Sensor fault diagnosis and repair system and method for smart city

Abstract

The invention discloses a sensor fault diagnosis and repair system and method for a smart city, and belongs to the technical field of smart city operation and maintenance management, and the method comprises the steps: carrying out the real-time monitoring of a sensor, and obtaining the output data of the sensor; carrying out continuity analysis and consistency analysis on the output data to obtain a continuity analysis result and a consistency analysis result; when the continuity analysis result and the consistency analysis result do not exceed the preset range in response, judging that the sensor has no fault, otherwise, marking the sensor as a potential fault; for the sensors marked as potential faults, the sensors are triggered to perform self-inspection, and the types of the potential faults are judged; and when the type of the potential fault is a hardware fault, feeding back sensor information of the potential fault to a remote management platform. Through fault diagnosis and remote repair of the sensor, the reliability and adaptability of the sensor are effectively improved, the operation reliability and sensing precision of the whole smart city are improved, and complex requirements of the smart city are met.

Description

Technical Field

[0001] The present invention relates to a sensor fault diagnosis and repair system and method for a smart city, belonging to the technical field of smart city operation and maintenance management. Background Art

[0002] A smart city is a new form of city built with the goal of more scientific development, more efficient management and better life, supported by information technology and communication technology. It improves urban operation efficiency, improves the level of public services and forms a low-carbon urban ecosystem through transparent and full information acquisition, extensive and secure information transmission and effective and scientific information processing.

[0003] With the rapid development of smart cities, various intelligent sensing devices are widely used in real-time monitoring and management of urban environments. These smart sensors are used to collect various environmental data such as air quality, temperature and humidity, noise level, traffic flow, etc., providing the necessary basic data for the intelligent operation of cities. However, the sensor systems currently used in smart cities still face many challenges in terms of adaptability to complex environments. Summary of the invention

[0004] The purpose of the present invention is to overcome the deficiencies in the prior art and to provide a sensor fault diagnosis and repair system and method for smart cities. In view of the complex needs of smart cities, the reliability and adaptability of sensors are effectively improved through sensor fault diagnosis and remote repair, and the problems of sensor failure and output data distortion are solved.

[0005] To achieve the above object, the present invention is implemented by adopting the following technical solutions:

[0006] In a first aspect, the present invention provides a sensor fault diagnosis and repair method for a smart city, comprising:

[0007] Monitor the sensor in real time and obtain the output data of the sensor;

[0008] Perform continuity analysis and consistency analysis on the output data to obtain continuity analysis results and consistency analysis results;

[0009] In response to the continuity analysis result and the consistency analysis result both not exceeding the preset range, the sensor is determined to be fault-free, otherwise the sensor is marked as a potential fault;

[0010] For sensors marked as potential faults, trigger the sensor to perform self-check to determine the type of potential fault;

[0011] In response to the type of the potential fault being a hardware fault, feeding back sensor information of the potential fault to the remote management platform;

[0012] In response to the type of potential fault being a software fault, the sensor self-repair is initiated, and redundant sensor data is collected at the same time, and data compensation and correction are performed on the sensor with the software fault in combination with the redundant sensor data.

[0013] Furthermore, the output data is subjected to continuity analysis and consistency analysis, wherein:

[0014] The continuity analysis includes: analyzing the continuity and consistency of data through a support vector machine and a time series model based on a long short-term memory network, and the formula is:

[0015] f(X)=sign(W·φ(X)+b)

[0016] Where W is the weight vector, φ(X) is the feature mapping function, b is the bias term, and f(X) is the output value;

[0017] The consistency analysis includes: comparing the output data with the data of multiple sensors in the same area horizontally to diagnose single point failures.

[0018] Furthermore, the sensor self-repair is started, and redundant sensor data is collected at the same time, and the redundant sensor data is combined to perform data compensation and correction on the sensor with software failure, wherein:

[0019] The sensor self-repair includes: recalibrating the sensor using standard reference data, and attempting to restart the sensor;

[0020] The collecting of redundant sensor data includes: continuing to collect data through redundant sensors, and fusing data of other redundant sensors for compensation using a redundant data processing module provided in an edge computing node;

[0021] The method of combining redundant sensor data to compensate and correct the sensor with software failure includes: using a multi-source data fusion algorithm to perform complementary analysis through data cross-validation between sensors, and, in the data fusion process, using a Kalman filter to fuse and filter the measurement results of different sensors, and modeling the environmental data through a machine learning model to perform dynamic inference when some data is missing. The formula is:

[0022]

[0023] in, is the predicted state, is the updated state, K k is the Kalman gain, z k is the measurement value, H is the observation matrix;

[0024]

[0025] in, is the predicted state, A is, Yes, B yes, u k yes.

[0026] Furthermore, the remote management platform uses the Internet of Things to connect sensors, and the remote management platform performs differentiated firmware updates on sensors through OTA according to the specific needs of different types of sensors. The remote management platform monitors the long-term data of sensors and uses machine learning to predict their possible failure time. In addition, based on the geographical location, current working status and failure possibility of the sensors, the remote management platform automatically generates the optimal maintenance tasks and notifies the operation and maintenance personnel to execute them.

[0027] Furthermore, the sensor adopts a modular system architecture, and each part of the sensor's perception system is a relatively independent module, and each module has a clear function and interface definition, wherein:

[0028] The edge computing nodes of the sensor use containerization technology to deploy applications, and automatically increase or decrease computing nodes according to real-time load conditions through the automatic expansion feature of Kubernetes container orchestration. The Kubernetes container orchestration expression is:

[0029]

[0030] x i ∈{0,1}

[0031] Among them, x i Indicates whether to add an instance on node i, c i represents the cost of running an instance on node i, r i is the amount of resources provided by node i, and D is the total amount of resources required;

[0032] When the system to which the sensor is connected detects that a new device is added or an old device is removed, the network topology is automatically updated to optimize the network. The network optimization formula is:

[0033]

[0034] f(u,v)∈{0,1}

[0035] Among them, V is the node set, E is the edge set, w is the edge weight function, and f(u, v) indicates whether there is an activated connection path from node u to node v.

[0036] Furthermore, the sensor is a multimodal sensor, and the multimodal sensors form a multimodal sensor fusion system through a multimodal data fusion algorithm. The multimodal sensor fusion system deploys a fusion device of a thermal imaging camera and a visible light camera, and designs a perception mode of laser radar and video data fusion. The multimodal data fusion algorithm includes an adaptive multimodal fusion formula, which is:

[0037]

[0038] Among them, the weight α i is the environmental adaptability weight, x i * is the data provided by sensor i, x fused is the data fusion result.

