Bridge crack real-time monitoring method based on edge calculation

Abstract

The invention relates to the technical field of bridge monitoring, in particular to a bridge crack real-time monitoring method based on edge calculation, which comprises the following steps of: acquiring bridge crack data by using a hyperspectral imaging camera, a thermal imaging device and a laser radar, adaptively adjusting edge calculation nodes by using an edge learning technology, and calculating the edge calculation nodes according to the edge learning technology. A deep learning model is used for crack analysis, an anomaly monitoring model is established, an augmented reality technology is introduced to assist field investigation, and through comprehensive application of technologies such as multi-source data acquisition, edge calculation, deep learning and augmented reality, the problems that a traditional bridge crack monitoring method is low in efficiency, low in precision, incapable of real-time monitoring and the like are solved. The system has the advantages of comprehensive data acquisition, real-time and efficient processing, strong adaptive ability, comprehensive abnormity monitoring, high system reliability and visual and efficient field investigation.

Description

Technical Field

[0001] The present invention relates to the technical field of bridge monitoring, and in particular to a real-time monitoring method for bridge cracks based on edge computing. Background Art

[0002] At present, with the aging of infrastructure and the influence of environmental factors, the safety and durability of bridge structures are facing severe challenges. Bridge cracks are a common structural damage that may lead to the degradation of bridge performance or even cause safety accidents. Therefore, real-time and accurate monitoring of bridge cracks is of great significance.

[0003] Traditional bridge crack monitoring methods mainly rely on manual inspections, which have problems such as low efficiency, low accuracy, and inability to monitor in real time. With the development of sensor technology and computer vision technology, some automated bridge crack monitoring methods have been proposed. These methods usually use image processing technology to analyze the bridge surface, but there are still some limitations.

[0004] First, a single data collection method is difficult to fully reflect the condition of bridge cracks. It is difficult to capture some hidden cracks and obtain the depth information of cracks by relying solely on visible light images. Secondly, the traditional centralized processing method has problems such as data transmission delay and insufficient computing resources, which makes it difficult to meet the needs of real-time monitoring. Furthermore, existing crack detection methods are often optimized for specific scenarios, lack of adaptive capabilities, and difficult to cope with complex and changing bridge environments.

[0005] In addition, existing technologies also have deficiencies in data fusion, abnormal monitoring, system reliability, etc. For example, there is a lack of effective fusion methods for multi-source heterogeneous data, making it difficult to fully utilize the advantages of different sensors; there is a lack of comprehensive abnormal monitoring mechanism for bridge structure status; the reliability and fault tolerance of the system need to be improved. At the same time, on-site surveys still mainly rely on the experience of engineers and lack intuitive and efficient auxiliary tools.

[0006] Therefore, a real-time monitoring method for bridge cracks based on edge computing is needed to solve the above problems. Summary of the invention

[0007] The present invention provides a real-time monitoring method for bridge cracks based on edge computing, which solves the problems of low efficiency, low accuracy, and inability to monitor in real time in traditional bridge crack monitoring methods. It has the advantages of comprehensive data collection, real-time and efficient processing, strong adaptability, comprehensive abnormal monitoring, high system reliability, and intuitive and efficient on-site investigation.

[0008] The technical solution adopted by the present invention to solve the above technical problems is: a real-time monitoring method for bridge cracks based on edge computing, including: S1: using a hyperspectral imaging camera, a thermal imaging device and a laser radar to collect bridge crack data, obtain bridge crack data and transmit the bridge crack data to an edge computing node; S2: based on the bridge crack data, using edge learning technology to adaptively adjust the edge computing node, including dynamic resource allocation and intelligent task scheduling; S3: using annotated bridge crack image data sets to train a deep learning model, and adaptively processing the deep learning model through meta-learning technology and a model compression method; S4: extracting features from the bridge crack data, using a fusion method to fuse the collected bridge crack data, and then inputting the fused bridge crack data into the deep learning model, and finally outputting the judgment and analysis results of the bridge cracks; S5: based on the judgment and analysis results output by S4, establishing an abnormal monitoring model to further monitor the abnormality of the structural state of the bridge.

[0009] Furthermore, the use of hyperspectral imaging, thermal imaging devices and lidar to collect bridge crack data in S1 includes: using a hyperspectral camera to collect hyperspectral images of the bridge surface; using a thermal imaging device to obtain a temperature distribution map of the bridge surface; scanning the bridge surface by lidar to generate three-dimensional point cloud data; and the collected hyperspectral images, temperature distribution maps, and three-dimensional point cloud data constitute a bridge crack data set.

[0010] Furthermore, in S2, based on the bridge crack data, edge learning technology is used to adaptively adjust the edge computing nodes, including dynamic resource allocation and intelligent task scheduling. In edge deployment, computing resources are dynamically adjusted to cope with uneven data loads.

[0011] Furthermore, the real-time monitoring method for bridge cracks is implemented based on a real-time monitoring system for bridge cracks. The real-time monitoring system for bridge cracks adopts a distributed architecture, including: distributed edge computing nodes, a cloud platform and a central server; the distributed edge computing nodes are distributed at different locations of the bridge, and are responsible for receiving and processing bridge crack data acquired by hyperspectral imaging cameras, thermal imaging devices and lidars; the cloud platform is used to process large-scale data uploaded by distributed edge computing nodes; the central server centrally manages task scheduling and data storage of all edge computing nodes.

