Community governance system combining image analysis and big data deep learning
By combining image analysis and big data deep learning technology, events in the community are monitored in real time, the problem of difficult events in traditional community governance in perceived complex social environments is solved, efficient event perception and feedback is achieved, and the quality of social governance and public services is improved. .
Patent Information
- Application Number
- CN202410485779.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-04-22
- Publication Date
- 2024-11-08
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
Traditional community governance methods are difficult to meet the modern society's demand for real-time perception of events in communities, especially in complex social environments, where efficient event perception and feedback mechanisms are lacking.
Combining image analysis and big data deep learning technology, by deploying cameras to real-time monitoring, image data is transmitted to the back-end processing center, abnormal behavior and event analysis, and real-time feedback is provided through text messages, emails, mobile APPs, etc., including face recognition, Traffic flow analysis and behavioral pattern analysis, and build a comprehensive management platform for smart communities.
It significantly improves the efficiency and accuracy of social governance, improves the quality and efficiency of public services, makes the community safer, harmonious and convenient, and provides new methods and tools for intelligent comprehensive governance.
Abstract
Description
Technical Field
[0001] The present invention relates to the fields of smart cities and public security, and in particular to a community governance system combining image analysis and big data deep learning. Background Art
[0002] In order to strengthen social governance, most community governance currently relies on daily monitoring, daily supervision and inspections by grid workers. Although colleges and universities, research institutes and Internet companies have developed a series of urban security monitoring and prevention platforms, with the increasing complexity of social cities and the deepening of social governance needs, traditional social governance methods have been unable to meet the needs of modern society, especially the real-time perception of various events in the community. Summary of the invention
[0003] The purpose of the present invention is to provide a community governance system that combines image analysis and big data deep learning, which has the advantage of real-time perception of various events in the community and feedback.
[0004] The above technical objectives of the present invention are achieved through the following technical solutions: a community governance system combining image analysis and big data deep learning, including a user management module, a device management module, a data acquisition and analysis module, a visual data display module, and a system security and performance optimization module; and also includes the following steps: S1, real-time monitoring, the system captures image data through cameras deployed in every corner of the community; S2, information transmission, image data is transmitted to the back-end processing center; S3, image analysis, analyzing abnormal behaviors or events; S4. Early warning mechanism, push notifications via SMS, email, and mobile APP, and send analysis results to the monitoring center screen.
[0005] Preferably, the abnormal behaviors or events in the image analysis in S3 include warning label events and fixed warning events.
[0006] Preferably, the fixed warning events include the entry and exit of key personnel; the warning tag events include Elderly care: testing residents aged 60 and above who have not been in or out of the community for 15 days; Frequent visitors, detecting visitors entering and leaving the community multiple times; Hiding during the day and coming out at night, the detection community personnel frequently enter and leave the community late at night; Smoke events: detect abnormal alarms issued by connected fire-fighting equipment and facilities; Suspected group rental housing: when residents in the community register, the average per capita housing area is less than 12 square meters; Boundary triggering detects alarms from connected IoT devices such as electronic fences.
[0007] Preferably, in the real-time monitoring step of the device management module in S1, the camera in the device management module will capture the face image; the S3 image analysis module includes face recognition and identity authentication, the system recognizes the face in the image taken for monitoring, and compares it with the registered information in the big database to achieve identity authentication.
[0008] Preferably, the S3 image analysis module includes image preprocessing technology in which the system first denoises and enhances the image to improve the accuracy of face recognition, then enters into deep learning of big data, extracts features of the face in the image, generates face feature vectors, and then compares the extracted face feature vectors with the information in the big database. The big database contains pre-collected and registered face data, including photos of suspects and records of people entering and leaving public places. Through the comparison algorithm, the system matches the face feature vectors to confirm the identity.
[0009] Preferably, the equipment management module includes traffic cameras inside the community and at the door. The traffic cameras are deployed at key sections and traffic nodes, intersections and highway entrances and exits. The S3 image analysis includes traffic flow analysis inside and outside the community. The image analysis technology begins to analyze the received image data. Through computer vision and image recognition algorithms, the image analysis can automatically extract vehicle information in the image, including the number of vehicles, vehicle models, driving speed, etc. The system can also judge the road congestion according to the vehicle's driving trajectory and density. Through real-time analysis of traffic flow, the system will feedback the information to the public and traffic managers in the early warning mechanism through mobile phone APP, traffic display screens, etc.
