Unmanned aerial vehicle-ground station-cloud three-level linkage power transmission line inspection system

Abstract

The invention relates to the field of unmanned aerial vehicles, and discloses an unmanned aerial vehicle-ground station-cloud three-level linkage power transmission line inspection system. Comprising a sensing acquisition node, a fog computing hub, a cloud decision node, a dual-mode communication link, a three-level computing power scheduling module, a closed-loop acquisition control module, a multi-mode enhancement module, a layered transmission module and a twinborn pre-judgment scheduling module. The sensing acquisition node acquires multi-dimensional data of a power transmission line, outputs the multi-dimensional data to the fog computing hub after being subjected to lightweight processing, and feeds back own state data at the same time; the fog computing hub receives the multi-dimensional data of the sensing acquisition nodes, and the multi-dimensional data is subjected to multi-modal fusion and defect preliminary screening. According to the invention, cross-node task reasonable allocation is realized through the three-level computing power scheduling module, tasks such as simple defect preliminary screening and complex defect research and judgment are respectively adapted to computing power characteristics of a fog computing hub and a cloud decision node, and meanwhile, the collection frequency of the unmanned aerial vehicle is dynamically adjusted according to the health state of a line, so that computing power waste and invalid collection are avoided.

Description

Technical Field

[0001] This invention relates to the field of unmanned aerial vehicle (UAV) technology, specifically to a three-level linkage system for power transmission line inspection involving UAVs, ground stations, and the cloud. Background Technology

[0002] Traditional power transmission line drone inspections often adopt a single-node processing mode. The front-end drone is only responsible for data collection and lacks local lightweight processing capabilities. This results in excessive bandwidth consumption when transmitting large amounts of raw data directly, and the reliance on a single communication link makes it prone to data loss due to signal interruption. Meanwhile, the back-end cloud centrally processes all tasks, which not only has the problem of large transmission latency, but also easily leads to resource waste due to uneven distribution of computing power, making it unable to adapt to the dynamic needs of complex inspection scenarios.

[0003] Existing inspection systems often employ fixed-weight strategies for multimodal data fusion, failing to dynamically adjust confidence levels based on the data signal-to-noise ratio. This results in insufficient accuracy of the fused data, leading to high rates of missed and false detections of defects. Furthermore, the systems lack in-depth analysis of indicators such as transmission line temperature trends and deformation accumulation, making it difficult to predict potential risk areas. They can only passively respond to existing defects, resulting in insufficient foresight and targeting in inspections. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides a three-level linkage system for power transmission line inspection, which integrates drones, ground stations, and the cloud. This system solves the problem that traditional drone inspections of power transmission lines often employ a single-node processing mode, which is prone to resource waste due to uneven distribution of computing power.

[0005] To achieve the above objectives, the present invention is implemented through the following technical solution: a three-level linkage power transmission line inspection system of UAV-ground station-cloud, including a sensing and acquisition node, a fog computing hub, a cloud decision node, a three-level computing power scheduling module, a closed-loop acquisition and control module, a multimodal enhancement module, a hierarchical transmission module, and a twin prediction and scheduling module; The sensing and acquisition node collects multi-dimensional data of the transmission line, including inspection data and operation status data. The inspection data is processed in a lightweight manner and then output to the fog computing hub. At the same time, the operation status data is output to the fog computing hub. The inspection data includes visual images, temperature distribution data and three-dimensional spatial point cloud data of the transmission line. The operation status data includes the temperature and acquisition frequency of the sensing and acquisition node. The fog computing hub receives multi-dimensional data from the sensing and acquisition nodes, and after multi-modal fusion and initial defect screening, synchronizes the multi-modal fusion results and the initial defect screening results to the cloud decision node. The cloud-based decision node receives the multimodal fusion results and initial defect screening results from the fog computing hub, combines them with historical inspection data to complete complex analysis and health assessment, and generates global scheduling instructions to be sent to the sensing and acquisition nodes. The three-level computing power scheduling module acquires the operating status data of the sensing and acquisition nodes and dynamically allocates inspection tasks to the sensing and acquisition nodes. The closed-loop acquisition and control module generates a supplementary acquisition instruction based on the initial screening results of the fog computing hub, sends it to the sensing acquisition node, commands the sensing acquisition node to perform supplementary acquisition, and synchronizes the supplementary acquisition data received from the sensing acquisition node to the cloud decision node. The multimodal enhancement module processes multi-dimensional data from the sensing and acquisition nodes and outputs optimized data to match the communication link bandwidth. The hierarchical transmission module classifies the inspection data according to confidence index and allocates transmission bandwidth according to priority. The twin prediction and scheduling module compares real-time inspection data with historical inspection data to generate risk prediction and resource scheduling instructions. Each component forms a complete inspection closed loop through a dual-mode communication link, which is used to optimize the resource utilization efficiency and defect identification completeness of power transmission line inspection.

