A method and system for batch processing of infrared inspection photos of a drone

Abstract

This application relates to the field of intelligent inspection and infrared image processing technology for power equipment, and discloses a method and system for batch analysis and processing of UAV infrared inspection photos. The method includes: image enhancement of batch-imported infrared photos; constructing a multi-dimensional relational database containing temperature, pixel coordinates, and physical distance based on shooting metadata; intelligently partitioning power equipment using an equipment structure model library; extracting the temperature of hot spots and ambient temperature; and filtering out the normal corresponding point temperatures under the same operating conditions based on electrical connection topology and pixel statistical characteristics; calculating the relative temperature difference; and finally, determining the defect level based on the equipment type using a dynamic threshold library and automatically generating a report. This invention, by introducing an environmentally corrected relative temperature difference model and physical space correlation technology, achieves automated and standardized processing of massive inspection data, effectively eliminating the interference of ambient temperature changes on thermal defect determination, and significantly improving the accuracy of fault identification and maintenance efficiency.

Description

Technical Field

[0001] This invention relates to the field of intelligent inspection of power equipment and infrared image processing technology, specifically to a method and system for batch analysis and processing of infrared inspection photos taken by unmanned aerial vehicles (UAVs). Background Technology

[0002] With the advancement of smart grid construction, using drones equipped with infrared thermal imagers to inspect transmission lines has become an important means for power operation and maintenance departments to detect equipment overheating defects and prevent power grid accidents. This non-contact detection method can efficiently acquire thermal field distribution information of power equipment and promptly detect potential overheating hazards caused by poor contact, insulation aging, or wire damage.

[0003] With the increasing frequency and expansion of drone inspection operations, the volume of infrared image data collected during inspections has exploded, posing a significant challenge to subsequent data processing and analysis. Currently, infrared inspection data processing relies primarily on manual image-by-image interpretation. Maintenance personnel must manually select equipment, identify hotspots, and enter data into the logbook from massive amounts of images. This traditional manual processing method is not only inefficient and unable to meet the real-time analysis needs of large-scale inspection data, but also prone to missed or incorrect judgments due to visual fatigue caused by the varying sensitivity of the human eye to grayscale or pseudo-color images. Furthermore, raw infrared images are often affected by sensor noise and background clutter, resulting in blurred edges and further increasing the difficulty of manually identifying equipment structures and locating faulty parts.

[0004] In the core defect assessment stage, existing technologies typically employ a simple absolute temperature difference method, which directly calculates the difference between the temperature of the heating point and a reference point to evaluate the severity of the defect. This method has significant limitations: firstly, the selection of the reference point (normal corresponding point) often relies on manual experience or simple regional low-temperature searches, lacking rigorous consideration of the electrical connections of the equipment and similar operating conditions, resulting in a lack of representativeness in the selected benchmark temperature; secondly, the absolute temperature difference method ignores the nonlinear influence of ambient temperature on the temperature rise characteristics of the equipment, failing to incorporate ambient temperature as a thermodynamic benchmark into the calculation system. Under different seasons, climates, or load conditions, the same absolute temperature difference often represents drastically different defect levels, leading to a lack of objectivity and comparability in the assessment results, making it difficult to truly reflect the health status of the equipment.

[0005] Different types of power equipment (such as tension clamps, splicing pipes, and insulators) have different heating mechanisms and allowable temperature rise standards. Existing automated analysis software often uses a one-size-fits-all threshold setting method, lacking dynamic judgment logic based on equipment type, making it difficult to achieve fine-grained classification of general defects, major defects, and emergency defects. This low level of standardization and reliance on subjective experience results in inconsistent quality of inspection reports, failing to provide accurate and quantitative data support for maintenance decisions. Therefore, there is an urgent need for a UAV infrared inspection photo analysis method that can achieve batch automated processing, has environmental temperature correction capabilities, and provides objective and unified judgment standards. Summary of the Invention

[0006] To address the shortcomings of existing technologies, this invention provides a method and system for batch analysis and processing of UAV infrared inspection photos. It solves the problems of low efficiency in manual processing of massive amounts of data in existing UAV infrared inspection, as well as the low accuracy and high misjudgment rate of thermal defect identification caused by the traditional temperature difference judgment method due to the lack of consideration for environmental temperature correction and the subjectivity of benchmark selection.

[0007] To achieve the above objectives, the present invention provides the following technical solution:

[0008] The first aspect of this invention provides a method for batch analysis and processing of infrared inspection photos taken by unmanned aerial vehicles (UAVs), the method comprising the following steps:

[0009] Step S1: Import the infrared inspection photos of power transmission lines collected by the drone into the analysis system in batches, and extract the shooting metadata of the infrared inspection photos;

[0010] Step S2: Based on the principle of infrared thermal imaging, perform pixel-by-pixel temperature analysis on the preprocessed infrared inspection photos, and establish an associated database in conjunction with the captured metadata;

[0011] Step S3: Call the preset power equipment structure model library, divide the power equipment in the infrared inspection photo into structural partitions, extract the highest temperature of the partition from the associated database as the heat point temperature for each partition, select the pixel area in normal operation, extract the temperature value of the pixel area as the normal corresponding point temperature, and at the same time obtain the ambient temperature of the area of ​​the tested equipment.

[0012] Step S4: Calculate the relative temperature difference based on the temperature of the heating point, the temperature of the normal corresponding point, and the ambient temperature, and combine it with the preset defect judgment criteria to determine the defect level of the equipment zone using the temperature of the heating point and the relative temperature difference.

