A power inspection system and satellite image processing method

Abstract

The present invention relates to the field of power facility monitoring technology, and in particular to a power inspection system and satellite image processing method. The system includes: a front-end perception module, which collects status data and establishes an inspection path after receiving an inspection instruction; an edge processing module, which receives status data and performs real-time processing and feature extraction according to the inspection path to generate key parameter information; a central analysis platform, which receives key parameter information and performs diagnostic analysis based on it to generate an inspection diagnosis report; a feedback optimization module, which updates the feature extraction rules of the edge processing module and the intelligent diagnostic analysis model of the central analysis platform based on the inspection diagnosis report and key parameter information; and an inspection result module, which sends the inspection diagnosis report to the operation and maintenance terminal. Through the present invention, the problems of low inspection efficiency, high safety risks and response delays caused by the inability of traditional inspection methods to achieve efficient, real-time and automated fault detection and status monitoring in remote or dangerous areas are effectively solved.

Description

Technical Field

[0001] The present invention relates to the technical field of power facility monitoring, and in particular to a power inspection system and a satellite image processing method. Background Art

[0002] The safe and efficient operation and maintenance of modern power networks urgently requires the construction of an inspection system that can implement real-time, automated, highly reliable status monitoring and intelligent diagnosis of widely distributed power equipment, especially those in remote, dangerous or inaccessible areas and their surrounding environmental hazards.

[0003] The current mainstream power inspection system faces fundamental system architecture bottlenecks and is seriously lacking in real-time response capabilities. The original massive data collected by remote platforms needs to be fully transmitted back for analysis. The lengthy data transmission and centralized processing processes lead to significant delays in generating effective alarms, which cannot meet the needs of immediate early warning. The end-to-end intelligence level is low and the processing efficiency is limited. The core analysis capabilities are highly concentrated in the background, and the front-end perception unit is only responsible for data collection, resulting in a serious separation between the perception and analysis links. This not only causes the amount of returned data to occupy the communication bandwidth, but also makes the accurate judgment of complex or subtle defects still rely on a large amount of manual screening. In addition, it has poor robustness to interference factors such as weather changes and lighting conditions, resulting in a significant reduction in the frequency of effective monitoring under harsh conditions and significant fluctuations in the accuracy and stability of the output results. This especially restricts the ability to provide continuous and reliable monitoring in remote and dangerous areas with harsh environments.

[0004] The information disclosed in this background technology section is only intended to deepen the understanding of the overall background technology of the present disclosure and should not be regarded as an admission or any form of suggestion that the information constitutes the prior art known to those skilled in the art. Summary of the Invention

[0005] The present invention provides an electric power inspection system and a satellite image processing method, which can effectively solve the problems in the background technology.

[0006] In order to achieve the above object, the technical solution adopted by the present invention is:

[0007] A power inspection system, comprising:

[0008] The front-end sensing module collects status data of power equipment in the specified area and establishes an inspection route after receiving the inspection instruction;

[0009] An edge processing module receives the status data and performs real-time processing and feature extraction based on the inspection path to generate key parameter information representing the operating status of the power equipment;

[0010] A central analysis platform receives the key parameter information and performs diagnostic analysis of abnormal conditions of the power equipment based on the key parameter information, and generates an inspection and diagnosis report;

[0011] A feedback optimization module, which updates the feature extraction rules of the edge processing module and the intelligent diagnosis and analysis model of the central analysis platform based on the inspection diagnosis report and the key parameter information;

[0012] Inspection result module, sending the inspection diagnosis report to the operation and maintenance terminal;

[0013] Among them, the data volume of the key parameter information generated by the edge processing module is smaller than the data volume of the original status data collected by the front-end perception module.

[0014] Furthermore, the edge processing module is deployed on at least one of a satellite, a drone or a substation edge computing node.

[0015] Furthermore, the edge processing module performs real-time processing and feature extraction of the state data, including:

[0016] Perform power equipment target detection on optical or infrared images to generate an initial detection box containing the equipment coordinates and the corresponding confidence score;

[0017] Filtering the initial detection frames based on a preset confidence threshold, retaining the initial detection frames with the confidence scores meeting the criteria, and extracting feature vectors of abnormal regions in the initial detection frames;

[0018] Obtain laser point cloud data and perform compression encoding to generate the three-dimensional structural parameters of the equipment;

[0019] Spatially aligning the abnormal region feature vector with the three-dimensional structural parameters of the device to generate a spatial spectrum feature code;

[0020] The key parameter information is generated based on the spatial spectrum feature code, and the data integrity is verified, and only the key parameter information that passes the verification is output.