[0006] Preferably, the sensing and acquisition node includes a data acquisition unit, a status sensing unit, a lightweight processing unit, and a communication transmission unit; the data acquisition unit acquires real-time inspection data of the transmission line, the lightweight processing unit compresses the real-time inspection data to reduce transmission bandwidth usage, the communication transmission unit sends the compressed inspection data to the fog computing hub, and the status sensing unit acquires the operating status data of the data acquisition unit and synchronously feeds it back to the three-level computing power scheduling module to provide a basis for cross-node computing power scheduling.

[0007] Preferably, the cloud-based decision node includes a big data storage unit, a global computing unit, and a decision scheduling unit. The big data storage unit stores twin model data and historical inspection data. The global computing unit performs complex defect analysis based on the multimodal fusion results received from the fog computing hub and the initial defect screening results. The decision scheduling unit combines the analysis results and line status to generate a global scheduling strategy that includes task allocation rules, resource scheduling priorities, and inspection route optimization parameters, providing guidance for subsequent comparison of real-time inspection data and historical inspection data.

[0008] Preferably, the fog computing hub specifically includes: a data receiving and caching unit, a fog computing processing unit, and an instruction generation unit. The data receiving and caching unit is used to receive inspection data and cache compressed data. The fog computing processing unit is used to perform inspection data fusion and initial defect screening. The instruction generation unit generates control instructions based on the initial defect screening results. The instruction generation unit adjusts the acquisition frequency of the data acquisition unit based on the global scheduling instructions issued by the cloud decision node.

[0009] Preferably, the three-level computing power scheduling module acquires the operating status data fed back by the status sensing unit of the sensing and acquisition node in real time, and matches the inspection data transmitted by the sensing and acquisition node with the twin model data and historical inspection data stored in the big data storage unit to determine whether there are any suspected defects. If the number of defects is less than the preset defect threshold, it is determined to be a defect initial screening task; If the defect is larger than the preset defect threshold, it is judged as a complex recognition task; When a task requires inspection of data storage and archiving, as well as subsequent traceability and verification, it is determined to be a data tracing task. The defect preset threshold is a specific value manually entered by the operator based on experience.

[0010] Preferably, the three-level computing power scheduling module calculates task priorities using a weighted summation method, constructs a multi-objective optimization model to match cross-node computing power, prioritizes the allocation of initial defect screening tasks to the fog computing hub, allocates complex identification tasks to cloud decision nodes, and adopts an asynchronous allocation mode for data tracing tasks. At the same time, it adjusts the acquisition frame rate of the sensing acquisition nodes according to the global scheduling strategy.

[0011] Preferably, when the fog computing hub detects a suspected defect, the closed-loop acquisition and control module receives the lidar point cloud coordinates and the UAV's real-time positioning data, generates a supplementary acquisition command including the camera shooting angle and infrared thermal imager sampling frequency adjustment parameters, sends it to the sensing acquisition node, receives the supplementary acquisition data, and synchronizes it to the cloud decision node.

