A method and system for aerial orbital thermal imaging temperature inspection

Abstract

This invention relates to the field of thermal imaging technology, and more particularly to an aerial orbital thermal imaging temperature inspection method and system. The method includes the following steps: identifying evaluation areas in the area to be inspected to obtain a list of evaluation areas; refining the thermally sensitive areas based on the list to obtain refined regions of interest; generating a topological interest map from the refined regions of interest to obtain a topological interest map; extracting inspection points from the topological interest map and calculating the target attitude to obtain a target attitude sequence; allocating motion parameters based on the target attitude sequence to obtain segmented motion parameters; generating a motion control sequence from the segmented motion parameters to obtain a motion control sequence; and generating a spatiotemporal thermal spectrum array based on the motion control sequence to obtain a spatiotemporal thermal spectrum array. This invention, by combining artificial intelligence and thermal imaging technology, achieves safer, more efficient, intelligent, and accurate inspection.

Description

Technical Field

[0001] This invention relates to the field of thermal imaging technology, and in particular to an aerial orbital thermal imaging temperature inspection method and system. Background Technology

[0002] Hotspot or thermal anomaly detection is crucial for large, complex industrial equipment (e.g., transformers, electrical panels, piping). Traditional methods, such as manual inspections, are inefficient, costly, time-consuming, labor-intensive, and pose safety hazards. They often miss subtle issues, requiring inspectors to personally visit each piece of equipment, typically in challenging environments. This consumes significant time and labor resources, leading to high operating costs. Scheduling inspections disrupts operations, and their frequency is limited by cost and resource constraints. Traditional methods, including basic automation, typically process thermal data in isolation, lacking contextual understanding and intelligent analysis of thermal anomalies. This results in: slightly elevated temperatures on non-critical components triggering alarms and wasting resources on unnecessary investigations; similar temperature increases on critical components within broad, generic thresholds being ignored. Simply identifying a hotspot does not provide insight into the root cause or the potential severity of the problem. Summary of the Invention

[0003] Therefore, it is necessary to provide an aerial orbital thermal imaging temperature inspection method to solve at least one of the above-mentioned technical problems.

[0004] To achieve the above objectives, an aerial orbital thermal imaging temperature inspection method includes the following steps: Step S1: Identify the evaluation areas of the area to be inspected to obtain a list of areas to be evaluated; refine the heat-sensitive areas based on the list of areas to be evaluated to obtain refined regions of interest; generate a topological interest map from the refined regions of interest to obtain a topological interest map. Step S2: Extract inspection points from the topological interest map and calculate the target attitude to obtain the target attitude sequence; allocate motion parameters according to the target attitude sequence to obtain segmented motion parameters; generate motion control sequences from the segmented motion parameters to obtain motion control sequences. Step S3: Acquire continuous radiation data according to the motion control sequence and provide real-time feedback on the orbital position to obtain time-stamped thermal images and time-stamped position information; generate a spatiotemporal thermal spectrum array from the time-stamped thermal images and time-stamped position information to obtain the spatiotemporal thermal spectrum array. Step S4: Calculate the local temperature deviation based on the spatiotemporal thermal spectrum array, and perform preliminary anomaly region identification to obtain a preliminary anomaly mask; extract the contextualized anomaly mask from the preliminary anomaly mask, and generate priority anomaly registration data. Step S5: Generate a maintenance action plan based on the priority anomaly registration data to obtain optional maintenance plans; optimize the maintenance schedule based on the optional maintenance plans and generate an optimized maintenance work order to obtain an optimized maintenance work order.

[0005] This invention achieves precise definition and priority allocation of inspection areas by automatically identifying assets, refining thermally sensitive areas, and generating topological interest maps. This improves inspection efficiency, reduces unnecessary scanning, and provides accurate foundational information for subsequent path planning and anomaly detection. Based on the topological interest map and target attitude, it intelligently plans inspection paths and motion parameters, enabling adaptive trajectory adjustment and prioritizing high-risk areas. This improves inspection efficiency and the quality of inspections in critical areas while reducing unnecessary movement and energy consumption of the track system. Continuous acquisition of thermal imaging data, real-time feedback of track position information, and time synchronization calibration ensure the integrity, accuracy, and spatiotemporal consistency of the acquired data, providing a high-quality data foundation for subsequent spatiotemporal thermal spectrum analysis and anomaly detection. By constructing a baseline temperature model, calculating local temperature deviations, fusing contextual information, and scoring and filtering anomalies, it achieves accurate identification and assessment of thermal anomalies, improving the accuracy and reliability of anomaly detection and effectively reducing false alarm rates. Through risk assessment, trend prediction, resource demand assessment, and scheduling optimization, it achieves intelligent scheduling and management of maintenance tasks, thereby improving maintenance efficiency, reducing maintenance costs, and effectively avoiding potential equipment failures and safety risks. Therefore, this invention provides an aerial orbital thermal imaging temperature inspection method, which solves the problems of inefficiency and high cost of traditional manual inspections, as well as the lack of contextual understanding and intelligent analysis of thermal anomalies. By combining artificial intelligence and thermal imaging technology, a safer, more efficient, intelligent, and accurate inspection is achieved, and the accuracy of anomaly detection is significantly improved through contextual understanding and intelligent analysis.

[0006] Preferably, step S1 includes the following steps: Step S11: Perform preliminary asset identification in the area to be inspected to obtain an equipment list; Step S12: Associate spatial information based on the equipment list to obtain a spatialized equipment list; Step S13: Based on the spatialized equipment list, integrate the equipment priority information to obtain the list of areas to be evaluated; Step S14: Refine the heat-sensitive areas based on the list of areas to be evaluated to obtain refined regions of interest; Step S15: Construct and optimize the spatial topology of the refined region of interest to obtain the labeled region of interest; Step S16: Generate a topological interest map for the labeled regions of interest to obtain the topological interest map.

[0007] This invention utilizes automated image recognition technology to rapidly identify assets within an inspection area, generating an equipment list. This reduces manual identification workload, improves efficiency, and avoids human error, providing an accurate equipment information foundation for subsequent steps. By associating the equipment in the list with their precise locations in three-dimensional space, a spatialized equipment list is generated. This provides the necessary spatial information for subsequent refinement of heat-sensitive areas and path planning, enabling the system to accurately locate targets requiring inspection. By integrating static priority, historical maintenance data, and real-time operational status data, each piece of equipment is assigned a more precise priority, allowing the system to prioritize high-risk and critical equipment, improving inspection efficiency and anomaly detection accuracy. The area to be evaluated is refined to include critical components and areas prone to thermal anomalies, generating refined regions of interest. This improves the targeting of inspections, avoids ineffective scanning of non-critical areas, saves time and resources, and improves anomaly detection efficiency. By constructing spatial topological relationships between regions of interest and optimizing paths based on the physical constraints of the track system, inspection paths can be effectively planned, reducing track system travel distance and time, improving inspection efficiency, and ensuring effective coverage of all critical areas. The system generates a visualized topological interest map containing the spatial location, priority, equipment information, and topological relationships of all regions of interest. This provides users with an intuitive overview of the inspection area, making it easier for them to understand the inspection scope and priorities. It also supports interactive adjustments and optimizations, improving the system's flexibility and usability.

[0008] Preferably, step S14 includes the following steps: Step S141: Obtain the 3D model based on the list of areas to be evaluated, and perform model preprocessing to obtain preprocessed 3D model data; Step S142: Perform intelligent template matching on the preprocessed 3D model data and the preset hotspot template library to obtain matching hotspot templates; Step S143: Based on the preprocessed 3D model data, perform local parameterization adjustment on the matching hotspot template to obtain the locally adjusted hotspot region; Step S144: Obtain equipment operating parameters; perform thermal simulation-driven hotspot supplementation based on preprocessed 3D model data and equipment operating parameters to obtain simulated supplementary hotspots; Step S145: Generate a refined region of interest based on the local hotspot adjustment area and the simulated supplementary hotspot, thus obtaining the refined region of interest.

[0009] This invention acquires and preprocesses a 3D model of the device to be evaluated, simplifying the model structure, removing noise and redundant information, reducing the complexity of subsequent calculations, improving the algorithm's efficiency, and ensuring model accuracy. This provides a high-quality 3D data foundation for subsequent hotspot identification and analysis. By intelligently matching the preprocessed 3D model with a pre-defined hotspot template library, standard hotspot areas on the device can be quickly identified. Utilizing existing knowledge bases improves the efficiency and accuracy of hotspot identification, laying the foundation for the generation of refined regions of interest. Local parameter adjustments are made to the matched hotspot templates to better match the size and shape of the actual device, improving the accuracy of hotspot area identification and avoiding errors caused by template mismatches. By acquiring device operating parameters and performing thermal simulation, dynamic hotspots generated by changes in device operating status can be identified, supplementing hotspot areas that may not be included in the template library, making hotspot identification more comprehensive and accurate, thereby improving the early warning capability for potential thermal anomalies. By fusing hotspot regions identified through template matching and thermal simulation, redundant information is removed, generating a final refined region of interest. This approach leverages the advantages of different methods, improving the accuracy and completeness of hotspot identification and providing reliable data support for subsequent inspection path planning and thermal anomaly analysis.