[0013] Step S5: Summarize the analysis results of all infrared inspection photos and generate an infrared inspection defect report.

[0014] Preferably, in step S1, the specific operations for image noise reduction and edge enhancement preprocessing of the infrared inspection photos include: smoothing the infrared inspection photos using a Gaussian filtering algorithm to eliminate image noise; and performing convolution operations on the smoothed infrared inspection photos using the Sobel operator to enhance the edge contour features of the power equipment components. The shooting metadata includes at least the shooting time, the drone's shooting altitude, and the infrared camera's focal length.

[0015] Preferably, in step S2, the specific operation of establishing the association database includes: calculating the actual physical distance between adjacent pixels in the infrared inspection photo based on the drone's shooting altitude and the infrared camera's focal length; traversing each pixel in the infrared inspection photo, mapping and associating the pixel's parsed temperature value, pixel coordinates, and corresponding actual physical distance, and storing them in the association database.

[0016] Preferably, in step S3, the selection logic for the normal corresponding point temperature is as follows: based on the power equipment structure model library, identify the body area or conductor area within the equipment partition; exclude background interference pixels and abnormally high temperature pixels in the body area or conductor area, and determine the remaining pixel set as the pixel point area in normal operation; select the pixel point temperature that can characterize the normal temperature rise state of the equipment under the current operating load as the normal corresponding point temperature.

[0017] Preferably, in step S4, the calculation logic for the relative temperature difference is as follows: calculate the difference between the temperature of the heating point and the normal corresponding point temperature as the temperature rise difference caused by the equipment defect; calculate the difference between the temperature of the heating point and the ambient temperature as the overall temperature rise of the heating point relative to the environment; calculate the ratio of the temperature rise difference to the overall temperature rise, and determine the ratio as the relative temperature difference.

[0018] Preferably, in step S4, the preset defect judgment criteria include a judgment rule library corresponding to different types of power equipment; before judging the defect level of the equipment partition using the heating point temperature and the relative temperature difference, the method further includes: identifying the equipment type of the current analysis object according to the result of the structural partition; and calling the corresponding temperature threshold set and relative temperature difference threshold set from the judgment rule library according to the equipment type as the judgment basis.

[0019] Preferably, in step S4, determining the defect level of the equipment partition specifically includes: determining whether the temperature of the heating point falls within the temperature threshold range, and simultaneously determining whether the relative temperature difference falls within the relative temperature difference threshold range; based on the combined determination result of the heating point temperature and the relative temperature difference, classifying the equipment defect level into a general defect, a major defect, or an emergency defect.

[0020] In one specific embodiment, when the power equipment is a hardware component, the defect level determination logic is as follows: if the relative temperature difference is greater than or equal to 35% and less than 80%, and the temperature of the heating point is less than 90°C, it is determined to be a general defect; if the temperature of the heating point is greater than or equal to 90°C and less than or equal to 130°C, or the relative temperature difference is greater than or equal to 80% and the temperature of the heating point is less than 90°C, it is determined to be a major defect; if the temperature of the heating point is greater than 130°C, or the relative temperature difference is greater than or equal to 95% and the temperature of the heating point is greater than 90°C, it is determined to be an emergency defect.

[0021] Preferably, steps S4 and S5 further include: identifying the heating location based on the spatial distribution characteristics of the heating point temperature within the equipment partition, and matching the defect type as current-induced heating, voltage-induced heating, or combined heating based on the heating location and the numerical range of the heating point temperature; the infrared inspection defect report includes the photo shooting time, equipment name, equipment partition information, the heating point temperature, the normal corresponding point temperature, the ambient temperature, the relative temperature difference, the defect type, and the defect level.

[0022] A second aspect of the present invention provides a batch analysis and processing system for UAV infrared inspection photos, used to execute the batch analysis and processing method for UAV infrared inspection photos described in the first aspect above. The system includes the following modules that are sequentially connected in communication:

[0023] The photo import and preprocessing module is configured to import infrared inspection photos in batches, perform image noise reduction and edge enhancement processing, and output the preprocessed infrared inspection photos and extracted shooting metadata.

[0024] The temperature analysis and association module is configured to receive the preprocessed infrared inspection photos and the captured metadata, perform pixel-by-pixel temperature analysis on the infrared inspection photos, and establish an association database containing temperature information, pixel coordinate information, and physical distance information.

[0025] The equipment partitioning and temperature measurement module is configured to partition the equipment based on the associated database, call the power equipment structure model library, and extract the temperature of the heating point, the normal corresponding point, and the ambient temperature of each partition.

[0026] The defect determination module is configured to acquire the temperature of the heating point, the temperature of the normal corresponding point, and the ambient temperature; calculate the relative temperature difference based on the temperature rise of the heating point relative to the environment and the temperature rise of the normal corresponding point relative to the environment; and determine the equipment defect level by combining the temperature of the heating point and the relative temperature difference.

[0027] The report generation module is configured to receive the analysis results output by the defect determination module, summarize them, and generate an infrared inspection defect report.

