Power distribution network fault positioning method and system based on unmanned aerial vehicle inspection

Abstract

The invention discloses a power distribution network fault positioning method and system based on unmanned aerial vehicle inspection, and relates to the technical field of power system fault positioning, and the method comprises the steps: controlling an unmanned aerial vehicle to synchronously collect infrared temperature, electrical measurement voltage, pose and depth data; carrying out radiation correction on the infrared frame to obtain a temperature rise diagram, and extracting an umbrella skirt root resident annulus based on an axial temperature rise difference; calculating electrical measurement signal energy to determine an electrical activity peak frame, and locking a fault orientation point on the annulus in combination with the depth data; and calculating three-dimensional coordinates of a fault point according to the pose and distance measurement, and projecting the three-dimensional coordinates to a line broken line model to determine a tower section to which the three-dimensional coordinates belong. According to the method, the accuracy of power distribution network fault positioning is effectively improved through geometric anchoring of the resident ring belt and electrical logging energy orientation.

Description

Technical Field

[0001] This invention relates to the field of power system fault location technology, and in particular to a method and system for fault location in distribution networks based on unmanned aerial vehicle (UAV) inspection. Background Technology

[0002] During long-term operation, the external insulation components of power distribution lines are susceptible to leakage current or partial discharge due to environmental factors such as moisture and dirt. This is particularly true at the base of the insulator skirts, where dry arcing can easily form, manifesting as localized high-temperature anomalies. Currently, using drones equipped with infrared imaging components for hovering or circumferential inspections of power distribution lines has become an important maintenance method, detecting these thermal faults by collecting infrared apparent temperature data. However, insulators are typically rotationally symmetric structures, and the dry arcing hotspots at the base of their skirts exhibit a directional distribution along the circumference. During drone observation, the apparent location of the hotspots observed in the infrared image shifts geometrically with changes in the imaging angle and the projection relationship of the target geometry, making it difficult to determine the precise location of the actual fault point in three-dimensional space.

[0003] Current technologies for processing such infrared inspection data typically extract the point of highest temperature rise in the infrared image as the fault point, or perform simple spatial mapping based solely on two-dimensional image coordinates. This approach often ignores the visual projection error of the hot spot onto a three-dimensional surface, leading to unstable location results. For example, when calculating the fault location based solely on an infrared hot spot from a single viewpoint, visual bias caused by the viewpoint may misjudge the fault point to the wrong insulator location or an adjacent tower section. Furthermore, relying solely on thermal imaging data makes it difficult to accurately capture the instantaneous peak of arc activity, and the lack of effective correlation between electrical measurement signals and spatial geometric information results in a systematic deviation between the final output three-dimensional fault coordinates and the actual discharge location, failing to meet the needs of precise emergency repairs and refined maintenance. Summary of the Invention

[0004] The purpose of this invention is to address the shortcomings of existing technologies that rely solely on infrared hotspot image coordinates for location, which are susceptible to geometric drift due to the influence of perspective projection, and lack guidance from electrical measurement signals, resulting in low accuracy of three-dimensional inverse calculation of fault points and difficulty in accurately pinpointing specific tower sections. Therefore, this invention proposes a method and system for fault location in power distribution networks based on unmanned aerial vehicle (UAV) inspection.

[0005] To address the problems existing in the prior art, the present invention adopts the following technical solution: A method for fault location in power distribution networks based on unmanned aerial vehicle (UAV) inspection includes: S1. Control the drone to hover and inspect the insulator targets of the power distribution network, and simultaneously collect infrared apparent temperature frame sequences, electrical voltage sequences, camera pose data and depth ranging data. S2. Perform radiometric correction and background removal on the infrared apparent temperature frame sequence to obtain the corresponding temperature rise map; S3. In the temperature rise diagram, extract the stationary ring at the root of the insulator skirt and generate a ring sampling point set on the outline of the stationary ring; S4. Calculate the electrical measurement energy of the electrical measurement voltage sequence, determine the peak frame corresponding to the time window when the electrical measurement energy reaches its peak, and combine the depth ranging data to lock the fault orientation point in the ring sampling point cluster; S5. Determine the three-dimensional spatial coordinates of the fault point based on the image pixel coordinates of the fault orientation point, the camera pose data corresponding to the peak frame, and the depth ranging data. S6, determine the section to which the fault belongs based on the three-dimensional spatial coordinates of the fault point.

[0006] Preferably, the infrared apparent temperature frame sequence is radiometrically corrected and background is removed to obtain the corresponding temperature rise map, including: Obtain the target emissivity of the insulator material and the reflected equivalent temperature of the environment; According to the law of thermal radiation, the reflected component determined by the reflected equivalent temperature is subtracted from the radiative energy of the infrared apparent temperature frame, and normalized based on the target emissivity to obtain the corrected temperature. The average temperature within the neighborhood of the corrected temperature is used as the background temperature, and the difference between the corrected temperature and the background temperature is calculated to obtain the temperature rise map.