[0010] Preferably, step S144 specifically includes: The simulation model is automatically constructed based on the preprocessed 3D model data and equipment operating parameters to obtain the initial simulation model. The initial simulation model is meshed and optimized to obtain an optimized simulation model. Steady-state thermal simulation calculations were performed based on the optimized simulation model and equipment operating parameters to obtain steady-state temperature field data. Transient thermal simulation analysis was performed based on the optimized simulation model to obtain transient temperature change data. Steady-state temperature field data and transient temperature change data are superimposed onto preprocessed 3D model data to identify hotspots and obtain simulated predicted hotspot regions. By associating the simulated hotspots with the template hotspots in the simulated predicted hotspot regions and the locally adjusted hotspot regions, we can obtain the simulated supplementary hotspots.

[0011] This invention automates the construction of simulation models, directly importing pre-processed 3D model data and equipment operating parameters into simulation software and automatically setting boundary conditions and material properties. This avoids the tediousness and errors of manual modeling, improves simulation efficiency, and ensures consistency between the simulation model and the actual equipment. Meshing and optimizing the initial simulation model improves simulation accuracy and efficiency. Appropriate mesh density captures temperature field details more accurately, while optimized models reduce computational load and shorten simulation time. Steady-state thermal simulation provides the temperature distribution of the equipment under stable operating conditions, crucial for identifying persistent hotspots and evaluating the equipment's heat dissipation performance, providing benchmark data for subsequent hotspot identification. Transient thermal simulation analysis captures temperature changes under different operating conditions, such as startup, shutdown, and load variations, helping to identify dynamic hotspots caused by changes in operating conditions, thus providing a more comprehensive understanding of the equipment's thermal characteristics. Overlaying steady-state and transient temperature field data onto the 3D model allows for more intuitive identification of hotspot regions. Combining this with indicators such as average temperature and temperature fluctuation range allows for a more accurate assessment of the severity of hotspots, improving the accuracy and reliability of hotspot identification. By associating the hotspot areas predicted by simulation with the hotspot areas matched by templates, it is possible to distinguish between fixed hotspots caused by equipment structure and dynamic hotspots caused by operating status. This allows for more targeted maintenance and management, and ultimately identifies the hotspot areas that need to be supplemented, avoiding redundant information.

[0012] Preferably, step S2 includes the following steps: Step S21: Extract inspection points from the topological interest map and sort the inspection points to obtain the inspection point sequence; Step S22: Calculate the target attitude of the inspection point sequence to obtain the target attitude sequence; Step S23: Assign motion parameters according to the target posture sequence to obtain segmented motion parameters; Step S24: Generate motor commands from the segmented motion parameters to obtain a discrete motor command set; Step S25: Perform timestamp allocation and synchronization on the discrete motor instruction set to obtain timestamped motor instructions; Step S26: Generate a motion control sequence from the timestamped motor commands to obtain the motion control sequence.

[0013] This invention ensures the completeness and efficiency of the inspection path by extracting and sorting inspection points from the topological interest map. The TSP algorithm is used to plan the shortest path, reducing inspection time and energy consumption, and prioritizing the inspection of high-priority areas, thus improving the inspection efficiency of critical areas. Target attitude calculation determines the optimal observation angle of the thermal imager at each inspection point, ensuring that the imager can capture complete and clear thermal image information of the target area, thereby improving the accuracy of thermal anomaly detection. Based on the target attitude sequence and the priority of the interest region, the motion parameters of the track system are allocated, enabling differentiated inspection of different areas. For example, in high-priority areas, the speed is reduced and the dwell time is increased to obtain more detailed thermal imaging data, thereby improving inspection efficiency and anomaly detection accuracy. Segmented motion parameters are converted into discrete motor instruction sets, transforming abstract motion planning into specific motor control commands, providing a foundation for precise motion control of the track system. Each motor command is assigned a precise timestamp and synchronized with the thermal imager's data acquisition, ensuring the accurate correspondence between thermal image data and position information, laying the foundation for subsequent spatiotemporal data fusion and analysis. The timestamped motor commands are arranged in chronological order to generate the final motion control sequence, providing the track system with complete motion control commands. This ensures that the track system can move precisely according to the predetermined trajectory and speed, completing the entire inspection process.

[0014] Preferably, step S23 includes the following steps: Step S231: Set the priority-weighted velocity reference according to the target attitude sequence to obtain the regional velocity reference; Step S232: Adjust the regional velocity benchmark for uncertainty correction to obtain the uncertainty-adjusted velocity; Step S233: Adjust the speed according to the uncertainty to perform multi-objective optimization speed planning and obtain the optimized segmented speed; Step S234: Adjust the optimized segmentation speed in real time using image quality feedback to obtain the feedback adjustment speed; Step S235: Calculate the segmented acceleration and deceleration parameters based on the feedback adjustment speed and target attitude sequence to obtain the segmented motion parameters.

[0015] This invention achieves differentiated speed control for areas of varying importance by setting a speed benchmark based on the priority of the target area. This allows the system to operate at a lower speed in high-priority areas, thereby acquiring more detailed thermal imaging data and improving the inspection quality of critical areas. Adjusting the speed based on the frequency of historical anomalies improves inspection reliability. For areas historically prone to anomalies, reducing the inspection speed allows for more thorough examination, increasing the anomaly detection rate and reducing the risk of missed detections. A multi-objective optimization algorithm is employed to minimize inspection time while ensuring data quality, thus improving inspection efficiency. By balancing inspection time and data quality, an optimal compromise is found. Adjusting the speed through real-time image quality feedback allows for dynamic adjustment based on actual conditions, ensuring the quality of the acquired thermal image data. When image quality is poor, reducing the speed improves image clarity; when image quality is good, the speed can be appropriately increased to improve inspection efficiency. By adjusting the speed and target attitude sequence based on feedback and calculating the segmented acceleration and deceleration parameters, the track system can smoothly transition between different inspection points. This ensures inspection efficiency while avoiding equipment vibration and image blurring caused by excessive speed changes, thus improving inspection stability and data quality.

[0016] Preferably, step S3 includes the following steps: Step S31: Perform precise execution of track motion according to the motion control sequence to obtain the real-time track status; Step S32: Based on the real-time orbital status, acquire continuous radiation data using a thermal imager to obtain time-stamped thermal image frames; Step S33: Provide real-time feedback on the track position based on the real-time track status to obtain time-stamped position information; Step S34: Perform time synchronization calibration on the time-stamped thermal image frame and the time-stamped location information to obtain the calibrated synchronized data stream; Step S35: Perform spatiotemporal data fusion on the calibrated synchronized data stream and construct spatiotemporal thermal spectrum data to obtain preliminary spatiotemporal thermal spectrum data; Step S36: Generate a spatiotemporal thermal spectrum array from the preliminary spatiotemporal thermal spectrum data to obtain the spatiotemporal thermal spectrum array.

[0017] This invention ensures the orbital system operates along a predetermined path and speed by precisely executing motion control sequences, acquiring accurate real-time orbital state information. This provides accurate position and attitude information for subsequent thermal imaging data acquisition and spatiotemporal data fusion, forming the foundation for the accuracy and reliability of the entire system. Continuously acquiring radiation data and adding precise timestamps ensures the integrity and temporal accuracy of the thermal imaging data, providing the raw data foundation for subsequent temperature analysis and anomaly detection. Real-time feedback of orbital position information and adding timestamps ensures the real-time nature and accuracy of the position information, providing a precise spatial reference for subsequent data fusion, ensuring that each temperature data point corresponds to precise three-dimensional spatial coordinates. Time synchronization calibration eliminates the time deviation between thermal imaging data and position information, ensuring the accuracy of data fusion and avoiding errors caused by time asynchrony, providing reliable data assurance for subsequent spatiotemporal analysis. Fusing the time-synchronized thermal imaging data and position information constructs preliminary spatiotemporal thermal spectrum data containing temperature, spatial location, and time information, providing a richer and more comprehensive data foundation for subsequent thermal anomaly detection and analysis. The generation of spatiotemporal thermal spectrum arrays transforms discrete spatiotemporal thermal spectrum data into a structured three-dimensional data format, facilitating subsequent spatiotemporal data analysis and visualization, and providing data support for a deeper understanding of the thermal behavior and abnormal development trends of equipment.