[0039] Furthermore, the edge node of the sensor is deployed with an adaptive algorithm based on environmental perception, and the adaptive algorithm includes adaptive adjustment of exposure time and dynamic white balance adjustment. The adaptive adjustment formula of the exposure time is:

[0040]

[0041] Where: T is the exposure time of the camera, I is the ambient light intensity, k is the adjustment factor, and β is the smoothing constant;

[0042] The dynamic white balance adjustment formula is:

[0043] WB new =f(L,B current )

[0044] Here: WB new is the adjusted white balance setting, L is the light level obtained from the ambient light sensor, and B current is the current white balance setting.

[0045] In a second aspect, the present invention provides a sensor fault diagnosis and repair system for a smart city, comprising:

[0046] Data monitoring module: monitor the sensor in real time and obtain the output data of the sensor;

[0047] Data analysis module: performs continuity analysis and consistency analysis on the output data to obtain continuity analysis results and consistency analysis results;

[0048] Fault determination module: in response to the continuity analysis result and the consistency analysis result not exceeding the preset range, the sensor is determined to be fault-free, otherwise the sensor is marked as a potential fault;

[0049] Self-check trigger module: for sensors marked as potential faults, trigger the sensor to perform self-check to determine the type of potential faults;

[0050] Fault handling module: In response to a potential fault of a hardware fault, the sensor information of the potential fault is fed back to the remote management platform; in response to a potential fault of a software fault, the sensor self-repair is started, and redundant sensor data is collected at the same time, and the redundant sensor data is combined to perform data compensation and correction on the sensor with the software fault.

[0051] In a third aspect, the present invention provides a sensor fault diagnosis and repair device for a smart city, including a processor and a storage medium;

[0052] The storage medium is used to store instructions;

[0053] The processor is used to operate according to the instructions to execute the steps of any of the above methods.

[0054] In a fourth aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, which implements the steps of any of the above methods when executed by a processor.

[0055] Compared with the prior art, the present invention has the following beneficial effects:

[0056] 1. This solution can effectively improve the reliability and adaptability of sensors through sensor fault diagnosis and remote repair, and can better deal with problems such as sensor failure and output data distortion, thereby improving the operational reliability and perception accuracy of the entire smart city. It has significant innovations in the stability, flexibility and continuous reliability of perception equipment, and adapts to the complex needs of smart cities;

[0057] 1. This solution designs innovative data fusion and self-repair: Compared with traditional perception systems that rely on only a single sensor to collect data, this solution introduces a multi-sensor fusion algorithm, combined with technologies such as Kalman filter, to provide an intelligent dynamic compensation mechanism when data anomalies occur, ensuring data accuracy and continuity;

[0058] 2. This solution enhances adaptability to complex environments through multimodal perception. By fusing multimodal sensors (such as thermal imaging, lidar, and visible light cameras), the system's perception capabilities are greatly improved under complex environmental conditions. This multimodal fusion technology can effectively make up for the limitations of a single sensor and can still provide high-quality data at night, in fog and haze, and other conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] The accompanying drawings constituting a part of the present invention are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the accompanying drawings:

[0060] Figure 1 A flow chart of a sensor fault diagnosis and repair method for a smart city provided in Embodiment 1 of the present invention. DETAILED DESCRIPTION

[0061] The present invention will be described in detail below with reference to the accompanying drawings and in combination with embodiments. It should be noted that the embodiments and features in the embodiments of the present invention can be combined with each other without conflict.

[0062] The following detailed description is an exemplary description, which is intended to provide further detailed description of the present invention. Unless otherwise specified, all technical terms used in the present invention have the same meaning as those generally understood by those skilled in the art to which the present invention belongs. The terms used in the present invention are only for describing specific embodiments, and are not intended to limit the exemplary embodiments according to the present invention.

[0063] Embodiment 1:

[0064] Sensor equipment operates in outdoor environments for a long time, and the quality of data is easily affected by bad weather, pollution and equipment aging. Once data distortion or sensor failure occurs, the data integrity of the entire monitoring system will be seriously threatened. Traditional sensor systems usually lack effective fault diagnosis and self-repair mechanisms, resulting in the equipment being unable to recover in time when problems occur and having to rely on manual maintenance, which greatly increases operating costs.

[0065] In order to ensure that sensors in smart cities can operate normally and provide high-quality data under various environmental conditions, this embodiment proposes a sensor fault diagnosis and repair method for smart cities, and designs the following measures for data quality and sensor reliability:

[0066] Intelligent sensor fault diagnosis and self-repair mechanism: Design an intelligent sensor fault diagnosis module, based on machine learning anomaly detection technology (such as support vector machine SVM or LSTM-based time series analysis), to monitor the working status of the sensor in real time, and detect sensor failures by analyzing the continuity and consistency of the collected data. For example, when a temperature sensor does not match the data of other surrounding sensors, the self-check mechanism can be automatically triggered. In addition, by integrating a self-repair module, the sensor can be automatically calibrated, or in some cases, it can be restored to normal working state by restarting. At the same time, design a redundant mechanism, such as multiple sensors covering the same area, so that when a sensor fails, data can continue to be collected through other redundant devices to ensure the stability of the perception system. The intelligent sensor fault diagnosis module has the following features and functions:

[0067] 1. Abnormal analysis based on multi-level data:

[0068] Real-time data monitoring: Utilize the time series data generated by sensors, such as temperature, humidity, gas concentration, etc., to continuously monitor the real-time data of each sensor.

[0069] Machine learning model: The continuity and consistency of the data are analyzed through the support vector machine (SVM) and the time series model based on the long short-term memory network (LSTM). The analysis formula is:

[0070] f(X)=sign(W·φ(X)+b)

[0071] Where W is the weight vector, φ(X) is the feature mapping function, and b is the bias term. When the output value of f(X) deviates from the expected range, the model detects data anomalies and flags potential faults.

[0072] These models can learn the normal operating patterns of sensor data and identify potential failures when the data is abnormal. For example, when the data of a temperature sensor deviates from the consistency of the data of surrounding sensors, the system can automatically determine that the sensor may be faulty.