[0012] Furthermore, the real-time monitoring system for bridge cracks adopts reinforcement learning to optimize the network structure configuration, and the method of optimizing the network structure configuration by reinforcement learning includes: a. using simulated annealing and genetic algorithm for optimization, and finding a near-optimal network structure through simulated annealing; b. using genetic algorithm and simulation to automatically adjust the resource allocation of the network; c. for the detection of anomalies in the real-time monitoring system for bridge cracks, using machine learning models to quickly detect anomalies in real time; the anomalies include system performance, availability, security and resource utilization; d. using exception handling strategies to handle the detected anomalies, including: when an anomaly is detected, automatically selecting an exception handling strategy according to the type of anomaly, severity level and historical processing records.

[0013] Furthermore, it also includes S6: redundancy configuration for edge computing nodes, including: designing a redundant backup mechanism in the edge computing nodes, synchronously storing the monitoring system data and deep learning model parameters among multiple nodes to ensure that when one of the nodes fails, the other nodes are still available; the edge computing nodes adopt a high-speed network redundant backup design to ensure the reliability of the communication link.

[0014] Furthermore, the data collaboration and data sharing between the edge computing nodes are realized based on the federated learning technology. The federated learning technology adopts the following method: each edge node has its own local data set, parameter exchange and gradient aggregation are performed between edge nodes, data processing and analysis tasks are performed between edge nodes, and data calculation and collaboration are performed with the assistance of cloud servers.

[0015] Furthermore, it also includes S7: introducing augmented reality technology to assist engineering personnel in conducting on-site surveys.

[0016] Furthermore, the augmented reality technology assists engineering personnel in conducting on-site surveys, including: S71: collecting on-site condition data of bridge cracks; S72: making an augmented reality model and integrating it with the real scene; S73: using augmented reality technology to display the augmented reality effect through the terminal device; S74: engineering personnel determine the on-site condition of the bridge cracks through the augmented reality effect and conduct real-time on-site surveys; S75: after determining the on-site condition of the cracks, upload the data to the edge computing node.

[0017] Furthermore, in S74, the staff uses AR glasses to superimpose the real scene on the virtual reality scene to assist in real-time on-site investigation, including the location of bridge cracks, morphological assessment and recording of crack severity.

[0018] The advantages of the present invention are: The present invention provides a real-time monitoring method for bridge cracks based on edge computing, including using a hyperspectral imaging camera, a thermal imaging device and a lidar to collect bridge crack data, using edge learning technology to adaptively adjust edge computing nodes, using a deep learning model to analyze cracks, establishing an abnormal monitoring model, and introducing augmented reality technology to assist on-site investigation. Through the comprehensive application of multi-source data collection, edge computing, deep learning and augmented reality technologies, the method solves the problems of low efficiency, low accuracy and inability to monitor in real time in traditional bridge crack monitoring methods. The method has the advantages of comprehensive data collection, real-time and efficient processing, strong adaptability, comprehensive abnormal monitoring, high system reliability and intuitive and efficient on-site investigation. DETAILED DESCRIPTION

[0019] The technical solution of the present invention is described clearly and completely below. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0020] Embodiment 1: The present invention provides a real-time monitoring method for bridge cracks based on edge computing. At present, when data is collected during the traditional bridge crack monitoring process, the data processing and analysis efficiency is low due to the large scale and complex structure of the collected data. Data acquisition equipment is distributed at different locations of the bridge, and data transmission needs to pass through multiple nodes, which increases the pressure on the network bandwidth and makes it difficult to ensure real-time performance. In addition, during the data processing process, the unreasonable allocation of computing resources leads to a reduced response speed of the monitoring system, and the problem of bridge cracks cannot be discovered and warned in time. Specifically, in the crack monitoring system of a cross-river bridge, the hyperspectral imaging camera generates 500MB of image data per second, the thermal imaging device generates 200MB of temperature distribution data per second, and the laser radar generates 300MB of point cloud data per second. These data need to be transmitted to a remote server in real time for processing, but the existing network bandwidth is only 1Gbps, which cannot meet the data transmission requirements, resulting in a serious reduction in the real-time performance of the monitoring data. At the same time, when the centralized remote server processes data from multiple monitoring points, it often suffers from overload, causing the response time of the monitoring system to exceed the preset 300ms threshold. The present application proposes a real-time monitoring method for bridge cracks based on edge computing. This method aims to optimize the data processing process, reduce network bandwidth pressure, and reasonably allocate computing resources through edge computing technology. In the data acquisition stage, hyperspectral imaging cameras, thermal imaging devices, and lidar are used to obtain bridge crack data. These devices collect hyperspectral images, temperature distribution maps, and three-dimensional point cloud data on the bridge surface, respectively, and transmit the data to the nearest edge computing node for preliminary processing. The adaptive adjustment of edge computing nodes is one of the key innovations of this method. Through edge learning technology, the monitoring system can dynamically allocate computing resources and perform intelligent task scheduling. Edge computing nodes can automatically adjust computing resources according to real-time data flow and processing requirements, thereby effectively solving the problem of unreasonable computing resource allocation. In the data processing stage, the collected multi-source data is input into the optimized deep learning model after feature extraction and fusion, and the judgment and analysis results of bridge cracks are output. The training and optimization of deep learning models are the key to improving the accuracy of crack detection. This method uses annotated bridge crack image datasets to train deep learning models, and uses meta-learning technology to enable the model to quickly adapt to new crack types or environmental changes.