[0018] Preferably, step S4 includes the following steps: Step S41: Construct a baseline temperature model based on the spatiotemporal thermal spectrum array to obtain the regional baseline temperature; Step S42: Calculate the local temperature deviation of the spatiotemporal thermal spectrum array and the regional baseline temperature to obtain a local temperature deviation map; Step S43: Perform preliminary anomaly region identification on the local temperature deviation map to obtain a preliminary anomaly mask; Step S44: Perform contextual information fusion and enhancement on the preliminary anomaly mask and topological interest map to obtain a contextualized anomaly mask; Step S45: Contextualize the anomaly mask with anomaly scoring to obtain a list of anomalies with scores; Step S46: Perform anomaly filtering and noise reduction based on the rated anomaly list to obtain a refined anomaly list; Step S47: Generate priority exception registration data by refining the exception list.

[0019] This invention constructs a baseline temperature model to learn the temperature distribution patterns of each region of interest under normal operating conditions, providing a reference benchmark for subsequent anomaly detection, thereby improving the accuracy of anomaly identification and effectively reducing false alarms. Calculating local temperature deviations and comparing measured temperatures with baseline temperatures allows for more sensitive detection of anomalous temperature changes, even small ones, thus improving the sensitivity to subtle anomalies. Threshold segmentation and connected component analysis effectively identify potential anomalous regions from the local temperature deviation map, providing a foundation for subsequent contextual information fusion and anomaly scoring. Fusing the initial anomaly mask with the topological interest map introduces contextual information such as device type, location, and importance, enabling anomaly detection to consider not only temperature deviations but also the characteristics and importance of the device itself, thus more accurately assessing the risk level of anomalies and avoiding misjudgments of non-critical areas. Scoring the contextualized anomaly mask comprehensively considers the area of ​​the anomaly, temperature deviation, and contextual information, quantitatively assessing the severity of the anomaly and providing a more refined basis for subsequent anomaly filtering and maintenance decisions. By setting scoring thresholds and applying filtering rules, noise and unimportant anomalies can be effectively removed, false alarms reduced, and the accuracy of anomaly identification improved. This allows maintenance personnel to focus on handling truly critical anomalies. Recording information from the refined anomaly list into a priority anomaly registration database facilitates unified anomaly management and tracking, provides data support for subsequent intelligent maintenance scheduling, and achieves structured storage and management of anomaly information.

[0020] Preferably, step S5 includes the following steps: Step S51: Perform anomaly risk assessment on the priority anomaly registration data to obtain the anomaly risk level; Step S52: Based on the spatiotemporal thermal spectrum array, predict the anomaly development trend of the priority anomaly registration data to obtain anomaly evolution prediction data; Step S53: Generate maintenance action plans based on the anomaly risk level and anomaly evolution prediction data to obtain optional maintenance plans; Step S54: Assess the resource requirements of the optional maintenance schemes to obtain a resource requirement list; Step S55: Optimize maintenance scheduling based on optional maintenance schemes and resource requirement lists to obtain an optimized maintenance sequence; Step S56: Generate an optimization maintenance work order based on the priority anomaly registration data, resource requirement list, and optimization maintenance sequence.

[0021] This invention classifies anomalies into risk levels using a risk assessment matrix, allowing for a more comprehensive consideration of the severity of the anomaly and the criticality of the affected equipment. This leads to a more accurate assessment of the anomaly's potential risks, providing crucial information for subsequent maintenance decisions. By analyzing historical temperature data and predicting future temperature trends, the development of anomalies can be anticipated in advance, supporting the development of more effective maintenance strategies. For example, for anomalies predicted to worsen, maintenance can be scheduled ahead of time to avoid greater losses. Based on the anomaly's risk level and development trend, optional maintenance action plans are automatically generated, improving the efficiency and scientific rigor of maintenance decisions, avoiding the subjectivity of human experience, and ensuring that the maintenance plan matches the actual anomaly situation. Resource requirement assessments for each optional maintenance plan allow for advance understanding of the various resources required for maintenance, such as spare parts, personnel, and tools, providing a basis for subsequent resource scheduling and optimization, and preventing maintenance delays due to insufficient resources. Through scheduling optimization algorithms, all maintenance tasks can be coordinated, resource allocation optimized, downtime and maintenance costs minimized, maintenance efficiency improved, and resource conflicts avoided. Detailed optimization and maintenance work orders are generated, containing all necessary maintenance information such as task description, execution time, and required resources. This facilitates the execution by maintenance personnel, improves the standardization and efficiency of maintenance work, and enables digital management of maintenance information.

[0022] Preferably, the present invention also provides an aerial orbital thermal imaging temperature inspection system for performing the aerial orbital thermal imaging temperature inspection method described above, the aerial orbital thermal imaging temperature inspection system comprising: The target domain definition module is used to identify the evaluation area of ​​the area to be inspected and obtain a list of areas to be evaluated; the thermally sensitive areas are refined based on the list of areas to be evaluated to obtain refined regions of interest; and a topological interest map is generated from the refined regions of interest to obtain a topological interest map. The adaptive trajectory synthesis module is used to extract inspection points from the topological interest map and calculate the target attitude to obtain the target attitude sequence; it then allocates motion parameters based on the target attitude sequence to obtain segmented motion parameters; and finally generates a motion control sequence from the segmented motion parameters to obtain the motion control sequence. The synchrotron radiation flow acquisition module is used to acquire continuous radiation data according to the motion control sequence and provide real-time feedback of orbital position to obtain timestamped thermal image frames and timestamped position information; and to generate a spatiotemporal thermal spectrum array from the timestamped thermal image frames and timestamped position information. The contextualized thermal anomaly extraction module is used to calculate local temperature deviations based on the spatiotemporal thermal spectrum array, identify preliminary anomaly regions, and obtain preliminary anomaly masks; it then extracts contextualized anomaly masks from the preliminary anomaly masks and generates priority anomaly registration data. The intelligent maintenance scheduling module is used to generate maintenance action plans based on priority anomaly registration data to obtain optional maintenance plans; optimize maintenance scheduling based on optional maintenance plans, and generate optimized maintenance work orders. Attached Figure Description

[0023] Figure 1 A schematic diagram of the steps involved in an aerial orbital thermal imaging temperature inspection. Figure 2 This is a detailed flowchart illustrating the implementation steps of step S1 in this invention. Figure 3 This is a detailed flowchart illustrating the implementation steps of step S2 in this invention.

[0024] The objectives, features, and advantages of this invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0025] The technical method of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0026] Furthermore, the accompanying drawings are merely illustrative of the invention and are not necessarily drawn to scale. The same reference numerals in the drawings denote the same or similar parts, and therefore repeated descriptions of them will be omitted. Some block diagrams shown in the drawings are functional entities and do not necessarily correspond to physically or logically independent entities. These functional entities can be implemented in software, in one or more hardware modules or integrated circuits, or in different network and / or processor methods and / or microcontroller methods.

[0027] It should be understood that although the terms "first," "second," etc., may be used herein to describe various units, these units should not be limited by these terms. These terms are used merely to distinguish one unit from another. For example, without departing from the scope of the exemplary embodiments, a first unit may be referred to as a second unit, and similarly, a second unit may be referred to as a first unit. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0028] To achieve the above objectives, please refer to Figures 1 to 3 An aerial orbital thermal imaging temperature inspection method includes the following steps: Step S1: Identify the evaluation areas of the area to be inspected to obtain a list of areas to be evaluated; refine the heat-sensitive areas based on the list of areas to be evaluated to obtain refined regions of interest; generate a topological interest map from the refined regions of interest to obtain a topological interest map. Step S2: Extract inspection points from the topological interest map and calculate the target attitude to obtain the target attitude sequence; allocate motion parameters according to the target attitude sequence to obtain segmented motion parameters; generate motion control sequences from the segmented motion parameters to obtain motion control sequences. Step S3: Acquire continuous radiation data according to the motion control sequence and provide real-time feedback on the orbital position to obtain time-stamped thermal images and time-stamped position information; generate a spatiotemporal thermal spectrum array from the time-stamped thermal images and time-stamped position information to obtain the spatiotemporal thermal spectrum array. Step S4: Calculate the local temperature deviation based on the spatiotemporal thermal spectrum array, and perform preliminary anomaly region identification to obtain a preliminary anomaly mask; extract the contextualized anomaly mask from the preliminary anomaly mask, and generate priority anomaly registration data. Step S5: Generate a maintenance action plan based on the priority anomaly registration data to obtain optional maintenance plans; optimize the maintenance schedule based on the optional maintenance plans and generate an optimized maintenance work order to obtain an optimized maintenance work order.