[0007] Preferably, in the temperature rise map, the stationary loop at the root of the insulator skirt is extracted, and a set of loop sampling points is generated on the outline of the stationary loop, including: Set a pixel displacement of one to three pixels along the axial direction of the insulator; The temperature rise graph is used to traverse different candidate axial positions and generate corresponding candidate annular contours. For each candidate annular contour, the average value of the absolute difference in temperature rise between upstream and downstream pixels at the pixel displacement is calculated as the gradient score of the candidate axial position. The candidate axial position with the largest gradient score is selected as the position of the stationary ring, and a preset number of pixels are selected on the contour of the stationary ring at equal arc length intervals to form the ring sampling point set.

[0008] Preferably, calculating the electrical energy of the electrical measurement voltage sequence includes: The electrical voltage sequence is segmented based on the frame time window corresponding to the infrared apparent temperature frame. Statistical analysis is performed on the electrical measurement voltage sequence within each time window to calculate the noise mean and noise standard deviation; The noise mean plus a preset multiple of the noise standard deviation is used as the adaptive threshold; Identify valid signal segments in the electrical voltage sequence whose absolute values ​​exceed an adaptive threshold; Obtain the equivalent resistance of the receiver and calculate the electrical energy based on the voltage within the effective signal segment and the equivalent resistance of the receiver.

[0009] Preferably, determining the peak frame corresponding to the time window when the electrical measurement energy reaches its peak, and combining this with depth ranging data to pinpoint the fault location point within the ring-zone sampling points, includes: Compare the electrical energy calculated from all frame time windows, and determine the infrared apparent temperature frame where the frame time window corresponding to the maximum electrical energy value is located as the peak frame. Obtain the depth ranging data corresponding to the peak frame; Compare the depth ranging data corresponding to each sampling point in the ring sampling point set, and take the sampling point with the smallest depth ranging data as the forward point; The forward point is determined as the fault orientation point locked in this inspection.

[0010] Preferably, the three-dimensional spatial coordinates of the fault point are determined based on the image pixel coordinates of the fault orientation point, the camera pose data corresponding to the peak frame, and the depth ranging data, including: Based on the camera intrinsic parameter matrix, the pixel coordinates of the fault orientation point are converted into a normalized line-of-sight vector in the camera coordinate system. The normalized line-of-sight vector is transformed to the world coordinate system based on the camera rotation matrix and translation vector corresponding to the peak frame to obtain the world direction vector. Starting from the camera optical center position of the peak frame, the vector extends along the world direction, with the extension length being the depth ranging value corresponding to the fault orientation point. The resulting endpoint position is used as the three-dimensional spatial coordinates of the fault point.

[0011] Preferably, determining the section to which the fault belongs based on the three-dimensional spatial coordinates of the fault point includes: The three-dimensional spatial coordinates of the fault point are projected onto the polyline data of the distribution network line, which consists of line segments connecting the towers; For each line segment, calculate the vertical projection of the three-dimensional spatial coordinates of the fault point onto the straight line containing the line segment; Determine whether the vertical projection point is located between the two ends of the line segment. If it is located between the two ends of the line segment, then take the vertical projection point as the corrected projection point; otherwise, take the line segment endpoint closest to the fault point as the corrected projection point. Calculate the Euclidean distance between the fault point and the corrected projection point; Compare the Euclidean distances of all line segments and identify the line segment with the smallest Euclidean distance as the faulty segment.

[0012] To address the aforementioned problems, the present invention also provides a power distribution network fault location system based on unmanned aerial vehicle (UAV) inspection, the system comprising: The data acquisition module is used to control the UAV to hover and inspect the insulator targets of the power distribution network, and simultaneously collect infrared apparent temperature frame sequences, electrical voltage sequences, camera pose data and depth ranging data. The radiation correction module is used to perform radiation correction on the infrared apparent temperature frame sequence and remove the background to obtain the corresponding temperature rise map. The ring extraction module is used to extract the stationary ring at the root of the insulator skirt in the temperature rise map and generate a ring sampling point set on the outline of the stationary ring; The electrical orientation module is used to calculate the electrical energy of the electrical voltage sequence, determine the peak frame corresponding to the time window when the electrical energy reaches its peak, and lock the fault orientation point in the ring sampling points by combining the depth ranging data. The three-dimensional positioning module is used to determine the three-dimensional spatial coordinates of the fault point based on the image pixel coordinates of the fault orientation point, the camera pose data corresponding to the peak frame, and the depth ranging data. The section point module is used to determine the section to which the fault belongs based on the three-dimensional spatial coordinates of the fault point.

[0013] Compared with the prior art, the beneficial effects of the present invention are: 1. This invention extracts the stationary ring at the root of the insulator skirt by utilizing the axial temperature rise difference feature in the temperature rise map, and constructs a geometric anchor point that remains stable with the change of viewing angle. This effectively overcomes the problem of hot spot geometric drift caused by the rotational symmetry structure of the insulator and the difference in projection of the observation viewing angle. By using the ring as a whole instead of a single hot spot as the positioning basis, it avoids the positioning instability and misjudgment caused by relying solely on the infrared apparent maximum point, and significantly improves the robustness of fault extraction under complex inspection perspectives.

