A method and system for identifying data defects in substation equipment

Abstract

The present invention discloses a method and system for identifying data defects in substation equipment, relating to the technical field of substation component defect detection. The method includes a porcelain sleeve state analysis step, in which a camera mounted on an unmanned aerial vehicle scans a target porcelain sleeve to generate a three-dimensional point cloud model, thereby determining the porcelain sleeve state and flight environment; a shooting angle coarse adjustment step, in which the shooting angle of the thermal imager is adjusted according to the porcelain sleeve state; a flight path planning step, in which shooting points are predetermined based on the three-dimensional point cloud model and a circular flight path is planned; an image acquisition and splicing step, in which thermal images are acquired and spliced; and a temperature defect analysis step, in which anomalies are determined based on temperature differences in the spliced ​​images. The advantages of the present invention include strategies such as porcelain sleeve state analysis, point addition and obstacle avoidance in flight path planning, and panoramic completion in image splicing. This ensures that complete and effective thermal imaging information can be obtained even in complex situations, such as when the porcelain sleeve is upright or tilted, or in the presence of circular flight obstacles, thereby accurately determining porcelain sleeve abnormalities and abnormal areas.

Description

Technical Field

[0001] The present invention relates to the technical field of substation component defect detection, and more particularly to a substation equipment data defect identification method and system. Background Art

[0002] With the rapid development of smart grid construction, the demand for automated inspection and intelligent defect identification of substation equipment is becoming increasingly prominent. Currently, substation defect detection relies primarily on manual inspections and traditional image processing methods. These methods suffer from low detection efficiency, poor environmental adaptability, and insufficient recognition accuracy, making them difficult to meet the requirements for real-time monitoring and high-precision diagnosis of power equipment. In particular, under the influence of interference factors such as complex lighting conditions and equipment surface contamination, the feature extraction capabilities and detection stability of existing algorithms are significantly reduced, seriously restricting the accuracy and practicality of defect identification.

[0003] Current computer vision-based substation defect detection technologies mostly use a single target detection model, which has limited ability to identify small defects and complex defect types, and lacks a multi-source data fusion mechanism; traditional methods usually only analyze visible light image data and fail to fully utilize multimodal information such as infrared thermal imaging and ultrasonic detection, resulting in incomplete defect feature extraction. In addition, although the application of drone technology has partially solved the shortcomings of manual inspections, existing drone inspection systems still have significant limitations. Such systems are usually equipped with only a single visible light sensor, and their effectiveness is greatly reduced when lighting conditions are poor or when detecting hidden defects, making it difficult to meet application scenarios with high real-time requirements such as drone inspections.

[0004] In addition, after searching, a ceramic bushing detection system, method, device, equipment and storage medium with publication number CN117589793B was disclosed, and the publication date is April 16, 2024. The patent obtains the appearance image of the ceramic bushing through an image acquisition device, and combines an ultrasonic oscillation device and an infrared thermal imaging device to detect external and internal defects of the ceramic bushing; this technical solution realizes multi-source data fusion and improves the effectiveness of ceramic bushing defect detection. However, this technical solution is mainly aimed at the specific application scenarios of ceramic bushings, and fails to fully consider the special needs of substation porcelain bushings in complex environments. For example, under strong electromagnetic interference, the stability and reliability of its ultrasonic oscillation device and infrared thermal imaging device may be affected, resulting in a decrease in detection accuracy. In addition, the setting position of the substation porcelain bushing is not taken into account. For example, when the porcelain bushing is tilted, it is difficult for the detection equipment to obtain a complete image of the outer ring of the porcelain bushing, which affects the detection accuracy of the porcelain bushing. Summary of the Invention

[0005] In view of the shortcomings of the existing technology, the purpose of the present invention is to provide a method and system for identifying data defects in substation equipment. Through strategies such as porcelain sleeve status analysis, point addition and obstacle avoidance in flight path planning, and panoramic completion in image stitching, it is ensured that complete and effective thermal imaging information can still be obtained in complex situations such as when the porcelain sleeve is upright or tilted, or there are obstacles in circular flight, so as to achieve accurate judgment of abnormal conditions and abnormal areas of the porcelain sleeve, and ultimately improve the reliability and practicality of substation porcelain sleeve defect identification.

[0006] To achieve the above object, the present invention provides the following technical solutions:

[0007] A method for identifying data defects in substation equipment includes the following steps:

[0008] a porcelain sleeve state analysis step, wherein the target porcelain sleeve is scanned by a camera carried by the UAV to generate a three-dimensional point cloud model, and the porcelain sleeve state and flight environment are determined based on the three-dimensional point cloud model, wherein the porcelain sleeve state includes upright or tilted, and the flight environment includes whether circumferential flight is allowed or circumferential flight is obstacle-free;

[0009] a shooting angle coarse adjustment step, selecting whether to trigger a porcelain sleeve angle analysis condition according to the porcelain sleeve state; if so, calculating the inclination angle of the porcelain sleeve according to the porcelain sleeve spatial coordinate information in the three-dimensional point cloud model, and adjusting the shooting angle of the thermal imager carried by the UAV according to the inclination angle;

[0010] A flight path planning step includes presetting a plurality of shooting points on the periphery of the porcelain sleeve based on the three-dimensional point cloud model, and then selecting whether to trigger shooting point adjustment conditions based on the flight environment. If so, marking the reserved points among the shooting points based on preset UAV flight safety information, and planning additional points based on the angular interval constraints between two adjacent shooting points and the distance constraints between the shooting points and the porcelain sleeve, and planning a circular flight path based on the reserved points and the additional points.