[0029] In this embodiment of the invention, reference Figure 1 The diagram shown is a flowchart illustrating the steps of the aerial orbital thermal imaging temperature inspection method of the present invention. In this example, the aerial orbital thermal imaging temperature inspection method includes the following steps: Step S1: Identify the evaluation areas of the area to be inspected to obtain a list of areas to be evaluated; refine the heat-sensitive areas based on the list of areas to be evaluated to obtain refined regions of interest; generate a topological interest map from the refined regions of interest to obtain a topological interest map. Step S2: Extract inspection points from the topological interest map and calculate the target attitude to obtain the target attitude sequence; allocate motion parameters according to the target attitude sequence to obtain segmented motion parameters; generate motion control sequences from the segmented motion parameters to obtain motion control sequences. Step S3: Acquire continuous radiation data according to the motion control sequence and provide real-time feedback on the orbital position to obtain time-stamped thermal images and time-stamped position information; generate a spatiotemporal thermal spectrum array from the time-stamped thermal images and time-stamped position information to obtain the spatiotemporal thermal spectrum array. Step S4: Calculate the local temperature deviation based on the spatiotemporal thermal spectrum array, and perform preliminary anomaly region identification to obtain a preliminary anomaly mask; extract the contextualized anomaly mask from the preliminary anomaly mask, and generate priority anomaly registration data. Step S5: Generate a maintenance action plan based on the priority anomaly registration data to obtain optional maintenance plans; optimize the maintenance schedule based on the optional maintenance plans and generate an optimized maintenance work order to obtain an optimized maintenance work order.

[0030] Preferably, step S1 includes the following steps: Step S11: Perform preliminary asset identification in the area to be inspected to obtain an equipment list; Step S12: Associate spatial information based on the equipment list to obtain a spatialized equipment list; Step S13: Based on the spatialized equipment list, integrate the equipment priority information to obtain the list of areas to be evaluated; Step S14: Refine the heat-sensitive areas based on the list of areas to be evaluated to obtain refined regions of interest; Step S15: Construct and optimize the spatial topology of the refined region of interest to obtain the labeled region of interest; Step S16: Generate a topological interest map for the labeled regions of interest to obtain the topological interest map.

[0031] As an example of the present invention, reference is made to Figure 2 As shown, in this example, step S1 includes: Step S11: Perform preliminary asset identification in the area to be inspected to obtain an equipment list; In this embodiment of the invention, a panoramic image of the area to be inspected is acquired. An image recognition algorithm, such as YOLOv5, analyzes the panoramic image to identify predefined types of industrial equipment appearing in the image, such as transformers, switchgear, and pipelines. The algorithm classifies the identified equipment based on features in the training dataset and assigns a unique ID to each identified device. The recognition results are stored in JSON format containing the device ID, device type, and bounding box coordinates, forming a device list. For example, a transformer is identified and recorded as: {"id":"transformer_1","type":"transformer","bbox":[100,200,300,400]}.

[0032] Step S12: Associate spatial information based on the equipment list to obtain a spatialized equipment list; In this embodiment of the invention, a lidar is used to acquire 3D point cloud data of the area to be inspected. A point cloud registration algorithm, such as the Iterative Closest Point (ICP) algorithm, is used to register the acquired point cloud data with a pre-constructed 3D model of the factory building to obtain the precise position of the equipment in the global coordinate system. Based on the bounding box information provided in the equipment list in step S11, a subset of the 3D point cloud corresponding to the equipment is extracted from the registered point cloud data. The equipment ID, equipment type, 3D point cloud subset, and center point coordinates in the global coordinate system are stored as a spatialized equipment list in a PLY file format, containing metadata such as the equipment ID and center point coordinates.

[0033] Step S13: Based on the spatialized equipment list, integrate the equipment priority information to obtain the list of areas to be evaluated; In this embodiment of the invention, an initial priority is assigned to each device in the spatialized equipment list according to predefined equipment type priority rules, such as transformers having a higher priority than switchgear, and switchgear having a higher priority than pipelines. The historical maintenance record database of the devices is read to count the number of failures and maintenance visits for each device. Based on the statistical results, the initial priority of the devices is adjusted, with devices having a higher failure frequency having a higher priority. Real-time equipment operating status data, such as transformer load rate and switchgear current, is acquired. If the operating parameters of a device exceed a preset threshold, the priority of that device is increased. The device ID, spatial information, initial priority, priority adjusted based on historical maintenance data, and final priority adjusted based on real-time operating data are integrated to generate a list of areas to be evaluated, which is stored in CSV format.

[0034] Step S14: Refine the heat-sensitive areas based on the list of areas to be evaluated to obtain refined regions of interest; In this embodiment of the invention, for each device in the list of areas to be evaluated, its corresponding CAD model is retrieved from the device's 3D model database. Based on the device type and expert knowledge base, the heat-sensitive areas of the device are determined, such as transformer connection points and heat sinks. These heat-sensitive areas are marked in the CAD model and assigned corresponding priorities. The marked heat-sensitive areas are extracted from the CAD model to generate refined region of interest data containing the region's 3D coordinates, its associated device ID, and priority, and stored in STL format.

[0035] Step S15: Construct and optimize the spatial topology of the refined region of interest to obtain the labeled region of interest; In this embodiment of the invention, a graph database, such as Neo4j, is used to construct the spatial topology of refined regions of interest (ROIs). Each refined ROI is stored as a node in the graph database, with node attributes including the region's 3D coordinates, its associated device ID, and priority. The spatial distance between any two ROIs is calculated; if the distance is less than a preset threshold, an edge is established between the two nodes, representing a spatial adjacency relationship. A graph algorithm, such as Dijkstra's algorithm, is used to calculate the shortest path from the track system to each ROI, and the ROIs are reordered based on the path length to optimize the inspection order. The optimized inspection order information is added to the attributes of each node in the graph database, generating labeled ROI data.

[0036] Step S16: Generate a topological interest map from the labeled regions of interest to obtain the topological interest map; In this embodiment of the invention, labeled regions of interest (ROIs) data, including node attributes and edge relationships, are exported from a graph database. The data is converted to GeoJSON format, where each ROI is represented as a geometric object (point, line, or polygon) with attributes. The GeoJSON data is loaded into a Geographic Information System (GIS) platform, such as ArcGIS, to generate a visualized topological ROI map. The topological ROI map contains the spatial location, priority, associated device information, and topological relationships of all ROIs, displayed in an overlay format for easy viewing and editing. The final topological ROI map is saved in Shapefile format for use in subsequent steps.

[0037] Preferably, step S14 includes the following steps: Step S141: Obtain the 3D model based on the list of areas to be evaluated, and perform model preprocessing to obtain preprocessed 3D model data; Step S142: Perform intelligent template matching on the preprocessed 3D model data and the preset hotspot template library to obtain matching hotspot templates; Step S143: Based on the preprocessed 3D model data, perform local parameterization adjustment on the matching hotspot template to obtain the locally adjusted hotspot region; Step S144: Obtain equipment operating parameters; perform thermal simulation-driven hotspot supplementation based on preprocessed 3D model data and equipment operating parameters to obtain simulated supplementary hotspots; Step S145: Generate a refined region of interest based on the local hotspot adjustment area and the simulated supplementary hotspot, thus obtaining the refined region of interest.

[0038] In this embodiment of the invention, based on the ID and type of each device in the list of areas to be evaluated, the corresponding 3D model file is retrieved from a pre-established 3D model database. The model format is STL. The STL file is read, and the 3D model data is loaded into the 3D modeling software Blender. In Blender, a model simplification operation is performed to reduce the number of facets and lower computational complexity; for example, the Decimal Modifier is used to reduce the number of facets to 50% of the original model. The simplified model is then smoothed to remove noise and sharp edges from the model surface, for example, using the Smooth Modifier. The preprocessed 3D model data is exported in OBJ format, and the device ID is added as metadata for subsequent identification and association.

[0039] The preprocessed 3D model data is loaded into the point cloud registration software CloudCompare. A pre-defined hotspot template library (containing 3D models of hotspot areas for various standard equipment components, such as transformer bushings and motor bearings, in PLY format) is also loaded into CloudCompare. For each preprocessed 3D model, all templates in the hotspot template library are traversed, and point cloud registration is performed using the ICP algorithm. The root mean square error (RMSE) after registration is calculated; templates with RMSE values ​​below a preset threshold are considered matching hotspot templates. The 3D coordinates, corresponding device IDs, and RMSE values ​​of the matching hotspot templates are stored as matching hotspot template data, saved in TXT format.

[0040] The matched hotspot template and corresponding preprocessed 3D model data are loaded into the parametric modeling software FreeCAD. Based on the actual size and shape of the preprocessed 3D model data, key parameters of the matched hotspot template are adjusted. For example, for a transformer bushing template, the template size is adjusted according to the actual bushing diameter and height. The adjusted hotspot template needs to undergo Boolean operations (intersection) with the preprocessed 3D model data to ensure that the adjusted hotspot area is accurately located on the equipment surface. The 3D coordinates of the adjusted hotspot area, its corresponding equipment ID, and adjustment parameters are stored as local adjusted hotspot area data in STEP format.

[0041] Real-time operating parameters of the equipment, such as current, voltage, and load, are acquired through an industrial control system (e.g., SCADA system). The pre-processed 3D model data is imported into the finite element analysis software ANSYS. Boundary conditions and loads are set for the ANSYS simulation model based on the acquired equipment operating parameters. For example, the heat source of the model is set according to the current magnitude. Steady-state thermal simulation analysis is performed to obtain the temperature distribution of the equipment under steady-state conditions. Regions where the temperature exceeds a preset threshold are identified and marked as simulation prediction hotspots. The 3D coordinates of the simulation prediction hotspots, their associated equipment IDs, and corresponding temperature values ​​are stored as supplementary simulation hotspot data in VTK format.