[0010] Preferably, the S3 image analysis also includes behavior pattern analysis. The deep learning library needs to collect a large amount of surveillance video data samples. The data samples come from cameras in various areas of the city. The cameras cover parks, squares, commercial areas, transportation hubs, schools and hospitals. The image analysis algorithm analyzes these collected images and samples in the large database to identify individual and group behaviors in the crowd, including walking routes, gathering situations and communication methods. By comparing and analyzing crowd behavior data in different time periods and different areas, abnormal behaviors or potential social problems can be discovered.
[0011] Preferably, when a social event or a criminal case occurs, the system can trace back and analyze the image data captured by the surveillance camera to provide solid evidence and valuable clues for the investigation of the event. After the event occurs, the relevant investigators can use the image analysis system in S3 to trace back the surveillance video of a specific time period. The video includes detailed scenes and participant behaviors before and after the event. The system automatically or manually analyzes the video frame by frame to find key information related to the event. During the tracing process, the image analysis system uses algorithms and models to identify the target object in the video and track its movement trajectory, analyze the posture and action behavior characteristics of the target object, and feed back the object's role and intention in the event as a structure to the S4 early warning mechanism; the system also includes identification and analysis of buildings and road sign elements in the scene. The early warning mechanism can extract key frames or video clips related to the event and generate a report including the timeline of the event, the behavioral characteristics of the participants and scene information.
[0012] Preferably, the device management module includes cameras deployed in public places, which continuously transmit image data to the back-end processing center. The image data includes people's facial expressions, body movements and social interactions. The image analysis algorithm in S3 processes the data to identify key information related to public needs and feedback. The system analyzes people's stay time and activity frequency in public places, and analyzes which areas or services are most popular or have problems. The system analyzes people's satisfaction and feelings about public services by obtaining people's facial expressions and body language.
[0013] As a preference, a health scoring mechanism is also included. Through self-service data analysis, the results of image analysis are summarized, and the cell data in each project is comprehensively evaluated to achieve a visual health score. In addition, notifications / emails are automatically sent to alert operation and maintenance personnel.
[0014] To sum up, combining image analysis technology with social governance through algorithms and big data technology can greatly improve the efficiency and accuracy of social governance, and can also improve the quality and efficiency of public services. The platform-level product based on the combination of image analysis and community governance, the smart community comprehensive management platform, through the empowerment of image analysis and big data, can effectively serve and supervise the side, making the community safer, more harmonious and more convenient; by building a cloud-edge collaborative architecture realized by the system, using multi-camera based image analysis technology, big data analysis visualization technology and deep learning algorithms, it can effectively realize the intelligent comprehensive governance of community scenes, and ensure large-scale application in cities, providing new ideas, new methods and new tools for social governance. DETAILED DESCRIPTION
[0015] This specific embodiment is merely an explanation of the present invention and is not a limitation of the present invention. After reading this specification, those skilled in the art may make non-creative modifications to the present embodiment as needed. However, as long as they are within the scope of the claims of the present invention, they are protected by the patent law.
[0016] The collected data involved in this application does not infringe on personal privacy, complies with the "Personal Information Protection Law of the People's Republic of China" and other legal requirements, does not infringe on the personal information rights and interests of natural persons, and fully follows the principles of legality, legitimacy, necessity and good faith in the processing of personal information. The collection of data abides by relevant laws and regulations to ensure that personal privacy is not infringed and personal information is processed legally, legitimately and necessary.
[0017] A community governance system combining image analysis and big data deep learning, including a user management module, a device management module, a data collection and analysis module, a visual data display module, and a system security and performance optimization module; and also includes the following steps: S1, real-time monitoring, the system captures image data through cameras deployed in every corner of the community; S2, information transmission, image data is transmitted to the back-end processing center; S3. Image analysis: analyzing abnormal behaviors or events. Abnormal behaviors or events in image analysis include warning label events and fixed warning events. Fixed warning events include the entry and exit of key personnel. The warning label events include Elderly care: testing residents aged 60 and above who have not been in or out of the community for 15 days; Frequent visitors, detecting visitors entering and leaving the community multiple times; Hiding during the day and coming out at night, the detection community personnel frequently enter and leave the community late at night; Smoke events: detect abnormal alarms issued by connected fire-fighting equipment and facilities; Suspected group rental housing: when residents in the community register, the average per capita housing area is less than 12 square meters; Boundary triggering, detecting connected electronic fences and other IoT devices to send out alarms; S4. Early warning mechanism, push notifications via SMS, email, and mobile APP, and send analysis results to the monitoring center screen.