[0011] An image acquisition and stitching step is to obtain thermal images captured by the thermal imager under the circular flight path after adjusting the shooting angle, and stitch them together to obtain a stitched image;

[0012] The temperature defect analysis step determines the abnormal condition and abnormal area of ​​the porcelain sleeve according to the temperature difference in the splicing diagram.

[0013] Furthermore, the image acquisition and stitching step also includes a panoramic completion strategy, which includes a missing area identification step and a missing area completion step;

[0014] The missing area identification step performs pixel confidence analysis on the stitched panorama to identify the missing area, extracts its outline and boundary coordinates through an edge detection algorithm, and extracts valid pixels within a set range outside the missing area. When the number of valid pixels is insufficient, the screening radius is expanded until the number is met;

[0015] In the missing area completion step, the temperature of each point in the missing area is inferred from the temperature value of the valid pixel point through the Kriging interpolation method according to the external structure of the porcelain sleeve.

[0016] Furthermore, the flight path planning step includes a point supplementation strategy, which includes a quantization constraint step and a supplementation point coordinate calculation step;

[0017] The quantitative constraint step includes arbitrarily marking a point between two adjacent reserved points as an additional point, wherein the central angle between the additional point and the adjacent reserved point is less than or equal to a preset angle threshold, and the straight-line distance between the additional point and the outer surface of the porcelain sleeve is within a preset distance range;

[0018] The step of calculating the coordinates of the additional points is to construct a cylindrical coordinate system with one end point of the central axis of the porcelain sleeve as the origin. The three axes of the cylindrical coordinate system include the direction perpendicular to the origin and the central axis of the porcelain sleeve, the circumferential direction rotating around the central axis of the porcelain sleeve, and the direction coinciding with the central axis of the porcelain sleeve. The coordinate values ​​of the additional points are calculated based on the coordinate values ​​of the retained points in the cylindrical coordinate system and the porcelain sleeve size data.

[0019] Furthermore, the point supplementation strategy also includes a supplementary point obstacle avoidance verification step, which includes constructing a three-dimensional detection box at the supplementary point, and at the same time judging whether the three-dimensional detection box has volume overlap with the obstacle in the cylindrical coordinate system, and simulating the overlap rate between the thermal image collected when the supplementary point is added and the obstacle. When there is volume overlap and / or the overlap rate is greater than or equal to a preset threshold, the supplementary point is determined to be invalid and the supplementary point coordinate calculation step is performed again.

[0020] Furthermore, it also includes a shooting angle fine-tuning step, identifying the pixel coordinates of the same calibration point based on any two thermal images taken at different shooting points, and converting the pixel coordinates into image physical coordinates, and then looking up the theoretical physical coordinates of the calibration point in the three-dimensional point cloud model, comparing the image physical coordinates with the theoretical physical coordinates, and if there is a deviation, calculating the angle correction amount through the image physical coordinates and the theoretical physical coordinates, and adjusting the shooting angle of the thermal imager carried by the drone with the angle correction amount.

[0021] Furthermore, the shooting angle coarse adjustment step includes a porcelain sleeve angle calculation strategy, which includes extracting the spatial coordinate information of the porcelain sleeve in the three-dimensional point cloud model, and the spatial coordinate information includes the coordinates of the two endpoints of the porcelain sleeve axis, connecting the two endpoint coordinates and calculating the angle with the horizontal plane as the inclination angle.

[0022] Furthermore, the porcelain sleeve state analysis step includes a morphological analysis strategy, which includes extracting several feature points of the porcelain sleeve in the three-dimensional point cloud model, the feature points including a porcelain sleeve boundary point set, connecting each boundary point in the boundary point set for straight line fitting, and judging whether the fitting straight line coincides with a vertical line constructed by any boundary point in the boundary point set, and according to the judgment result, determining whether the porcelain sleeve is in an upright state or an inclined state.

[0023] Furthermore, the porcelain sleeve state analysis step includes a flight analysis strategy, which includes fitting the porcelain sleeve point cloud in the three-dimensional point cloud model, and synchronously segmenting the obstacle point cloud around the porcelain sleeve point cloud, calculating the minimum distance between any point in the porcelain sleeve point cloud and any point in the obstacle point cloud, and then comparing the minimum distance with the preset UAV flight safety information to obtain whether the flight environment allows circular flight or circular flight obstacles.

