conflict between the optimal face recognition algorithm and the face recognition algorithm being executed in the core subprocess. If there is no dynamic library conflict, schedule the core subprocess to execute the SDK of the optimal face recognition algorithm and perform face recognition on the photo; if there is a dynamic library conflict, schedule an independent subprocess to execute the SDK of the optimal face recognition algorithm. Retrieve the face recognition results returned by the core subprocess or independent subprocess, respond to the caller's request in a standard format, and record the call log, performance metrics, and recognition results.

2. The intelligent recognition method based on multi-vendor face recognition algorithm process scheduling according to claim 1, characterized in that, The algorithm preference information includes at least one of the following: business scenario type, accuracy requirement, response time requirement, and cost constraint; The optimal face recognition algorithm is selected based on the caller's algorithm preference information, including: Based on multi-dimensional algorithm preference information, a weighted scoring model is used to calculate the comprehensive fit score of face recognition algorithms from multiple vendors. Based on the comprehensive fit score, the algorithm is sorted from high to low, and the optimal face recognition algorithm is selected to perform the face recognition task.

3. The intelligent recognition method based on multi-vendor face recognition algorithm process scheduling according to claim 1, characterized in that, The optimal face recognition algorithm performs face recognition on photos, including: Perform face detection on the photo to obtain images of the face regions; Feature extraction and image quality detection are performed on facial region images; Determine whether the image quality meets the preset quality threshold. If it does, compare the extracted facial features with the locally stored registered facial feature data and search for similar individuals.

4. The multi-protocol intelligent recognition system based on multi-vendor face recognition algorithm process scheduling according to claim 3, characterized in that, Determine if the optimal face recognition algorithm conflicts with the face recognition algorithm being executed in the core subprocess using dynamic libraries, including: Pre-obtain the dynamic library conflict relationships between face recognition algorithms from multiple vendors and construct an algorithm conflict matrix; After selecting the optimal face recognition algorithm, query the dynamic library conflict relationship between the optimal similarity algorithm in the algorithm conflict matrix and the face recognition algorithm being executed in the core subprocess.

5. A multi-protocol intelligent recognition system based on multi-vendor face recognition algorithm process scheduling according to claim 2, 3, or 4, characterized in that, Before selecting the optimal face recognition algorithm from multiple vendors' face recognition algorithms based on the caller's algorithm preference information, the quality detection subprocess is called to execute the image quality detection algorithm to perform a preliminary quality assessment of the input image. The assessment indicators include at least one of sharpness, brightness, and face size. If the initial quality assessment meets the feature extraction quality threshold, the optimal face recognition algorithm will be selected from multiple vendors' face recognition algorithms based on the caller's algorithm preference information; otherwise, the face recognition process will end and the result of poor image quality will be returned to the caller.

6. A multi-protocol intelligent recognition system based on multi-vendor face recognition algorithm process scheduling according to claim 2, 3, or 4, characterized in that, When the initial quality assessment meets the feature extraction quality threshold, and the optimal face recognition algorithm is selected from multiple vendors' face recognition algorithms based on the caller's algorithm preference information, the optimal face recognition algorithm can also be selected from multiple vendors' face recognition algorithms by combining the initial quality assessment results.

7. A multi-protocol intelligent recognition system based on multi-vendor face recognition algorithm process scheduling, characterized in that, include: A unified API gateway is configured to expose a unified interface to receive and obtain recognition requests submitted by callers, the recognition requests including photos and algorithm preference information; The main control process is configured to perform algorithm task distribution, child process management, and caller request forwarding based on algorithm preference information. The core subprocess is configured to run multiple face recognition algorithm SDKs that do not conflict with dynamic libraries; An independent algorithm subprocess is configured to load and run a face recognition algorithm that has dynamic library conflicts. The communication framework is configured to enable remote procedure calls between processes via the RPCX framework. The algorithm feature pool is configured to store registered facial feature data locally for feature comparison and similar person retrieval during the execution of the facial recognition algorithm.

8. A multi-protocol intelligent recognition system based on multi-vendor face recognition algorithm process scheduling according to claim 6, characterized in that, The algorithm preference information includes at least one of the following: business scenario requirements, security, accuracy, speed, and cost requirements; the communication protocol of the identification request includes HTTP, NATS, RPCX, or gRPC protocol. The main control process also includes a plugin management module, which continuously monitors a preset plugin directory. When a new face recognition algorithm SDK is added, the algorithm dynamic library and description file are placed in the plugin directory. The plugin management module of the main control process detects that the specified directory has changed, extracts key metadata from the description file, and registers it in the algorithm service registry. When the face recognition algorithm SDK is removed, the corresponding registration information is removed from the algorithm service registry. The main control process also monitors the running status of the child process in real time through RPCX. When the child process runs abnormally, it terminates the abnormal child process and then starts a new child process according to the configuration information of the abnormal child process. According to claim 6, a multi-protocol intelligent recognition system based on multi-vendor face recognition algorithm process scheduling is characterized in that multiple node servers are set up, each node server runs an independent algorithm service instance, and the multiple node servers construct a distributed cluster through NATS; one of the node services is the master control scheduling node, which is responsible for receiving requests and intelligent distribution, and distributes tasks to different node servers according to the load.

9. A computer device, characterized in that: It includes a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other through the communication bus; Memory, used to store computer programs; A processor, when executing a program stored in memory, implements the intelligent recognition method as described in any one of claims 1-6.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the intelligent recognition method as described in any one of claims 1-6.

Images