[0028] The technical solution provided by this invention preprocesses and extracts metadata from batch-imported infrared inspection photos, constructs a relational database containing temperature, coordinates, and physical distance based on infrared thermal imaging principles, and automatically partitions the equipment using a power equipment structural model library, extracting the temperature of hot spots, normal corresponding points, and ambient temperatures respectively. Based on this, a relative temperature difference calculation logic including the normal corresponding point temperature and ambient temperature is employed, combined with a pre-set judgment standard library for different equipment types (such as hardware components), to determine the level of equipment defects. This solution achieves batch and automated processing of UAV infrared inspection data, improves the accuracy and consistency of thermal defect identification in power equipment by introducing ambient temperature parameters to correct the relative temperature difference calculation logic and setting specific defect judgment thresholds for specific equipment components.

[0029] This invention provides a method and system for batch analysis and processing of infrared inspection photos taken by unmanned aerial vehicles (UAVs). It has the following beneficial effects:

[0030] 1. This invention integrates image noise reduction, edge enhancement preprocessing, and a power equipment structure model library to achieve batch import and automatic structural partitioning of infrared inspection photos. This automated processing flow replaces the traditional manual image-by-image analysis mode. By using Gaussian filtering and the Sobel operator to enhance the edge contour features of the equipment, it can quickly and accurately locate and extract temperature data of different equipment partitions in batch data, shortening the analysis cycle and reducing the labor intensity of manual processing.

[0031] 2. This invention constructs a relative temperature difference calculation logic that includes the normal corresponding point temperature and the ambient temperature. By utilizing the established database of temperature, pixel coordinates and physical distance association, it can accurately select a normal reference point under the same working conditions as the heating point. This method corrects the temperature rise difference caused by equipment defects by calculating the overall temperature rise of the heating point relative to the environment, effectively eliminating the influence of ambient temperature changes and background heat source interference on the temperature measurement results, and ensuring the objectivity and stability of the defect judgment index under different seasons or climate conditions.

[0032] 3. This invention adopts a dynamic judgment rule library for different types of power equipment. In particular, it sets a combination judgment logic with clear temperature threshold range and relative temperature difference threshold range for hardware components. The system can automatically identify the equipment type and call the corresponding judgment standard according to the structural partitioning results, realizing the automated classification of general defects, major defects and emergency defects, avoiding the subjective bias of human experience judgment, and ensuring the standardization and consistency of the judgment of various defect types such as current-induced heating. Attached Figure Description

[0033] Figure 1 This is a flowchart of the method of the present invention;

[0034] Figure 2 This is a schematic diagram of the system module connections of the present invention;

[0035] Figure 3 This is a schematic diagram of the automatic partitioning of the fittings device of the present invention;

[0036] Figure 4 This is a schematic diagram of the infrared image temperature analysis results of the present invention;

[0037] Figure 5 This is a schematic diagram of the defect report structure of the present invention. Detailed Implementation

[0038] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0039] See attached document Figure 2 This invention provides a batch analysis and processing system for UAV infrared inspection photos. The system is configured to perform batch data processing, temperature field analysis, and automatic defect identification of UAV infrared inspection photos.

[0040] like Figure 1 As shown, the UAV infrared inspection photo batch analysis and processing system includes: a photo import and preprocessing module, a temperature analysis and correlation module, an equipment zoning and temperature measurement module, a defect judgment module, and a report generation module. These modules are sequentially connected via a data bus or communication interface to form a data flow processing link.

[0041] The photo import and preprocessing module is configured to receive externally input UAV infrared inspection photo datasets. This module features batch reading capabilities and supports infrared image files in JPG, TIFF, or R-JPEG formats. Internally, it integrates an image processing algorithm unit configured to perform image noise reduction and edge enhancement processing on each input infrared inspection photo. Noise reduction employs a Gaussian filtering algorithm to smooth the pixel value distribution and suppress sensor noise; edge enhancement uses the Sobel operator for convolution operations to highlight the geometric contours of electrical equipment components in the image. Furthermore, the module includes a metadata parsing unit to extract shooting metadata from the header information of the infrared image files. This metadata includes at least the shooting time, UAV shooting altitude, and infrared camera focal length. The module outputs the preprocessed infrared inspection photo and the extracted shooting metadata.

[0042] The temperature analysis and correlation module communicates with the output of the photo import and preprocessing module. This module is configured to receive preprocessed infrared inspection photos and their metadata. Based on the principle of infrared thermal imaging radiation thermometry, it converts the grayscale or pseudo-color values ​​of the infrared image into specific temperature values, achieving pixel-by-pixel temperature analysis. Simultaneously, based on the drone's shooting altitude and the infrared camera's focal length, the module uses trigonometric relationships to calculate the actual physical distance between adjacent pixels in the infrared inspection photo within the real-world scene. The module maps and binds the analyzed temperature value, pixel coordinates in the image system, and corresponding actual physical distance for each pixel, constructing a correlation database.

[0043] The equipment zoning and temperature measurement module has its input end communicating with the output end of the temperature analysis and correlation module. Internally, this module stores a pre-defined power equipment structural model library, which includes infrared feature templates for transformers, circuit breakers, and transmission line fittings. The module is configured to call upon this library and, based on an image feature matching algorithm, identify and categorize the power equipment in infrared inspection photos into different structural zones. For example, fittings may be categorized into terminal block areas, body areas, and connecting bolt areas.

[0044] For each defined equipment zone, the equipment zone and temperature measurement module are configured to perform the following operations:

[0045] Iterate through the associated database to find the temperature of all pixels within the device partition, extract the maximum value as the temperature of the heating point, and denot it as... ;

[0046] Within the device partition or an area electrically connected to the device partition, pixel areas in normal operating condition are selected based on preset normal state characteristics. The temperature value of these areas is extracted as the normal corresponding point temperature, denoted as . ;

[0047] Obtain the ambient temperature of the area where the device under test is located, and denot it as... .