[0073] 2. Multi-level diagnosis of fault types:

[0074] Data consistency analysis: By comparing the data of multiple sensors in the same area horizontally, possible single point failures can be diagnosed. For example, by statistically analyzing the data differences between different sensors, if the data of a sensor deviates significantly from other sensors, the sensor can be marked as suspicious.

[0075] Multi-level alarm mechanism: Set different levels of alarm mechanism to grade according to the severity of the fault. For example, a minor reading anomaly can be marked as "Caution", while a significant data inconsistency will be marked as "Warning" and trigger further diagnosis.

[0076] 3. Self-check mechanism and fault feedback:

[0077] Self-check mechanism triggering: When it is detected that the data of a certain sensor is seriously inconsistent with that of other sensors, the system will automatically trigger the sensor's self-check function. The self-check may include checking the sensor's voltage, current and other parameters to confirm whether there is a hardware problem.

[0078] Fault feedback and logging: All fault and diagnostic information will be automatically recorded in the system log and fed back to the remote management platform for further analysis and management.

[0079] This solution is also equipped with self-repair and redundancy mechanisms, including:

[0080] 1. Automatic calibration and troubleshooting:

[0081] Automatic calibration function: For some software-level problems, the system can be repaired through remote firmware calibration. For example, the sensor can be recalibrated using standard reference data to ensure the accuracy of the data it returns.

[0082] Automatic restart mechanism: After detecting an abnormality, the system can automatically restart the sensor to eliminate some problems caused by software or temporary failures. The restart operation is performed under the control of the remote platform to ensure that the system returns to normal in the shortest time.

[0083] 2. Redundant data collection mechanism:

[0084] Multi-sensor coverage: Multiple types of sensors are deployed in key areas to ensure that redundant sensors can continue to collect data when a sensor fails. For example, an important air quality monitoring point can be equipped with multiple air quality sensors, as well as temperature and humidity sensors, to provide backup data when any sensor fails.

[0085] Redundant data processing: In the edge computing node, a redundant data processing module is set up. When the data of a certain sensor fails, it can be compensated by fusing the data of other redundant sensors to ensure the continuity and integrity of system perception.

[0086] Data compensation mechanism based on multi-sensor fusion: Design a multi-sensor fusion algorithm to perform data compensation and correction by combining data from multiple types of sensors (such as temperature, humidity, and gas sensors). When the data of a certain sensor is missing or distorted, the data of other sensors are used for compensation. For example, in the case of air quality sensor failure, the air quality can be estimated by a combination of temperature, humidity, and wind speed. Use Kalman filters to fuse and filter multi-source data to eliminate data noise and dynamically correct abnormal data generated by sensors. This method helps to improve the system's adaptability to sudden environmental changes. The compensation mechanism based on multi-sensor data fusion specifically includes:

[0087] 1. Multi-source data fusion:

[0088] Data complementary analysis: Design a multi-source data fusion algorithm to achieve complementary analysis through data cross-validation between sensors such as temperature, humidity, and gas concentration. For example, when a gas concentration sensor fails, the system can use the data from the temperature and humidity sensors to infer the air quality status.

[0089] Application of Kalman filter: In the data fusion process, Kalman filter is used to fuse and filter the measurement results of different sensors to improve the accuracy of data and reduce noise. This technology enables the system to provide smooth and reliable data output in the face of sudden environmental changes (such as sudden changes in weather). The formulas for fusion and filtering include:

[0090] The measurement update equation is:

[0091]

[0092] in, is the predicted state, is the updated state, K k is the Kalman gain, z k is the measured value, and H is the observation matrix. This formula is used to fuse and filter sensor data to enhance the stability and accuracy of the data.

[0093] The state prediction equation is:

[0094]

[0095] in, is the predicted state, A is, Yes, B yes, u k yes

[0096] 2. Intelligent data compensation mechanism:

[0097] Dynamic compensation algorithm: Model environmental data through machine learning models, and dynamically infer possible values ​​when some data is missing. For example, when air quality data in a certain area is missing, dynamic compensation can be performed by combining historical data and current temperature and humidity data, thereby reducing the impact of missing data on overall perception.

[0098] Remote monitoring and maintenance platform: Design a remote monitoring and maintenance platform to remotely manage sensors using IoT connections, including status monitoring, fault detection, and updates and upgrades. Through regular remote firmware updates, the functions of sensors can be enhanced, and their operating parameters can be adjusted in time when failures or performance degradation occur, thereby extending the service life of the equipment. The specific functions of the remote monitoring and maintenance platform include:

[0099] 1. Real-time status monitoring:

[0100] IoT remote connection: Using IoT technology, all sensors are connected to a remote monitoring and maintenance platform. The system can view the status of each sensor in real time, including parameters such as data collection quality, operating time, and hardware voltage.

[0101] Remote fault detection: The platform analyzes the status of the sensor and detects abnormalities in real time. By linking with the fault diagnosis module on the sensor side, the system can quickly identify potential problems with the sensor and provide suggestions for repairs.

[0102] 2.Automated firmware upgrade:

[0103] Regular remote updates: Design an automated remote update mechanism so that each sensor can perform remote firmware upgrades via OTA (Over-the-Air) to ensure that it can cope with changing environmental requirements and improve sensor functionality.

[0104] Differentiated upgrades: Based on the specific needs of different types of sensors, the platform will perform differentiated firmware updates on the devices. Through remote adjustment and upgrades, the sensor's acquisition frequency, sensitivity and other parameters are optimized to better adapt to environmental changes.

[0105] 3. Maintenance and life management:

[0106] Intelligent life prediction: By monitoring the long-term data of sensors, machine learning is used to predict their possible failure time and arrange maintenance or replacement in advance. The remaining useful life (RUL) prediction algorithm is used to calculate the remaining life of the sensor based on its operating status and historical data, and replacement or maintenance is planned in advance to reduce the risk of system interruption.

[0107] Maintenance task scheduling: The platform will automatically generate the optimal maintenance task based on the sensor's geographical location, current working status and fault possibility, and notify the operation and maintenance personnel to execute it. Through intelligent scheduling, equipment maintenance is more efficient and downtime caused by faults is reduced.

[0108] Innovative data fusion and self-repair: Compared with traditional perception systems that rely on only a single sensor to collect data, this solution introduces a multi-sensor fusion algorithm, combined with technologies such as the Kalman filter, to provide an intelligent dynamic compensation mechanism when data anomalies occur, ensuring data accuracy and continuity.