[0021] Based on the output judgment and analysis results, this method establishes an abnormal monitoring model to further monitor the structural status of the bridge. This model can predict the development trend of cracks based on historical data and current status, and trigger an alarm when the cracks reach a preset threshold.

[0022] The method of the present application significantly reduces the amount of data transmission and improves the real-time performance of the system by performing data processing and analysis at the edge node. At the same time, the dynamic resource allocation mechanism of the edge node can flexibly adjust the computing resources according to the changes in the amount of data, ensuring that the system response time always remains within the preset threshold. By implementing the method of the present application, the data processing efficiency of the bridge crack monitoring system is significantly improved, the network bandwidth pressure is greatly reduced, and the computing resource allocation is more reasonable, which helps to timely discover and warn potential structural safety hazards.

[0023] The present application further proposes a method for collecting bridge crack data using hyperspectral imaging, thermal imaging devices, and lidar. Specifically, the present application uses a hyperspectral camera to collect hyperspectral images of the bridge surface. Hyperspectral imaging technology can capture subtle spectral changes that cannot be recognized by the human eye, which helps to find early or hidden cracks. Secondly, a thermal imaging device is used to obtain a temperature distribution map of the bridge surface. Abnormal temperature distribution may indicate potential structural problems or crack locations. Finally, the bridge surface is scanned by lidar to generate three-dimensional point cloud data. This high-precision three-dimensional data can accurately locate and measure the geometric characteristics of cracks. These three data collection methods have their own advantages and complement each other. Hyperspectral images provide rich spectral information, which helps to identify different types of cracks and material changes. Temperature distribution maps can reveal thermal anomalies caused by cracks, especially at night or during periods of large temperature changes. Three-dimensional point cloud data provides accurate geometric information that can be used to quantify the width, depth, and length of cracks. The collected hyperspectral images, temperature distribution maps, and three-dimensional point cloud data constitute a comprehensive bridge crack data set. This multimodal data set provides a solid foundation for subsequent analysis and monitoring. For example, hyperspectral data can be fused with temperature data to identify potential crack areas that are not obvious under visible light but have temperature anomalies. At the same time, 3D point cloud data can be used to verify and accurately locate anomalies found in the other two types of data. Furthermore, this multi-source data collection method can also be adjusted according to specific bridge structures and environmental conditions. For example, for large bridges, drones can be used to carry out comprehensive scans with these sensors. For underwater bridge piers, underwater sonar technology can be considered to supplement data collection.

[0024] As a preferred implementation, an automated data collection system can be designed. The system can automatically start the data collection process periodically or as needed, and transmit the collected data to the edge computing node in real time for preliminary processing and analysis. This automated method can greatly improve the efficiency and frequency of data collection, thereby achieving continuous and real-time monitoring of bridge cracks.

[0025] Therefore, by comprehensively utilizing hyperspectral imaging, thermal imaging, and lidar technology, the present application can obtain more comprehensive and accurate bridge crack data. This multimodal data acquisition method not only improves the quality and reliability of the data, but also provides a richer source of information for subsequent data analysis and crack monitoring. Compared with the traditional single data acquisition method, the method of the present application can detect potential crack problems earlier and more accurately assess the severity of cracks, thereby providing stronger support for bridge maintenance and safety management.

[0026] The present application further proposes to use edge learning technology to adaptively adjust edge computing nodes based on bridge crack data, including dynamic resource allocation and intelligent task scheduling. In edge deployment, computing resources are dynamically adjusted to cope with uneven data loads. The technical solution proposed in this application aims to solve the problem of low resource utilization efficiency of edge computing nodes. By introducing edge learning technology, adaptive adjustment of edge computing nodes is achieved, thereby improving resource utilization efficiency and better coping with data load fluctuations during bridge crack monitoring. Specifically, the technical solution of this application includes the following key features: First, edge learning is performed based on bridge crack data. This step can establish a data load model by analyzing historical data and real-time data to predict future data processing needs. Secondly, edge learning technology is used to adaptively adjust edge computing nodes. This process can optimize resource allocation strategies through machine learning algorithms such as reinforcement learning or deep learning. Thirdly, dynamic resource allocation is performed. According to changes in data load, the system can adjust the allocation of computing resources in real time. For example, computing resources are increased when the data volume is large, and excess resources are released when the data volume is small. Finally, intelligent task scheduling is achieved. The system can dynamically adjust the execution order and allocation method of tasks according to the priority, complexity and current resource status of the tasks. These features are interrelated and work together to achieve efficient operation of edge computing nodes. Edge learning technology provides decision-making basis for dynamic resource allocation and intelligent task scheduling, which in turn provide more optimization space for edge learning.