[0024] Furthermore, the temperature defect analysis step includes a model building strategy, which includes training, testing and verifying the preset substation defect recognition initial model with historical thermal imaging images to obtain a substation defect recognition model, and then inputting the spliced ​​image into the substation defect recognition model to obtain the porcelain sleeve abnormality and abnormal area based on the temperature difference analysis.

[0025] A substation equipment data defect identification system, comprising:

[0026] The porcelain sleeve state analysis module uses the camera onboard the drone to scan the target porcelain sleeve to generate a three-dimensional point cloud model, and determines the porcelain sleeve state and flight environment based on the three-dimensional point cloud model. The porcelain sleeve state includes upright or tilted, and the flight environment includes whether circumferential flight is allowed or circumferential flight is obstacle-free.

[0027] a shooting angle coarse adjustment module, which selects whether to trigger the porcelain sleeve angle analysis condition according to the porcelain sleeve state; if so, calculates the inclination angle of the porcelain sleeve according to the porcelain sleeve spatial coordinate information in the three-dimensional point cloud model, and adjusts the shooting angle of the thermal imager carried by the drone based on the inclination angle;

[0028] A flight path planning module predetermines a number of shooting points on the periphery of the porcelain sleeve based on the three-dimensional point cloud model, and then selects whether to trigger shooting point adjustment conditions based on the flight environment. If so, it marks the reserved points among the shooting points according to the preset UAV flight safety information, and plans additional points based on the angular interval constraints between two adjacent shooting points and the distance constraints between the shooting points and the porcelain sleeve, and plans a circular flight path based on the reserved points and the additional points;

[0029] An image acquisition and stitching module acquires thermal images captured by the thermal imager under the circular flight path after adjusting the shooting angle, and stitches them together to obtain a stitched image;

[0030] The temperature defect analysis module determines the abnormal condition and abnormal area of ​​the porcelain sleeve according to the temperature difference in the splicing diagram.

[0031] The beneficial effects of the present invention are as follows: 1. The tilt angle is calculated through the three-dimensional point cloud model and the shooting angle of the thermal imager is adjusted to ensure that accurate thermal imaging data can still be obtained when the posture of the porcelain sleeve changes, and the problem of circumferential flight obstacles in the flight environment is taken into consideration. By retaining point screening and supplementing point planning, combined with angle interval, distance constraint and obstacle avoidance verification, the optimal circumferential flight path is planned to avoid shooting omissions due to obstacles and ensure the comprehensiveness of the peripheral image of the porcelain sleeve. In addition, a panoramic completion strategy is introduced to complete the temperature information after edge detection and identification of the missing areas in the spliced ​​image, further improving the image integrity and providing a complete data basis for subsequent defect analysis.

[0032] 2. Through the "coarse adjustment + fine adjustment" shooting angle adjustment mechanism, coarse adjustment is based on the tilt angle of the porcelain sleeve, and fine adjustment is calculated by calculating the correction amount through the deviation between the image physical coordinates and the theoretical physical coordinates of the calibration point. This ensures the geometric accuracy of the thermal image and reduces the impact of angle deviation on temperature analysis. In the temperature defect analysis stage, the substation defect recognition model is combined to use temperature differences to accurately judge abnormal conditions and areas, fully explore the defect characteristics in the thermal imaging information, and improve the recognition ability of small and complex defects. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 It is the overall flow chart of the present invention;

[0034] Figure 2 It is a three-dimensional point cloud model diagram of the substation in the present invention;

[0035] Figure 3 This is a diagram showing the porcelain sleeve in an upright state in the present invention;

[0036] Figure 4 This is a thermal image of multiple sets of porcelain sleeves at the equipment station in the present invention;

[0037] Figure 5This is a thermal image of a single porcelain sleeve in the present invention;

[0038] Figure 6 This is a flow chart of the steps for coarse adjustment of the shooting angle in the present invention;

[0039] Figure 7 It is a flow chart of the flight path planning steps in the present invention. DETAILED DESCRIPTION

[0040] The present invention will be described in further detail below with reference to the accompanying drawings and embodiments. Identical components are denoted by the same reference numerals. It should be noted that the terms "front," "rear," "left," "right," "upper," and "lower" used in the following description refer to directions in the accompanying drawings, and the terms "bottom," "top," "inner," and "outer" refer to directions toward or away from the geometric center of a particular component, respectively.