[0109] Intelligent remote management platform: Integrates remote monitoring, fault diagnosis, firmware upgrades, and life prediction functions to ensure that sensors can operate efficiently and intelligently, reducing human intervention. This intelligent management greatly reduces the cost of equipment maintenance.

[0110] Highly reliable self-repair mechanism: Through the linkage between the intelligent fault diagnosis module and the self-repair module, the system can take proactive repair measures after detecting an anomaly, such as automatic restart and recalibration, thereby improving the stability and reliability of the system and reducing the need for sensor maintenance.

[0111] Through the above technical solutions, the intelligent perception system can better deal with equipment failures and data distortion problems, and improve the operational reliability and perception accuracy of the entire smart city. This solution has significant innovations in the stability, flexibility and continuous reliability of perception equipment, and adapts to the complex needs of smart cities.

[0112] In the face of the continuous expansion of the scale of smart cities and the continuous access of new devices, existing sensor systems often lack sufficient scalability and flexibility. Due to the lack of modularity in the system architecture design, the lack of unified standardized interfaces between devices, and the difficulty in integrating devices from different manufacturers, the complexity of system expansion and upgrade has greatly increased. In addition, with the increase in the number of sensors and the amount of data, the scheduling of computing and storage resources has also become an important challenge. The traditional static resource configuration method can no longer meet the needs of high-frequency data changes in smart cities.

[0113] In order to cope with the continuous expansion of smart cities and the access of new devices, a flexible and scalable perception system architecture is designed so that the system can adapt to the increase of devices, the expansion of functions and the needs of new environments.

[0114] Modular architecture design and standardized interfaces: A modular system architecture is adopted to design each part of the perception system (such as the sensor layer, edge computing layer, and cloud management layer) as relatively independent modules, each of which has clear functions and interface definitions. In this way, when adding new devices or functions, you only need to add the corresponding modules to the existing system without making large-scale changes to the entire system. In addition, open standardized interfaces (such as RESTAPI and MQTT protocols) are used to connect different modules to ensure that newly added devices can be quickly integrated with the existing system. This design allows the system to be easily expanded and adapt to new device types and communication protocols in the future.

[0115] 1. Modular system architecture:

[0116] Independent functional modularization: The entire perception system is designed as multiple independent functional modules, including the sensor layer (responsible for data collection), the edge computing layer (responsible for local data processing) and the cloud management layer (responsible for centralized management and complex calculations of the overall system). Each module has independent functions and clear interfaces.

[0117] Module interoperability: These modules are interconnected through standardized interfaces. For example, the sensor layer can send data to the edge computing layer through the edge computing interface, while the cloud management layer collects data from the edge computing layer through the API interface for unified processing. This modular design allows each part to be updated and expanded independently. For example, the cloud management system can be replaced with new technology without affecting the operation of the sensor.

[0118] 2. Application of standardized interface:

[0119] Open standard interfaces: By using open standard interfaces such as REST API and MQTT protocol, modules can communicate with each other and share data. For example, the sensor pushes data to the edge node using the lightweight MQTT protocol, while the edge node uploads the processed data to the cloud management platform using REST API. These standardized interfaces make the system widely compatible and can access sensors and devices from different manufacturers and protocols.

[0120] Adaptive protocol conversion: Design a protocol conversion module that can convert incompatible protocols to adapt to the standard interface of the existing system when a new device or sensor is connected to the system. This protocol conversion mechanism enables the system to dynamically support new devices, even if their native protocols do not match, they can be integrated through this module.

[0121] Data collection and processing in a smart city environment are highly dynamic, and the amount of data will fluctuate greatly with the number of devices and the activity of the city. To meet this challenge, an elastic computing and storage resource management mechanism is introduced to dynamically manage computing and storage requirements.

[0122] Elastic computing and storage resources: By introducing cloud-native technologies, such as Kubernetes container orchestration, computing and storage resources in the management and scheduling system can be dynamically expanded on demand. For example, computing nodes are automatically added when the system load increases, and idle resources are released when it decreases to save energy and operating costs. In addition, by using the elastic computing model that combines edge computing with cloud computing, edge computing can dynamically add edge nodes when needed to meet local computing needs. When the system pressure is low, active edge nodes can be reduced to reduce system energy consumption.

[0123] 1. Cloud native technology and containerized management:

[0124] Kubernetes container orchestration: Use Kubernetes as a container orchestration tool to manage and schedule the computing and storage resources of the system. Kubernetes resource scheduling can be expressed as an optimization problem:

[0125]

[0126] x i ∈{0,1}

[0127] Among them, x i Indicates whether to add an instance on node i, c i represents the cost of running an instance on node i, r i is the amount of resources provided by node i, and D is the total amount of resources required. The model aims to minimize the cost while meeting the resource requirements.

[0128] For example, edge computing nodes can use containerization technology to deploy applications, and through the automatic expansion feature of Kubernetes, automatically increase or decrease computing nodes according to real-time load conditions. In this way, during peak data load periods, the system can automatically add more edge nodes to share computing pressure, and reduce resource allocation when the load decreases, achieving the purpose of optimizing performance and saving costs.

[0129] Microservice architecture: By containerizing each functional module and managing it based on a microservice architecture, different services can be scaled and updated independently. For example, data processing and model prediction services can be divided into different containers to meet different processing requirements. This ensures the system's elastic adaptability to the increasing number of sensors or processing requirements.

[0130] 2. Elastic combination of edge computing and cloud computing:

[0131] Elastic expansion of edge nodes: In response to local computing needs, an edge computing elastic node management system is introduced to add edge nodes in certain special areas (such as crowded areas) to handle real-time perception and analysis. The system can automatically and dynamically increase or decrease the number of edge computing nodes according to changes in data traffic and computing needs, ensuring that the system can operate effectively under different load conditions.

[0132] Centralized cloud resource management: The computing and storage resources in the cloud are managed through a dynamic scheduling mechanism. For example, when the computing power of the edge computing node is insufficient, some tasks can be transferred to the cloud computing for processing; when the cloud computing pressure is too high, low-priority tasks can be postponed, thereby achieving reasonable resource scheduling and optimization.