[0012] Preferably, the multimodal enhancement module employs a fusion algorithm of visible light features, infrared temperature features, and lidar spatial features to perform dimensionality reduction and compression on real-time inspection data.

[0013] Preferably, the layered transmission module first receives the initial defect screening results and calculates the confidence level, and then divides the data according to the global scheduling strategy evaluated by the cloud decision node and the timeliness index calculated based on the data acquisition rate of the sensing and acquisition node. When the initial confidence level of the defect is high, and the corresponding line is evaluated as a core line by the cloud decision node, it is considered an urgent defect data. If the initial confidence level of the defect is medium, and the health level of the line is assessed as being at the level of concern by the cloud decision node, then it is considered suspicious abnormal data. When the initial confidence level of the defect screening is low, and the line health level is assessed as normal by the cloud decision node, it is considered normal status data. The hierarchical transmission module allocates transmission bandwidth according to the priority of emergency defect data, suspicious abnormal data, and normal status data.

[0014] Preferably, the twin prediction and scheduling module receives real-time inspection data from the sensing and acquisition nodes, compares the real-time inspection data with historical twin model data in the big data storage unit of the cloud decision node, and calculates the degradation trend index of line components. When the degradation trend exceeds the preset degradation threshold, it generates and issues a joint scheduling instruction for increasing inspection density, pre-allocating computing resources, and adjusting task priority. The preset degradation threshold is manually input by human experience.

[0015] This invention provides a three-level linkage system for power transmission line inspection, integrating drones, ground stations, and the cloud. It offers the following advantages: This invention achieves reasonable allocation of cross-node tasks through a three-level computing power scheduling module, adapting tasks such as simple defect screening and complex defect analysis to the computing power characteristics of the fog computing hub and cloud decision nodes respectively. At the same time, it dynamically adjusts the drone collection frequency according to the health status of the line, avoiding waste of computing power and ineffective collection, and improving resource utilization.

[0016] This invention achieves efficient fusion of visual images, temperature data, and 3D point clouds by dynamically allocating confidence weights based on the data signal-to-noise ratio through a multimodal enhancement module. Combined with the precise supplementary acquisition mechanism of the closed-loop acquisition control module, it effectively reduces missed and false defects, ensuring the comprehensiveness and accuracy of defect identification.

[0017] This invention relies on a twin prediction and scheduling module to construct a digital twin model of transmission lines, extract temperature trends and deformation accumulation indicators to predict potential risk areas, and adjust the computing power reserve ratio and inspection density in advance, transforming traditional post-event remedial inspections into pre-event early warning inspections, thereby reducing the line fault rate. Attached Figure Description

[0018] Figure 1 This is a flowchart of the present invention. Detailed Implementation

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

[0020] Please see the appendix Figure 1This invention provides a three-level linkage system for power transmission line inspection, consisting of a drone, a ground station, and a cloud platform. The system comprises three core functional nodes: a sensing and acquisition node, a fog computing hub, and a cloud decision-making node. It also includes a dual-mode communication link and five collaborative modules. This system optimizes resource utilization efficiency and defect identification completeness in power transmission line inspection. Each of the sensing and acquisition node, fog computing hub, and cloud decision-making node is equipped with independent data processing, status awareness, and communication transmission units. The five collaborative modules include a three-level computing power scheduling module, a closed-loop acquisition and control module, a multi-modal enhancement module, a hierarchical transmission module, and a twin prediction and scheduling module. Each module relies on the dual-mode communication link to achieve data interaction and command transmission, forming a complete inspection closed loop.

[0021] The sensing and acquisition node, as the core of front-end data acquisition, includes a data acquisition unit, a status sensing unit, a lightweight processing unit, and a communication transmission unit. The data acquisition unit is equipped with three types of sensors: a visible light camera, an infrared thermal imager, and a lidar, to simultaneously acquire visual images, temperature distribution data, and three-dimensional spatial point cloud data of the transmission line. The status sensing unit collects real-time node operating status indicators, including remaining power, processor load rate, remaining inspection range, and data acquisition rate. The lightweight processing unit is equipped with an embedded processor with a main frequency of no less than 1.8GHz and memory of no less than 4GB to perform sensor calibration and raw data compression. The communication transmission unit supports a dual-mode communication protocol to realize data uploading and command reception.