[0027] In practical applications, the technical solution of the present application can significantly improve the resource utilization efficiency of edge computing nodes. For example, when the bridge crack data acquisition equipment generates a large amount of data in a certain period of time, the system can quickly increase the computing resources of the corresponding nodes and give priority to processing these data. On the contrary, in periods with less data, the system can allocate excess resources to other tasks or enter low-power mode, thereby achieving rational utilization of resources.

[0028] In addition, the technical solution of the present application can also effectively cope with uneven data loads. In the process of bridge crack monitoring, data collection may experience sudden and periodic fluctuations. Through dynamic resource allocation and intelligent task scheduling, the monitoring system can better adapt to such fluctuations and ensure the continuity and accuracy of monitoring. As a preferred implementation, the present application can adopt the following specific steps to achieve adaptive adjustment of edge computing nodes: Data analysis: Continuously collect and analyze bridge crack data, including indicators such as data volume, data type, and processing time. Load prediction: Based on historical data and current trends, use time series analysis or machine learning models to predict future data loads. Resource evaluation: Real-time monitoring of resource usage of edge computing nodes, including CPU, memory, storage, and network bandwidth. Dynamic resource allocation: Dynamically adjust the computing resources of each node based on load prediction and resource evaluation results. This may include operations such as the creation or destruction of virtual machines and the expansion and contraction of containers. Task prioritization: Prioritize the tasks to be processed based on the urgency, importance, and resource requirements of the tasks. Intelligent task scheduling: Consider task priority, node load, and network conditions, and assign tasks to the most suitable edge computing nodes for execution. Performance monitoring and feedback: Continuously monitor system performance, including indicators such as task completion time and resource utilization, and feed this information back to the edge learning model to further optimize the decision-making process. In this way, the technical solution of this application can achieve efficient operation of edge computing nodes, improve resource utilization, and flexibly respond to changes in data load, thereby providing more reliable computing support for real-time monitoring of bridge cracks.

[0029] Compared with the prior art, the technical solution of the present application has the following advantages: Dynamic adaptability: Compared with the static resource allocation method, the present application can dynamically adjust resources according to the real-time data load, thereby improving the flexibility and adaptability of the system. Intelligent decision-making: By introducing edge learning technology, the present application can make more intelligent resource allocation and task scheduling decisions based on historical data and current conditions, which is more efficient than the traditional rule-based method. Improved resource utilization: Through dynamic resource allocation and intelligent task scheduling, the present application can significantly improve the resource utilization of edge computing nodes and reduce resource waste. Performance optimization: Through more reasonable resource allocation and task scheduling, the present application can reduce task processing delays and improve the overall performance of the system. Scalability: The technical solution of the present application is easy to expand and can adapt to bridge crack monitoring systems of different scales and complexities. In general, by introducing edge learning technology, the present application realizes the adaptive adjustment of edge computing nodes, effectively solves the problem of inefficient resource utilization, and provides more reliable and efficient computing support for real-time monitoring of bridge cracks.

[0030] The real-time monitoring method for bridge cracks based on edge computing in this application is implemented based on a real-time monitoring system for bridge cracks. The real-time monitoring system for bridge cracks adopts a distributed architecture, including distributed edge computing nodes, a cloud platform and a central server. The distributed edge computing nodes are distributed at different locations of the bridge (piers, main beams, supports, bridge decks, abutments), and are responsible for receiving and processing bridge crack data acquired by hyperspectral imaging cameras, thermal imaging devices and lidars. The cloud platform is used to process large-scale data uploaded by distributed edge computing nodes. The central server centrally manages the task scheduling and data storage of all edge computing nodes. The design of this distributed architecture is intended to solve the problems of data processing efficiency and system reliability in real-time monitoring of bridge cracks. By deploying edge computing nodes at different locations of the bridge, data can be processed nearby, data transmission delays can be reduced, and system response speed can be improved. At the same time, the introduction of the cloud platform enables the system to process large-scale data and provide support for the training and optimization of deep learning models. The setting of the central server ensures unified management and coordination of the entire system. Specifically, the distributed edge computing nodes can use high-performance embedded devices or small servers, which are equipped with sufficient computing power and storage space to process data from hyperspectral imaging cameras, thermal imaging devices and lidars in real time. Each node can independently perform tasks such as data preprocessing and feature extraction, thereby reducing the burden on the central server. The cloud platform can adopt an elastic computing architecture to dynamically adjust computing resources according to data processing needs. For example, when the amount of data is large, the computing nodes can be automatically expanded; when the amount of data is small, the excess resources can be released to achieve efficient use of resources. As the brain of the entire system, the central server is responsible for formulating a global task scheduling strategy. It can dynamically adjust task allocation according to factors such as the load of each edge node and the network status to ensure the overall performance of the system is optimal. At the same time, the central server also assumes the function of data storage, and can use distributed database technology to ensure the security and reliability of data. In practical applications, the real-time monitoring system for bridge cracks of this application can operate as follows: First, the bridge surface data is continuously collected through a hyperspectral imaging camera, a thermal imaging device and a lidar. These data will be transmitted to the nearest edge computing node. After receiving the data, the edge computing node immediately performs preliminary processing, including data cleaning, feature extraction, etc. Part of the processed data is stored locally, and part is uploaded to the cloud platform. The cloud platform receives data from each edge node and performs large-scale data analysis and model training. The trained model can be sent to each edge node again to achieve continuous optimization of the model. The central server is responsible for monitoring the operating status of the entire system. It can view the workload of each edge node in real time and adjust the task allocation as needed. For example, if the load of an edge node is too high, the central server can allocate some tasks to nodes with lower load to ensure the balanced operation of the monitoring system.Through this distributed architecture, the real-time monitoring system for bridge cracks of the present application can achieve efficient data processing and analysis. The local processing of edge computing nodes reduces the delay in data transmission and improves the real-time performance of the monitoring system. The introduction of the cloud platform provides powerful computing support for complex data analysis and model training. The unified management of the central server ensures the coordinated operation of the entire monitoring system. Compared with the traditional centralized architecture, the distributed architecture of the present application has higher scalability and reliability. Even if an edge node fails, other nodes can continue to work without affecting the operation of the entire system. In addition, this architecture is also more suitable for processing large-scale real-time data, and can better meet the needs of real-time monitoring of bridge cracks.