[0048] Ambient temperature This is obtained by reading environmental parameters from the shooting metadata or by temperature statistics of non-device background areas in the image.

[0049] The defect detection module has its input terminal connected to the output terminal of the equipment partition and the temperature measurement module. The defect detection module is configured to acquire the temperature of the heating point. Normal corresponding point temperature and ambient temperature .

[0050] The defect determination module integrates a relative temperature difference calculation unit, which is configured to calculate the relative temperature difference based on the temperature rise of the heating point in a relatively dry environment and the temperature rise of the corresponding normal point in a relatively dry environment.

[0051] Specifically, the defect determination module calculates the relative temperature difference based on the following formula:

[0052] ;

[0053] In the formula, Indicates relative temperature difference. Indicates the temperature of the heating point. This indicates the normal corresponding temperature. Indicates ambient temperature.

[0054] In this embodiment, relative temperature difference refers to the normalization of the temperature rise difference of the equipment's heating point relative to the current ambient temperature and the total temperature rise amplitude, thereby eliminating the influence of ambient temperature and load fluctuations and obtaining a dimensionless, objective percentage indicator.

[0055] The defect determination module is also connected to a defect determination standard library, which stores sets of temperature thresholds and relative temperature difference thresholds for different equipment types. The defect determination module is configured to retrieve the corresponding thresholds based on the type of the currently analyzed equipment, determine the range in which the heating point temperature and relative temperature difference fall, and output the defect level of the equipment accordingly. The defect levels include general defects, major defects, and emergency defects.

[0056] The report generation module's input is communicatively connected to the output of the defect assessment module. The report generation module is configured to receive the analysis results output by the defect assessment module and summarize the analysis data of all batch-processed photos. The report generation module is also configured to generate an infrared inspection defect report by arranging the photo capture time, equipment name, equipment partition information, hot spot temperature, normal corresponding point temperature, ambient temperature, relative temperature difference, defect type, and defect level into a structured document according to a preset report template.

[0057] At the hardware implementation level, the aforementioned UAV infrared inspection photo batch analysis and processing system runs on a computer device. This computer device includes a memory, a processor, and computer programs stored in the memory and executable on the processor. The memory stores a power equipment structural model library, a defect judgment standard library, and a related database. When the processor executes the computer program, it implements the functions of the photo import and preprocessing module, the temperature analysis and correlation module, the equipment zoning and temperature measurement module, the defect judgment module, and the report generation module. For batch processing of massive amounts of image data, the computer device is equipped with a graphics processing unit (GPU) to provide parallel matrix operation capabilities, accelerating image convolution operations and temperature field analysis processes.

[0058] See attached document Figure 1 The batch analysis and processing method for UAV infrared inspection photos provided by this invention mainly includes the following specific implementation steps and technical details regarding image preprocessing and the construction of a multidimensional association database:

[0059] In the infrared image batch import and preprocessing stage of step S1, the system first establishes a queue of tasks to be processed and sequentially reads the infrared inspection image file data streams collected by the UAV. For each frame of the original infrared image, the system performs a smooth convolution operation based on a Gaussian kernel function to eliminate the inherent shot noise and random thermal noise of the thermal imaging sensor.

[0060] Specifically, a two-dimensional Gaussian filtering algorithm is employed, scanning the image using a convolution kernel of a preset size (e.g., 3×3 or 5×5). The numerical distribution of this convolution kernel conforms to a two-dimensional Gaussian distribution function. A weighted average is performed using this convolution kernel and the gray values ​​of each pixel in the image and its neighboring pixels to recalculate the gray value of that pixel. This process preserves low-frequency thermal distribution information while filtering out high-frequency noise signals, preventing noise points from being misidentified as high-temperature anomalies in subsequent steps.

[0061] After smoothing and denoising, the system then performs edge enhancement to address the common issues of edge blurring and low contrast in infrared images. Specifically, the system uses the Sobel operator to calculate the spatial gradient of the smoothed image. This process involves two 3×3 convolution kernels, used to calculate the spatial gradient of the image in the horizontal direction (…). grayscale gradient (direction) and vertical direction ( grayscale gradient (direction) .

[0062] For each pixel in the image, the system uses the formula Calculate the gradient magnitude at this point. If the calculated gradient magnitude is greater than a preset threshold, the pixel is determined to be an edge point and is preserved or enhanced. Through the above Sobel operator convolution operation, the thermal contour boundary between power equipment (such as conductors, insulator strings, and hardware) and the background environment is significantly sharpened, providing clear geometric feature boundaries for subsequent equipment structure partitioning.

[0063] Simultaneously, the system parses the header data segment (such as the Exif information segment) of the infrared image file to extract the shooting metadata. This metadata forms the basis for subsequent physical distance calculations; specific extracted fields include: the timestamp of the shooting time and the relative flight altitude of the drone at the time of shooting. (Unit: meters) Focal length of infrared camera lens (Unit: mm) and the pixel size of the infrared detector (Unit: micrometers)

[0064] In the temperature analysis and associated database construction phase of step S2, the system first performs pixel-by-pixel temperature inversion based on the principle of infrared thermal imaging. The system reads the raw radiation data from the infrared image file and, according to Planck's blackbody radiation law and the camera's radiometric calibration curve, converts the radiation intensity value recorded for each pixel into a Celsius temperature value. This process calls upon atmospheric transmittance, emissivity, and reflectance temperature parameters from the metadata for radiometric correction, thereby obtaining the true surface temperature of the object corresponding to that pixel. .