[0133] Automatic device registration and self-configuration: Design an automatic device registration and configuration system. Once a new sensor is connected to the system, it can automatically communicate with the management platform through the network, automatically register and obtain the assigned work tasks and network addresses. Through the self-configuration function, new devices can work normally after being connected without complex manual intervention, thereby greatly improving the efficiency of device expansion.

[0134] In order to adapt to the ever-expanding sensor network in smart cities, an automatic device registration and self-configuration system is designed to ensure that new devices can be seamlessly connected to the entire smart sensing system.

[0135] 1. Automatic device registration mechanism:

[0136] Self-organizing network access: Once a new sensor is connected to the system, it can automatically establish a connection with other devices in the network through self-organizing network protocols (such as Zigbee, LoRaWAN, etc.). Each newly added device will send an access request to the system by broadcasting. After receiving the request, the system management platform will verify the device and automatically complete the registration after confirming its legitimacy.

[0137] Centralized management platform registration: A centralized device management platform is designed. New devices are automatically registered through the platform and obtain unique identifiers, which are recorded in the device database for subsequent management and tracking. At the same time, the management platform automatically allocates configuration information for each device, including communication parameters, sampling frequency, etc.

[0138] 2. Automatic configuration and task allocation:

[0139] Self-configuration mechanism: Once the device is successfully registered, it will immediately obtain the corresponding work tasks and network configuration through the remote configuration system. According to the type of device and the working area, the system automatically assigns the corresponding work tasks and the node address for data upload to ensure that the new device can be put into operation quickly.

[0140] Adaptive configuration based on the environment: The system also has a set of adaptive configuration mechanisms based on the environment. When the device is first connected, the system will automatically adjust its working parameters according to the characteristics of the environment. For example, in a noisy environment, the sampling frequency of the acoustic sensor will be automatically reduced to reduce energy consumption and improve data quality. Through this adaptive mechanism, the working parameters of the new device can be intelligently optimized according to different application scenarios.

[0141] 3. Automatic topology update and maintenance:

[0142] Dynamic topology adjustment: When new devices are added or old devices are removed, the system automatically updates the network topology to ensure the optimal path for data transmission. For example, when the density of sensors in a certain area becomes too high, the system automatically optimizes the transmission path to balance the data load between nodes.

[0143] Assume that the network can be represented as a graph G = (V, E, W), where V is the node set, E is the edge set, and w is the edge weight function, which represents the connection cost. The network optimization problem can be expressed as:

[0144]

[0145] f(u,v)∈{0,1}

[0146] Among them, f(u, v) indicates whether there is an activated connection path from node u to node v. The model aims to find the lowest cost network connection method to optimize the overall data transmission efficiency.

[0147] Automatic processing of faulty nodes: When a device fails or is no longer in use, the system will automatically mark the device as offline and update the topology. The system management platform will automatically allocate more resources to other devices in the area to ensure that the monitoring function of the area will not be affected by the failure of a single device.

[0148] High modularity and standardized interfaces: The system adopts a modular design concept and uses open standardized interfaces to achieve communication between modules. This makes the entire system highly scalable and compatible, allowing for easy access to new devices from different manufacturers and types, while reducing the need for large-scale changes.

[0149] Elastic resource management and load optimization: Introducing cloud-native technologies such as Kubernetes to achieve elastic resource management and optimized scheduling, so that the system can dynamically adjust computing resources according to real-time load requirements. This approach can effectively reduce operating costs while ensuring system performance and adapt to the ever-changing data processing needs in the smart city environment.

[0150] Automated device management and adaptive configuration: The design of the automatic device registration and self-configuration system enables new devices to quickly and seamlessly join the network and automatically obtain configuration and work tasks. Combined with the function of environmental adaptive configuration, new devices can automatically optimize according to the actual working environment after access, reducing the complexity of manual operations and improving the efficiency and adaptability of deployment.

[0151] Through the above design, the system not only has high scalability and management efficiency, but also can cope with the challenges of increasing number of sensors and data traffic in smart cities, ensuring the efficiency, flexibility and reliability of the entire system. This flexible expansion mechanism and elastic computing resource management system are significantly innovative and practical in intelligent perception systems.

[0152] Existing intelligent sensor systems have limited perception capabilities under complex environmental conditions (such as at night, rainy and foggy weather, etc.). Single-type sensors are prone to data loss and noise interference in harsh environments, resulting in insufficient stability and accuracy of the system in environmental perception. These problems seriously affect the quality of intelligent decision-making based on perception data in smart cities.

[0153] In order to improve the adaptability of the system in different environments, especially the reliability and perception ability in bad weather, at night and other complex scenarios, the following technical solutions are proposed:

[0154] Multimodal fusion enhances perception capabilities: In certain environments (such as at night, rainy and foggy weather), a single sensor often faces the problem of degraded data quality. To this end, a multimodal sensor fusion system is designed. For example, in low-light conditions, the data from thermal imaging cameras and traditional cameras are combined to improve the accuracy of target recognition. In environments with severe air pollution, video data and lidar measurement data are combined to improve the overall perception of environmental conditions.

[0155] 1. Deployment of multimodal sensors

[0156] Fusion of thermal imaging and visible light: In low-light conditions (such as at night), the image quality of traditional cameras will drop significantly, making it difficult to accurately identify targets. To solve this problem, the system deploys a fusion device that combines thermal imaging cameras with visible light cameras. Thermal imaging cameras can detect infrared radiation from objects and can identify targets without lighting conditions. By fusing the two types of data, the recognition accuracy can be significantly improved at night or in low-light conditions.

[0157] LiDAR and video fusion: In haze weather, the performance of visual sensors is greatly reduced, resulting in a decrease in the clarity of video images. In order to improve the perception of the environment, the system has designed a perception mode that integrates LiDAR and video data. LiDAR can accurately detect the position of objects by measuring the reflection of laser pulses. Combined with video data, it can provide high-precision environmental perception under conditions with poor visual effects.

[0158] 2. Multimodal data fusion algorithm

[0159] Data synchronization and spatial alignment: A set of multi-sensor data synchronization and spatial alignment algorithms are designed to ensure that multiple data sources such as thermal imaging, video, and lidar are processed in the same time and space framework to achieve true multimodal fusion. The ICP (Iterative Closest Point) algorithm is used to align lidar and camera data to ensure the accuracy of data fusion.