[0021] Furthermore, the feature extraction rules of the updated edge processing module include:

[0022] Dynamically adjusting the confidence threshold according to the false alarm rate in the inspection diagnosis report;

[0023] According to the utilization rate of the key parameter information in the diagnostic analysis, the spatial alignment strategy is optimized. If the utilization rate of the three-dimensional structure parameters is lower than the set value, the original point cloud registration algorithm is replaced by an alignment method based on feature point matching;

[0024] Based on the verification result of the data integrity, the dimensions of the spatial spectral feature encoding are added or deleted.

[0025] Furthermore, the alignment method based on feature point matching includes:

[0026] Extract SIFT feature points from the acquired image and generate normal vector features at the corresponding positions of the point cloud data;

[0027] Matching the SIFT feature points with the normal vector features to establish a point mapping relationship;

[0028] The three-dimensional coordinates of the abnormal area feature vector are bound based on the image point mapping relationship.

[0029] Furthermore, the intelligent diagnostic analysis model constructs a multi-source decision network based on equipment structural characteristics, thermal anomaly characteristics and semantic recognition results, among which semantic recognition parses the equipment name entities and status description words in the patrol personnel's voice instructions, and dynamically associates them with the three-dimensional coordinates and thermal map of the corresponding equipment to generate diagnostic labels.

[0030] Furthermore, the feedback optimization module triggers an update when the false alarm rate of the inspection diagnosis report exceeds a threshold or the quality score of the key parameter information is lower than a standard score.

[0031] Furthermore, the edge processing module constructs a feature vector with spatiotemporal evolution characteristics based on the equipment degradation characteristics in the historical retrieval data and the abnormal areas in the real-time detection data, wherein the historical retrieval data is dynamically associated with the three-dimensional coordinates of the current detection target based on equipment code matching.

[0032] A satellite image processing method, the method comprising:

[0033] Receive multispectral satellite time-series images and generate a surface coverage map of the area to be analyzed based on cloud filtering and terrain correction;

[0034] Matching the transmission tower positioning points in the surface coverage map based on the device spatial coordinate library, and constructing a corridor buffer with adjustable width based on the positioning points;

[0035] Identify wildfire hotspots within the corridor buffer zone, detect vegetation encroachment distances, and track surface deformation displacement;

[0036] The wildfire intensity, vegetation encroachment rate, and surface displacement are encapsulated into a spatiotemporal evolution matrix and transmitted to the edge processing module of the power inspection system.

[0037] Furthermore, the processing of wildfire hotspots, detecting vegetation encroachment distance, and tracking surface deformation and displacement includes:

[0038] The vegetation encroachment distance detection compares historical images of the same season to generate a dynamic change threshold, and triggers an alarm when the real-time vegetation index increase exceeds the historical fluctuation range constructed by the dynamic change threshold;

[0039] The identification of the wildfire hotspot corrects the thermal radiation risk level according to real-time meteorological parameters, and pushes an early warning instruction when the thermal radiation risk level exceeds a critical threshold;

[0040] The surface deformation displacement is obtained by analyzing the interference phase data of the associated area of ​​the equipment, filtering out non-related deformation components, and obtaining a displacement acceleration trend, and generating a settlement warning based on the displacement acceleration trend.

[0041] The technical solution of the present invention can achieve the following technical effects:

[0042] It effectively solves the problems of low inspection efficiency, high safety risks and delayed response caused by the inability of traditional inspection methods to achieve efficient, real-time and automated fault detection and status monitoring in remote or dangerous areas.

[0043] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments recorded in the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0045] Figure 1 This is a structural diagram of a power inspection system;

[0046] Figure 2 This is a conceptual diagram of satellite-based power inspection processing;

[0047] Figure 3 Schematic diagram of the feature extraction process;

[0048] Reference numerals: 100, transmission tower; 200, power inspection unit; 300, satellite; 400, signal receiving device; 500, analysis platform. DETAILED DESCRIPTION

[0049] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments.

[0050] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the art to which this invention pertains. The terms used in this specification are for the purpose of describing specific embodiments only and are not intended to limit the invention. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0051] Example 1

[0052] like Figure 1 As shown, the present application provides a power inspection system, the system comprising:

[0053] The front-end sensing module collects status data of power equipment in the specified area and establishes an inspection route after receiving the inspection instruction;

[0054] The edge processing module receives status data and performs real-time processing and feature extraction based on the inspection path to generate key parameter information that characterizes the operating status of power equipment;

[0055] The central analysis platform receives key parameter information and performs diagnostic analysis of abnormal conditions of power equipment based on it, generating inspection and diagnosis reports;

[0056] The feedback optimization module updates the feature extraction rules of the edge processing module and the intelligent diagnosis and analysis model of the central analysis platform based on inspection and diagnosis reports and key parameter information;

[0057] Inspection result module sends inspection diagnosis report to the operation and maintenance terminal;

[0058] Among them, the data volume of the key parameter information generated by the edge processing module is smaller than the data volume of the original status data collected by the front-end perception module.