[0018] In the real-time monitoring step of the device management module in S1, the camera in the device management module will capture the face image; the S3 image analysis module includes face recognition and identity authentication. The system recognizes the face in the image taken for monitoring and compares it with the registered information in the big database to achieve identity authentication; the S3 image analysis module also includes image preprocessing technology. The system will first denoise and enhance the image to improve the accuracy of face recognition, and then enter into deep learning of big data to extract features of the face in the image and generate a face feature vector. The extracted face feature vector is then compared with the information in the big database. The big database contains pre-collected and registered face data, including photos of suspects and records of people entering and leaving public places. Through the comparison algorithm, the system matches the face feature vector to confirm the identity.
[0019] The equipment management module includes traffic cameras inside the community and at the door. Traffic cameras are deployed at key sections and traffic nodes, intersections and highway entrances and exits. S3 image analysis includes traffic flow analysis inside and outside the community. Image analysis technology begins to analyze the received image data. Through computer vision and image recognition algorithms, image analysis can automatically extract vehicle information in the image, including the number of vehicles, model, driving speed, etc. The system can also judge the road congestion based on the vehicle's driving trajectory and density. Through real-time analysis of traffic flow, the system will feedback information to the public and traffic managers in the early warning mechanism through mobile phone APP, traffic display screens, etc.
[0020] S3 image analysis also includes behavior pattern analysis. The deep learning library needs to collect a large amount of surveillance video data samples. The data samples come from cameras in various areas of the city. The cameras cover parks, squares, commercial areas, transportation hubs, schools and hospitals. The image analysis algorithm analyzes these collected images and samples in the large database to identify individual and group behaviors in the crowd. Behaviors include walking routes, gatherings and communication methods. By comparing and analyzing crowd behavior data in different time periods and different areas, abnormal behaviors or potential social problems can be discovered. When a social event or a criminal case occurs, the system can trace back and analyze the image data captured by the surveillance camera to provide conclusive evidence and valuable clues for the investigation of the incident. After the incident, relevant investigators can use the image analysis system in S3 to trace back the surveillance video of a specific time period. The video includes detailed scenes and participant behaviors before and after the incident. The system automatically or manually analyzes the video frame by frame to find key information related to the incident. During the retrospective process, the image analysis system can identify the target object in the video and track its movement trajectory through algorithms and models, analyze the posture and action behavior characteristics of the target object, and feed the object's role and intention in the event as a structure back to the S4 early warning mechanism; the system also includes the identification and analysis of buildings and road sign elements in the scene. The early warning mechanism can extract key frames or video clips related to the event and generate a report including the timeline of the event, the behavioral characteristics of the participants and the scene information; the equipment management module includes cameras deployed in public places, which continuously transmit image data back to the back-end processing center. The image data includes people's facial expressions, body movements and social interactions. The image analysis algorithm in S3 processes the data and identifies key information related to public needs and feedback. The system analyzes people's stay time and activity frequency in public places, and analyzes which areas or services are most popular or have problems. The system analyzes people's satisfaction and feelings about public services by obtaining people's facial expressions and body language.
[0021] The S3 image analysis results are aggregated into a health scoring mechanism. Through self-service data analysis, the image analysis results are summarized, and a comprehensive evaluation is performed on the cell data in each project to achieve a visual health score. In addition, notifications / emails are automatically sent to alert operation and maintenance personnel.
[0022] The user management module, device management module, data collection and analysis module, visual data display module, and system security and performance optimization module are based on a cloud-edge collaborative architecture. This system breaks through the limitations of the centralized architecture of the existing social governance information platform, and makes full use of cutting-edge hot technologies to build a platform framework for edge computing and cloud computing collaborative modes. It uses big data cloud services and high-performance edge computing to combine edge large-scale camera image preprocessing with cloud-based deep learning neural network algorithms. It can better utilize real-time contextual information, improve the response speed and result accuracy of intelligent reasoning decisions, reduce the platform computing cost, and provide a strong guarantee for the realization of more comprehensive community governance.