[0014] 2. This invention introduces electrical measurement voltage sequences and calculates electrical measurement energy to lock the peak frame of electrical activity. Combined with depth ranging data, it determines the forward point on the ring and performs three-dimensional inverse calculation and line polygonal projection, realizing a precise mapping from two-dimensional image features to three-dimensional spatial physical landing point. It corrects the positioning basis from a single thermal characterization to a spatiotemporal consistent characterization of thermal and electrical data, eliminating the systematic deviation between pure image positioning and the actual discharge location, ensuring that the fault point can be accurately located to the specific tower section, and providing accurate spatial location basis for the refined operation and maintenance and emergency repair of the distribution network. Attached Figure Description

[0015] The accompanying drawings, which are included to provide a further understanding of the invention and form part of this application, illustrate exemplary embodiments of the invention and, together with their description, serve to explain the invention and do not constitute an undue limitation thereof. In the drawings: Figure 1 This is a flowchart illustrating a method for locating power distribution network faults based on unmanned aerial vehicle (UAV) inspection, as provided in an embodiment of the present invention. Figure 2 This is a functional block diagram of a power distribution network fault location system based on unmanned aerial vehicle (UAV) inspection, provided as an embodiment of the present invention. Detailed Implementation

[0016] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0017] Example: This example provides a method for fault location in power distribution networks based on unmanned aerial vehicle (UAV) inspection. See [link / reference]. Figure 1 Specifically, including: S1. Control the drone to hover and inspect the insulator targets of the power distribution network, and simultaneously collect infrared apparent temperature frame sequences, electrical voltage sequences, camera pose data and depth ranging data. In an embodiment of the present invention, a drone is controlled to hover and inspect insulator targets in a power distribution network, simultaneously collecting infrared apparent temperature frame sequences, electrical voltage sequences, camera pose data, and depth ranging data, including: Control the drone to hover and inspect the insulators of the power distribution network, and simultaneously collect infrared apparent temperature frame sequences, electrical voltage sequences, camera pose data and depth ranging data; Specifically, the UAV, carrying infrared imaging components, electrical sensor components, attitude and positioning components, and depth ranging components, takes off from the takeoff and landing point. The flight control system plans and executes the flight path to the target tower area. During flight, the positioning module corrects the flight trajectory in real time to ensure flight accuracy. After arriving at the preset airspace of the target tower, the UAV uses its IMU component to sense its own attitude and adjust its flight attitude to hover. The hovering position is set 3-5 meters to the side of the target insulator string. This distance balances the clarity of infrared imaging with the safety of UAV flight, avoiding interference from the power grid's electromagnetic environment or the risk of collision if the flight is too close, while ensuring the base area of ​​the insulator skirts is protected. The entire area is within the infrared imaging field of view. After hovering and stabilizing (attitude fluctuation ≤ 0.5°), the UAV gimbal is adjusted so that the optical axis of the infrared imaging component's lens is aligned with the root area of ​​the insulator skirt. The gimbal's horizontal azimuth adjustment range is ±15° relative to the center of the insulator string, and the pitch adjustment range is -10° to 20° to ensure that the stationary ring at the root of the skirt is completely presented in the infrared field of view and is located in the center of the field of view, improving the accuracy of subsequent ring extraction. The data synchronization acquisition trigger mechanism is activated, using the frame exposure signal of the infrared imaging component as the synchronization reference, triggering the synchronous operation of the electrical measurement sensor component, pose and positioning component, and depth ranging component, among which the infrared imaging component is preferred. Infrared apparent temperature frame sequences are acquired at a frame rate of 25-30fps. This frame rate range effectively captures the instantaneous temperature changes of the dry-band arc. The infrared image resolution is set to 1280×1024 to ensure clear imaging of fine hot spots at the base of the umbrella skirt. The electrical measurement sensor assembly synchronously acquires electrical measurement voltage sequences, with a sampling rate preferably between 100MHz and 1GHz. This sampling rate range is determined based on the frequency characteristics of the partial discharge / arc electrical activity signals, and can fully cover the electrical signal characteristics of the high-frequency / UHF bands. The equivalent resistance of the receiver is selected as a 50Ω standard impedance to match the high-frequency signal transmission link, reduce signal attenuation, and ensure the integrity and accuracy of the electrical signal acquisition. Pose and positioning... The bit component synchronously outputs camera pose data (camera rotation matrix and translation vector) corresponding to the exposure time of each frame of infrared image. The positioning module ensures that the positioning accuracy of the pose data reaches the centimeter level, and the sampling rate is consistent with the frame rate of the infrared imaging component to achieve accurate correlation between each frame of infrared image and pose data. The depth ranging component can use a ToF depth camera to synchronously output depth ranging data that matches the resolution of the infrared image. The depth data sampling rate is synchronized with the infrared imaging frame rate. The acquired infrared apparent temperature frame sequence, electrical voltage sequence, camera pose data and depth ranging data are associated and stored according to a unified timestamp to ensure the frame-level matching accuracy of subsequent different types of data.