[0059] Specifically, first of all, the power inspection system consists of the following modules: front-end perception module, edge processing module, central analysis platform, feedback optimization module, and inspection result module; after receiving the inspection instruction, the front-end perception module is responsible for real-time collection of power equipment status data in the specified area. These data may include information read by sensors, such as voltage, current, temperature and other parameters. In order to ensure the accuracy and integrity of data collection, the front-end perception module should adopt high-precision sensing equipment and combine satellite and ground networks for automatic path planning to establish reasonable inspection routes to cover all target equipment; then, the data is transmitted to the edge processing module. The edge processing module is responsible for real-time processing and feature extraction in the system architecture and performs rapid feature extraction to identify the operating status of power equipment. The key to the edge processing module is to use optimized algorithms to simplify a large amount of raw data into key parameter information to ensure that the data volume is reduced and the content is complete. This not only improves the system transmission efficiency, but also avoids data redundancy, laying the foundation for subsequent analysis; after receiving the key parameter information, the central analysis platform conducts in-depth diagnostic analysis The central analysis platform uses artificial intelligence technology and big data analysis methods to identify and analyze abnormal conditions of transmitted parameters. By establishing a set of intelligent diagnostic models, it updates and optimizes diagnostic rules in real time to generate accurate inspection and diagnosis reports. The inspection and diagnosis reports detail the current health status of the equipment, potential faults, and even predict the service life. To ensure the accuracy of diagnostic analysis and system iterative evolution, the feedback optimization module uses information from the inspection and diagnosis reports and continuously adjusts the feature extraction rules in the edge processing module and the diagnostic model of the central analysis platform through machine learning technology, gradually optimizing the analysis accuracy and reliability of the entire system. Finally, the inspection result module is responsible for the distribution and display of inspection and diagnosis reports. The inspection result module transmits the final diagnostic results to the operation and maintenance terminal so that operation and maintenance personnel can take immediate maintenance measures. In addition, the inspection result module can generate a historical database to provide reference for subsequent inspection work. The mutual collaboration of these modules enables the system to perform real-time inspections of power equipment in complex environments, especially in remote and dangerous areas, ensuring the safety and efficiency of inspections.

[0060] The technical solution of the present invention effectively solves the problems of low inspection efficiency, high safety risks and response delays caused by the inability of traditional inspection methods to achieve efficient, real-time and automated fault detection and status monitoring in remote or dangerous areas.

[0061] Further, if Figure 2 As shown, the edge processing module is deployed on at least one of a satellite, a drone, or a substation edge computing node.

[0062] As a preferred embodiment of the above, the first is satellite deployment. The edge processing module is deployed on the satellite 300, so that the power equipment in large areas and remote areas can be monitored in real time over a wide area. The satellite 300 has the ability to cruise at high altitude, which enables it to cover a large number of ground equipment such as transmission towers 100. Through the high-precision sensors and data collection equipment carried by the satellite 300, the edge processing module can receive and process a large amount of raw data in real time. The satellite-side edge computing can directly perform data simplification and preliminary analysis in space, identify potential faults or abnormal conditions in time, and prevent unnecessary data from being transmitted back to the ground, thereby reducing communication bandwidth occupancy and latency problems; the next is drone deployment. The edge processing module is deployed on the drone to provide flexible mobility, which is particularly suitable for rapid inspections of small areas or substations and lines that need to be observed in key areas. For example, in an emergency, the drone can quickly reach a specific location for deployment. For precise inspections, the maneuverability and flexibility of drones allow them to approach equipment and acquire data from various angles. The edge processing module onboard a drone offers a significant advantage in that it can continuously process data during flight, reducing the burden of post-flight data processing. Under appropriate conditions, processed critical data can be instantly uploaded to the analysis platform 500 via the signal receiving device 400, enabling faster analysis and decision-making regarding equipment status. Finally, substation edge computing node deployment involves deploying edge processing modules within the substation. The multiple power inspection units 200 contained within the edge processing module enable rapid processing and analysis of status data for all equipment within the substation. Due to the relatively stable environment within the substation, the edge computing module operates efficiently, not only monitoring current equipment status but also storing more historical data to support long-term status trend analysis. This enables more comprehensive real-time monitoring and fault prediction, maintaining the continuity and safety of substation operations. In any practical application scenario, the edge processing module is highly integrated with the existing front-end sensing module to adapt to data collection needs in various environments. By utilizing edge computing at various deployment locations, local data processing is accelerated, improving the efficiency and responsiveness of the entire inspection system.