[0041] Due to the complicated location of components in the substation, it is difficult for the inspection equipment to collect data for analysis during the inspection. The present invention mainly detects defects in the porcelain bushings in the substation, and the inspection equipment is carried out by drones. A method for identifying data defects in substation equipment is designed. Figure 1 As shown, it includes a porcelain sleeve state analysis step, a shooting angle coarse adjustment step, a flight path planning step, an image acquisition and splicing step, and a temperature defect analysis step;

[0042] The porcelain bushing status analysis step is to use the 3D line scan camera carried by the UAV to perform a full-scale scan of the target porcelain bushing, collect the original point cloud data of the porcelain bushing and the surrounding environment, and perform denoising and filtering on the original point cloud data (such as removing discrete noise points caused by environmental interference and scanning errors) to generate a 3D point cloud model including the porcelain bushing area and surrounding substation components, such as Figure 2 As shown;

[0043] Based on the three-dimensional point cloud model, the point geometric feature recognition algorithm is used to extract the feature point set of the porcelain sleeve (porcelain sleeve point cloud), focusing on extracting the boundary point set of the porcelain sleeve (including the upper and lower edges of the porcelain sleeve, surface contour boundaries and other key geometric feature points). The straight line fitting algorithm is used to fit the boundary points in the boundary point set to obtain a fitting line representing the central axis of the porcelain sleeve. A boundary point is selected from the boundary point set, and a vertical line (a straight line perpendicular to the horizontal plane) is constructed with this point as the starting point. It is determined whether the fitting line coincides with the vertical line. If so, the porcelain sleeve is determined to be in an "upright state", such as Figure 3 As shown, the porcelain sleeve is in an upright state. If not, the porcelain sleeve is judged to be in a "tilted state". Figure 2 As shown, most of the porcelain sleeves are in a tilted state.

[0044] Based on the 3D point cloud model, the segmentation algorithm of geometric features (such as the differences in spatial distribution, curvature, normal vector, etc. of the point cloud) is used to separate the porcelain sleeve point cloud from the surrounding environment point cloud, and the porcelain sleeve point cloud and the surrounding obstacle point cloud (such as multiple sets of porcelain sleeve settings, other power equipment, brackets, trees, etc.) are segmented. Figure 4 As shown in the figure, multiple groups of porcelain sleeves are tilted to form a "V" shape, which makes it difficult for the UAV to fly between two groups of porcelain sleeves. The minimum spatial distance between any point in the porcelain sleeve point cloud and any point in the obstacle point cloud (that is, the closest distance between the porcelain sleeve and the surrounding obstacles) is calculated, and the preset UAV flight safety information (including the safety distance threshold, such as the minimum safe distance between the UAV and the obstacle) is called. The above minimum distance is compared with the safety distance threshold: if the minimum distance is greater than the safety distance threshold, the flight environment is determined to be "circular flight allowed"; if the minimum distance is less than or equal to the safety distance threshold, the flight environment is determined to be "circular flight obstacle", which ensures the safety of the UAV flight path planning, avoids shooting omissions due to obstacle obstruction, and ensures the integrity of subsequent image acquisition.

[0045] Steps for coarse adjustment of shooting angle, such as Figure 6 As shown, the porcelain sleeve state (upright or tilted) output by the porcelain sleeve state analysis step is received to determine whether the porcelain sleeve angle analysis condition is triggered: if the porcelain sleeve state is "tilted", the analysis condition is triggered; if it is "upright", it is not triggered (the default thermal imager shooting angle is the initial vertical angle). When the analysis condition is triggered, the spatial coordinate information of the porcelain sleeve is extracted from the three-dimensional point cloud model, focusing on obtaining the three-dimensional coordinates of the two endpoints of the porcelain sleeve center axis or the three-dimensional coordinates of the two endpoints of the boundary fitting line, which are recorded as endpoints and endpoints ,in, is the horizontal coordinate, is the vertical height coordinate, according to the endpoint and endpoints The coordinates of the porcelain sleeve are used to calculate the angle between the center axis of the porcelain sleeve and the horizontal plane (i.e. the inclination angle ), ,in, For endpoints With endpoint In the vertical direction ( The absolute value of the distance from the axis; For endpoints With endpoint The straight-line distance between the two spaces is calculated, and then the angle adjustment instruction is sent to the drone control system to control the rotation of the thermal imager. Angle, so that the shooting direction of the thermal imager remains perpendicular to the central axis of the porcelain sleeve, completing the coarse adjustment of the shooting angle. After adjustment, the thermal imager can be vertically aligned with the surface of the porcelain sleeve, reducing the temperature information collection error caused by angle deviation, providing accurate raw data for subsequent temperature defect analysis, and the unified shooting angle makes the geometric scale of the thermal imaging images of each point consistent, reducing the difficulty of alignment when stitching images, and indirectly improving the integrity of the stitching image.

[0046] Flight path planning steps, such as Figure 7 As shown in the figure, based on the 3D point cloud model, several shooting points are preset on the circumference of the porcelain sleeve, and the circumference is evenly divided into Equal parts, (such as the default , corresponding to 12 points), determined by the diameter of the porcelain sleeve, the present invention Take 6, the initial angle interval between adjacent points is , and the distance between each point and the outer surface of the porcelain sleeve is the preset standard shooting distance (determined based on porcelain bushing size data);

[0047] Receive flight environment information (circular flight allowed or circular flight obstacle). If the flight environment is "circular flight obstacle", the shooting point adjustment condition is triggered; if it is "circular flight allowed", the predetermined point is directly used as the basic point of the circular flight path, and the preset drone flight safety information (including the safety distance threshold) is called. ), perform safety check on the predetermined points: calculate the minimum distance between each predetermined shooting point and the obstacle point cloud ,like , marked as "reserved point"; if , marked as "point to be deleted".