[0018] S2. Perform radiometric correction and background removal on the infrared apparent temperature frame sequence to obtain the corresponding temperature rise map; In an embodiment of the present invention, radiometric correction and background removal are performed on the infrared apparent temperature frame sequence to obtain the corresponding temperature rise map, including: Obtain the target emissivity of the insulator material and the reflected equivalent temperature of the environment; Specifically, insulator material refers to the dielectric material that constitutes the body and skirts of the insulator in the power distribution network. Under the action of an electric field, it provides electrical isolation and determines the energy level of external radiation through its surface thermal radiation capacity and thermal conductivity during heat exchange. Target emissivity refers to the proportionality coefficient of the material surface's ability to emit thermal radiation relative to an ideal blackbody at the same temperature. Its value reflects the efficiency of the surface in releasing internal energy in the form of infrared radiation. Reflection equivalent temperature refers to the infrared reflection effect produced by the surrounding environment on the target surface using an equivalent temperature, thereby representing the radiation input brought by environmental reflection in the form of radiation intensity corresponding to the temperature.

[0019] Specifically, the insulator material type and surface condition information are read and matched in the emissivity parameter table to obtain the target emissivity. The target emissivity reflects the efficiency of the insulator surface in emitting internal energy outward in the form of infrared radiation and is used to correct for differences in radiation intensity between different materials at the same temperature. The target emissivity is preferably set to 0.92 on the surface of the silicone rubber jacket or skirt of the composite insulator, and preferably set to 0.95 on the surface of the porcelain or glass insulator. The above values ​​are based on the fact that the emissivity of common external insulation materials in the 8 to 14 micrometer thermal imaging band is usually in the range of 0.9 to 0.98, and the on-site positioning task aims to reduce the temperature back-calculation deviation caused by emissivity error. The reflection equivalent temperature is used to characterize the reflected radiation input generated by the surrounding environment on the target surface. When acquiring it, after the UAV hovers, the infrared imager is controlled to be aimed at the background area close to the insulator and without local heat sources to collect multiple frames of apparent temperature and take the statistical median as the reflection equivalent temperature. It is preferred to use the median of 30 frames to suppress occasional reflected highlights and noise. The temperature corresponding to the median value can represent the stable level of environmental radiation intensity.

[0020] According to the law of thermal radiation, the reflected component determined by the reflected equivalent temperature is subtracted from the radiative energy of the infrared apparent temperature frame, and normalized based on the target emissivity to obtain the corrected temperature. Specifically, radiant energy refers to the infrared radiation power or its equivalent that reaches the infrared detector per unit time; the reflection component refers to the energy contribution that is reflected from the target surface by ambient radiation and enters the detector; and the corrected temperature refers to the true temperature of the target surface obtained by back-calculating from the thermal radiation relationship after deducting the environmental reflection contribution and correcting it according to the target emissivity.

[0021] Specifically, in the radiation correction stage, after the infrared imager converts the detected radiation intensity into apparent temperature, it still includes the superposition components of the target's own radiation and environmental reflection. To separate the two, the temperature needs to be mapped to radiation and reflection subtraction and emissivity correction need to be performed. The corrected temperature is calculated using the following formula: In the formula, For pixel position The corrected temperature at that location The infrared apparent temperature at the same pixel. For target emissivity, The reflection equivalent temperature is used; the fourth power relationship is adopted because the intensity of thermal radiation is proportional to the fourth power of the absolute temperature. The radiation corresponding to the infrared apparent temperature can be regarded as the superposition of the target radiation term and the reflection term. The reflection term is formed by the reflection of ambient radiation by the target surface, and its ratio corresponds to one minus the target emissivity. Therefore, multiplying the fourth power of the reflection equivalent temperature by one minus the target emissivity yields the reflection component, which is then subtracted from the apparent radiation. The remaining part is then divided by the target emissivity to normalize the radiation to the emission level under the target's own radiation capability. Finally, the correction temperature is obtained by back-calculation using the fourth root.

[0022] The average temperature within the neighborhood of the corrected temperature is used as the background temperature, and the difference between the corrected temperature and the background temperature is calculated to obtain the temperature rise map. Specifically, the neighborhood background region refers to the region that is in the same observation frame as the target and is spatially adjacent to the target but does not contain local abnormal hot spots. It is used to provide a temperature benchmark under the same environmental conditions. The temperature rise map refers to the distribution result formed in the pixel space by the difference, thereby expressing the degree of increase of local temperature relative to the background in image form and using it for subsequent ring extraction and anomaly localization.

[0023] Specifically, in the temperature correction frame, a background ring or rectangular region is constructed centered on the projection area at the root of the insulator skirt. The region size is preferably extended by 20 to 40 pixels in both the horizontal and vertical directions, avoiding the pixel band where the dwelling ring is located. This ensures that the background region is in the same atmospheric path and similar reflection conditions as the target, but does not contain local abnormal hot spots. The background temperature is taken as the arithmetic mean of the correction temperatures of all pixels in the background region to reduce the influence of thermal image noise and individual pixel anomalies. The temperature rise map is generated by calculating the difference between the temperature of each pixel in the temperature correction frame and the background temperature. The difference directly represents the temperature increment of the pixel relative to the reference thermal state of the same frame and environment. This highlights the local heating caused by dry-line arcing or leakage current from the overall temperature field and provides a stable input for subsequent dwelling ring extraction.