[0065] Subsequently, to achieve quantitative analysis of the equipment size and the impact range of thermal defects, the system performs spatial geometric mapping calculations to establish the correspondence between pixel coordinates and actual physical space. The system utilizes the pinhole imaging principle and the triangle similarity theorem to calculate the spatial resolution of the image, i.e., the ground sampling distance (GSD). The calculation logic is as follows:

[0066] ;

[0067] in, Represents the width or length (unit: meters / pixel) of a pixel in an image in actual physical space. The altitude for drone photography. For sensor pixel size, It is the focal length.

[0068] Based on calculations The system can calculate the numerical values ​​of any two adjacent pixels or any two pixels in an image. and The actual physical distance between The calculation formula is:

[0069] ;

[0070] Furthermore, as an alternative implementation method for calculating physical distance based on metadata, the system is also equipped with distance calibration logic based on reference object features. This is useful when metadata is missing or when relative altitude is affected by errors in the drone's GPS signal. When the data is unreliable, the system automatically initiates the reference calibration procedure.

[0071] The system identifies standard power components (such as single insulators in an insulator string) in infrared images and reads the physical diameter of the standard component stored in the model library. (For example, the diameter of a standard disc suspension insulator is 255mm). The system calculates the pixel diameter of this insulator in the image. (Unit: pixels), and then the ground sampling distance is calculated.

[0072] ;

[0073] The system uses this back-reasoning to obtain Replace the original formula Continue executing subsequent physical distance calculations. The calculation. This dual verification mechanism ensures that the spatial information in the associated database remains accurate and reliable even in the event of metadata anomalies.

[0074] Finally, the system iterates through every pixel of the infrared inspection photos to construct a multidimensional relational database. This database uses a matrix or hash table data structure, with the two-dimensional coordinates of the pixels as the data source. Using the index key, the parsed temperature value of the pixel, its position in the image coordinate system, and the physical spatial scale factor (GSD) calculated based on the above formula, or its physical distance to other key feature points, are mapped, associated, and stored as attribute values. This step transforms the planar infrared image into a digital structure containing three-dimensional information about the temperature, geometric, and spatial fields, providing data support for subsequent determination of electrical connections and thermal defect levels at the physical spatial scale.

[0075] See attached document Figure 3In step S3, the present invention calls a preset power equipment structure model library to realize the automatic identification and structured partitioning of key components of power equipment in complex backgrounds, and extracts feature temperature values ​​for subsequent calculations based on specific selection logic.

[0076] Specifically, the power equipment structure model library pre-stores standard infrared feature templates, geometric contour descriptors, and topological connection data for various standard power equipment (such as tension clamps, suspension clamps, transformer bushings, circuit breaker contacts, etc.).

[0077] The system utilizes image recognition algorithms (such as deep learning-based semantic segmentation networks or gradient-based contour matching algorithms) to perform traversal matching between preprocessed infrared inspection photos and features in the model library. When a specific device target is identified, the system automatically generates a device partition mask according to the model definition, dividing the target device into several independent logical partitions.

[0078] For example, for a tension clamp, the system divides it into a crimping pipe area, a drain plate area, and a connecting wire area. For each generated device partition, the system performs a pixel-level temperature traversal operation, compares the resolved temperature values ​​of all pixels within the partition's mask area, determines the temperature with the highest value as the heating point temperature, and records the pixel coordinates of that point.

[0079] After determining the temperature of the heating point, the system executes the selection process for the critical normal corresponding temperature.

[0080] This process does not simply select the lowest temperature within the region, but rather aims to find a reference temperature that represents the normal operating condition of the equipment under the current load. The specific selection logic is as follows:

[0081] First, based on the topological connection relationships in the power equipment structure model library, the system identifies the main body area or extended conductor area that has a direct electrical connection relationship with the current heat-generating equipment zone and sets it as a candidate sampling area.

[0082] Next, the system performs statistical analysis on the pixel temperature distribution histogram within the candidate sampling area and executes bidirectional threshold filtering:

[0083] Pixels with temperatures below the environmental reference threshold are removed to eliminate the impact of background interference pixels such as sky, ground, or vegetation on lowering the reference temperature.

[0084] High-temperature abnormal pixels whose temperature values ​​are within a certain gradient range of the heating point temperature (e.g., above 80% of the heating point temperature) are removed to prevent the heat conduction area from interfering with the determination of the normal reference value.

[0085] After the above exclusion process, the system uses statistical methods (such as calculating the median or mode) to extract a temperature value from the remaining set of valid pixels that can characterize the steady-state distribution of the region, and uses it as the normal corresponding point temperature.

[0086] This selection method ensures that the corresponding point temperature and the heating point temperature are under the same current load and environmental conditions, making the relative temperature difference calculated subsequently physically comparable.

[0087] Meanwhile, the system obtains the ambient temperature of the area of ​​the tested equipment through multiple methods.

[0088] In one implementation, the system directly parses the real-time temperature data recorded by the UAV's onboard meteorological sensor from the metadata of the infrared inspection photos.