[0031] The real-time monitoring system for bridge cracks in this application adopts reinforcement learning to optimize the network structure configuration. The method based on reinforcement learning to optimize the network structure configuration includes using simulated annealing and genetic algorithm for optimization, finding a near-optimal network structure through simulated annealing; using genetic algorithm and simulation to automatically adjust the resource allocation of the network; using machine learning model to detect abnormalities in the real-time monitoring system for bridge cracks, and using abnormality handling strategy to handle the monitored abnormalities. This reinforcement learning optimization network structure configuration method of this application aims to improve the performance and reliability of the real-time monitoring system for bridge cracks. By combining advanced technologies such as simulated annealing, genetic algorithm, and machine learning, this method can adaptively optimize the network structure and improve the overall efficiency of the system. Specifically, the simulated annealing algorithm is used to find a network structure close to the global optimal in a complex solution space. This method can effectively avoid falling into a local optimal solution, thereby finding a better network configuration. The combination of genetic algorithm and simulation is used to automatically adjust the network resource allocation. By simulating different resource allocation schemes and optimizing using genetic algorithm, the resource allocation strategy that best suits the current monitoring task can be found. For abnormality detection, this method uses a machine learning model for rapid real-time detection. The anomalies here include system performance, availability, security, and resource utilization. The machine learning model can learn from historical data and identify potential anomaly patterns, thereby improving the accuracy and efficiency of detection. In terms of anomaly handling, the reinforcement learning optimization network structure configuration method introduces an intelligent anomaly handling strategy. When an anomaly is detected, the system automatically selects the most suitable handling strategy based on the type, severity, and historical processing records of the anomaly. This method can greatly improve the automation and response speed of the system.

[0032] Through this reinforcement learning optimization network structure configuration method, the performance of the bridge crack real-time monitoring system has been significantly improved. The network structure is more optimized, resource allocation is more reasonable, and anomaly detection and processing are more efficient. This not only improves the overall efficiency of the system, but also enhances the reliability and stability of the system.

[0033] As a preferred implementation, the following specific implementation may be considered: First, the simulated annealing algorithm is used to preliminarily optimize the network structure. The algorithm starts with an initial network structure and generates a new structure through random perturbations. In each iteration, if the new structure has better performance, it is accepted; if the performance is poor, it is accepted with a certain probability. As the "temperature" decreases, the probability of accepting a poor solution gradually decreases, and finally converges to a near-optimal network structure. Next, genetic algorithms and simulations are used to optimize resource allocation. Different resource allocation schemes are encoded as "chromosomes" and new schemes are generated through crossover and mutation operations. The performance of each scheme is evaluated using simulations, and selection is made based on performance, retaining the excellent schemes to enter the next generation. After multiple generations of evolution, an optimized resource allocation strategy is finally obtained.

[0034] For anomaly detection, a deep learning-based anomaly detection model is trained. The model receives various indicators as input, including CPU usage, memory usage, network traffic, etc. By learning from normal operation data and known anomaly data, the model can identify potential abnormal patterns.

[0035] In terms of exception handling, a decision tree model is established. The model automatically selects a handling strategy based on the type, severity, and historical handling records of the exception. For example, for minor performance anomalies, resources may be reallocated; for serious safety anomalies, the affected nodes may be immediately isolated. In this way, the network structure of the bridge crack real-time monitoring system is optimized, resource allocation is more reasonable, and anomaly detection and processing are more efficient. This not only improves the overall performance of the monitoring system, but also enhances its ability to cope with various complex situations.

[0036] Compared with the prior art, the method of the present application has the following advantages: First, by combining a variety of advanced algorithms, adaptive optimization of network structure and resource allocation is achieved, which can better adapt to the dynamically changing monitoring environment compared to the traditional static configuration method. Secondly, the introduction of anomaly detection and intelligent processing mechanisms based on machine learning greatly improves the automation level and response speed of the monitoring system. Finally, the various components of the present method cooperate with each other to form a closed-loop optimization system that can continuously improve the performance of the monitoring system. These advantages make the method of the present application significantly progressive in improving the efficiency and reliability of the real-time monitoring system for bridge cracks.