[0024] S3. In the temperature rise diagram, extract the stationary ring at the root of the insulator skirt and generate a ring sampling point set on the outline of the stationary ring; In an embodiment of the present invention, in the temperature rise map, the stationary loop at the root of the insulator skirt is extracted, and a set of loop sampling points is generated on the outline of the stationary loop, including: Set a pixel displacement of one to three pixels along the axial direction of the insulator; Specifically, after obtaining the temperature rise map, the axial direction of the insulator is first determined in the temperature rise map. This is done by extracting the slender contour of the insulator body from the temperature rise map and fitting the contour with the principal direction to obtain the axial unit direction vector. This ensures that the forward and reverse movement of any pixel in the temperature rise map along this direction corresponds to the upstream and downstream of the insulator's axial direction. The pixel displacement is set using the discrete step length along the axial unit direction vector. The displacement range is limited to one to three pixels to take into account both types of constraints. A one-pixel displacement can maintain sensitivity to sudden changes in local temperature rise at the root of the shed, while a three-pixel displacement can provide a more stable differential component when there is infrared noise and slight jitter. The preferred value is two pixels to achieve a balance between resolution and noise resistance. This pixel displacement is recorded in the ring search parameters of this frame.

[0025] The temperature rise graph is used to traverse different candidate axial positions and generate corresponding candidate annular contours. Specifically, the traversal of candidate axial positions uses the axial projection range of the insulator in the temperature rise map as the search interval. This interval is obtained by determining the minimum and maximum coordinates of the insulator's outer contour in the axial direction, and scanning is performed within this interval at a preset axial step size. The axial step size is preferably one to two pixels to ensure that it does not cross the narrow band area at the root of the skirt and miss detection. When scanning each candidate axial position, the generation of the candidate ring contour is aimed at the circumferential boundary at that axial position. Specifically, gradient enhancement is performed on the temperature rise map to highlight the ring response of the skirt edge, and a line is selected near the candidate axial position along the axial direction. A near-vertical narrow strip is used as the search band, and pixel trajectories with local maxima of gradient magnitude are searched along the circumference within the search band to form closed or near-closed contour lines, thereby obtaining the candidate ring contour corresponding to the candidate axial position. When the insulator is slightly tilted in the field of view, the candidate ring contour is synchronously translated and updated along the axial unit direction vector as the axial position advances, so that the candidate ring contour generated at different candidate axial positions remains geometrically consistent with the circumference of the insulator skirt, providing repeatable contour input for subsequent calculation of axial temperature rise differential features on the candidate ring contour and determination of the stationary ring position.

[0026] For each candidate annular contour, the average value of the absolute difference in temperature rise between upstream and downstream pixels at the pixel displacement is calculated as the gradient score of the candidate axial position. The candidate axial position with the largest gradient score is selected as the position of the stationary ring, and a preset number of pixels are selected on the contour of the stationary ring at equal arc length intervals to form the ring sampling point set; Specifically, the stationary ring zone refers to a target strip-shaped area located near the root of the insulator skirt and distributed in a ring shape along the circumference of the skirt in the temperature rise map. This strip-shaped area is selected by statistically analyzing the temperature rise difference of the ring zone profile at different candidate axial positions. The selection criterion is that the average absolute temperature rise difference between the pixels on the ring zone profile at that axial position and the pixels upstream and downstream along the axial direction reaches its maximum. This characterizes the ring zone position where the axial temperature rise change is most significant and more consistent with the dry strip boundary and local electrothermal effect at the skirt root. It serves as a reference point during inspection and subsequent orientation processes. Stable geometric anchoring bands are used; the ring sampling point set refers to a set of discrete pixels formed by selecting a preset number of points from the contour pixels along the contour of the stationary ring at equal arc length intervals after the stationary ring is determined. Each sampling point has corresponding image pixel coordinates and can establish a one-to-one correspondence with depth ranging data in subsequent processing to select forward points. At the same time, it can also serve as an index carrier for binding electrical measurement energy with spatial location, so that the fault orientation point corresponding to the peak electrical measurement energy can be limited to the stationary ring, thereby improving the spatial consistency of fault location.

[0027] Specifically, the candidate ring contour corresponding to a certain candidate axial position is read in the temperature rise map. The candidate ring contour is represented as a sequence of pixels arranged in contour order. For each contour pixel in the sequence, its upstream and downstream pixels along the insulator axial direction are determined. The upstream and downstream pixels are obtained by moving the pixel displacement in the forward and reverse directions along the insulator axial direction, respectively. The pixel displacement is one to three pixels, preferably two pixels, to suppress differential fluctuations caused by infrared noise and slight jitter while retaining the sensitivity of local temperature rise abrupt changes at the umbrella skirt root. For each contour pixel, the temperature rise value at the upstream pixel and the temperature rise value at the downstream pixel are read respectively. The difference between the two is calculated and the absolute value is taken to characterize the temperature rise change amplitude of the contour point in the axial upstream and downstream directions. Then, the absolute difference of the temperature rise of all contour pixels on the candidate ring contour is statistically averaged to obtain the gradient score of the candidate axial position. The gradient score numerically reflects the overall intensity of the temperature rise abrupt change along the axial direction at the candidate ring, thus more closely approximating the temperature rise transition caused by local electrothermal effect near the dry strip boundary at the umbrella skirt root.