[0063] Further, if Figure 3 As shown in the figure, the edge processing module performs real-time processing and feature extraction of status data, including:

[0064] Perform power equipment target detection on optical or infrared images to generate an initial detection box containing the equipment coordinates and the corresponding confidence score;

[0065] Filter the initial detection frames based on the preset confidence threshold, retain the initial detection frames with qualified confidence scores, and extract the feature vectors of abnormal areas in the initial detection frames;

[0066] Obtain laser point cloud data and perform compression encoding to generate the three-dimensional structural parameters of the equipment;

[0067] Spatially align the feature vector of the abnormal area with the three-dimensional structural parameters of the equipment to generate a spatial spectrum feature code;

[0068] Generate key parameter information based on spatial spectral feature encoding, verify data integrity, and only output key parameter information that passes verification.

[0069] As a preferred embodiment of the above, the edge processing module first performs target detection on the collected optical or infrared image. Through the convolutional neural network model based on deep learning, it can automatically identify the power equipment target in the image and generate an initial detection frame and its corresponding confidence score. The data of the detection frame includes the coordinate information of the equipment, which lays the foundation for subsequent data processing; then, the edge processing module filters the initial detection frame according to the preset confidence threshold. The confidence scores of all detection frames must meet the predetermined standard to be retained. This process ensures that subsequent analysis focuses on high-confidence areas and improves the accuracy of anomaly detection. In the retained detection frame, the abnormal area will be feature-encoded by feature vector extraction technology to form a feature vector. The feature vector can describe the image features of the area, such as texture or color changes, and will provide data support for spatial spectral feature encoding; at the same time, the edge processing module obtains the point cloud data provided by the laser equipment, and generates the three-dimensional structural parameters of the power equipment through compression coding technology. The implementation of compression coding technology It helps to reduce the amount of data and improve processing efficiency while ensuring accuracy. The three-dimensional structure of the equipment is extremely critical in practical applications because the spatial structure of the power equipment directly affects its function and safety. The next step is to spatially align the previously extracted abnormal area feature vector with the three-dimensional structural parameters of the equipment to generate a complete spatial spectral feature code. The spatial alignment process needs to consider the actual geographic coordinates and structural characteristics of the equipment to ensure the consistency of the feature code with the equipment's own status. This step combines image information and three-dimensional structural information to achieve accurate positioning and identification of equipment abnormalities. Finally, based on the generated spatial spectral feature code, the edge processing module generates the system's key parameter information. The key parameter information is not only concise but also contains the most important equipment status data. Verifying data integrity is the last step of the preferred implementation plan: the edge processing module performs data verification to ensure that all output key parameter information has been accurately detected and verified. Only verified information will be output, providing a data basis for further analysis by the central analysis platform.

[0070] Specifically, the feature extraction rules for updating the edge processing module include:

[0071] Dynamically adjust the confidence threshold based on the false alarm rate in the inspection diagnosis report;

[0072] Based on the utilization rate of key parameter information in diagnostic analysis, the spatial alignment strategy is optimized. If the utilization rate of the 3D structural parameters is lower than the set value, the original point cloud registration algorithm is replaced by an alignment method based on feature point matching.

[0073] Based on the verification results of data integrity, the dimensions of spatial spectral feature encoding are added or deleted.

[0074] As a preferred embodiment of the above, first, the feature extraction rules of the edge processing module are updated with the goal of reducing the false alarm rate. During the operation of the system, the inspection diagnosis report provides detailed false alarm rate data. Based on this data, the edge processing module dynamically adjusts the confidence threshold. This is a key link. The adjustment process is based on historical data and current false alarm conditions. If the false alarm rate is high, the edge processing module will moderately increase the confidence threshold to reduce false alarms, and vice versa. Such adjustments can improve the accuracy of the evaluation target in subsequent inspections; then, the edge processing module will evaluate the utilization of key parameter information in the diagnostic analysis. If the utilization rate of the three-dimensional structure parameters is lower than the set value, this indicates that the current point cloud registration algorithm may be too complex or inappropriate. To adapt to actual scenarios, the system will switch to an alignment strategy based on feature point matching. This method achieves more efficient alignment by identifying and matching feature points. It is especially suitable for situations where the point cloud data quality is poor or the scene is too complex. This optimization strategy ensures efficient data processing while maintaining the validity of key parameter information. In addition, the edge processing module also continuously adjusts the dimensions of spatial spectral feature encoding to adapt to different data integrity requirements. This is achieved by analyzing the data integrity results during the inspection process. For features that are found to be ignored during the verification process, the edge processing module can add corresponding dimensions to capture more information, and for features that are redundant or noisy, the relevant dimensions are deleted in a timely manner to optimize the overall quality and simplicity of the data.