[0160] Adaptive multimodal fusion formula: Assume that we have observation data x from different sensors 1 * ,x 2 * ...x n * , where the data provided by each sensor i is x i * , the corresponding environmental adaptability weight is α i These weights reflect the reliability and accuracy of each sensor in the current environment. Data fusion can be performed by weighted averaging, and the calculation formula is as follows:

[0161]

[0162] Among them, x fuse d is the data fusion result, weight α iThe weights are dynamically adjusted based on the sensor's performance under specific environmental conditions. For example, the weights can be adjusted based on the sensor's error rate, signal-to-noise ratio, or detection efficiency. These weights can be dynamically updated through real-time performance feedback or a predetermined environmental adaptation model. The weights are adaptively adjusted based on sensor errors to ensure reliable sensor-led fusion results in complex environments.

[0163] In a smart city environment, different external conditions (such as light changes, noisy environment, etc.) may have a significant impact on the sensor's acquisition effect. In order to ensure the perception quality in various environments, an environment-aware adaptive algorithm is designed: an adaptive algorithm based on environment perception is deployed on edge devices, such as using a light intensity sensor to detect the current ambient brightness, and automatically adjusting the camera parameters (such as ISO, exposure time) according to the brightness to ensure the shooting quality. At the same time, a sensitivity adjustment mechanism for the audio sensor is designed to dynamically adjust its sampling rate according to the background noise situation to improve the clarity and accuracy of the audio data.

[0164] 1. Ambient light perception and camera parameter self-adjustment

[0165] For the light intensity perception and camera parameter self-adjustment part, we can define the adaptive adjustment formula of exposure time in detail. Considering that the camera exposure time should be inversely proportional to the ambient light intensity, we can use the following formula:

[0166]

[0167] Where: T is the exposure time of the camera, I is the ambient light intensity, k is the adjustment factor used to adjust the baseline of the exposure time, and β is the smoothing constant used to adjust the sensitivity of the exposure time to changes in light intensity. This formula ensures that the camera can automatically adjust its exposure time to maintain image quality when the ambient light intensity changes. This solution introduces the adjustment factor and smoothing constant to achieve fine-grained adaptive illumination adjustment.

[0168] Dynamic white balance adjustment: Design a dynamic white balance adjustment mechanism to adaptively adjust the camera's white balance settings based on ambient light information to address color distortion caused by light changes and ensure true image restoration in different lighting environments. In the dynamic white balance adjustment part, although mathematical formulas are usually not directly involved, a conceptual model can be used to describe its adjustment mechanism:

[0169] WB new =f(L,B current )

[0170] Here: WB new is the adjusted white balance setting, L is the light level obtained from the ambient light sensor, and B currentis the current white balance setting and f is a function that adjusts the white balance based on light levels to ensure natural color rendering.

[0171] 2. Adaptive sampling rate adjustment of audio sensors

[0172] Noise detection and sampling rate control: In a noisy environment, audio data may be disturbed, resulting in a decrease in the quality of collected data. The system has designed a dynamic sampling rate adjustment mechanism based on noise detection. The noise detection module is deployed in the edge computing node to automatically adjust the sampling rate of the audio sensor through real-time analysis of the environmental background noise. For example, in a low-noise environment, the sampling rate is increased to capture more details; in a high-noise environment, the sampling rate is reduced to filter out noise, thereby improving the signal-to-noise ratio of the audio.

[0173] Application of environmental robustness models: In order to improve the adaptability of intelligent perception systems to various complex environments, a series of environmental robustness deep learning models have been designed. These models have added harsh environment data sets during the training process. Through data enhancement, domain adaptation and other technical means, the models have strong generalization capabilities for different environments (such as strong light, night, fog, etc.). For example, the generative adversarial network (GAN) is used to generate training data simulating harsh environments to enhance the adaptability of the model under various conditions and ensure that the model has good generalization capabilities in different harsh environments.

[0174] 1. Environmental robustness model training

[0175] Data enhancement for harsh environments: During the model training process, a large number of harsh environment data sets are introduced, such as data on strong light, night, fog, rainy days, etc., and data enhancement techniques (such as random brightness adjustment, noise addition, blur processing, etc.) are used to improve the model's adaptability to different environments.

[0176] Generate simulated data using Generative Adversarial Networks (GANs): Generate various simulated harsh environment data using Generative Adversarial Networks (GANs) to supplement the model’s training data in these scenarios. For example, GANs are used to generate simulated foggy and nighttime scenes, enabling the model to better detect and identify targets under these conditions.

[0177] 2. Domain Adaptation Technology

[0178] Domain adaptive deep learning model: Domain adaptive technology is used to enable models trained in normal environments to be adaptively applied to harsh environments. For example, through reverse adversarial training, the distribution difference between the source domain (good environment) and the target domain (harsh environment) of the model is reduced, so that the model can maintain high accuracy and robustness in different environments.

[0179] In complex environments, the computing power of traditional edge devices may not be sufficient to process large amounts of high-precision data. To this end, edge AI is designed to enhance complex scene detection: deploy AI coprocessors at the edge to enhance computing power, such as using neural network accelerators (such as NVIDIA Jetson) or Google's TPU, to improve computing power in complex scenes with low power consumption, so that the system can still ensure efficient event detection and recognition capabilities in extreme environments.

[0180] 1. Introduction of AI coprocessor

[0181] Deployment of AI coprocessors: Integrate AI coprocessors on edge computing nodes, such as NVIDIA JetsonNano or Google Edge TPU. These accelerators are designed for inference tasks of deep learning models. They can provide high computing power with low power consumption and are suitable for real-time event detection in complex environments.

[0182] Deep neural network acceleration: By running deep neural network models (such as the YOLO model for target detection and OpenPose for pose estimation) through the AI ​​coprocessor, the inference speed and recognition accuracy of edge nodes in complex scenarios can be significantly improved. For example, in areas with dense traffic, the AI ​​coprocessor can quickly detect information such as vehicle type and driving direction to ensure the system's real-time analysis of traffic conditions.

[0183] 2. Collaborative reasoning between edge and cloud

[0184] Hierarchical reasoning mechanism: In complex scenarios, edge nodes complete preliminary event detection and identification tasks through AI coprocessors, and transmit key feature data to the cloud for more detailed analysis. For example, when abnormal behavior is detected, the edge node can make a preliminary judgment on the behavior, and send possible abnormal event data to the cloud for higher-precision secondary reasoning to ensure the accuracy of detection.