[0042] Locally adjusted hotspot data and simulation-supplemented hotspot data are loaded into the 3D data processing software MeshLab. The two types of data are registered, mapping the simulation-supplemented hotspots onto the preprocessed 3D model. The locally adjusted hotspots and simulation-supplemented hotspots are merged, overlapping areas are removed, and the final refined regions of interest (ROIs) are generated. Each refined ROI is assigned a unique ID, and its 3D coordinates, associated device ID, source (template matching or simulation supplementation), and priority (set based on temperature values ​​or expert knowledge) are recorded. The final refined ROI data is saved in OBJ format, including ROI ID and priority information.

[0043] Preferably, step S144 specifically includes: The simulation model is automatically constructed based on the preprocessed 3D model data and equipment operating parameters to obtain the initial simulation model. The initial simulation model is meshed and optimized to obtain an optimized simulation model. Steady-state thermal simulation calculations were performed based on the optimized simulation model and equipment operating parameters to obtain steady-state temperature field data. Transient thermal simulation analysis was performed based on the optimized simulation model to obtain transient temperature change data. Steady-state temperature field data and transient temperature change data are superimposed onto preprocessed 3D model data to identify hotspots and obtain simulated predicted hotspot regions. By associating the simulated hotspots with the template hotspots in the simulated predicted hotspot regions and the locally adjusted hotspot regions, we can obtain the simulated supplementary hotspots.

[0044] In this embodiment of the invention, preprocessed 3D model data (OBJ format) is imported into COMSOL Multiphysics simulation software. Based on the device type and operating parameters, the built-in physics interface of COMSOL, such as the heat transfer module, is automatically selected. Boundary conditions for the simulation model are set according to the device operating parameters, such as current, voltage, and ambient temperature. For example, the current value is used as the heat source input, and the ambient temperature is used as the external temperature boundary condition. Based on the device material property database, corresponding material properties, such as thermal conductivity and specific heat capacity, are specified for different components in the model. An initial simulation model containing the geometric model, material properties, boundary conditions, and physics interface is generated.

[0045] In COMSOL, the initial simulation model is meshed using a free tetrahedral meshing algorithm. Mesh generation parameters, such as maximum and minimum element size, are set to control mesh density and quality. For critical regions, such as those near heat sources or geometrically complex areas, local mesh refinement is performed to improve computational accuracy. Mesh quality checks are performed, such as checking element skewness and aspect ratio, to ensure the mesh quality meets simulation requirements. Based on the check results, the mesh is adjusted and optimized, for example, by adjusting mesh generation parameters or performing local mesh re-meshing. Finally, an optimized simulation model that meets the required simulation accuracy is obtained.

[0046] In COMSOL, select the steady-state solver to solve the optimization simulation model. Set the solver parameters, such as convergence accuracy and maximum number of iterations. Run the solver to calculate the temperature distribution of the model in a steady state. Export the calculation results as a VTK format file, containing the temperature values ​​of each node in the model, i.e., the steady-state temperature field data.

[0047] In COMSOL, select the transient solver to solve the optimized simulation model. Set the simulation time step and total simulation time, for example, a 1-second time step to simulate temperature changes over 10 minutes. Set initial conditions, such as the initial temperature of the equipment. Run the solver to calculate the temperature distribution of the model at different time steps. Export the calculation results as an HDF5 file, containing the temperature values ​​of each node in the model at different time steps, i.e., the transient temperature change data.

[0048] Import preprocessed 3D model data (OBJ format), steady-state temperature field data (VTK format), and transient temperature change data (HDF5 format) into ParaView data visualization software. Map the steady-state temperature field data and transient temperature change data onto the preprocessed 3D model. Use ParaView's calculator function to calculate the average temperature and temperature fluctuation range for each node. Set temperature thresholds and mark areas where the average temperature or temperature fluctuation range exceeds the threshold as simulation prediction hotspots. Store the 3D coordinates, average temperature, and temperature fluctuation range of the simulation prediction hotspots as CSV files.

[0049] Import the simulation-predicted hotspot region data (CSV format) and the locally adjusted hotspot region data (STEP format) into the 3D modeling software Blender. For each simulation-predicted hotspot region, calculate its distance to all locally adjusted hotspot regions. If the distance between a simulation-predicted hotspot region and a locally adjusted hotspot region is less than a preset threshold, they are considered associated. Simulation-predicted hotspot regions not associated with locally adjusted hotspot regions are marked as simulation supplementary hotspots. Store the simulation supplementary hotspot data, including 3D coordinates, average temperature, and temperature fluctuation range, as a PLY format file.

[0050] Preferably, step S2 includes the following steps: Step S21: Extract inspection points from the topological interest map and sort the inspection points to obtain the inspection point sequence; Step S22: Calculate the target attitude of the inspection point sequence to obtain the target attitude sequence; Step S23: Assign motion parameters according to the target posture sequence to obtain segmented motion parameters; Step S24: Generate motor commands from the segmented motion parameters to obtain a discrete motor command set; Step S25: Perform timestamp allocation and synchronization on the discrete motor instruction set to obtain timestamped motor instructions; Step S26: Generate a motion control sequence from the timestamped motor commands to obtain the motion control sequence.

[0051] As an example of the present invention, reference is made to Figure 3 As shown, step S2 in this example includes: Step S21: Extract inspection points from the topological interest map and sort the inspection points to obtain the inspection point sequence; In this embodiment of the invention, a topological interest map (Shapefile format) is read, and the coordinates of the center point of each interest region are extracted as initial inspection points. Each inspection point is associated with its corresponding interest region ID and priority. A Traveling Salesman Problem (TSP) solving algorithm, such as a genetic algorithm, is used to calculate the shortest path connecting all inspection points. During path planning, physical constraints of the orbital system, such as maximum acceleration and maximum velocity, are considered. Based on the TSP algorithm calculation results, the inspection points are sorted to generate an inspection point sequence, which is saved in CSV format containing the inspection point ID, coordinates, associated interest region ID, and priority.

[0052] Step S22: Calculate the target attitude of the inspection point sequence to obtain the target attitude sequence; In this embodiment of the invention, for each inspection point in the inspection point sequence, the optimal observation attitude required by the thermal imager is calculated based on the geometry and size of its associated region of interest (ROI). The attitude calculation considers the thermal imager's field of view to ensure that the ROI is completely within the field of view. If the ROI is planar, an attitude vector perpendicular to the plane is calculated; if the ROI has a complex shape, an attitude calculation method based on the center point of the ROI and the thermal imager's field of view is used. The target attitude (including pitch, yaw, and roll angles) of each inspection point is associated with its corresponding inspection point ID to generate a target attitude sequence, which is saved in JSON format containing the inspection point ID and attitude angles.

[0053] Step S23: Assign motion parameters according to the target posture sequence to obtain segmented motion parameters; In this embodiment of the invention, the time required for the orbital system to move between two points is calculated based on the distance between adjacent inspection points in the target attitude sequence and the difference in target attitude. Based on the calculated motion time and the maximum speed and acceleration limits of the orbital system, a trapezoidal velocity planning algorithm is used to calculate the velocity curve for each motion segment, including acceleration, constant speed, and deceleration segments. The dwell time in a region of interest is adjusted according to its priority. Higher priority regions have longer dwell times to ensure sufficient thermal imaging data is acquired. The parameters for each motion segment, including the starting inspection point ID, ending inspection point ID, motion time, velocity curve parameters, and dwell time, are stored as segmented motion parameters and saved in XML format.

[0054] Step S24: Generate motor commands from the segmented motion parameters to obtain a discrete motor command set; In this embodiment of the invention, the target position and velocity of the track system motor in each control cycle are calculated based on the velocity curve of each segment of motion in the segmented motion parameters. The control cycle is determined according to the control frequency of the track system, for example, 10ms. Based on the motor type and control method of the track system, the target position and velocity are converted into corresponding motor control commands, such as pulse count and frequency. Each motor control command is associated with its corresponding inspection point ID and time to generate a discrete motor command set, which is stored in binary format.

[0055] Step S25: Perform timestamp allocation and synchronization on the discrete motor instruction set to obtain timestamped motor instructions; In this embodiment of the invention, a precise timestamp is assigned to each instruction in the discrete motor instruction set using a system clock synchronized with the thermal imager. The timestamp accuracy is in the millisecond range. The execution time of each instruction is calculated based on the start time and duration of each motion segment in the segmented motion parameters. The timestamp and execution time are added to each instruction to generate a timestamped motor instruction, which is then saved in binary format.