[0028] Specifically, the gradient scores of each candidate axial position obtained through traversal are compared. The candidate axial position with the largest gradient score is determined as the location of the stationary ring, and the corresponding candidate ring contour is output as the stationary ring contour. This is because the largest gradient score means that the axial temperature rise change at that ring is the most significant and can more stably characterize the boundary zone features of the stationary ring of the directional dry zone hot spot at the root of the umbrella skirt. When generating the ring sampling point set, the stationary ring contour is parameterized by arc length parameterization. The pixel distances between adjacent points in the contour pixel sequence are accumulated to form the cumulative arc length. The total arc length of the contour is used to set sampling positions on the cumulative arc length axis at equal arc length intervals and backtrack to the corresponding contour pixel points as sampling points, thereby ensuring that the sampling points are evenly distributed in the circumferential direction of the contour. The preset number is preferably 24 to 72, and more preferably 36. This value range can control the amount of calculation while ensuring circumferential resolution coverage and adapt to the circumferential pixel length of common insulator skirts in the image. Finally, the set of pixel coordinates of the selected sampling points is output as a ring sampling point set, which can be used to lock the fault orientation point and complete the three-dimensional positioning by combining depth ranging data.

[0029] S4. Calculate the electrical measurement energy of the electrical measurement voltage sequence, determine the peak frame corresponding to the time window when the electrical measurement energy reaches its peak, and combine the depth ranging data to lock the fault orientation point in the ring sampling point cluster; In an embodiment of the present invention, the electrical energy of the electrical measurement voltage sequence is calculated, the peak frame corresponding to the time window when the electrical energy reaches its peak value is determined, and the fault orientation point is located in the ring sampling point set by combining depth ranging data, including: The electrical voltage sequence is segmented based on the frame time window corresponding to the infrared apparent temperature frame. Statistical analysis is performed on the electrical measurement voltage sequence within each time window to calculate the noise mean and noise standard deviation; The noise mean plus a preset multiple of the noise standard deviation is used as the adaptive threshold; Specifically, during the synchronous acquisition phase, the infrared apparent temperature frame sequence and the electrical voltage sequence are written with the same source timestamp. The exposure center time of each infrared apparent temperature frame is recorded as the frame time, and the difference between the frame times of two adjacent frames is used as the frame period. The frame time window corresponding to each infrared apparent temperature frame is then set as a time interval centered on the frame time and with the frame period as its width, ensuring that the frame time window covers the observation time range represented by that frame. After the electrical voltage sequence is sorted and stored according to the sampling time, it is truncated using the start and end times of each frame time window as boundaries. This truncated sequence obtains the set of electrical voltage sampling points falling within the frame time window, thus completing the segmentation of the electrical voltage sequence into frame-by-frame segments and ensuring that each segment of the electrical voltage sequence corresponds one-to-one with an infrared apparent temperature frame. When calculating the noise mean and noise standard deviation, the electrical voltage sampling points within the frame time window are sorted by absolute value, and a subset of sampling points with smaller amplitudes are selected as noise samples to avoid localized noise. Electric or dry arc pulses cause a high bias in noise statistics. The noise samples are preferably 70% of the sampling points with the smallest amplitude within the frame time window. This proportion can effectively eliminate the influence of pulse peaks while retaining a sufficient number of samples. The noise mean is the arithmetic mean of the noise samples, and the noise standard deviation is the square root of the dispersion of the noise samples relative to the noise mean, thereby obtaining the baseline level and fluctuation amplitude of the electrical measurement voltage within the frame time window. The adaptive threshold is obtained by linearly superimposing the noise mean and the noise standard deviation. The preset multiple is used to control the margin of the threshold relative to the noise fluctuation. The preset multiple is preferably four, and can be in the range of three to five. When it is four, it can significantly reduce the probability of noise false triggering under common approximate Gaussian noise conditions while still maintaining the ability to detect medium amplitude pulses. This allows the threshold to be automatically adjusted according to changes in the distance between the UAV and the target, changes in attitude, and changes in electromagnetic background, thereby maintaining stable detection of abnormal pulses and suppressing noise false triggering under different noise conditions.

[0030] Identify valid signal segments in the electrical voltage sequence whose absolute values ​​exceed an adaptive threshold; Specifically, an effective signal segment refers to a continuous time interval or a set of discrete sampling points selected from the electrical measurement voltage sequence within a certain frame time window. This interval or set satisfies the criterion that the absolute value of the electrical measurement voltage sequence exceeds the adaptive threshold, thus representing the electrical activity response portion that is significantly prominent relative to the background noise. The effective signal segment is used to carry voltage change information caused by partial discharge or dry-circuit arcing.