[0185] Task priority scheduling: According to the urgency of different tasks and the computing power of edge nodes, the scheduling algorithm allocates reasoning tasks between the edge and the cloud, ensuring that urgent tasks are prioritized at the edge, while less urgent tasks can be deeply analyzed by the cloud. In this way, the system can effectively respond to diverse needs in complex environments.

[0186] In summary, this embodiment can realize real-time fault diagnosis and self-repair of sensor equipment, and support multi-sensor data fusion to improve data quality and system reliability. In addition, the system needs to have high scalability and environmental adaptability to cope with the ever-changing urban environment and diverse application scenarios, so as to ensure the stable operation and efficient management of the intelligent perception network in the smart city, and has the following technical effects:

[0187] 1. Dynamic compensation and repair of sensor data quality: Through intelligent fault diagnosis and self-repair mechanisms, the reliability of sensors can be effectively improved. The application of multi-sensor data fusion and Kalman filtering further improves the system's adaptability to environmental changes and data accuracy. The combination of these mechanisms enables the system to maintain data quality in various emergencies;

[0188] 2. Efficient system expansion and resource management: Modular design, automatic device registration and configuration make system expansion more flexible and efficient. Cloud-native resource scheduling and elastic management can dynamically adjust computing resources on demand and optimize system performance. This automated expansion method significantly improves the system's ability to cope with changes in city scale;

[0189] 3. Multimodal perception enhances adaptability to complex environments: Through the fusion of multimodal sensors (such as thermal imaging, lidar, and visible light cameras), the system's perception ability is greatly improved under complex environmental conditions. This multimodal fusion technology can effectively make up for the limitations of a single sensor and can still provide high-quality data at night, in fog and haze, etc.

[0190] 4. Environmental perception adaptive capability: Based on the adaptive adjustment mechanism of ambient light perception and audio noise analysis, the sensor can dynamically adjust its working parameters to ensure the perception effect under various external conditions. This adaptive mechanism can effectively improve the working stability of the system and the accuracy of the data;

[0191] 5. Environmental robustness deep learning model: Through data enhancement, GAN to generate harsh environment data and domain adaptation technology, the system can maintain good detection results in different harsh environments. This innovative model training method enables the system to be widely used in diverse urban environments;

[0192] 6. Enhancement of edge AI coprocessors: AI coprocessors are introduced to improve the computing power of edge nodes, ensuring that the system can perform real-time analysis and decision-making in complex scenarios. This layered reasoning mechanism between the edge and the cloud enables the system to respond to real-time needs while ensuring the depth and accuracy of analysis.

[0193] Through the above technical solutions, the intelligent perception system has a higher ability to adapt to complex environments, can effectively cope with various harsh environmental conditions, and provide high-quality perception and event detection. This solution has significant innovation and application value in enhancing perception capabilities, dynamic adaptive adjustment, and AI applications in complex environments.

[0194] Embodiment 2:

[0195] A sensor fault diagnosis and repair system for a smart city can implement a sensor fault diagnosis and repair method for a smart city described in Embodiment 1, including:

[0196] Data monitoring module: monitor the sensor in real time and obtain the output data of the sensor;

[0197] Data analysis module: performs continuity analysis and consistency analysis on the output data to obtain continuity analysis results and consistency analysis results;

[0198] Fault determination module: in response to the continuity analysis result and the consistency analysis result not exceeding the preset range, the sensor is determined to be fault-free, otherwise the sensor is marked as a potential fault;

[0199] Self-check trigger module: for sensors marked as potential faults, trigger the sensor to perform self-check to determine the type of potential faults;

[0200] Fault handling module: In response to a potential fault of a hardware fault, the sensor information of the potential fault is fed back to the remote management platform; in response to a potential fault of a software fault, the sensor self-repair is started, and redundant sensor data is collected at the same time, and the redundant sensor data is combined to perform data compensation and correction on the sensor with the software fault.

[0201] Embodiment three:

[0202] The embodiment of the present invention further provides a sensor fault diagnosis and repair device for a smart city, which can implement the sensor fault diagnosis and repair method for a smart city described in the first embodiment, including a processor and a storage medium;

[0203] The storage medium is used to store instructions;

[0204] The processor is used to operate according to the instructions to perform the steps of the following method:

[0205] Monitor the sensor in real time and obtain the output data of the sensor;

[0206] Perform continuity analysis and consistency analysis on the output data to obtain continuity analysis results and consistency analysis results;

[0207] In response to the continuity analysis result and the consistency analysis result both not exceeding the preset range, the sensor is determined to be fault-free, otherwise the sensor is marked as a potential fault;

[0208] For sensors marked as potential faults, trigger the sensor to perform self-check to determine the type of potential fault;

[0209] In response to the type of the potential fault being a hardware fault, feeding back sensor information of the potential fault to the remote management platform;

[0210] In response to the type of potential fault being a software fault, the sensor self-repair is initiated, and redundant sensor data is collected at the same time, and data compensation and correction are performed on the sensor with the software fault in combination with the redundant sensor data.

[0211] Embodiment 4:

[0212] The embodiment of the present invention further provides a computer-readable storage medium, which can implement the sensor fault diagnosis and repair method for smart cities described in the first embodiment, and stores a computer program thereon, which implements the steps of the following method when the program is executed by a processor:

[0213] Monitor the sensor in real time and obtain the output data of the sensor;

[0214] Perform continuity analysis and consistency analysis on the output data to obtain continuity analysis results and consistency analysis results;

[0215] In response to the continuity analysis result and the consistency analysis result both not exceeding the preset range, the sensor is determined to be fault-free, otherwise the sensor is marked as a potential fault;

[0216] For sensors marked as potential faults, trigger the sensor to perform self-check to determine the type of potential fault;

[0217] In response to the type of the potential fault being a hardware fault, feeding back sensor information of the potential fault to the remote management platform;

[0218] In response to the type of potential fault being a software fault, the sensor self-repair is initiated, and redundant sensor data is collected at the same time, and data compensation and correction are performed on the sensor with the software fault in combination with the redundant sensor data.

[0219] It is known from common technical knowledge that the present invention can be implemented by other embodiments that do not deviate from its spirit or essential features. Therefore, the above disclosed embodiments are only illustrative in all respects and are not exclusive. All changes within the scope of the present invention or within the scope equivalent to the present invention are included in the present invention.