[0037] The present application further proposes to configure redundancy for edge computing nodes, including: designing a redundant backup mechanism in the edge computing node, and synchronously storing the monitoring system data and deep learning model parameters among multiple nodes to ensure that when one of the nodes fails, other nodes are still available; the edge computing node adopts a high-speed network redundant backup design to ensure the reliability of the communication link. This redundancy configuration method is intended to improve the reliability and stability of the monitoring system. By synchronously storing the monitoring system data and deep learning model parameters among multiple edge computing nodes, the present application can effectively prevent data loss or system paralysis caused by single point failure. Even if a node fails, other nodes can continue to work normally to ensure the continuity and accuracy of bridge crack monitoring. Specifically, the redundant backup mechanism can be implemented in the following ways: First, a master-slave backup strategy can be adopted. In this strategy, a master node is designated to handle all data and computing tasks, and the data is synchronized to one or more slave nodes in real time. When the master node fails, the monitoring system can quickly switch to the slave node to ensure that the monitoring work is not interrupted. Secondly, distributed storage technology can be implemented. This method stores data in multiple nodes in a dispersed manner, and each data block has multiple copies. In this way, even if a node fails, the data can still be recovered from other nodes. Furthermore, data mirroring technology can be used. This method maintains exactly the same copies of data in different physical locations, ensuring that when data is lost at one location, it can be quickly restored from another location. For high-speed network redundant backup design, a dual-link or multi-link configuration can be used. For example, each edge computing node can be connected to two or more independent networks at the same time, such as a wired network and a wireless network. When a network fails, it can be immediately switched to a backup network to ensure the continuity of communication. In addition, the redundancy configuration method of the present application may also include regular data consistency checks. The monitoring system can regularly compare data on different nodes to ensure that the data of all nodes remains synchronized. If inconsistencies are found, the monitoring system can automatically repair the data.

[0038] By implementing these redundancy configuration methods, the present application significantly improves the reliability and robustness of the bridge crack real-time monitoring system. Even under complex environmental conditions, or in the face of hardware failures and network interruptions, the monitoring system can still maintain efficient and stable operation. This not only ensures the integrity and continuity of the monitoring data, but also greatly reduces the demand and cost of system maintenance.

[0039] As a preferred implementation, the present application can deploy a local database and a synchronization management module on each edge computing node. The local database is used to store the data collected and processed by the node, while the synchronization management module is responsible for data synchronization with other nodes. The synchronization process can adopt an incremental synchronization strategy to transmit only the changed data to reduce the network load. In addition, the present application can also introduce an intelligent load balancing mechanism. When it is detected that the load of a node is too high, the system can automatically assign some tasks to nodes with lighter loads to optimize the overall performance and prevent a single node from being overloaded. Through these technical means, the present application not only solves the problem of possible failure or data loss of edge computing nodes, but also further improves the performance and reliability of the entire bridge crack real-time monitoring system. Compared with traditional centralized monitoring systems, the distributed architecture and redundant design of the present application greatly reduce the risk of system failures and improve the availability and integrity of data, thereby providing more reliable technical support for bridge safety management.

[0040] This application further proposes a scheme to introduce augmented reality technology to assist engineering personnel in conducting on-site surveys, aiming to improve the data acquisition and analysis capabilities of engineering personnel during on-site surveys. The application of augmented reality technology includes the collection of on-site condition data, the integration of augmented reality model production and real-world scenes, the display of augmented reality effects through terminal devices, and real-time surveys and data uploads by engineering personnel through augmented reality effects. In terms of technical understanding, augmented reality technology can help engineers more intuitively identify and evaluate the status of bridge cracks by superimposing virtual information on real scenes. The specific implementation method includes using AR glasses for on-site surveys, and the staff can see the location, morphological evaluation and severity records of cracks in real scenes, thereby improving the efficiency and accuracy of surveys. The introduction of augmented reality technology makes the on-site survey process more efficient, and engineers can obtain and analyze data in real time to ensure the timeliness and accuracy of monitoring information. At the same time, combined with the data processing capabilities of edge computing nodes, it can play a greater role in data collaboration and sharing, solving the problem of information processing lag in traditional monitoring methods. In a specific embodiment, engineers wear AR glasses to obtain real-time crack information when conducting on-site surveys of bridges. By collecting the status data of the cracks on site, making the corresponding augmented reality model, and integrating it with the real scene, the augmented reality effect is finally displayed through the terminal device. Engineers can quickly judge the status of the cracks on site and upload the relevant data to the edge computing node to ensure real-time update and sharing of information. The application of augmented reality technology not only improves the efficiency of on-site surveys, but also enhances the visualization of information, reduces the errors of manual judgment, and ensures the accuracy and timeliness of monitoring data. Compared with existing technologies, traditional on-site survey methods often rely on manual observation and recording, which is inefficient and prone to errors. After the introduction of augmented reality technology, the on-site survey method has become more intelligent and efficient, and can realize the acquisition and analysis of real-time data, thereby effectively improving the overall effect and accuracy of monitoring. By combining virtual information with the actual environment, engineers can observe and analyze the situation of bridge cracks more intuitively.