[0048] Set the maximum allowable angle interval threshold for adjacent reserved points And the distance constraint between the additional point and the porcelain sleeve, the column coordinate system is constructed with one end point of the porcelain sleeve center axis as the origin ( ),in, is the distance between the origin and the center axis of the porcelain sleeve in the vertical direction (i.e. shooting distance), is the circumferential angle of rotation around the central axis of the porcelain sleeve, is the coordinate of the direction coinciding with the center axis of the porcelain sleeve, and the angles of the two adjacent reserved points are and , if the angle interval , then the angle of the additional point needs to be calculated , ,in, is the actual angle interval (satisfying ), To add the number of points, and satisfy ; The coordinates of the additional points are , is the axial coordinate consistent with the reserved point.

[0049] To add points Construct a three-dimensional detection box whose spatial range is: , It is the redundancy set based on the size of the drone. If there is a volume overlap between the stereo detection box and the obstacle point cloud, or the overlap rate between the thermal image taken at that point and the obstacle is simulated and calculated , then the supplementary point is determined to be invalid and the supplementary point coordinate calculation step is repeated.

[0050] Connect all retained points and valid additional points to form a continuous circular flight path. The path must meet the smoothness constraints of the UAV flight, that is, the flight trajectory between adjacent points is a curved transition. The flight path planning step uses quantitative constraints and coordinate calculations to ensure that the angular interval and distance of the additional points meet the shooting requirements, avoid shooting blind spots caused by obstacles, and provide complete data for subsequent image stitching. The obstacle avoidance verification mechanism combines spatial and image dual verification to ensure the safety of the UAV and avoid the loss of temperature data caused by obstacles.

[0051] Image acquisition and stitching steps: Control the UAV to fly along the planned circular flight path, and use the thermal imager with completed angle coarse adjustment to take thermal images of the porcelain sleeve at each retained point and additional point (such as Figure 4 and Figure 5 As shown in the figure, ensure that each thermal image covers the corresponding area around the porcelain sleeve, and that the thermal images of adjacent points have a preset overlapping area. Preprocess the collected thermal images, including removing image noise and unifying the image size and temperature scale. Then use the feature point matching stitching algorithm to stitch the preprocessed thermal images, extract the temperature feature points of each thermal image (such as the edge of the high-temperature area, the temperature mutation point, etc.), identify the common feature points in adjacent thermal images through the feature point matching algorithm (such as the SIFT algorithm), calculate the geometric transformation matrix between the images, map all thermal images to a unified coordinate system according to the transformation matrix, fuse the temperature values ​​of the overlapping areas (take the average or weighted value), and generate a complete panoramic thermal imaging stitching image of the porcelain sleeve periphery.

[0052] Due to image loss due to obstacle occlusion or stitching, the image acquisition and stitching steps also include a panorama completion strategy. Specifically, pixel confidence analysis is performed on the stitched panorama. The outline and boundary coordinates of the panorama are extracted using an edge detection algorithm (such as the Canny algorithm). Areas where temperature information is missing due to obstacle occlusion or shooting angle deviation are located. An initial screening radius is set outside the missing area, and valid pixels within this range (pixels with complete temperature values) are extracted. If the number of valid pixels is less than the preset threshold, the screening radius is gradually expanded until the number requirement is met.

[0053] According to the external structure of the porcelain sleeve, the Kriging interpolation method is used to fill in the temperature values ​​of the missing area. The temperature values ​​of valid pixels are used as samples to construct a temperature spatial distribution model. The spatial correlation between the sample points and each point in the missing area is calculated through the semivariogram, and the temperature value of each pixel in the missing area is inferred to finally generate a complete panoramic thermal image.

[0054] Through feature point matching stitching and panoramic completion strategies, the problem of image loss caused by circumferential flight obstacles and shooting angle deviations is solved, and a complete thermal image covering the entire circumference of the porcelain sleeve is generated, providing comprehensive data for subsequent temperature defect analysis. In addition, the preprocessing step unifies the temperature scale, and the temperature values ​​of overlapping areas are fused during stitching. The completion step infers missing temperatures based on spatial correlation, reducing data errors and ensuring the continuity and reliability of temperature distribution. Moreover, edge detection and Kriging interpolation methods are combined with the external structure of the porcelain sleeve to accurately complete temperature information even in the case of severe occlusion or incomplete shooting, improving the method's adaptability to complex substation environments.