[0056] Step S26: Generate a motion control sequence from the timestamped motor commands to obtain the motion control sequence; In this embodiment of the invention, the timestamped motor instructions are sorted according to their execution time. The sorted instruction sequence is converted into an instruction format that the track system's motor controller can recognize, such as G-code. Checksums and end markers are added to ensure the integrity and correctness of the instructions. The final motion control sequence is saved in text format for the track system to execute.

[0057] Preferably, step S23 includes the following steps: Step S231: Set the priority-weighted velocity reference according to the target attitude sequence to obtain the regional velocity reference; Step S232: Adjust the regional velocity benchmark for uncertainty correction to obtain the uncertainty-adjusted velocity; Step S233: Adjust the speed according to the uncertainty to perform multi-objective optimization speed planning and obtain the optimized segmented speed; Step S234: Adjust the optimized segmentation speed in real time using image quality feedback to obtain the feedback adjustment speed; Step S235: Calculate the segmented acceleration and deceleration parameters based on the feedback adjustment speed and target attitude sequence to obtain the segmented motion parameters.

[0058] In this embodiment of the invention, the target attitude sequence (JSON format) is read to obtain the priority of the region of interest associated with each inspection point. Based on a predefined priority-velocity mapping table, a velocity reference value is set for each inspection point. For example, the velocity reference value for a high-priority region is 0.1 m / s, for a medium-priority region it is 0.3 m / s, and for a low-priority region it is 0.5 m / s. The velocity reference value of each inspection point is associated with its corresponding inspection point ID to generate regional velocity reference data, which is saved in CSV format.

[0059] Read historical inspection data and calculate the historical anomaly frequency for each region of interest. Based on the historical anomaly frequency, calculate the uncertainty coefficient for each region of interest. For example, use the formula: Uncertainty coefficient = 1 / (1+exp(-k)). The historical anomaly frequency is calculated, where k is an adjustment coefficient, for example, k=5. The regional velocity baseline value is multiplied by the uncertainty coefficient of the corresponding region of interest to obtain the uncertainty adjustment velocity. The uncertainty adjustment velocity of each inspection point is associated with its corresponding inspection point ID to generate uncertainty adjustment velocity data, which is saved in CSV format.

[0060] Uncertainty-adjusted speeds are used as initial speed values ​​input into a multi-objective optimization algorithm, such as the NSGA-II algorithm. The optimization objectives are to minimize the total inspection time and maximize the inspection time of high-priority areas. Constraints are set for the optimization algorithm, such as the maximum speed and maximum acceleration of the orbital system. The optimization algorithm is run to obtain a Pareto optimal solution set, where each solution represents a set of optimized segmented speeds. Based on actual requirements, a suitable solution is selected from the Pareto optimal solution set as the final optimized segmented speed. The optimized speed of each segment is associated with its corresponding start and end inspection point IDs to generate optimized segmented speed data, which is saved in XML format.

[0061] During the operation of the orbital system, thermal imagers acquire images in real time. Image sharpness assessment algorithms, such as the Laplacian operator, are used to calculate a sharpness score for each frame. A sharpness score threshold is set, for example, 80 points. If the image sharpness score is below the threshold, the motion speed of the current segment is reduced, for example, by 10%. If the image sharpness score is above the threshold, the motion speed of the current segment is maintained or slightly increased, for example, by 5%, but not exceeding the maximum speed set for the optimized segment. The real-time adjusted speed value is associated with its corresponding inspection point ID and timestamp to generate feedback adjustment speed data, which is saved in binary format.

[0062] Read the feedback adjustment velocity data and target attitude sequence data. Calculate the time required for the orbital system to move between two adjacent checkpoints based on the distance between them and the feedback adjustment velocity. Using the calculated motion time, feedback adjustment velocity, and the orbital system's maximum acceleration limit, calculate the acceleration and deceleration curves for each motion segment using an S-shaped velocity planning algorithm. Based on the attitude differences between adjacent checkpoints in the target attitude sequence, calculate the attitude adjustment parameters required for the orbital system during motion. Store the parameters for each motion segment, including the starting checkpoint ID, ending checkpoint ID, motion time, velocity curve parameters, acceleration curve parameters, deceleration curve parameters, dwell time, and attitude adjustment parameters, as segmented motion parameters in XML format.

[0063] Preferably, step S3 includes the following steps: Step S31: Perform precise execution of track motion according to the motion control sequence to obtain the real-time track status; Step S32: Based on the real-time orbital status, acquire continuous radiation data using a thermal imager to obtain time-stamped thermal image frames; Step S33: Provide real-time feedback on the track position based on the real-time track status to obtain time-stamped position information; Step S34: Perform time synchronization calibration on the time-stamped thermal image frame and the time-stamped location information to obtain the calibrated synchronized data stream; Step S35: Perform spatiotemporal data fusion on the calibrated synchronized data stream and construct spatiotemporal thermal spectrum data to obtain preliminary spatiotemporal thermal spectrum data; Step S36: Generate a spatiotemporal thermal spectrum array from the preliminary spatiotemporal thermal spectrum data to obtain the spatiotemporal thermal spectrum array.

[0064] In this embodiment of the invention, the track control system receives a motion control sequence (text format) and parses the specific parameters of each instruction, including target position, velocity, acceleration, etc. The controller converts the parsed parameters into motor drive signals to control the motor movement. A high-precision encoder measures the motor's rotation angle in real time and converts the angle data into the track system's position coordinates (x, y, z) and attitude angles (pitch, yaw, roll) in three-dimensional space. Based on the position information fed back by the encoder and the target position in the motion control sequence, the controller adjusts the motor drive signals in real time to ensure that the track system moves precisely along the planned trajectory. The real-time position coordinates, attitude angles, and timestamps of the track system are packaged into real-time track status data, stored in binary format, and transmitted to the data processing center via a network.

[0065] The thermal imager is set to continuous acquisition mode and pre-configured with acquisition parameters such as frame rate (25 frames / second), resolution (640x480 pixels), and temperature range (-20°C to 120°C). The imager's internal clock provides timestamps with microsecond-level precision. Whenever a frame of thermal image data is acquired, the imager immediately adds a timestamp to that frame and converts the data into a temperature matrix, where each element represents the temperature value of the corresponding pixel. The temperature matrix and timestamps are packaged into timestamped thermal image frame data, stored in H.264 video stream format, and transmitted over the network to the data processing center.

[0066] Extract the real-time position coordinates (x, y, z) and attitude angles (pitch, yaw, roll) of the orbital system from the real-time orbital status data obtained in step S31. Using the same system clock as the thermal imager, add a timestamp with the same precision as the thermal image frame timestamp to each set of position information. Package the position coordinates, attitude angles, and timestamps into timestamped position information data, store it in CSV format, and transmit it to the data processing center via the network.

[0067] Arrange the timestamped thermal image frame data and timestamped location information data in timestamp order. Calculate the time difference between each thermal image frame timestamp and the nearest location information timestamp. Use a linear interpolation method to interpolate the location information based on the time difference, ensuring precise alignment with the thermal image frame timestamps. Associate the time-synchronized and calibrated thermal image frame data and location information data to generate a calibrated and synchronized data stream, saved in a custom binary format containing a temperature matrix, location coordinates, attitude angles, and a unified timestamp.

[0068] The system reads the synchronized data stream after calibration and associates the temperature matrix of each thermal image frame with its corresponding orbital position information (coordinates and attitude angle) and timestamp. Based on the orbital position information and the thermal imager's field of view parameters, the coordinates of each pixel in three-dimensional space are calculated. The three-dimensional coordinates, temperature value, and timestamp of each pixel are integrated to generate preliminary spatiotemporal thermal spectrum data, which is stored in HDF5 format for convenient and efficient subsequent access and processing.

[0069] Read the initial spatiotemporal thermal spectrum data (HDF5 format). Divide the continuous spatiotemporal thermal spectrum data into time series according to a preset time interval (e.g., 1 second). Spatially resample the data in each time series and convert it into a fixed-resolution three-dimensional temperature matrix, where each element represents the temperature value at a specific time and spatial location. Arrange the three-dimensional temperature matrices of all time series in chronological order to generate the final spatiotemporal thermal spectrum array, which is stored in NetCDF format for subsequent spatiotemporal analysis and visualization.

[0070] Preferably, step S4 includes the following steps: Step S41: Construct a baseline temperature model based on the spatiotemporal thermal spectrum array to obtain the regional baseline temperature; Step S42: Calculate the local temperature deviation of the spatiotemporal thermal spectrum array and the regional baseline temperature to obtain a local temperature deviation map; Step S43: Perform preliminary anomaly region identification on the local temperature deviation map to obtain a preliminary anomaly mask; Step S44: Perform contextual information fusion and enhancement on the preliminary anomaly mask and topological interest map to obtain a contextualized anomaly mask; Step S45: Contextualize the anomaly mask with anomaly scoring to obtain a list of anomalies with scores; Step S46: Perform anomaly filtering and noise reduction based on the rated anomaly list to obtain a refined anomaly list; Step S47: Generate priority exception registration data by refining the exception list.