[0089] In another implementation (e.g., when metadata is missing), the system executes an image background analysis algorithm to identify low-temperature regions (such as shadowed backgrounds) that are large, connected areas with uniform temperature distribution in the edge regions of the infrared image, and calculates the average temperature of the region as a substitute for the ambient temperature.

[0090] Through the above steps, the system has completed the accurate conversion from the original image to structured feature temperature data, providing reliable data input for subsequent defect level determination.

[0091] See attached document Figure 1 In step S4, the core of this invention lies in using a relative temperature difference algorithm corrected for ambient temperature and dynamic multidimensional judgment logic for specific equipment types to achieve standardized automatic classification of thermal defects in power equipment.

[0092] Specifically, the system first calls the relative temperature difference calculation unit integrated within the defect judgment module to perform a temperature rise analysis based on environmental parameter corrections. This differs from traditional infrared analysis, which only calculates the absolute temperature difference between the hot spot and the normal point. This invention introduces ambient temperature. As a thermodynamic benchmark, the relative temperature difference is calculated using the aforementioned formula. .

[0093] In this calculation logic, the numerator represents the additional temperature rise caused by internal defects in the equipment (such as excessive contact resistance or increased insulation dielectric loss); the denominator represents the total temperature rise of the heat source relative to the current ambient thermal equilibrium state.

[0094] By calculating the ratio of the two, the system can eliminate the influence of different ambient temperature conditions (such as low temperature in winter and high temperature in summer) on the absolute operating temperature of the equipment. At the same time, it also normalizes the influence of the current line load current on the absolute value of temperature rise to a certain extent, so that the calculated relative temperature difference can more objectively reflect the severity of equipment defects.

[0095] After calculating the relative temperature difference, the system enters the defect level determination stage. This stage relies on a pre-set defect determination standard library, which employs a dynamic calling mechanism that maps equipment type indexes to threshold sets. The system first confirms the equipment type of the current analysis object (e.g., conductor splice pipe, tension clamp, transformer bushing, etc.) based on the structural partitioning results in step S3, and then retrieves the corresponding specific determination rules from the database.

[0096] Taking the most common hardware components in transmission lines (such as tension clamps and splicing pipes) as an example, this embodiment of the invention sets up a strict joint judgment logic based on both surface temperature and relative temperature difference, which is specifically executed as follows:

[0097] General Defect Judgment: The system determines whether the relative temperature difference is within the closed-open range of 35% to 80%, and simultaneously checks whether the temperature of the heating point is less than 90℃. Only when both conditions are met simultaneously, the system marks the defect level of the equipment zone as a general defect.

[0098] Such defects typically correspond to initial oxidation or slight loosening of the equipment's contact surfaces and do not pose an immediate threat to operational safety.

[0099] Major defect assessment: The system performs a logical OR operation. If the detected temperature of the hot spot is within a closed range of 90℃ to 130℃;

[0100] Alternatively, if a relative temperature difference of 80% or greater is detected, but the temperature of the heating point is still less than 90°C.

[0101] If any of the above conditions are met, the system will upgrade the defect level to a critical defect. This type of defect indicates a significant increase in contact resistance and requires repair in the near future.

[0102] Emergency Defect Judgment: The system also performs a logical OR operation. If the detected hot spot temperature exceeds 130℃;

[0103] Or if the relative temperature difference is detected to be as high as 95% or more, and the temperature of the heating point exceeds 90°C.

[0104] If any of the above conditions are met, the system will mark the defect level as the highest-level emergency defect. This type of defect means that the equipment has an extremely high risk of burn-out and wire breakage, and the system will immediately trigger an alarm mechanism.

[0105] In addition, after determining the defect level, the system further combines the spatial distribution characteristics of the temperature of the heating point within the equipment partition to identify the defect type.

[0106] The system calculates the gradient distribution of the thermal image of the heating area. If the heating center is highly concentrated at the metal connection point (such as the bolt crimping) and the thermal attenuation gradient is large, the system determines it to be a current-induced heating type defect (usually caused by poor contact).

[0107] If the heating area is diffusely distributed throughout the system or located in the insulating medium (such as the insulator skirt), and the temperature is correlated with the voltage level, the system classifies it as a voltage-induced heating defect (usually caused by insulation degradation). If both characteristics are present, it is classified as a combined heating defect. This automatic classification based on physical characteristics provides a direct basis for subsequent maintenance strategy development.

[0108] To quantitatively distinguish between current-induced and voltage-induced defects, the defect determination module further performs feature analysis based on the thermal gradient decay rate. The system uses the temperature of the heating point... A one-dimensional temperature distribution curve is established with the pixel at the center along the direction of the wire extension. The system calculates the temperature decay gradient within a specified physical distance (e.g., 10cm) extending to both sides of the heating center. :

[0109] ;

[0110] in, Distance from the fever center The temperature at that location.

[0111] If the calculated gradient If the gradient value is greater than the preset gradient threshold (e.g., 5℃ / cm), it indicates that the heat is highly concentrated and the heat source is resistance heating from the metal contact surface. The system determines that it is current-induced heating.

[0112] If gradient If the temperature is less than the preset gradient threshold and the coverage area of ​​the heating zone exceeds the preset length ratio of the insulating medium, it indicates that the heat is diffusely distributed, and the system determines it to be voltage-induced heating.

[0113] By introducing quantitative calculation of physical gradients, this invention effectively solves the problem of subjective uncertainty in classifying hot spot shapes based solely on human visual observation.