[0055] In the temperature defect analysis step, the panoramic thermal imaging mosaic image (including the missing areas after completion) generated in the image acquisition and stitching step is preprocessed, including temperature value standardization and background noise filtering. Then, historical thermal imaging data is collected, including labeled data of known defect types, such as overheating defects and local low temperature anomalies, to build a training set. The target detection algorithm is used to build an initial substation defect recognition model, and the historical thermal imaging images are input into the model for training, testing and verification. In the training phase, the model parameters are optimized through the back propagation algorithm to minimize the error between the predicted defect area and the labeled area. In the testing and verification phase, the test set is used to evaluate the recognition accuracy and recall rate of the model, and the model parameters are adjusted through cross-validation to finally obtain the optimized substation defect recognition model.

[0056] The pre-processed panoramic thermal image mosaic is fed into the substation defect recognition model. The model analyzes the temperature difference features in the image (such as hot spots where the temperature is more than 5°C higher than normal, and blocky areas with uneven temperature distribution) and outputs the following results:

[0057] Determination of abnormal conditions of the porcelain bushing (such as whether there is overheating defect, temperature abnormality caused by insulation aging, etc.);

[0058] The location of the abnormal area (marking the coordinate range of the abnormal area in the mosaic image by a bounding box) and the temperature value range (such as the maximum temperature and average temperature of the abnormal area).

[0059] Finally, the output results will be recorded and fed back so that the back-end terminal can view and make corresponding response plans.

[0060] The defect analysis of the present invention uses a substation defect recognition model trained with historical data to accurately capture subtle temperature differences in thermal images, overcoming the problem of traditional manual analysis missing minor defects. In particular, the recognition ability of hidden defects such as local overheating and abnormal temperature distribution is significantly improved. In addition, it can integrate multiple models to combine temperature characteristics and porcelain sleeve structure information to distinguish different types of defects (such as overheating defects and insulation defects), providing targeted basis for subsequent maintenance.

[0061] Since the coarse adjustment step only adjusts the shooting angle based on the tilt angle of the porcelain sleeve, it does not take into account the subtle angle deviations caused by factors such as the UAV's flight attitude fluctuations, 3D point cloud model errors, and thermal imager installation errors. The angle deviation of the thermal image will directly affect the accuracy of the temperature information. For example, the angle tilt may cause the thermal image of the same area at different points to show different temperature distribution characteristics, or cause the boundary positioning of the high-temperature area to shift. Therefore, the present invention also includes a shooting angle fine adjustment step, which specifically includes:

[0062] ①, calibration point recognition, in the thermal images collected at different shooting points of the circular flight path, the same calibration point is identified by feature matching algorithm (such as matching based on temperature gradient or geometric contour) (the obvious feature points on the surface of the porcelain sleeve or the feature points on other parts on the top of the porcelain sleeve can be selected), and the pixel coordinates of the calibration point in the two thermal images are recorded as and ( is the horizontal pixel coordinate, is the vertical pixel coordinate).

[0063] ② Image physical coordinate conversion, according to the internal parameters of the thermal imager (focal length , pixel size ), convert the pixel coordinates of the calibration points into image physical coordinates (two-dimensional coordinates with the optical center of the thermal imager as the origin), where , , for the same calibration point in the two thermal images, calculate the physical coordinates of the image and ,in, is the image physical coordinate of the calibration point, is the pixel coordinate of the calibration point, is the center pixel coordinate of the thermal image, are the horizontal and vertical pixel sizes of the thermal imager.

[0064] ③. Extract theoretical physical coordinates and locate the spatial coordinates of the calibration point in the 3D point cloud model , and project it onto the imaging plane of the thermal imager (based on the posture information of the drone at the corresponding shooting point) to obtain the theoretical physical coordinates of the calibration point .

[0065] ④. Angle deviation calculation: compare the image physical coordinates of the calibration point with the theoretical physical coordinates, and calculate the deviation values ​​in the horizontal and vertical directions: , , calculate the angle correction value (horizontal correction angle) according to the deviation value and vertical correction angle ), calculated as follows: , ,in, is the focal length of the thermal imager.

[0066] ⑤、Fine-tune the shooting angle according to the calculated horizontal correction angle and vertical correction angle , send angle adjustment instructions to the drone control system to control the thermal imager to rotate horizontally and vertically respectively and , complete the fine adjustment of the shooting angle.

[0067] By calculating the coordinate deviation of the calibration points, the shooting angle of the thermal imager is accurately corrected, which compensates for the angular error that may exist in the coarse adjustment step and ensures the geometric accuracy of the thermal image. In addition, based on the comparison between the physical coordinates of the image and the theoretical physical coordinates, the thermal imaging data and three-dimensional spatial information are directly linked, which reduces the temperature distribution distortion caused by angular deviation and provides a more reliable image basis for subsequent temperature defect analysis.

[0068] This invention designs a data defect identification system for substation equipment, replacing the traditional model that relies on manual inspections. The system uses drones for autonomous scanning, path planning, image acquisition, and intelligent analysis to automate the process of porcelain sleeve defect detection. This eliminates the need for manual climbing or close-range operation, reducing labor costs and operation time. It is particularly suitable for batch inspections of large-scale substations and meets the efficiency requirements of real-time monitoring of power equipment. Specifically, it includes:

[0069] The porcelain sleeve state analysis module uses the camera onboard the drone to scan the target porcelain sleeve to generate a 3D point cloud model. The module then determines the porcelain sleeve state and flight environment based on the 3D point cloud model. The porcelain sleeve state includes upright or tilted, and the flight environment includes whether circumferential flight is allowed or obstacle-free.