[0075] Furthermore, alignment methods based on feature point matching include:

[0076] Extract SIFT feature points from the acquired image and generate normal vector features at the corresponding positions of the point cloud data;

[0077] Match SIFT feature points and normal vector features to establish a point mapping relationship;

[0078] The three-dimensional coordinates of the feature vectors of the abnormal area are bound based on the point mapping relationship.

[0079] As a preferred embodiment of the above, by extracting scale-invariant feature transformation, namely SIFT feature points, from the collected image, as image visual features, these feature points can capture unique and stable key points in the image, and can maintain consistency regardless of scaling or rotation. The extraction of SIFT feature points is achieved through a specific algorithm, which identifies significant edges and texture features in the image, thereby providing a solid foundation for subsequent matching and binding; at the same time, corresponding to the position of the acquired point cloud data, the edge processing module generates normal vector features, which are generated by reconstructing the three-dimensional information of the point cloud. Each normal vector represents a part of the geometric shape of the point cloud surface. The step of generating normal vectors not only eliminates noise, but also effectively highlights the geometric characteristics of the device surface, laying the foundation for subsequent point mapping relationships; then, the edge processing module performs feature point matching and establishes a point mapping relationship. This process is achieved by matching SIFT features. Compare the features of the feature points and normal vectors, identify the correspondence between the two, and establish the mapping relationship based on the spatial consistency of the feature points. By creating a certain mapping matrix to connect the spatial positions of the image and point cloud, this mapping relationship is like a bridge, helping to accurately locate the two-dimensional image features in the three-dimensional space of the equipment; finally, based on the point mapping relationship, the three-dimensional coordinates of the abnormal area feature vectors are bound. At this time, the abnormal area feature vectors are assigned specific three-dimensional coordinates to complete the spatial binding, making anomaly detection more accurate and meaningful. The three-dimensional coordinate binding process not only considers the spatial position of each feature point, but also considers the correlation between the feature points and the neighborhood data to ensure the accuracy of anomaly detection; through this implementation plan, the use of feature point matching and point mapping relationship improves the reliability and accuracy of abnormal area detection, reduces errors in data processing, and optimizes the overall efficiency and benefits of the power inspection system.

[0080] Furthermore, the intelligent diagnostic analysis model constructs a multi-source decision network based on the equipment structure characteristics, thermal anomaly characteristics and semantic recognition results. The semantic recognition parses the equipment name entities and status description words in the patrol personnel's voice instructions, and dynamically associates them with the three-dimensional coordinates and thermal map of the corresponding equipment to generate diagnostic labels.

[0081] As a preferred embodiment of the above, the intelligent diagnostic analysis model combines the equipment structural characteristics, thermal anomaly characteristics and semantic recognition results to construct a multi-source decision network. First, the equipment structural characteristics define the physical properties and spatial distribution of the equipment through point cloud data and three-dimensional modeling, providing a basic framework to support the analysis model. At the same time, through infrared or thermal imaging technology, the thermal anomaly characteristics of the equipment are extracted to identify possible problems with the equipment, such as overheating parts or poorly cooled areas. These characteristics are combined to form a multi-dimensional data foundation, providing a variety of information sources for subsequent diagnosis; on this basis, semantic recognition and analysis play a key role. During on-site operations, inspectors may use voice equipment to describe the detected equipment status or problems. Through natural language processing technology, the diagnostic model can intelligently parse the equipment name entity and status description words in the voice command, and the parsing results are dynamically associated with the three-dimensional coordinates of the equipment. As well as thermal maps, accurate diagnostic labels are generated for equipment anomalies. This process also reflects that semantic recognition technology, as a cutting-edge application in intelligent diagnostic analysis, effectively integrates manual input and automated analysis; next, the multi-source decision network will unify the equipment structural characteristics, thermal anomaly characteristics and semantic recognition results to perform a comprehensive analysis. This part uses complex algorithms and machine learning technologies to integrate and analyze multiple data sources to ensure the accuracy of the generated diagnostic results. The decision rules and parameter settings in the network are self-adjusted in a timely manner. Through this intelligent mechanism, the system can continuously optimize the diagnostic strategy and output quality during operation; Combining these technical implementation details, the intelligent diagnostic analysis model demonstrates the ability to process multi-source information and its application potential in power equipment inspection. Through multi-angle feature fusion, the multi-source decision network can provide support for accurate diagnosis and preventive maintenance of equipment inspection, while achieving dynamic adaptation and real-time response.

[0082] Furthermore, the feedback optimization module triggers an update when the false alarm rate of the inspection diagnosis report exceeds a threshold or the quality score of the key parameter information is lower than the standard score.