[0114] See attached document Figure 1 Appendix Figure 4 and attached Figure 5 This section will use a specific application scenario to illustrate in detail how the system completes the workflow from raw data input to final report generation, in order to further verify the feasibility and logical rigor of the technical solution of this invention.

[0115] In this embodiment, it is assumed that the dataset to be processed is a folder of infrared images of a drone inspection of a 220kV high-voltage transmission line. The system starts a batch processing task and reads an infrared photo with the file name "IR_20251118_Line220_Tower056.jpg".

[0116] First, in the photo import and preprocessing stage, the system reads the photo file. The photo import and preprocessing module automatically extracts the shooting metadata from the Exif information in the file header, identifying the shooting time as 14:30:00 and the drone's shooting altitude. The focal length of the infrared camera is 30 meters. The diameter is 19 mm, and the recorded ambient temperature of the airborne sensor is 28 degrees Celsius.

[0117] Meanwhile, the image processing unit performs Gaussian smoothing filtering on the original image to remove high-frequency random thermal noise in the background, and uses the Sobel operator to calculate the gradient of the image. In the processed image, the contrast between the outline of the metal fittings of the transmission line and the background sky is significantly enhanced, forming a clear edge feature mask.

[0118] Next, the equipment partitioning and feature extraction stage begins. The equipment partitioning and temperature measurement module is based on a power equipment structure model library. Through feature matching, it identifies the main body of the equipment in the center of the image as a tension clamp (C phase) and automatically divides it into three logical partitions: the current-draining plate, the crimping pipe, and the connecting wire.

[0119] The system traverses the pixel matrix, detects the pixel with the highest radiation temperature within the compression pipe partition, and extracts it as the temperature of the heating point. .

[0120] In this embodiment, the system parses The value was 98℃. Subsequently, within the area of ​​the connecting wire directly connected to the crimping pipe, the system removed background pixels with temperatures below ambient temperature and heat-conducting pixels near the crimping pipe, selecting the median value of the remaining pixel set as the normal corresponding temperature. .

[0121] In this embodiment, the parsed The value is 45℃. Simultaneously, the system reads the ambient temperature from the metadata. The temperature is 28℃.

[0122] Subsequently, in the core calculation and defect determination phase, the defect determination module performs precise calculations of the relative temperature difference. The system substitutes the extracted temperature parameters into the relative temperature difference calculation formula:

[0123] ;

[0124] The relative temperature difference in the tension wire clamping nozzle area was calculated. Approximately 75.7%.

[0125] The system then invokes the defect judgment standard library for hardware components to perform logical verification:

[0126] General defect inspection: Judgment condition is 35% ≤ <80% and In this example, although It meets the interval requirements, but Since the temperature does not meet the requirement of being less than 90℃, it is not classified as a general defect.

[0127] Major defect verification: The judgment condition is 90℃≤ ≤130℃ or ( ≥80% and <90℃T). In this example, The temperature is 98℃, falling entirely within the closed range of 90℃ to 130℃. Although the relative temperature difference does not reach 80%, according to the absolute temperature threshold judgment rules, the equipment condition has triggered the judgment conditions for a major defect.

[0128] Emergency defect verification: The judgment condition is as follows >130℃, etc. This condition is not met in this example.

[0129] Based on the above logical operations, the system ultimately classified the defect level of the tension clamp as a major defect. Furthermore, the system analyzed the heat source. Based on the spatial gradient, it was found that the heat was highly concentrated in the middle of the press-fit pipe, which is consistent with the thermal characteristics caused by excessive contact resistance. Therefore, the defect type was marked as current-induced heating type.

[0130] Finally, during the report generation phase, the report generation module automatically generates a structured defect record:

[0131] Equipment Name: 220kV XX Line #056 Tower C Phase Tension Clamp

[0132] Defective area: crimped pipe

[0133] Measured temperatures: Heating point 98℃ / Normal point 45℃ / Ambient temperature 28℃

[0134] Technical specifications: Relative temperature difference 75.7%

[0135] Defect level: Major defect

[0136] Recommended action: It is recommended to arrange a power outage for maintenance or infrared retest within two weeks.

[0137] This record is written into the final infrared inspection defect report document. Through the above embodiments, the invention fully demonstrates how it utilizes an ambient temperature correction model and multi-dimensional judgment logic to accurately identify complex defect situations that may lead to misjudgments based on a single indicator (such as the case in this example where the relative temperature difference does not meet the standard but the absolute temperature exceeds the standard), verifying the accuracy and practicality of the technical solution.

Claims

1. A method for batch analysis and processing of infrared inspection photos taken by unmanned aerial vehicles (UAVs), characterized in that, Includes the following steps: Step S1: Import the infrared inspection photos of power transmission lines collected by the drone into the analysis system in batches, and extract the shooting metadata of the infrared inspection photos; Step S2: Based on the principle of infrared thermal imaging, perform pixel-by-pixel temperature analysis on the preprocessed infrared inspection photos, and establish an associated database in conjunction with the captured metadata; Step S3: Call the preset power equipment structure model library, divide the power equipment in the infrared inspection photo into structural partitions, extract the highest temperature of the equipment partition from the associated database as the heat point temperature for each equipment partition, select the pixel area in normal operation, extract the temperature value of the pixel area as the normal corresponding point temperature, and at the same time obtain the ambient temperature of the tested equipment area. Step S4: Calculate the relative temperature difference based on the temperature of the heating point, the temperature of the normal corresponding point, and the ambient temperature; use the temperature of the heating point and the relative temperature difference to determine the defect level of the equipment partition. Step S5: Summarize the analysis results of all infrared inspection photos and generate an infrared inspection defect report.