[0070] The shooting angle coarse adjustment module selects whether to trigger the porcelain sleeve angle analysis condition according to the porcelain sleeve status. If so, it calculates the porcelain sleeve's tilt angle based on the porcelain sleeve's spatial coordinate information in the 3D point cloud model and adjusts the shooting angle of the thermal imager on the drone based on the tilt angle;

[0071] The flight path planning module predetermines several shooting points around the porcelain sleeve based on the 3D point cloud model, and then determines whether to trigger the shooting point adjustment conditions based on the flight environment. If so, it marks the reserved points among the shooting points according to the preset UAV flight safety information, and plans additional points based on the angular interval constraints between two adjacent shooting points and the distance constraints between the shooting points and the porcelain sleeve. The circular flight path is planned based on the reserved points and the additional points.

[0072] An image acquisition and stitching module acquires thermal images captured by the thermal imager under the circular flight path after adjusting the shooting angle, and stitches them together to obtain a stitched image;

[0073] The temperature defect analysis module determines the abnormal conditions and abnormal areas of the porcelain sleeve based on the temperature differences in the splicing diagram.

[0074] The above are merely preferred embodiments of the present invention. The scope of protection of the present invention is not limited to the above embodiments. All technical solutions based on the principles of the present invention are within the scope of protection of the present invention. It should be noted that improvements and modifications that do not depart from the principles of the present invention are within the scope of protection of the present invention.

Claims

1. A method for identifying data defects in substation equipment, characterized by: The steps include: a porcelain sleeve state analysis step, wherein the target porcelain sleeve is scanned by a camera carried by the UAV to generate a three-dimensional point cloud model, and the porcelain sleeve state and flight environment are determined based on the three-dimensional point cloud model, wherein the porcelain sleeve state includes upright or tilted, and the flight environment includes whether circumferential flight is allowed or circumferential flight is obstacle-free; a shooting angle coarse adjustment step, selecting whether to trigger a porcelain sleeve angle analysis condition according to the porcelain sleeve state; if so, calculating the inclination angle of the porcelain sleeve according to the porcelain sleeve spatial coordinate information in the three-dimensional point cloud model, and adjusting the shooting angle of the thermal imager carried by the UAV according to the inclination angle; A flight path planning step includes presetting a plurality of shooting points on the periphery of the porcelain sleeve based on the three-dimensional point cloud model, and then selecting whether to trigger shooting point adjustment conditions based on the flight environment. If so, marking the reserved points among the shooting points based on preset UAV flight safety information, and planning additional points based on the angular interval constraints between two adjacent shooting points and the distance constraints between the shooting points and the porcelain sleeve, and planning a circular flight path based on the reserved points and the additional points. An image acquisition and stitching step is to obtain thermal images captured by the thermal imager under the circular flight path after adjusting the shooting angle, and stitch them together to obtain a stitched image; A temperature defect analysis step, determining abnormal conditions and abnormal areas of the porcelain sleeve based on temperature differences in the splicing diagram; It also includes a shooting angle fine-tuning step, identifying the pixel coordinates of the same calibration point based on any two thermal images taken at different shooting points, and converting the pixel coordinates into image physical coordinates, then looking up the theoretical physical coordinates of the calibration point in the three-dimensional point cloud model, comparing the image physical coordinates with the theoretical physical coordinates, and if there is a deviation, calculating the angle correction amount through the image physical coordinates and the theoretical physical coordinates, and adjusting the shooting angle of the thermal imager carried by the drone with the angle correction amount.

2. A method for identifying substation equipment data defects according to claim 1, characterized in that: The image acquisition and stitching step also includes a panoramic completion strategy, which includes a missing area identification step and a missing area completion step; The missing area identification step performs pixel confidence analysis on the stitched panorama to identify the missing area, extracts its outline and boundary coordinates through an edge detection algorithm, and extracts valid pixels within a set range outside the missing area. When the number of valid pixels is insufficient, the screening radius is expanded until the number is met; In the missing area completion step, the temperature of each point in the missing area is inferred from the temperature value of the valid pixel point through the Kriging interpolation method according to the external structure of the porcelain sleeve.

3. The method for identifying substation equipment data defects according to claim 2, characterized in that: The flight path planning step includes a point supplementation strategy, which includes a quantization constraint step and a supplementation point coordinate calculation step; The quantitative constraint step includes arbitrarily marking a point between two adjacent reserved points as an additional point, wherein the central angle between the additional point and the adjacent reserved point is less than or equal to a preset angle threshold, and the straight-line distance between the additional point and the outer surface of the porcelain sleeve is within a preset distance range; The step of calculating the coordinates of the additional points is to construct a cylindrical coordinate system with one end point of the central axis of the porcelain sleeve as the origin. The three axes of the cylindrical coordinate system include the direction perpendicular to the origin and the central axis of the porcelain sleeve, the circumferential direction rotating around the central axis of the porcelain sleeve, and the direction coinciding with the central axis of the porcelain sleeve. The coordinate values ​​of the additional points are calculated based on the coordinate values ​​of the retained points in the cylindrical coordinate system and the porcelain sleeve size data.