[0083] As a preferred embodiment of the above, the core of the feedback optimization module lies in its continuous monitoring and flexible adjustment of system performance. When the false alarm rate of the inspection diagnosis report exceeds the preset threshold or the quality score of the key parameter information is lower than the standard score, the feedback optimization module triggers an update process. This update mechanism is to ensure that the power inspection system can respond promptly to conditions of poor data quality or high diagnostic errors to improve the overall inspection efficiency and accuracy. First, the system strictly monitors the false alarm rate and identifies situations in the inspection report where the false alarm frequency exceeds the threshold. After identifying this situation, the feedback optimization module will automatically re-evaluate the parameter settings and algorithm strategies of the current edge processing module and the central analysis platform. The false alarm rate is a test of diagnostic accuracy. By dynamically adjusting the confidence threshold or other diagnostic rules, it can effectively reduce The false alarm rate enables the system to quickly restore to an accurate detection state; secondly, the quality score of key parameter information is the evaluation standard for data integrity and validity. A quality score report is generated for each inspection. If the score is lower than the preset standard, the feedback optimization module will analyze potential problems in the data collection and processing process and explore possible parameter improvement strategies. This process may involve fine-tuning of feature extraction algorithms, data compression coding and other technologies to ensure that the output key parameter information meets the standards in quality and practicality, and adopts advanced machine learning technology to achieve adaptive changes to ensure that the quality of future inspections continues to improve; through this automated update and optimization mechanism, the feedback optimization module not only improves the agility and adaptability of the power inspection system, but also ensures that the complex needs of smart power equipment management are met through continuous technological innovation.

[0084] Furthermore, the edge processing module constructs a feature vector with spatiotemporal evolution characteristics based on the equipment degradation characteristics in the historical retrieval data and the abnormal areas in the real-time detection data, where the historical retrieval data is dynamically associated with the three-dimensional coordinates of the current detection target based on device code matching.

[0085] As a preferred embodiment of the above, the edge processing module obtains the equipment degradation characteristics in the historical retrieval data. These characteristics are accumulated in the long-term monitoring and data analysis of the equipment, including the service life of the equipment, environmental impact and specific degradation manifestations such as corrosion and wear. The historical retrieval data is matched through the equipment code to ensure dynamic association with the current detection target. The equipment code is a unique identifier that can accurately locate the specific equipment to ensure the accuracy of the association process; at the same time, the edge processing module also continuously processes abnormal areas in the real-time detection data. These abnormal areas are detected by high-precision sensors, including voltage fluctuations, temperature anomalies, deformation, etc., representing instant feedback on the current status of the equipment. Real-time detection technology can quickly capture dynamic changes and provide the information required for real-time analysis; the innovation of the edge processing module lies in combining historical retrieval data with real-time detection data, through data fusion Generate a feature vector with spatiotemporal evolution characteristics because it can simultaneously describe the historical degradation trend and current abnormal performance of the equipment. This enables the feature vector to not only reflect the health status of the equipment, but also predict potential risks in the future. This spatiotemporal characteristic provides important data support for equipment maintenance and risk warning. In the process of constructing the feature vector, the edge processing module dynamically associates the three-dimensional coordinates of the equipment. The three-dimensional coordinates provide the positioning information of the equipment in space, ensuring that historical data and real-time data can be integrated at the correct spatial position. This process involves complex coordinate alignment technology to ensure the precise combination of spatiotemporal data, thereby improving the accuracy and reliability of data analysis. Through this implementation method, the edge processing module realizes the timeliness monitoring and predictive analysis of power equipment, which not only supports more efficient equipment maintenance plans, but also provides an accurate information basis for the system's intelligent diagnosis.

[0086] Example 2

[0087] Based on the same inventive concept as the power inspection system in the aforementioned embodiment, the present invention further provides a satellite image processing method, the method comprising:

[0088] Receive multispectral satellite time-series images and generate a surface coverage map of the area to be analyzed based on cloud filtering and terrain correction;

[0089] Based on the equipment spatial coordinate library, the transmission tower positioning points in the surface coverage map are matched, and a corridor buffer with adjustable width is constructed based on the positioning points.

[0090] Identify wildfire hotspots within corridor buffers, detect vegetation encroachment distances, and track surface deformation and displacement;

[0091] The wildfire intensity, vegetation encroachment rate, and surface displacement are encapsulated into a spatiotemporal evolution matrix and transmitted to the edge processing module of the power inspection system.