[0022] The fog computing hub, serving as the central collaborative core, comprises a data receiving and caching unit, a status awareness unit, a fog computing processing unit, an instruction generation unit, and a communication and transmission unit. The data receiving and caching unit is equipped with a local storage module with a capacity of at least 1TB to temporarily store compressed data transmitted by the sensing and acquisition nodes. The status awareness unit collects node operating status indicators, including CPU utilization, memory utilization, remaining cache space, and the number of currently processed tasks. The fog computing processing unit utilizes an industrial-grade server configuration with at least 8 CPU cores, at least 32GB of memory, and at least 8GB of GPU memory. It is equipped with an edge computing framework to perform multimodal data fusion, initial defect screening, and lightweight model operation. The instruction generation unit generates control instructions such as adjusting acquisition parameters and optimizing flight paths based on the initial defect screening results and cloud scheduling strategies. The communication and transmission unit serves as a bidirectional data relay hub, enabling bidirectional communication with the sensing and acquisition nodes and cloud decision-making nodes.

[0023] The cloud-based decision-making node serves as the core of backend decision-making, comprising a big data storage unit, a status awareness unit, a global computing unit, a decision scheduling unit, and a communication transmission unit. The big data storage unit is built on a distributed server cluster with no fewer than 10 nodes and a storage capacity of no less than 100TB, storing historical inspection data, real-time synchronized data, and twin model data. The status awareness unit collects cluster operation status indicators, including task queue length, average task processing latency, cluster computing power utilization, and remaining storage space. The global computing unit has a total computing power of no less than 500 TOPS, deploys a big data analysis platform, and performs complex defect analysis, line health status assessment, and twin model construction. The decision scheduling unit generates global task allocation, resource scheduling, and inspection strategy optimization instructions based on the analysis results. The communication transmission unit is responsible for data synchronization and instruction interaction with the fog computing hub.

[0024] The dual-mode communication link provides transmission support for the collaboration of the three major nodes, including a main communication link and a backup communication link. The main communication link is based on 5G standalone (SA) networking mode and uses network slicing technology to divide a dedicated communication channel for power line inspection, ensuring an uplink bandwidth of no less than 100Mbps and a downlink bandwidth of no less than 200Mbps, with transmission latency controlled within 20ms. The backup communication link uses a LoRa dedicated communication module, with a communication distance of no less than 5km and a transmission latency of no more than 100ms. It automatically switches when the 5G signal is interrupted or the bandwidth is lower than the threshold, ensuring the continuous transmission of core instructions and critical data. The communication process uses an improved MQTT protocol. The message format includes five parts: message header, message type, payload, checksum, and message tail. The checksum is calculated using the CRC16 algorithm and is used to verify the integrity of data transmission. The message header is simplified to 4 bytes to reduce transmission overhead.

[0025] The three-level computing power scheduling module relies on the status perception units of each node to acquire data, and achieves resource optimization through task priority calculation and cross-node computing power matching. The status perception units of each node collect indicator data with a sampling period of 0.5 seconds. The remaining power of the perception and acquisition nodes is collected through the battery management system, the load rate is obtained through the embedded processor monitoring module, the remaining inspection range is calculated based on GPS positioning data and preset routes, the data acquisition rate is statistically analyzed through sensor output frame rate, the CPU and memory usage of the fog computing hub is obtained through server monitoring software, the remaining cache space is statistically analyzed through the storage management module, the number of processed tasks is counted through the task scheduling module of the edge computing framework, the task queue length and average processing latency of the cloud decision nodes are statistically analyzed through the cluster scheduling system, the computing power utilization rate is calculated based on the CPU and GPU running status of each node, and the remaining storage space is obtained through the distributed storage management module.