[0023] Intelligent security system: Integrate video surveillance and intelligent recognition technology to achieve intelligent security monitoring, including face recognition, abnormal behavior detection and other functions to improve the safety of the community; Intelligent parking management: through license plate recognition, intelligent parking space management system and other technologies, improve the efficiency of community parking management and alleviate parking problems; Intelligent property management: Using Internet of Things technology to achieve functions such as remote monitoring of equipment, alarm processing, repair and maintenance, and improve property management efficiency; Smart community services: Establish a smart community service platform, including community activity releases, public facility reservations and other functions, to improve the convenience of residents’ lives; Intelligent information platform: Establish an information management platform, including intelligent access control system, visitor management system, etc., to improve the efficiency and security of community management; Intelligent emergency response: Establish an intelligent emergency response system, including fire warning, battery vehicle warning, etc., to improve the community's emergency response capabilities; Accurate and fine-grained recognition: This project can accurately identify abnormal behaviors or events, such as crowds, traffic accidents, fires, etc., and extract this information in real time to trigger the next early warning mechanism. After tens of thousands of event recognitions, the recognition accuracy can reach 95%, and in terms of face recognition, gas tank recognition, and fire hydrant recognition, the accuracy can reach 100%; High-level privacy and security protection: For sensitive information and system data such as face recognition, behavior analysis, and location tracking, data encryption technology, a four-level permission control system, and more than 200 detailed permission management functions are used to ensure that data is not leaked or improperly used; The system is stable and reliable: unit testing and integrated testing frameworks, asynchronous processing, multi-threading / multi-processing and other technologies are used to ensure the stability and reliability of the system. After nearly one year of operation, serving more than 10,000 people, the system runs stably and reliably.
[0024] User management module, user authentication and authorization, realize user login, registration, authorization and other functions, user authority management, study how to realize the authority control of different user roles, such as administrators, property personnel, residents, managers, etc. authority differentiation and management.
[0025] Commonly used identity authentication and permission management solutions, such as JWT, OAuth, etc., design user data models and database table structures, including user information, roles, permissions, etc. Implement user registration, login and other functions, and add corresponding security measures in the front and back ends, such as password encryption, SQL injection prevention, etc.
[0026] Equipment management module, equipment access and communication, research on how to achieve access and data communication of various types of equipment, including cameras, sensors, smart door locks, etc., equipment monitoring and control: research on how to monitor equipment status in real time, collect equipment data, and achieve remote control and management.
[0027] Commonly used IoT communication protocols and platforms, such as third-party SDK, GA1400, etc., select the appropriate scenario solution, device access and communication protocol, and consider factors such as device type, data format, and security. Implement functions such as device registration, data collection, and status monitoring, and provide corresponding APIs for front-end calls.
[0028] Data collection and analysis module, the data collection method collects various types of data in the community, including environmental data, user behavior data, etc. Data processing and analysis: study how to process and analyze a large amount of collected data and discover the correlation and rules between the data.
[0029] Determine the type and source of data that needs to be collected, such as sensor data, user operation data, etc., and select appropriate data collection equipment or technologies, such as sensors, data interfaces, and algorithms and tools for data processing and analysis, such as data cleaning, feature extraction, and machine learning models.
[0030] Visual data display modules, such as charts, maps, dashboards, etc. Customized display of relevant data based on user needs and roles. Commonly used data visualization tools and libraries, such as D3.js, ECharts, Mapbox, etc., select tools suitable for the project, design data display interfaces and interactive methods based on user needs and system functions, such as responsive design, data filtering, timeline selection, etc. The front-end page uses the selected visualization tools to implement data display functions, and connects with the back-end interface to achieve real-time data updates and interactive operations.
[0031] The system security and performance optimization module includes data encryption, access control, vulnerability patching, etc. It studies how to improve system performance, including code optimization, resource utilization optimization, cache technology application, security risk assessment, discover potential security vulnerabilities, and take corresponding security measures, such as encrypted transmission, firewall configuration, etc. It also conducts system performance testing, analyzes system bottlenecks and performance bottlenecks, and performs targeted optimization, such as code refactoring, database index optimization, cache configuration, etc.