[0071] In this embodiment of the invention, spatiotemporal thermal array data (NetCDF format) is read, and temperature data for each region of interest at different time points are extracted. A Gaussian mixture model (GMM) is used to model the historical temperature data of each region of interest, learning its normal temperature distribution. The parameters of the GMM model, such as mean, variance, and weights, are estimated using the expectation-maximization (EM) algorithm. The GMM model parameters for each region of interest are stored in JSON format, forming the regional baseline temperature data.

[0072] Read the spatiotemporal thermal spectrum array data and regional baseline temperature data. For each data point in the spatiotemporal thermal spectrum array, determine its region of interest based on its spatial location, and calculate the probability density value of that point using the GMM model corresponding to that region. Convert the probability density value into a temperature deviation value, for example, using the formula: Temperature Deviation = μ - σ sqrt(-2 ln(probability density) is used, where μ and σ are the mean and standard deviation of the GMM model, respectively. The temperature deviation value of each data point is stored as a new NetCDF file to form a local temperature deviation map.

[0073] Read the local temperature deviation map (NetCDF format). Set a temperature deviation threshold, for example, 3σ, where σ is the standard deviation of the GMM model corresponding to the region of interest. Mark data points with temperature deviation values ​​exceeding the threshold as outliers. Perform connected component analysis on the outliers, aggregating adjacent outliers into outlier regions. Save the binary image of the outlier regions (1 indicates outlier, 0 indicates normal) as a TIFF file to form a preliminary outlier mask.

[0074] Read the preliminary anomaly mask (TIFF format) and the topological interest map (Shapefile format). Perform spatial overlay analysis on the preliminary anomaly mask and the topological interest map to determine the region of interest to which each anomaly region belongs, as well as its corresponding device type, priority, and other information. Based on predefined rules, such as device importance and historical failure rate, anomalies in different regions of interest are weighted, with higher-priority regions receiving greater weight. The weighted anomaly information is added to the preliminary anomaly mask to generate a contextualized anomaly mask, which is then saved in GeoTIFF format.

[0075] Read the contextualized anomaly mask (GeoTIFF format). Calculate a comprehensive score for each anomaly region based on factors such as area, temperature deviation, and weighting coefficients. For example, use the formula: Comprehensive Score = Area Temperature deviation Weighting coefficients. The ID, location, area, temperature deviation, weighting coefficient, and overall score of each anomalous region are stored in CSV format to form a list of anomalies with scores.

[0076] Read the list of rated anomalies (CSV format). Set a rating threshold, such as the average rating plus a standard deviation. Filter out anomalies with a composite score below the threshold, considering them noise or unimportant. For the remaining anomalies, further filter them according to predefined rules, such as excluding known, stable hotspots. Save the filtered anomaly information as a CSV file, forming a refined anomaly list.

[0077] Read the refining anomaly list (CSV format). Record each anomaly in the refining anomaly list into the priority anomaly registration database. Each record in the database contains information such as anomaly ID, location, area, temperature deviation, weighting coefficient, comprehensive score, associated equipment, equipment type, priority, and discovery time. Sort the anomalies according to their comprehensive score and the priority of their associated equipment to generate priority anomaly registration data.

[0078] Preferably, step S5 includes the following steps: Step S51: Perform anomaly risk assessment on the priority anomaly registration data to obtain the anomaly risk level; Step S52: Based on the spatiotemporal thermal spectrum array, predict the anomaly development trend of the priority anomaly registration data to obtain anomaly evolution prediction data; Step S53: Generate maintenance action plans based on the anomaly risk level and anomaly evolution prediction data to obtain optional maintenance plans; Step S54: Assess the resource requirements of the optional maintenance schemes to obtain a resource requirement list; Step S55: Optimize maintenance scheduling based on optional maintenance schemes and resource requirement lists to obtain an optimized maintenance sequence; Step S56: Generate an optimization maintenance work order based on the priority anomaly registration data, resource requirement list, and optimization maintenance sequence.

[0079] In this embodiment of the invention, priority anomaly registration data is read to obtain the severity score, equipment type, and equipment criticality of each anomaly. Based on a predefined risk assessment matrix, the severity score and equipment criticality of the anomaly are mapped to corresponding risk levels. The risk assessment matrix is ​​divided according to equipment type. For example, for transformers, anomalies with both high severity scores and high equipment criticality are defined as high-risk, while for ordinary pipelines, even with high severity scores, their risk level is defined as medium-risk due to low equipment criticality. The risk level of each anomaly is stored in a database to form anomaly risk level data.

[0080] Read spatiotemporal thermal array data and priority anomaly registration data. For each anomaly, extract its temperature data over a past period, such as the past week. Use time series analysis methods, such as the ARIMA model, to model the temperature data and predict temperature change trends over a future period, such as the temperature change trend for the next day. Store the prediction results, including predicted temperature values ​​and confidence intervals, in a database to form anomaly evolution prediction data.

[0081] Read anomaly risk level data and anomaly evolution prediction data. Based on a predefined maintenance strategy rule base, generate corresponding maintenance action plans according to the anomaly's risk level and predicted temperature change trend. For example, for high-risk anomalies with a predicted continuous temperature increase, immediate shutdown and maintenance are recommended; for medium-risk anomalies with a predicted stable temperature, scheduled periodic inspections are recommended; for low-risk anomalies with a predicted temperature decrease, continued monitoring is recommended. Store the optional maintenance plans for each anomaly in the database to form optional maintenance plan data.

[0082] Read the available maintenance plan data. Based on the specific details of each maintenance plan, assess the required resources, such as spare parts, personnel, and tools. For example, if the maintenance plan involves replacing a component, the required spare part model and quantity need to be determined; if the maintenance plan involves on-site repair, the required number of personnel and their skill requirements need to be determined. Store the resource requirements information for each maintenance plan, including resource type, name, quantity, and specifications, in the database to form a resource requirements list.

[0083] Read the optional maintenance plan data and resource requirements list. Use a scheduling optimization algorithm, such as a genetic algorithm, to optimize the scheduling of all pending maintenance tasks. The optimization objective is to minimize downtime and maintenance costs while satisfying resource constraints, such as personnel and spare parts availability. The input parameters of the optimization algorithm include the estimated execution time, risk level, resource requirements, and resource availability for each maintenance task. The output of the optimization algorithm is an optimized maintenance sequence, which includes the execution time and order of each maintenance task.

[0084] Read priority anomaly registration data, resource requirement list, and optimized maintenance sequence. Based on this information, generate a detailed maintenance work order for each maintenance task. The maintenance work order includes a task description, execution time, required resources, responsible person, contact information, and anomaly information. Store the generated maintenance work orders in the database and send them to relevant maintenance personnel via email or other means, forming optimized maintenance work orders.

[0085] Preferably, the present invention also provides an aerial orbital thermal imaging temperature inspection system for performing the aerial orbital thermal imaging temperature inspection method described above, the aerial orbital thermal imaging temperature inspection system comprising: The target domain definition module is used to identify the evaluation area of ​​the area to be inspected and obtain a list of areas to be evaluated; the thermally sensitive areas are refined based on the list of areas to be evaluated to obtain refined regions of interest; and a topological interest map is generated from the refined regions of interest to obtain a topological interest map. The adaptive trajectory synthesis module is used to extract inspection points from the topological interest map and calculate the target attitude to obtain the target attitude sequence; it then allocates motion parameters based on the target attitude sequence to obtain segmented motion parameters; and finally generates a motion control sequence from the segmented motion parameters to obtain the motion control sequence. The synchrotron radiation flow acquisition module is used to acquire continuous radiation data according to the motion control sequence and provide real-time feedback of orbital position to obtain timestamped thermal image frames and timestamped position information; and to generate a spatiotemporal thermal spectrum array from the timestamped thermal image frames and timestamped position information. The contextualized thermal anomaly extraction module is used to calculate local temperature deviations based on the spatiotemporal thermal spectrum array, identify preliminary anomaly regions, and obtain preliminary anomaly masks; it then extracts contextualized anomaly masks from the preliminary anomaly masks and generates priority anomaly registration data. The intelligent maintenance scheduling module is used to generate maintenance action plans based on priority anomaly registration data to obtain optional maintenance plans; optimize maintenance scheduling based on optional maintenance plans, and generate optimized maintenance work orders.

[0086] Therefore, the embodiments should be considered as exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of the equivalents of the application are intended to be included within the invention.

[0087] The above description is merely a specific embodiment of the present invention, enabling those skilled in the art to understand or implement the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the present invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features of the invention herein.