[0220] It will be appreciated by those skilled in the art that embodiments of the present invention may be provided as methods, systems or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0221] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems) and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the process in the flowchart. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0222] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0223] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0224] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the relevant field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents. Any modification or equivalent replacement that does not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A sensor fault diagnosis and repair method for smart cities, characterized in that: include: Monitor the sensor in real time and obtain the output data of the sensor; Perform continuity analysis and consistency analysis on the output data to obtain continuity analysis results and consistency analysis results; In response to the continuity analysis result and the consistency analysis result both not exceeding the preset range, the sensor is determined to be fault-free, otherwise the sensor is marked as a potential fault; For sensors marked as potential faults, trigger the sensor to perform self-check to determine the type of potential fault; In response to the type of the potential fault being a hardware fault, feeding back sensor information of the potential fault to the remote management platform; In response to the type of potential fault being a software fault, the sensor self-repair is initiated, and redundant sensor data is collected at the same time, and data compensation and correction are performed on the sensor with the software fault in combination with the redundant sensor data.

2. The sensor fault diagnosis and repair method for smart city according to claim 1 is characterized in that: Perform continuity and consistency analysis on the output data, including: The continuity analysis includes: analyzing the continuity and consistency of data through a support vector machine and a time series model based on a long short-term memory network, and the formula is: f(X)=sign(W·φ(X)+b) Where W is the weight vector, φ(X) is the feature mapping function, b is the bias term, and f(X) is the output value; The consistency analysis includes: comparing the output data with the data of multiple sensors in the same area horizontally to diagnose single point failures.

3. The sensor fault diagnosis and repair method for smart city according to claim 1 is characterized in that: Start sensor self-repair and collect redundant sensor data at the same time. Combine the redundant sensor data to perform data compensation and correction on the sensor with software failure, including: The sensor self-repair includes: recalibrating the sensor using standard reference data, and attempting to restart the sensor; The collecting of redundant sensor data includes: continuing to collect data through redundant sensors, and fusing data of other redundant sensors for compensation using a redundant data processing module provided in an edge computing node; The method of combining redundant sensor data to compensate and correct the sensor with software failure includes: using a multi-source data fusion algorithm to perform complementary analysis through data cross-validation between sensors, and, in the data fusion process, using a Kalman filter to fuse and filter the measurement results of different sensors, and modeling the environmental data through a machine learning model to perform dynamic inference when some data is missing. The formula is: in, is the predicted state, is the updated state, K k is the Kalman gain, z k is the measurement value, H is the observation matrix; Among them, A is the state transfer matrix, which is used to describe the change of the system state from time k-1k-1k-1 to time kkk, reflecting the dynamic process of the system. Yes, B is the control matrix, describing the control input u k Impact on system status, u k It is the control input, which represents the external control amount on the system and is introduced during the state update process.

4. The sensor fault diagnosis and repair method for smart city according to claim 1 is characterized in that: The remote management platform uses the Internet of Things to connect sensors, and the remote management platform performs differentiated firmware updates on sensors through OTA according to the specific needs of different types of sensors. The remote management platform monitors the long-term data of sensors and uses machine learning to predict their possible failure time. In addition, based on the geographical location, current working status and failure possibility of the sensors, the remote management platform automatically generates the optimal maintenance tasks and notifies the operation and maintenance personnel to execute them.

5. The sensor fault diagnosis and repair method for smart city according to claim 1 is characterized in that: The sensor adopts a modular system architecture. Each part of the sensor's perception system is a relatively independent module. Each module has a clear function and interface definition, where: The edge computing nodes of the sensor use containerization technology to deploy applications, and automatically increase or decrease computing nodes according to real-time load conditions through the automatic expansion feature of Kubernetes container orchestration. The Kubernetes container orchestration expression is: x i ∈{0,1} Among them, x i Indicates whether to add an instance on node i, c i represents the cost of running an instance on node i, r i is the amount of resources provided by node i, and D is the total amount of resources required; When the system to which the sensor is connected detects that a new device is added or an old device is removed, the network topology is automatically updated to optimize the network. The network optimization formula is: f(u,v)∈{0,1} Among them, V is the node set, E is the edge set, w is the edge weight function, and f(u, v) indicates whether there is an activated connection path from node u to node v.

6. The sensor fault diagnosis and repair method for smart city according to claim 1 is characterized in that: The sensor is a multimodal sensor, and the multimodal sensors form a multimodal sensor fusion system through a multimodal data fusion algorithm. The multimodal sensor fusion system deploys a fusion device of a thermal imaging camera and a visible light camera, and designs a perception mode of laser radar and video data fusion. The multimodal data fusion algorithm includes an adaptive multimodal fusion formula, which is: Among them, the weight α i is the environmental adaptability weight, x i * is the data provided by sensor i, x fused is the data fusion result.

7. The sensor fault diagnosis and repair method for smart city according to claim 1 is characterized in that: The edge node of the sensor is deployed with an adaptive algorithm based on environmental perception. The adaptive algorithm includes adaptive adjustment of exposure time and dynamic white balance adjustment. The adaptive adjustment formula of exposure time is: Where: T is the exposure time of the camera, I is the ambient light intensity, k is the adjustment factor, and β is the smoothing constant; The dynamic white balance adjustment formula is: WB new =f(L,B current ) Here: WB new is the adjusted white balance setting, L is the light level obtained from the ambient light sensor, and B current is the current white balance setting.

8. A sensor fault diagnosis and repair system for smart cities, characterized in that: include: Data monitoring module: monitor the sensor in real time and obtain the output data of the sensor; Data analysis module: performs continuity analysis and consistency analysis on the output data to obtain continuity analysis results and consistency analysis results; Fault determination module: in response to the continuity analysis result and the consistency analysis result not exceeding the preset range, the sensor is determined to be fault-free, otherwise the sensor is marked as a potential fault; Self-check trigger module: for sensors marked as potential faults, trigger the sensor to perform self-check to determine the type of potential faults; Fault processing module: in response to the type of potential fault being a hardware fault, feeding back the sensor information of the potential fault to the remote management platform; In response to the type of potential fault being a software fault, the sensor self-repair is initiated, and redundant sensor data is collected at the same time, and data compensation and correction are performed on the sensor with the software fault in combination with the redundant sensor data.

9. A sensor fault diagnosis and repair device for smart cities, characterized in that: including processor and storage medium; The storage medium is used to store instructions; The processor is configured to operate according to the instructions to execute the steps of the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images