[0032] For example, for the dangerous source of "gas tanks on the stairs", the system extracts, analyzes and detects image features through monitoring screenshots, makes a decision on whether there are gas tanks, and realizes real-time early warning on the platform, as well as actual scenarios such as face recognition, electric vehicle recognition, and fire hydrant recognition.
[0033] Implementation technology 1: Target image detection technology based on deep learning Deep Learning Architecture: Convolutional Neural Network (CNN) Deep Learning Model: Yolov5 Other architectures and technologies: python, c++, java, cuda, cudnn, TensorRT, java JNI Python script commands python export.py --weights C: / models / yolov5s.pt --include onnx --opset 10 Export script, export onnx model script original model pt format, onnx general model python train.py --img 640 --epochs 3 --data dataset.yaml --weightsyolov5x.pt Training script --epochs 100-300 is recommended python detect.py --source C: / models / 1679388393957.jpg --weights C: / models / yolov5x.onnx --img 640 --conf-thres 0.75 recognition script Labeling tools: labelme or labelimg (labelme can export yolo format labels, labelimg requires you to write your own script to change the labels to yolo format) Description: Background projects are generally Java, so Java is needed to run the YOLO model, and AI requires computing power. Given the limited CPU computing power, GPU reasoning perfectly solves the speed problem. You only need to use the DNN module of OpenCV to run the ONNX model. Each GPU manufacturer has its own driver architecture. This project uses NVIDIA's CUDA architecture. NVIDIA's deep learning reasoning framework TensorRT perfectly adapts to its own graphics card, and TensorRT does not have a Java API, so you need to use C++ to implement reasoning, and use Java's JNI to let C++ provide a Java interface. After testing, the 4060 graphics card reasoning FP16 500FPS FP32300FPSTensorRT can quantize the model, the reasoning speed is faster, and the FP16 quantization can be doubled. At present, through this framework and model, face modeling, face recognition, electric vehicle recognition, gas tank recognition, fire hydrant recognition and other functions have been realized.
[0034] Through artificial intelligence algorithms and image analysis, we can identify the temperature warning of the carport in the community; realize smoke warning through efficient sensors; realize warning of battery vehicles entering the elevator through image behavior analysis; realize trajectory analysis of high-altitude objects through image analysis; through image recognition, we can realize real-time warning of personnel leaving their posts, gas tanks entering the building, fire fighting facilities moving, fire doors not closed, road occupation, garbage dumping, etc. This project has accumulated multiple algorithm application scenarios and applied them to the smart community platform, which greatly improved the sense of security and happiness in the community.
[0035] Through the actual implementation of LLM (Large Language Model) and image recognition technology, we have innovatively realized the functions required by many scenarios, solved the industry difficulties of community management in real life, and promoted the informatization development of my country's social governance. For example, an event tracking and trace system has been established for users, so that grassroots governance can electronically record, track, review, count, and improve the events that have occurred, which has improved the informatization and modernization level of grassroots governance.
Claims
1. A community governance system combining image analysis and big data deep learning, characterized by: Including user management module, equipment management module, data collection and analysis module, visual data display module, system security and performance optimization module; return The following steps are included: S1, real-time monitoring, the system captures image data through cameras deployed in every corner of the community; S2, information transmission, image data is transmitted to the back-end processing center; S3, image analysis, analyzing abnormal behaviors or events; S4. Early warning mechanism, push notifications via SMS, email, and mobile APP, and send analysis results to the monitoring center screen.
2. The community governance system combining image analysis and big data deep learning according to claim 1 is characterized by: The abnormal behaviors or events in the image analysis in S3 include warning label events and fixed warning events.
3. The community governance system combining image analysis and big data deep learning according to claim 2 is characterized by: The fixed warning events include the entry and exit of key personnel; the warning tag events include Elderly care: testing residents aged 60 and above who have not been in or out of the community for 15 days; Frequent visitors, detecting visitors entering and leaving the community multiple times; Hiding during the day and coming out at night, the detection community personnel frequently enter and leave the community late at night; Smoke events: detect abnormal alarms issued by connected fire-fighting equipment and facilities; Suspected group rental housing: when residents in the community register, the average per capita housing area is less than 12 square meters; Boundary triggering detects alarms from connected IoT devices such as electronic fences.