[0041] Specifically, the application of augmented reality technology in real-time monitoring of bridge cracks can include the following aspects: First, augmented reality technology can display the location and morphology of bridge cracks in real time. Through AR devices, engineers can intuitively see the exact location of cracks in the bridge structure, which helps to quickly locate the problem area. Secondly, augmented reality technology can provide historical data and change trends of cracks. Engineers can view the history of specific cracks through the AR interface, including information such as its formation time and expansion speed, so as to better assess the severity and potential risks of cracks. Furthermore, augmented reality technology can assist engineers in crack measurement. Through AR devices, data such as the length, width and depth of cracks can be displayed in real time to improve the accuracy and efficiency of measurement. In addition, augmented reality technology can also provide real-time data analysis results. For example, the analysis results based on the deep learning model can be directly displayed on the AR interface, helping engineers to quickly determine the type and severity of cracks. Finally, augmented reality technology can achieve remote collaboration. On-site engineers can communicate with remote experts in real time through AR devices, share on-site perspectives, and obtain professional advice, thereby improving the accuracy and timeliness of decision-making.

[0042] By introducing augmented reality technology, the real-time monitoring method for bridge cracks in this application can significantly improve the efficiency and accuracy of on-site surveys. Engineering personnel can understand the situation of bridge cracks more intuitively and comprehensively, so as to make more accurate assessments and decisions. This can not only save time and labor costs, but also improve the quality and safety of bridge maintenance.

[0043] As a preferred implementation, augmented reality technology can be used to assist engineers in conducting on-site surveys in the following steps: First, the on-site condition data of bridge cracks is collected. This includes using hyperspectral imaging cameras, thermal imaging devices, and lidar to obtain detailed data on the bridge surface. Next, an augmented reality model is created based on the collected data and integrated with the real scene. This step requires the digital crack information to be accurately mapped onto the actual bridge structure. Then, the augmented reality effect is displayed through terminal devices (such as AR glasses). Through these devices, engineers can see the actual bridge scene superimposed with virtual information. Engineers then use the augmented reality effect to determine the on-site condition of the bridge cracks and conduct real-time on-site surveys. This includes precise positioning of the crack location, detailed evaluation of the morphology, and accurate recording of the severity. For example, when engineers wear AR glasses to observe the bridge, an overlay will be displayed on the glasses to mark the location of the detected cracks. AR glasses can also capture the image of the bridge cracks in the engineer's field of view in real time, and superimpose key information of the cracks on the display screen, such as crack width, depth, direction and other parameters. Engineers can interact with the AR system through voice commands or gestures to record on-site survey data. In actual applications, the display interface of AR glasses can be divided into multiple functional areas: the central area is used to display real-time augmented reality scenes, the left area displays crack parameter data, and the right area provides operation control buttons. When an engineer looks at a crack, the system automatically identifies the crack characteristics and marks the relevant data on the display interface. Engineers can choose to perform specific operations such as location positioning, morphological assessment, or severity recording through touch or voice. Engineers can also use the virtual control buttons on the glasses interface to select different observation modes, such as thermal imaging mode or 3D point cloud mode, to obtain more information about the crack. At the same time, AR glasses can display the measurement data of the crack, such as length, width, and depth, in real time, and automatically assess the severity of the crack according to preset standards. Finally, after determining the on-site status of the crack, the relevant data will be uploaded to the edge computing node for further analysis and processing. In this way, augmented reality technology not only improves the efficiency and accuracy of on-site surveys, but also achieves seamless integration with edge computing systems, providing more comprehensive and reliable data support for real-time monitoring and analysis of bridge cracks.

[0044] Compared with traditional on-site survey methods, the solution of introducing augmented reality technology in this application has significant advantages. Traditional methods mainly rely on the experience of engineers and simple measuring tools, which are easily affected by subjective factors and have low efficiency. The application of augmented reality technology not only provides a more intuitive and accurate means of observation, but also combines the analysis results of deep learning models in real time, greatly improving the scientificity and efficiency of the survey. In addition, augmented reality technology also realizes the possibility of remote collaboration, so that on-site surveys are no longer limited to the capabilities of a single engineer, but can bring together the wisdom of experts from multiple parties to make a more comprehensive and accurate assessment.

[0045] In the on-site survey solution based on AR glasses, engineers can directly observe the actual situation of bridge cracks through AR glasses. AR glasses superimpose the collected bridge crack data with the virtual scene, presenting an augmented reality effect in the field of vision of engineers. In this way, engineers can observe and evaluate the status of bridge cracks more intuitively. In specific implementation, AR glasses can be equipped with high-definition cameras and display screens, and transmit data with edge computing nodes through wireless networks. As a specific embodiment, when crack detection is performed on a bridge, after the engineer wears AR glasses, the system first loads the three-dimensional model and historical detection data of the bridge. When the engineer observes on-site, the AR glasses collect images in the field of vision in real time, and align the collected actual crack images with the pre-loaded three-dimensional model. Through real-time calculation of the edge computing node, the outline annotation, size data and damage level assessment results of the cracks are superimposed on the display screen of the AR glasses. Engineers can save the crack data of interest through voice commands, and the system will automatically record the geographical coordinates, captured images and related parameters of the cracks. These data are then uploaded to the edge computing node for storage and analysis. This on-site survey solution based on AR glasses improves the work efficiency and accuracy of engineers in bridge crack detection. Through the application of augmented reality technology, engineers can observe and record crack conditions more intuitively, reducing subjective errors in traditional manual inspections. At the same time, it realizes real-time collection and uploading of inspection data, providing reliable data support for subsequent crack monitoring and analysis.