[0031] Specifically, within each frame time window, the electrical measurement voltage sequence segment corresponding to that time window and the corresponding adaptive threshold are read. The absolute value of each sampling point of the electrical measurement voltage sequence segment is calculated and compared with the adaptive threshold. Sampling points with an absolute value not less than the adaptive threshold are marked as candidate valid points. To avoid isolated false triggers caused by occasional noise spikes, a continuity constraint is applied to the candidate valid points. The time interval between adjacent candidate valid points is not greater than one sampling period and is merged into a continuous interval. Intervals with a length less than the minimum duration are removed. The minimum duration is set to three to five sampling points, preferably five sampling points. This value can suppress single-point spike noise without significantly losing the true pulse detection rate. At the same time, considering that partial discharge pulses may appear in clusters, bridging and merging are performed when the interval between two adjacent continuous intervals is less than the merging interval. The merging interval is set to two to ten sampling periods, preferably five sampling periods. This value can unify pulse clusters caused by the same electrical activity event into the same valid signal segment, ultimately obtaining one or more valid signal segments.

[0032] Obtain the equivalent resistance of the receiver and calculate the electrical energy based on the voltage within the effective signal segment and the equivalent resistance of the receiver. Specifically, the equivalent resistance of the receiver is given by the input impedance of the electrical sensor receiving link and is fixed in the system parameters. A commonly used value is fifty ohms to match coaxial transmission and standard RF measurement links, thus ensuring comparability of the energy calculation of the electrical voltage sequence across this equivalent resistance. The electrical energy is statistically analyzed within each frame time window according to the effective signal segment. It is obtained by squaring the voltage, normalizing it according to the receiver's equivalent resistance, and then accumulating or integrating it within the effective signal segment. The specific calculation is as follows: In the formula, The electrical energy measured within the frame time window. It is the set of sampling times contained in all valid signal segments within the frame time window. Sampling time The measured voltage value at the location, The equivalent resistance of the receiving end. The sampling period for the electrical measurement voltage sequence is determined by the square of the voltage. This is because the instantaneous power on a resistive load is proportional to the square of the voltage. By summing the power over time, we can obtain the energy characterization. Dividing by the equivalent resistance at the receiving end is used to convert the voltage amplitude under different impedance conditions into a comparable power scale. This ensures that the stronger the electrical activity, the longer the effective signal segment, or the higher the amplitude within the same time window, the greater the calculated electrical measurement energy. This provides a stable quantization basis for subsequently determining the peak frame where the electrical measurement energy reaches its peak value.

[0033] Compare the electrical energy calculated from all frame time windows, and determine the infrared apparent temperature frame where the frame time window corresponding to the maximum electrical energy value is located as the peak frame. Obtain the depth ranging data corresponding to the peak frame; Compare the depth ranging data corresponding to each sampling point in the ring sampling point set, and take the sampling point with the smallest depth ranging data as the forward point; The forward point is determined as the fault orientation point locked in this inspection; Specifically, the peak frame refers to the infrared apparent temperature frame corresponding to the frame time window corresponding to the infrared apparent temperature frame during hovering inspection. It is the frame where the measured electrical energy reaches its maximum value after extracting effective signal segments and performing energy-based statistical analysis of the electrical voltage sequence within that time window. Objectively, this frame reflects the strongest external insulation electrical activity within the observation time range covered by this frame. This means that phenomena such as sudden leakage current changes, partial discharges, or dry-charge arcing radiate the most significant electromagnetic disturbances outward within this time range. Therefore, this frame is more likely to be under the same excitation conditions as the actual location of abnormal heating. Suitable as the master reference frame for subsequent geometric calculations, the forward point refers to a sampling point located in the ring sampling point set. This sampling point has the smallest ranging value under the depth ranging data corresponding to the peak frame. Its objective meaning is that the sampling point is located in the front area of ​​the umbrella skirt root stationary ring closest to the camera optical center under the current camera view. This is equivalent to the surface normal being closer to the observation line of sight and the occlusion and oblique viewing errors being smaller. Therefore, using it as the fault orientation point can make the fault landing point have a clearer geometric orientation in space and reduce the positioning drift caused by the ring projection and viewpoint changes.