[0026] Inspection tasks are divided into three categories: initial defect screening, complex defect identification, and data storage and tracing. Their priorities are calculated using a weighted summation method, as shown in the formula: in Let be the weighting coefficient, satisfying The parameters are determined by the analytic hierarchy process and can be dynamically adjusted. S is the task urgency coefficient, I is the line importance coefficient, and U is the node resource urgency coefficient, all of which are calculated based on preset rules or collected data. P is the inspection task priority, used to measure task priority.

[0027] Cross-node computing power matching aims to minimize total task processing latency and resource consumption. A multi-objective optimization model is constructed, with the following formula: in, Represents transmission delay. Represents processing latency, determined by the resource consumption of the sensing and acquisition node (u), the fog computing hub (g), and the cloud decision-making node (c). ) multiplied by the corresponding weighting coefficient ( The weighted summation is then used to arrive at the final result, achieving precise matching between tasks and computing resources. Constraints include core threshold limits such as node load, power consumption, and task queue length. The model is solved using an improved particle swarm optimization algorithm. Initial defect screening tasks are preferentially assigned to the fog computing hub, while complex defect identification tasks are fixedly assigned to cloud decision nodes. Data storage and tracing tasks adopt an asynchronous allocation mode. Sensing and acquisition nodes adjust the acquisition frame rate according to the scheduling results using a formula: in The base sampling frequency is k, and k is an adjustment coefficient. The health status coefficient of the line issued by the cloud decision-making node.

[0028] The closed-loop acquisition and control module, based on the bidirectional transmission capability of the dual-mode communication link, achieves dynamic response in acquisition, analysis, and feedback. After the fog computing hub completes the initial defect screening, it performs spatial transformation on the defect pixel coordinates based on the lidar point cloud data, using the following formula: in, The vertical distance from the defect measured by the lidar to the UAV. and The intrinsic focal length of the camera. X-axis offset The offset is along the Y-axis, combined with the current position of the sensing and acquisition node. Calculate the optimal shooting angle: Among them, the yaw angle (θ) and the pitch angle This represents the best shooting angle. These represent the X, Y, and Z axis coordinates of the UAV's real-time positioning, respectively. Precise supplementary data collection commands are generated and sent to the sensing and data collection nodes. After the supplementary data is transmitted back, it is synchronized to the cloud decision-making node. In weak network environments, local caching and breakpoint resume mechanisms are activated, and data block identification and verification are used to ensure transmission integrity.

[0029] The multimodal enhancement module optimizes data through feature extraction, spatial correction, feature fusion, and lightweight compression. After extracting corresponding features from the three types of sensor data, an adaptive weighted fusion algorithm is used to calculate the final feature data after multimodal fusion. The formula is as follows: in The confidence weights are calculated based on the data signal-to-noise ratio. These represent the preprocessed visible light, infrared, and lidar sensor feature data, corresponding to the visible light feature extraction, infrared temperature feature extraction, and lidar spatial feature extraction of the multimodal enhancement module. F' represents the feature form after preprocessing such as correction and dimensionality reduction, satisfying... The fused data undergoes PCA dimensionality reduction and secondary compression, with the compression ratio adjusted using a formula: in The compression ratio is... The compressive strength coefficient, To reduce the dimensionality of PCA and adapt to different network bandwidth conditions, the hierarchical transmission module divides the inspection data into three levels, with the priority quantification formulas for each level as follows: Bandwidth is allocated to data at each level according to priority, using the following formula: in Allocate bandwidth for the k-th level of data. To ensure transmission efficiency and integrity, a differentiated transmission mode is adopted based on the current total available bandwidth.