[0092] Specifically, the initial step of the satellite image processing method is to receive multispectral satellite time-series images, which cover geographical area images of different time periods and spectra and have rich information layers. After receiving these images, cloud-covered areas in the satellite images are removed through cloud filtering technology to ensure accurate identification of surface features. In addition, the image is geometrically corrected using a terrain correction algorithm to ensure that various surface features are accurately mapped in the image. A surface coverage map of the area to be analyzed is then generated, which shows the coverage distribution of vegetation, exposed surface, etc., providing a basis for subsequent analysis. Then, the generated surface coverage map is matched with the device spatial coordinate library. By identifying and locating the precise position of the transmission tower and using the matched transmission tower positioning point as a benchmark, a corridor buffer with adjustable width is constructed to cover the key impact area around the transmission tower. The buffer width is adjusted to meet the needs of different geographical environments and analysis purposes, ensuring sufficient coverage of potential risk factors. Within the corridor buffer, image analysis technology is used to identify wildfire hotspots, detect the distance of vegetation encroachment, and track surface deformation and displacement. Wildfire hotspot identification is based on thermal imaging information and can quickly detect potential fire locations. Vegetation encroachment is analyzed using the greening index to identify vegetation expansion near transmission lines. Surface deformation is identified by comparing and analyzing time-series images to identify displacement changes. These identification and detection measures ensure that the main external risks faced by transmission lines are checked. The identification results, including wildfire intensity, vegetation encroachment rate and surface displacement, will be organized into a spatiotemporal evolution matrix. This matrix combines time, space and numerical characteristics to comprehensively display the changing trends of various risk factors. The spatiotemporal evolution matrix is ​​transmitted to the edge processing module of the power inspection system for further data integration and analysis to improve the efficiency and accuracy of the system's inspection decision-making. This satellite image processing method, by combining multispectral and time series data analysis, provides a highly efficient and accurate external environment monitoring method for power inspection, especially in applications over wide areas and complex terrain.

[0093] Specifically, the processing of wildfire hotspots, detecting vegetation encroachment distances, and tracking surface deformation displacements includes:

[0094] Vegetation encroachment distance detection compares historical images of the same season to generate a dynamic change threshold. When the real-time vegetation index increase exceeds the historical fluctuation range constructed by the dynamic change threshold, an alarm is triggered.

[0095] Identification of wildfire hotspots modifies the thermal radiation risk level based on real-time meteorological parameters, and pushes warning instructions when the thermal radiation risk level exceeds the critical threshold;

[0096] The surface deformation displacement is analyzed based on the interference phase data of the associated area of ​​the equipment. After screening out the non-related deformation components, the displacement acceleration trend is obtained, and a settlement warning is generated based on the displacement acceleration trend.

[0097] As a preferred embodiment of the above, first, vegetation encroachment distance detection is performed. The edge processing module extracts the vegetation index obtained in the current season from the satellite image and compares it with the index extracted from the historical image of the same season. By analyzing the historical data, a dynamic change threshold is generated. This threshold reflects the upper and lower limits of the vegetation index within the normal seasonal variation range. When the increase in the vegetation index detected in real time exceeds this historical fluctuation range, a corresponding alarm will be triggered to prevent vegetation from encroaching on power equipment. The next step is the identification of wildfire hotspots. This method combines real-time meteorological parameters and identifies wildfire hotspots by correcting the thermal radiation risk level. This correction process takes into account the impact of environmental factors on thermal radiation measurement to ensure that the risk level accurately reflects the risk level. For the current fire risk, once the revised thermal radiation risk level exceeds the set critical threshold, an early warning instruction will be immediately pushed to remind relevant personnel to take preventive measures; finally, the surface deformation displacement is tracked, and the interference phase data of the equipment-related area is obtained and analyzed to filter out non-related deformation components, which may include natural terrain changes or display errors. The filtered data is used to evaluate the acceleration trend of surface displacement, which is a key indicator for predicting potential surface subsidence. When the acceleration trend indicates that there is a large risk of subsidence in the equipment area, a corresponding subsidence warning is generated to provide a decision-making basis for preventing further geological disasters; through the precise extraction and mapping of feature points, efficient feature alignment and three-dimensional coordinate binding of abnormal areas are achieved.

[0098] Although the present application has been described with reference to specific features and embodiments thereof, it is apparent that various modifications and combinations may be made thereto without departing from the spirit and scope of the present application. Accordingly, this specification and drawings are merely illustrative of the present application as defined herein and are deemed to cover any and all modifications, variations, combinations or equivalents within the scope of the present application. Obviously, those skilled in the art may make various modifications and variations to the present application without departing from the scope of the present application. Thus, the present application is intended to include such modifications and variations as fall within the scope of the present application and its equivalents.