[0046] 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 aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein by equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A real-time monitoring method for bridge cracks based on edge computing, characterized in that: include: S1: Use hyperspectral imaging cameras, thermal imaging devices and lidar to collect bridge crack data, obtain bridge crack data and transmit the bridge crack data to edge computing nodes; S2: Based on bridge crack data, use edge learning technology to adaptively adjust edge computing nodes, including dynamic resource allocation and intelligent task scheduling; S3: Use annotated bridge crack image datasets to train a deep learning model, and adaptively process the deep learning model through meta-learning technology and model compression methods; S4: Extract features from bridge crack data, fuse the collected bridge crack data using a fusion method, then input the fused bridge crack data into a deep learning model, and finally output the judgment and analysis results of bridge cracks; S5: Based on the judgment and analysis results output by S4, establish an abnormal monitoring model to further monitor the structural status of the bridge.

2. The real-time monitoring method for bridge cracks based on edge computing according to claim 1 is characterized in that: The use of hyperspectral imaging, thermal imaging devices and lidar to collect bridge crack data in S1 includes: using a hyperspectral camera to collect hyperspectral images of the bridge surface; using a thermal imaging device to obtain a temperature distribution map of the bridge surface; scanning the bridge surface by lidar to generate three-dimensional point cloud data; and forming a bridge crack data set from the collected hyperspectral images, temperature distribution maps, and three-dimensional point cloud data.

3. The real-time monitoring method for bridge cracks based on edge computing according to claim 1 is characterized in that: In S2, based on the bridge crack data, edge learning technology is used to adaptively adjust the edge computing nodes, including dynamic resource allocation and intelligent task scheduling. In edge deployment, computing resources are dynamically adjusted to cope with uneven data loads.

4. The real-time monitoring method for bridge cracks based on edge computing according to claim 1 is characterized in that: The real-time monitoring method for bridge cracks is implemented based on a real-time monitoring system for bridge cracks. The real-time monitoring system for bridge cracks adopts a distributed architecture, including: distributed edge computing nodes, a cloud platform and a central server; the distributed edge computing nodes are distributed at different locations of the bridge and are responsible for receiving and processing bridge crack data acquired by a hyperspectral imaging camera, a thermal imaging device and a lidar; the cloud platform is used to process large-scale data uploaded by the distributed edge computing nodes; the central server centrally manages task scheduling and data storage of all edge computing nodes.

5. The real-time monitoring method for bridge cracks based on edge computing according to claim 4 is characterized in that: The bridge crack real-time monitoring system adopts reinforcement learning to optimize the network structure configuration. The method of optimizing the network structure configuration by reinforcement learning includes: a. using simulated annealing and genetic algorithm for optimization, and finding a near-optimal network structure through simulated annealing; b. using genetic algorithm and simulation to automatically adjust the resource allocation of the network; c. for the detection of anomalies in the bridge crack real-time monitoring system, using machine learning models to quickly detect anomalies in real time; the anomalies include system performance, availability, security and resource utilization; d. using anomaly handling strategies to handle the monitored anomalies, including: when an anomaly is detected, automatically selecting anomaly handling strategies according to the anomaly type, severity level and historical processing records.

6. The real-time monitoring method for bridge cracks based on edge computing according to claim 1 is characterized in that: It also includes S6: redundancy configuration for edge computing nodes, including: designing a redundant backup mechanism in the edge computing nodes, synchronously storing the monitoring system data and deep learning model parameters among multiple nodes to ensure that when one of the nodes fails, the other nodes are still available; the edge computing nodes adopt a high-speed network redundant backup design to ensure the reliability of the communication link.

7. The real-time monitoring method for bridge cracks based on edge computing according to claim 6 is characterized in that: The data collaboration and data sharing between the edge computing nodes are achieved based on federated learning technology. The federated learning technology adopts the following method: each edge node has its own local data set, parameter exchange and gradient aggregation are performed between edge nodes, data processing and analysis tasks are performed between edge nodes, and data calculation and collaboration are performed with the assistance of cloud servers.

8. The real-time monitoring method for bridge cracks based on edge computing according to claim 1 is characterized in that: Also includes S7: Introducing augmented reality technology to assist engineering personnel in on-site surveys.

9. The real-time monitoring method for bridge cracks based on edge computing according to claim 8 is characterized in that: The augmented reality technology assists engineering personnel in conducting on-site surveys, including: S71: collecting on-site condition data of bridge cracks; S72: making an augmented reality model and integrating it with the real scene; S73: using augmented reality technology to display the augmented reality effect through a terminal device; S74: engineering personnel determine the on-site condition of bridge cracks through the augmented reality effect and conduct real-time on-site surveys; S75: after determining the on-site condition of the cracks, upload the data to the edge computing node.

10. A real-time monitoring method for bridge cracks based on edge computing according to claim 9, characterized in that: In S74, the staff uses AR glasses to superimpose the real scene on the virtual reality scene to assist in real-time on-site inspection, including the location of bridge cracks, morphological assessment, and recording of crack severity.