[0030] The twin prediction and scheduling module builds a twin model based on collected data and historical data. The local model grid resolution is adjusted using a formula: in Given the lidar point cloud density, temperature trends and cumulative deformation indices are extracted based on the model. When the indices exceed a threshold, the area is identified as a potential risk zone, and the risk probability is calculated as follows: in, This serves as an indicator of the temperature degradation trend of circuit components. For the cumulative deformation index of line components, the computing power reserve ratio is calculated based on the risk probability. Inspection density ) and task priority adjustment coefficient ( ): The system generates and sends scheduling instructions to each node, enabling the advance planning and precise deployment of inspection resources.

[0031] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A three-level linkage power transmission line inspection system integrating UAV, ground station, and cloud, characterized in that: It includes a sensing and acquisition node, a fog computing hub, a cloud decision-making node, a three-level computing power scheduling module, a closed-loop acquisition and control module, a multi-modal enhancement module, a hierarchical transmission module, and a twin prediction and scheduling module; The sensing and acquisition node collects multi-dimensional data of the transmission line, including inspection data and operation status data. The inspection data is processed in a lightweight manner and then output to the fog computing hub. At the same time, the operation status data is output to the fog computing hub. The inspection data includes visual images, temperature distribution data and three-dimensional spatial point cloud data of the transmission line. The operation status data includes the temperature and acquisition frequency of the sensing and acquisition node. The fog computing hub receives multi-dimensional data from the sensing and acquisition nodes, and after multi-modal fusion and initial defect screening, synchronizes the multi-modal fusion results and the initial defect screening results to the cloud decision node. The cloud-based decision node receives the multimodal fusion results and initial defect screening results from the fog computing hub, combines them with historical inspection data to complete complex analysis and health assessment, and generates global scheduling instructions to be sent to the sensing and acquisition nodes. The three-level computing power scheduling module acquires the operating status data of the sensing and acquisition nodes and dynamically allocates inspection tasks to the sensing and acquisition nodes. The closed-loop acquisition and control module generates a supplementary acquisition instruction based on the initial screening results of the fog computing hub, sends it to the sensing acquisition node, commands the sensing acquisition node to perform supplementary acquisition, and synchronizes the supplementary acquisition data received from the sensing acquisition node to the cloud decision node. The multimodal enhancement module processes multi-dimensional data from the sensing and acquisition nodes and outputs optimized data to match the communication link bandwidth. The hierarchical transmission module classifies the inspection data according to confidence index and allocates transmission bandwidth according to priority. The twin prediction and scheduling module compares real-time inspection data with historical inspection data to generate risk prediction and resource scheduling instructions. Each component forms a complete inspection closed loop through a dual-mode communication link, which is used to optimize the resource utilization efficiency and defect identification completeness of power transmission line inspection.

2. The UAV-Ground Station-Cloud Three-Level Linkage Power Line Inspection System according to claim 1, characterized in that, The sensing and acquisition node includes a data acquisition unit, a status sensing unit, a lightweight processing unit, and a communication transmission unit. The data acquisition unit collects real-time inspection data of the transmission line, the lightweight processing unit compresses the real-time inspection data to reduce transmission bandwidth usage, the communication transmission unit sends the compressed inspection data to the fog computing hub, and the status sensing unit collects the operating status data of the data acquisition unit and synchronously feeds it back to the three-level computing power scheduling module to provide a basis for cross-node computing power scheduling.

3. The UAV-Ground Station-Cloud Three-Level Linkage Power Line Inspection System according to claim 2, characterized in that, The cloud-based decision node includes a big data storage unit, a global computing unit, and a decision scheduling unit. The big data storage unit stores twin model data and historical inspection data. The global computing unit performs complex defect analysis based on the multimodal fusion results received from the fog computing hub and the initial defect screening results. The decision scheduling unit combines the analysis results and line status to generate a global scheduling strategy that includes task allocation rules, resource scheduling priorities, and inspection route optimization parameters, providing guidance for subsequent comparison of real-time inspection data and historical inspection data.