4. The community governance system combining image analysis and big data deep learning according to claim 2 is characterized by: In the real-time monitoring step of the device management module in S1, the camera in the device management module will capture the face image; the S3 image analysis module includes face recognition and identity authentication. The system recognizes the face in the image taken for monitoring and compares it with the registered information in the big database to achieve identity authentication.
5. The community governance system combining image analysis and big data deep learning according to claim 2 is characterized by: The S3 image analysis module includes image preprocessing technology in which the system first denoises and enhances the image to improve the accuracy of face recognition, then enters into deep learning of big data, extracts features of the face in the image, generates face feature vectors, and then compares the extracted face feature vectors with the information in the big database. The big database contains pre-collected and registered face data, including photos of suspects and records of people entering and leaving public places. Through the comparison algorithm, the system matches the face feature vectors to confirm the identity.
6. The community governance system combining image analysis and big data deep learning according to claim 1, characterized in that: The equipment management module includes traffic cameras inside the community and at the door. Traffic cameras are deployed at key sections and traffic nodes, intersections and highway entrances and exits. The S3 image analysis includes traffic flow analysis inside and outside the community. Image analysis technology begins to analyze the received image data. Through computer vision and image recognition algorithms, image analysis can automatically extract vehicle information in the image, including the number of vehicles, vehicle models, driving speed, etc. The system can also judge the road congestion according to the vehicle's driving trajectory and density. Through real-time analysis of traffic flow, the system will feedback the information to the public and traffic managers in the early warning mechanism through mobile phone APP, traffic display screens, etc.
7. The community governance system combining image analysis and big data deep learning according to claim 1 is characterized by: The S3 image analysis also includes behavior pattern analysis. The deep learning library needs to collect a large amount of surveillance video data samples. The data samples come from cameras in various areas of the city. The cameras cover parks, squares, commercial areas, transportation hubs, schools and hospitals. The image analysis algorithm analyzes these collected images and samples in the large database to identify individual and group behaviors in the crowd. Behaviors include walking routes, gathering situations and communication methods. By comparing and analyzing crowd behavior data in different time periods and different areas, abnormal behaviors or potential social problems can be discovered.
8. The community governance system combining image analysis and big data deep learning according to claim 1, characterized in that: When a social event or criminal case occurs, the system can trace back and analyze the image data captured by the surveillance camera to provide solid evidence and valuable clues for the investigation of the incident. After the incident, the relevant investigators can use the image analysis system in S3 to trace back the surveillance video of a specific time period. The video includes detailed scenes and participants' behaviors before and after the incident. The system automatically or manually analyzes the video frame by frame to find key information related to the incident. During the tracing process, the image analysis system can identify the target object in the video and track its movement trajectory through algorithms and models, analyze the posture and action behavior characteristics of the target object, and feed back the role and intention of the object in the incident as a structure to the S4 early warning mechanism; the system also includes the identification and analysis of buildings and road sign elements in the scene. The early warning mechanism can extract key frames or video clips related to the event and generate a report including the timeline of the event, the behavioral characteristics of the participants and the scene information.
9. The community governance system combining image analysis and big data deep learning according to claim 1, characterized in that: The device management module includes cameras deployed in public places, which continuously transmit image data to the back-end processing center. The image data includes people's facial expressions, body movements and social interactions. The image analysis algorithm in S3 processes the data and identifies key information related to public needs and feedback. The system analyzes people's stay time and activity frequency in public places, and analyzes which areas or services are most popular or have problems. The system analyzes people's satisfaction and feelings about public services by obtaining people's facial expressions and body language.
10. The community governance system combining image analysis and big data deep learning according to claim 1, characterized in that: It also includes a health scoring mechanism, which summarizes the results of image analysis through self-service data analysis, conducts a comprehensive assessment of the community data in each project, achieves a visual health score, and automatically sends notifications / emails to alert operation and maintenance personnel.
Citation Information
Cited By
Community old-age care service management system and method based on risk analysis
CN120218594A
A community pension service management system and method based on risk analysis
CN120218594B
Community personnel multi-mode identification system and method based on image perception large model
CN120726674A
Community personnel multi-modal recognition system and method based on image perception large model
CN120726674B