Claims

1. A power inspection system, characterized in that: The system comprises: The front-end sensing module collects status data of power equipment in the specified area and establishes an inspection route after receiving the inspection instruction; The edge processing module receives the status data and performs real-time processing and feature extraction based on the inspection path to generate key parameter information representing the operating status of the power equipment, including: Perform power equipment target detection on optical or infrared images to generate an initial detection box containing the equipment coordinates and the corresponding confidence score; Filtering the initial detection frames based on a preset confidence threshold, retaining the initial detection frames with the confidence scores meeting the criteria, and extracting feature vectors of abnormal regions in the initial detection frames; Obtain laser point cloud data and perform compression encoding to generate the three-dimensional structural parameters of the equipment; Spatially aligning the abnormal region feature vector with the three-dimensional structural parameters of the device to generate a spatial spectrum feature code; Generate the key parameter information based on the spatial spectrum feature code, verify data integrity, and only output the key parameter information that passes the verification; A central analysis platform receives the key parameter information and performs diagnostic analysis of abnormal conditions of the power equipment based on the key parameter information, and generates an inspection and diagnosis report; A feedback optimization module updates the feature extraction rules of the edge processing module and the intelligent diagnosis and analysis model of the central analysis platform based on the inspection and diagnosis report and the key parameter information, wherein the updating of the feature extraction rules of the edge processing module includes: Dynamically adjusting the confidence threshold according to the false alarm rate in the inspection diagnosis report; According to the utilization rate of the key parameter information in the diagnostic analysis, the spatial alignment strategy is optimized. If the utilization rate of the three-dimensional structure parameters is lower than the set value, the original point cloud registration algorithm is replaced by an alignment method based on feature point matching; Based on the verification result of the data integrity, adding or deleting the dimension of the spatial spectral feature encoding; Inspection result module, sending the inspection diagnosis report to the operation and maintenance terminal; Among them, the data volume of the key parameter information generated by the edge processing module is smaller than the data volume of the original status data collected by the front-end perception module.

2. The power inspection system according to claim 1, characterized in that: The edge processing module is deployed on at least one of a satellite, a drone or a substation edge computing node.

3. The power inspection system according to claim 1, characterized in that: Alignment methods based on feature point matching include: Extract SIFT feature points from the acquired image and generate normal vector features at the corresponding positions of the point cloud data; Matching the SIFT feature points with the normal vector features to establish a point mapping relationship; The three-dimensional coordinates of the abnormal area feature vector are bound based on the image point mapping relationship.

4. The power inspection system according to claim 1, characterized in that: The intelligent diagnostic analysis model constructs a multi-source decision network based on equipment structural characteristics, thermal anomaly characteristics and semantic recognition results. Semantic recognition parses the equipment name entities and status descriptors in the inspector's voice instructions, and dynamically associates them with the three-dimensional coordinates and thermal map of the corresponding equipment to generate diagnostic labels.

5. The power inspection system according to claim 1, characterized in that: The feedback optimization module triggers an update when the false alarm rate of the inspection diagnosis report exceeds a threshold or the quality score of the key parameter information is lower than a standard score.

6. The power inspection system according to claim 1, characterized in that: The edge processing module constructs a feature vector with spatiotemporal evolution characteristics based on the equipment degradation characteristics in the historical retrieval data and the abnormal areas in the real-time detection data, wherein the historical retrieval data is dynamically associated with the three-dimensional coordinates of the current detection target based on equipment code matching.

7. A satellite image processing method, characterized in that: Applied to the power inspection system according to claim 1, the method comprises: Receive multispectral satellite time-series images and generate a surface coverage map of the area to be analyzed based on cloud filtering and terrain correction; Matching the transmission tower positioning points in the surface coverage map based on the device spatial coordinate library, and constructing a corridor buffer with adjustable width based on the positioning points; Identify wildfire hotspots within the corridor buffer zone, detect vegetation encroachment distances, and track surface deformation displacement; The wildfire intensity, vegetation encroachment rate, and surface displacement are encapsulated into a spatiotemporal evolution matrix and transmitted to the edge processing module of the power inspection system.

8. The satellite image processing method according to claim 7, wherein: Processing of wildfire hotspots, detecting vegetation encroachment distances, and tracking surface deformation and displacement, including: The vegetation encroachment distance detection compares historical images of the same season to generate a dynamic change threshold, and triggers an alarm when the real-time vegetation index increase exceeds the historical fluctuation range constructed by the dynamic change threshold; The identification of the wildfire hotspot corrects the thermal radiation risk level according to real-time meteorological parameters, and pushes an early warning instruction when the thermal radiation risk level exceeds a critical threshold; The surface deformation displacement is obtained by analyzing the interference phase data of the associated area of ​​the equipment, filtering out non-related deformation components, and obtaining a displacement acceleration trend, and generating a settlement warning based on the displacement acceleration trend.

Images