Claims

1. A method for aerial orbital thermal imaging temperature inspection, characterized in that, Includes the following steps: Step S1: Identify the areas to be inspected to obtain a list of areas to be evaluated; Based on the list of regions to be evaluated, the heat-sensitive regions are refined to obtain refined regions of interest; a topological interest map is generated from the refined regions of interest to obtain the topological interest map. Step S2: Extract inspection points from the topological interest map and calculate the target attitude to obtain the target attitude sequence; allocate motion parameters according to the target attitude sequence to obtain segmented motion parameters; generate motion control sequences from the segmented motion parameters to obtain motion control sequences. Step S3: Acquire continuous radiometric data according to the motion control sequence and provide real-time feedback on the orbital position to obtain time-stamped thermal image frames and time-stamped position information; Spatiotemporal thermal spectrum arrays are generated by combining time-stamped thermal image frames and time-stamped location information to obtain spatiotemporal thermal spectrum arrays. Step S4: Calculate the local temperature deviation based on the spatiotemporal thermal spectrum array, and perform preliminary anomaly region identification to obtain a preliminary anomaly mask; extract the contextualized anomaly mask from the preliminary anomaly mask, and generate priority anomaly registration data. Step S5: Generate a maintenance action plan based on the priority anomaly registration data to obtain optional maintenance plans; optimize the maintenance schedule based on the optional maintenance plans and generate an optimized maintenance work order to obtain an optimized maintenance work order.

2. The aerial orbital thermal imaging temperature inspection method according to claim 1, characterized in that, Step S1 includes the following steps: Step S11: Perform preliminary asset identification in the area to be inspected to obtain an equipment list; Step S12: Associate spatial information based on the equipment list to obtain a spatialized equipment list; Step S13: Based on the spatialized equipment list, integrate the equipment priority information to obtain the list of areas to be evaluated; Step S14: Refine the heat-sensitive areas based on the list of areas to be evaluated to obtain refined regions of interest; Step S15: Construct and optimize the spatial topology of the refined region of interest to obtain the labeled region of interest; Step S16: Generate a topological interest map for the labeled regions of interest to obtain the topological interest map.

3. The aerial orbital thermal imaging temperature inspection method according to claim 2, characterized in that, Step S14 includes the following steps: Step S141: Obtain the 3D model based on the list of areas to be evaluated, and perform model preprocessing to obtain preprocessed 3D model data; Step S142: Perform intelligent template matching on the preprocessed 3D model data and the preset hotspot template library to obtain matching hotspot templates; Step S143: Based on the preprocessed 3D model data, perform local parameterization adjustment on the matching hotspot template to obtain the locally adjusted hotspot region; Step S144: Obtain equipment operating parameters; perform thermal simulation-driven hotspot supplementation based on preprocessed 3D model data and equipment operating parameters to obtain simulated supplementary hotspots; Step S145: Generate a refined region of interest based on the local hotspot adjustment area and the simulated supplementary hotspot, thus obtaining the refined region of interest.

4. The aerial orbital thermal imaging temperature inspection method according to claim 3, characterized in that, Step S144 is as follows: The simulation model is automatically constructed based on the preprocessed 3D model data and equipment operating parameters to obtain the initial simulation model. The initial simulation model is meshed and optimized to obtain an optimized simulation model. Steady-state thermal simulation calculations were performed based on the optimized simulation model and equipment operating parameters to obtain steady-state temperature field data. Transient thermal simulation analysis was performed based on the optimized simulation model to obtain transient temperature change data. Steady-state temperature field data and transient temperature change data are superimposed onto preprocessed 3D model data to identify hotspots and obtain simulated predicted hotspot regions. By associating the simulated hotspots with the template hotspots in the simulated predicted hotspot regions and the locally adjusted hotspot regions, we can obtain the simulated supplementary hotspots.

5. The aerial orbital thermal imaging temperature inspection method according to claim 1, characterized in that, Step S2 includes the following steps: Step S21: Extract inspection points from the topological interest map and sort the inspection points to obtain the inspection point sequence; Step S22: Calculate the target attitude of the inspection point sequence to obtain the target attitude sequence; Step S23: Assign motion parameters according to the target posture sequence to obtain segmented motion parameters; Step S24: Generate motor commands from the segmented motion parameters to obtain a discrete motor command set; Step S25: Perform timestamp allocation and synchronization on the discrete motor instruction set to obtain timestamped motor instructions; Step S26: Generate a motion control sequence from the timestamped motor commands to obtain the motion control sequence.

6. The aerial orbital thermal imaging temperature inspection method according to claim 5, characterized in that, Step S23 includes the following steps: Step S231: Set the priority-weighted velocity reference according to the target attitude sequence to obtain the regional velocity reference; Step S232: Adjust the regional velocity benchmark for uncertainty correction to obtain the uncertainty-adjusted velocity; Step S233: Adjust the speed according to the uncertainty to perform multi-objective optimization speed planning and obtain the optimized segmented speed; Step S234: Adjust the optimized segmentation speed in real time using image quality feedback to obtain the feedback adjustment speed; Step S235: Calculate the segmented acceleration and deceleration parameters based on the feedback adjustment speed and target attitude sequence to obtain the segmented motion parameters.

7. The aerial orbital thermal imaging temperature inspection method according to claim 1, characterized in that, Step S3 includes the following steps: Step S31: Perform precise execution of track motion according to the motion control sequence to obtain the real-time track status; Step S32: Based on the real-time orbital status, acquire continuous radiation data using a thermal imager to obtain time-stamped thermal image frames; Step S33: Provide real-time feedback on the track position based on the real-time track status to obtain time-stamped position information; Step S34: Perform time synchronization calibration on the time-stamped thermal image frame and the time-stamped location information to obtain the calibrated synchronized data stream; Step S35: Perform spatiotemporal data fusion on the calibrated synchronized data stream and construct spatiotemporal thermal spectrum data to obtain preliminary spatiotemporal thermal spectrum data; Step S36: Generate a spatiotemporal thermal spectrum array from the preliminary spatiotemporal thermal spectrum data to obtain the spatiotemporal thermal spectrum array.

8. The aerial orbital thermal imaging temperature inspection method according to claim 1, characterized in that, Step S4 includes the following steps: Step S41: Construct a baseline temperature model based on the spatiotemporal thermal spectrum array to obtain the regional baseline temperature; Step S42: Calculate the local temperature deviation of the spatiotemporal thermal spectrum array and the regional baseline temperature to obtain a local temperature deviation map; Step S43: Perform preliminary anomaly region identification on the local temperature deviation map to obtain a preliminary anomaly mask; Step S44: Perform contextual information fusion and enhancement on the preliminary anomaly mask and topological interest map to obtain a contextualized anomaly mask; Step S45: Contextualize the anomaly mask with anomaly scoring to obtain a list of anomalies with scores; Step S46: Perform anomaly filtering and noise reduction based on the rated anomaly list to obtain a refined anomaly list; Step S47: Generate priority exception registration data by refining the exception list.

9. The aerial orbital thermal imaging temperature inspection method according to claim 1, characterized in that, Step S5 includes the following steps: Step S51: Perform anomaly risk assessment on the priority anomaly registration data to obtain the anomaly risk level; Step S52: Based on the spatiotemporal thermal spectrum array, predict the anomaly development trend of the priority anomaly registration data to obtain anomaly evolution prediction data; Step S53: Generate maintenance action plans based on the anomaly risk level and anomaly evolution prediction data to obtain optional maintenance plans; Step S54: Assess the resource requirements of the optional maintenance schemes to obtain a resource requirement list; Step S55: Optimize maintenance scheduling based on optional maintenance schemes and resource requirement lists to obtain an optimized maintenance sequence; Step S56: Generate an optimization maintenance work order based on the priority anomaly registration data, resource requirement list, and optimization maintenance sequence.

10. An aerial orbital thermal imaging temperature inspection system, characterized in that, For performing the aerial orbital thermal imaging temperature inspection method as described in claim 1, the aerial orbital thermal imaging temperature inspection system includes: The target domain definition module is used to identify the evaluation area of ​​the area to be inspected and obtain a list of areas to be evaluated; the thermally sensitive areas are refined based on the list of areas to be evaluated to obtain refined regions of interest; and a topological interest map is generated from the refined regions of interest to obtain a topological interest map. The adaptive trajectory synthesis module is used to extract inspection points from the topological interest map and calculate the target attitude to obtain the target attitude sequence; it then allocates motion parameters based on the target attitude sequence to obtain segmented motion parameters; and finally generates a motion control sequence from the segmented motion parameters to obtain the motion control sequence. The synchrotron radiation flow acquisition module is used to acquire continuous radiation data according to the motion control sequence and provide real-time feedback of orbital position to obtain timestamped thermal image frames and timestamped position information; and to generate a spatiotemporal thermal spectrum array from the timestamped thermal image frames and timestamped position information. The contextualized thermal anomaly extraction module is used to calculate local temperature deviations based on the spatiotemporal thermal spectrum array, identify preliminary anomaly regions, and obtain preliminary anomaly masks; it then extracts contextualized anomaly masks from the preliminary anomaly masks and generates priority anomaly registration data. The intelligent maintenance scheduling module is used to generate maintenance action plans based on priority anomaly registration data to obtain optional maintenance plans; optimize maintenance scheduling based on optional maintenance plans, and generate optimized maintenance work orders.

Images