2. The method for batch analysis and processing of UAV infrared inspection photos according to claim 1, characterized in that, Step S1 specifically includes: A Gaussian filtering algorithm is used to smooth the infrared inspection photos and eliminate image noise; The Sobel operator is used to perform convolution operations on the smoothed infrared inspection photos to enhance the edge contour features of power equipment components. The shooting metadata includes at least the shooting time, the drone's shooting altitude, and the infrared camera's focal length.

3. The method for batch analysis and processing of UAV infrared inspection photos according to claim 2, characterized in that, In step S2, establishing the associated database specifically includes: Based on the drone's shooting altitude and the infrared camera's focal length, calculate the actual physical distance between adjacent pixels in the infrared inspection photo; The process involves iterating through each pixel in the infrared inspection photos, mapping and associating the pixel's temperature value, pixel coordinates, and corresponding actual physical distance, and storing these associations in the associated database.

4. The method for batch analysis and processing of UAV infrared inspection photos according to claim 1, characterized in that, In step S3, the selection logic for the normal corresponding point temperature is as follows: Identify the body region or conductor region within the equipment partition based on the power equipment structure model library; Background interference pixels and abnormally high temperature pixels are excluded from the body area or the conductor area, and the remaining pixel set is determined as the pixel area in normal operation. The temperature of a pixel that can characterize the normal temperature rise of the device under the current operating load is selected as the normal corresponding point temperature.

5. The method for batch analysis and processing of UAV infrared inspection photos according to claim 1, characterized in that, In step S4, the calculation logic for the relative temperature difference is as follows: The difference between the temperature of the heating point and the temperature of the corresponding normal point is calculated as the temperature rise difference caused by the equipment defect; The difference between the temperature of the heating point and the ambient temperature is calculated as the overall temperature rise of the heating point relative to the environment. Calculate the ratio of the temperature rise difference to the overall temperature rise, and determine the ratio as the relative temperature difference.

6. The method for batch analysis and processing of UAV infrared inspection photos according to claim 5, characterized in that, In step S4, determining the defect level of the equipment partition specifically includes: Identify the device type of the current analysis object based on the results of the structural partitioning; Based on the device type, the corresponding temperature threshold set and relative temperature difference threshold set are retrieved from the judgment rule base as the judgment basis; Determine whether the temperature of the heating point falls within the temperature threshold range, and simultaneously determine whether the relative temperature difference falls within the relative temperature difference threshold range; Based on the combined determination result of the temperature of the heating point and the relative temperature difference, the equipment defect level is classified into general defect, major defect or emergency defect.

7. The method for batch analysis and processing of UAV infrared inspection photos according to claim 6, characterized in that, When the power equipment is a hardware component, the defect level determination logic is as follows: If the relative temperature difference is greater than or equal to 35% and less than 80%, and the temperature of the heating point is less than 90°C, it is determined to be a general defect; If the temperature of the heating point is greater than or equal to 90°C and less than or equal to 130°C, or if the relative temperature difference is greater than or equal to 80% and the temperature of the heating point is less than 90°C, it is determined to be a major defect. If the temperature of the heating point is greater than 130°C, or the relative temperature difference is greater than or equal to 95% and the temperature of the heating point is greater than 90°C, it is determined to be an emergency defect.

8. The method for batch analysis and processing of UAV infrared inspection photos according to claim 1, characterized in that, Step S4 further includes: Based on the spatial distribution characteristics of the temperature at the heating point within the equipment partition, the heating location is identified; Based on the temperature range of the heating part and the heating point, the defect type is matched as current-induced heating type, voltage-induced heating type, or combined heating type.

9. The method for batch analysis and processing of UAV infrared inspection photos according to claim 1, characterized in that, In step S5, the infrared inspection defect report includes the photo shooting time, equipment name, equipment partition information, the temperature of the hot spot, the temperature of the normal corresponding point, the ambient temperature, the relative temperature difference, the defect type, and the defect level.

10. A batch analysis and processing system for UAV infrared inspection photos, executing the batch analysis and processing method for UAV infrared inspection photos as described in any one of claims 1-9, characterized in that, include: The photo import and preprocessing module is used to import infrared inspection photos in batches, perform image noise reduction and edge enhancement processing, and output the preprocessed infrared inspection photos and extracted shooting metadata. The temperature analysis and association module is used to receive the preprocessed infrared inspection photos and the shooting metadata, perform pixel-by-pixel temperature analysis on the infrared inspection photos, and establish an association database containing temperature information, pixel coordinate information and physical distance information. The equipment partitioning and temperature measurement module is used to partition the equipment based on the associated database and call the power equipment structure model library, and extract the temperature of the heating point, the normal corresponding point temperature and the ambient temperature of each partition. The defect determination module is used to obtain the temperature of the heating point, the temperature of the normal corresponding point, and the ambient temperature. It calculates the relative temperature difference based on the temperature rise of the heating point relative to the environment and the temperature rise of the normal corresponding point relative to the environment, and determines the equipment defect level by combining the temperature of the heating point and the relative temperature difference. The report generation module is used to receive the analysis results output by the defect judgment module, summarize them, and generate an infrared inspection defect report.

Images