4. The UAV-Ground Station-Cloud Three-Level Linkage Power Line Inspection System according to claim 3, characterized in that, The fog computing hub specifically includes: a data receiving and caching unit, a fog computing processing unit, and an instruction generation unit. The data receiving and caching unit is used to receive inspection data and cache compressed data. The fog computing processing unit is used to perform inspection data fusion and initial defect screening. The instruction generation unit generates control instructions based on the initial defect screening results. The instruction generation unit adjusts the acquisition frequency of the data acquisition unit based on the global scheduling instructions issued by the cloud decision node.

5. The UAV-Ground Station-Cloud Three-Level Linkage Power Line Inspection System according to claim 4, characterized in that, The three-level computing power scheduling module acquires the operating status data fed back by the status sensing unit of the sensing and acquisition node in real time, and matches the inspection data transmitted by the sensing and acquisition node with the twin model data and historical inspection data stored in the big data storage unit to determine whether there are any suspected defects. If the number of defects is less than the preset defect threshold, it is determined to be a defect initial screening task; If the defect is larger than the preset defect threshold, it is judged as a complex recognition task; When a task requires inspection of data storage and archiving, as well as subsequent traceability and verification, it is determined to be a data tracing task. The defect preset threshold is a specific value manually entered by the operator based on experience.

6. The UAV-Ground Station-Cloud Three-Level Linkage Power Line Inspection System according to claim 5, characterized in that, The three-level computing power scheduling module calculates task priorities using a weighted summation method, constructs a multi-objective optimization model to match cross-node computing power, prioritizes the allocation of initial defect screening tasks to the fog computing hub, allocates complex identification tasks to cloud decision nodes, and adopts an asynchronous allocation mode for data tracing tasks. At the same time, it adjusts the acquisition frame rate of the sensing acquisition nodes according to the global scheduling strategy.

7. The UAV-Ground Station-Cloud Three-Level Linkage Power Line Inspection System according to claim 1, characterized in that, When the fog computing hub detects a suspected defect, the closed-loop acquisition and control module receives the lidar point cloud coordinates and the UAV's real-time positioning data, generates a supplementary acquisition command that includes the camera shooting angle and infrared thermal imager sampling frequency adjustment parameters, sends it to the perception acquisition node, receives the supplementary acquisition data, and synchronizes it to the cloud decision node.

8. The UAV-Ground Station-Cloud Three-Level Linkage Power Line Inspection System according to claim 1, characterized in that, The multimodal enhancement module employs a fusion algorithm of visible light features, infrared temperature features, and lidar spatial features to perform dimensionality reduction and compression on real-time inspection data.

9. The UAV-Ground Station-Cloud Three-Level Linkage Power Line Inspection System according to claim 4, characterized in that, The hierarchical transmission module classifies the inspection data according to confidence level indicators and allocates transmission bandwidth according to priority, specifically as follows: First, the initial defect screening results are received and the confidence level is calculated. At the same time, the data is classified according to the global scheduling strategy evaluated by the cloud decision-making node and the timeliness index calculated based on the data acquisition rate of the sensing and acquisition node. When the initial confidence level of the defect is high, and the corresponding line is evaluated as a core line by the cloud decision node, it is considered an urgent defect data. If the initial confidence level of the defect is medium, and the health level of the line is assessed as being at the level of concern by the cloud decision node, then it is considered suspicious abnormal data. When the initial confidence level of the defect screening is low, and the line health level is assessed as normal by the cloud decision node, it is considered normal status data. The hierarchical transmission module allocates transmission bandwidth according to the priority of emergency defect data, suspicious abnormal data, and normal status data.

10. The UAV-Ground Station-Cloud Three-Level Linkage Transmission Line Inspection System according to claim 2, characterized in that, The twin prediction and scheduling module receives real-time inspection data from the sensing and acquisition nodes, compares the real-time inspection data with historical twin model data in the big data storage unit of the cloud decision node, and calculates the degradation trend index of line components. When the degradation trend exceeds the preset degradation threshold, it generates and issues a joint scheduling instruction to increase inspection density, pre-allocate computing resources, and adjust task priority. The preset degradation threshold is manually input by human experience.

Images