4. A method for identifying substation equipment data defects according to claim 3, characterized in that: The point supplementation strategy also includes a supplementary point obstacle avoidance verification step, which includes constructing a three-dimensional detection box at the supplementary point, and at the same time judging whether the three-dimensional detection box has volume overlap with the obstacle in the cylindrical coordinate system, and simulating the overlap rate between the thermal image collected when the supplementary point is added and the obstacle. When there is volume overlap and / or the overlap rate is greater than or equal to a preset threshold, the supplementary point is determined to be invalid and the supplementary point coordinate calculation step is performed again.

5. The method for identifying substation equipment data defects according to claim 1, characterized in that: The shooting angle coarse adjustment step includes a porcelain sleeve angle calculation strategy, which includes extracting the spatial coordinate information of the porcelain sleeve from the three-dimensional point cloud model. The spatial coordinate information includes the coordinates of the two endpoints of the porcelain sleeve axis, connecting the two endpoint coordinates and calculating the angle with the horizontal plane as the inclination angle.

6. A method for identifying substation equipment data defects according to claim 5, characterized in that: The porcelain sleeve state analysis step includes a morphological analysis strategy, which includes extracting several feature points of the porcelain sleeve from a three-dimensional point cloud model, wherein the feature points include a porcelain sleeve boundary point set, connecting each boundary point in the boundary point set to perform straight line fitting, and judging whether the fitted straight line coincides with a vertical line constructed by any boundary point in the boundary point set, and determining whether the porcelain sleeve is in an upright state or an inclined state based on the judgment result.

7. The method for identifying substation equipment data defects according to claim 5, characterized in that: The porcelain sleeve state analysis step includes a flight analysis strategy, which includes fitting the porcelain sleeve point cloud in the three-dimensional point cloud model, and synchronously segmenting the obstacle point cloud around the porcelain sleeve point cloud, calculating the minimum distance between any point in the porcelain sleeve point cloud and any point in the obstacle point cloud, and then comparing the minimum distance with the preset UAV flight safety information to obtain whether the flight environment allows circular flight or circular flight obstacles.

8. The method for identifying substation equipment data defects according to claim 1, characterized in that: The temperature defect analysis step includes a model building strategy, which includes training, testing and verifying the preset substation defect recognition initial model with historical thermal imaging images to obtain a substation defect recognition model, and then inputting the spliced ​​image into the substation defect recognition model to obtain the porcelain sleeve abnormality and abnormal area based on the temperature difference analysis.

9. A substation equipment data defect identification system, characterized by: include: The porcelain sleeve state analysis module uses the camera onboard the drone to scan the target porcelain sleeve to generate a three-dimensional point cloud model, and determines the porcelain sleeve state and flight environment based on the three-dimensional point cloud model. The porcelain sleeve state includes upright or tilted, and the flight environment includes whether circumferential flight is allowed or circumferential flight is obstacle-free. a shooting angle coarse adjustment module, which selects whether to trigger the porcelain sleeve angle analysis condition according to the porcelain sleeve state; if so, calculates the inclination angle of the porcelain sleeve according to the porcelain sleeve spatial coordinate information in the three-dimensional point cloud model, and adjusts the shooting angle of the thermal imager carried by the drone based on the inclination angle; A flight path planning module predetermines a number of shooting points on the periphery of the porcelain sleeve based on the three-dimensional point cloud model, and then selects whether to trigger shooting point adjustment conditions based on the flight environment. If so, it marks the reserved points among the shooting points according to the preset UAV flight safety information, and plans additional points based on the angular interval constraints between two adjacent shooting points and the distance constraints between the shooting points and the porcelain sleeve, and plans a circular flight path based on the reserved points and the additional points; An image acquisition and stitching module acquires thermal images captured by the thermal imager under the circular flight path after adjusting the shooting angle, and stitches them together to obtain a stitched image; A temperature defect analysis module, which determines the abnormal condition and abnormal area of ​​the porcelain sleeve according to the temperature difference in the splicing diagram; It also includes a shooting angle fine-tuning module, which identifies the pixel coordinates of the same calibration point based on any two thermal images taken at different shooting points, and converts the pixel coordinates into image physical coordinates, and then searches for the theoretical physical coordinates of the calibration point in the three-dimensional point cloud model, compares the image physical coordinates with the theoretical physical coordinates, and if there is a deviation, calculates the angle correction amount through the image physical coordinates and the theoretical physical coordinates, and adjusts the shooting angle of the thermal imager carried by the drone with the angle correction amount.

Images