[0034] Specifically, during a hover inspection, a one-to-one correspondence is established between each frame of infrared apparent temperature and its corresponding frame time window electrical energy measurement, forming an electrical energy sequence sorted by frame index. To avoid misjudging peak frames due to a single abnormal spike, the electrical energy sequence is smoothed before comparison. The smoothing process uses a moving average of three or five frames, preferably three frames, to suppress occasional fluctuations while maintaining a fast response to sudden increases in actual electrical activity. The peak frame is determined on the smoothed electrical energy sequence by traversing each frame of electrical energy measurement and selecting the frame index corresponding to the maximum value as the peak frame index. This identifies the infrared apparent temperature frame containing the frame time window with the strongest electrical activity as the peak frame, and uses this peak frame index as the reference index for subsequent data alignment. The depth ranging data corresponding to the peak frame is retrieved from the synchronized cache using the peak frame index. The depth ranging data is directly read from the depth ranging data stream. When comparing the depth ranging data corresponding to each sampling point in the ring sampling point set, the pixel coordinates of each sampling point in the ring sampling point set are used to sample the depth ranging data corresponding to the peak frame to obtain the depth ranging value of that sampling point. In order to improve the robustness to ranging noise and local holes, the depth ranging value of each sampling point is corrected by the median filtering method of neighborhood sampling. The neighborhood window is preferably three times three pixels so that the single-point abnormal ranging will not dominate the forward point selection. After the depth ranging values ​​of each sampling point are made comparable, the sampling point with the smallest depth ranging value is selected as the forward point. The smallest depth ranging value means that the sampling point is closest to the camera optical center in the current view and is more likely to correspond to the front side area of ​​the umbrella skirt root stationed on the ring towards the observation direction, so that the fault landing point of the subsequent three-dimensional solution has a clearer geometric orientation.

[0035] S5. Determine the three-dimensional spatial coordinates of the fault point based on the image pixel coordinates of the fault orientation point, the camera pose data corresponding to the peak frame, and the depth ranging data. In embodiments of the present invention, determining the three-dimensional spatial coordinates of the fault point based on the image pixel coordinates of the fault orientation point, the camera pose data corresponding to the peak frame, and the depth ranging data includes: Based on the camera intrinsic parameter matrix, the pixel coordinates of the fault orientation point are converted into a normalized line-of-sight vector in the camera coordinate system. The normalized line-of-sight vector is transformed to the world coordinate system based on the camera rotation matrix and translation vector corresponding to the peak frame to obtain the world direction vector. Specifically, the camera intrinsic parameter matrix refers to the set of mapping relationships composed of the camera's imaging geometric intrinsic parameters. It describes the influence of lens focal length and imaging center position on pixel scale, enabling pixel coordinates in the image to establish a correspondence with the actual observation direction in the camera's field of view. The fault orientation point refers to the pixel point locked on the dwell ring corresponding to the peak frame to indicate the fault location. It represents the target part in the camera's field of view that is most consistent with the peak time of electrical activity. The camera coordinate system refers to a three-dimensional reference coordinate system established with the camera optical center as the origin and the camera optical axis and imaging plane direction as the reference. It is used to express the line-of-sight direction and spatial point position within the camera's own reference system. The normalized line-of-sight direction vector is calculated from the pixel coordinates using the camera intrinsic parameter matrix. The direction of the observed line of sight, after normalizing its length, is a directional quantity that only expresses the direction from the camera's optical center to the corresponding spatial direction of the pixel without including distance scale. The camera rotation matrix is ​​an attitude transformation quantity used to describe the orientation relationship between the camera coordinate system and the world coordinate system. It reflects how the camera's orientation in space rotates relative to the world reference coordinates. The translation vector is a displacement quantity used to describe the spatial position of the camera's optical center in the world coordinate system. It reflects the three-dimensional coordinates of the camera relative to the world reference origin. The world coordinate system is a global reference coordinate system used to uniformly represent the position of the UAV, the position of the tower, and the line polygonal data, so that the measurement results of different sensors and at different times can be aligned under the same spatial reference.

[0036] Specifically, during the calibration phase, the camera intrinsic parameter matrix corresponding to the infrared imager is obtained in advance and written into the device parameter table. The camera intrinsic parameter matrix includes focal length parameters and principal point parameters, and represents the definite mapping relationship between pixel coordinates and camera imaging geometry. It is obtained offline using a calibration board multi-pose imaging method and remains unchanged when the lens focus state is fixed, thereby ensuring the stability of the conversion from pixel coordinates to spatial line of sight. When the fault orientation point of the peak frame is locked, the pixel coordinates of the fault orientation point are read and a homogeneous pixel vector is constructed. By inverting the camera intrinsic parameter matrix, the pixel coordinates are removed from the influence of focal length scaling and principal point offset, resulting in an unnormalized line of sight direction in the camera coordinate system. This vector is then normalized according to its magnitude to obtain a normalized line of sight direction vector that only reflects direction and does not carry scale. The specific calculation process is as follows: In the formula, For the camera intrinsic parameter matrix, and These are the pixel coordinates of the fault orientation point, respectively. This is the normalized viewing direction vector in the camera coordinate system. The numerator represents the viewing direction of back-projecting the pixel coordinates onto the front of the camera imaging plane, and the denominator represents the magnitude of the direction vector. The reason for using the inverse of the camera intrinsic parameter matrix is ​​that the pixel coordinates are two-dimensional observations under the camera imaging model. The intrinsic parameter matrix includes the focal length and principal point position. Inversion can restore the pixel coordinates to the directional scale relationship in the camera coordinate system, thus mapping the two-dimensional pixels to the three-dimensional viewing direction. The reason for using normalization is that the subsequent three-dimensional position calculation needs to combine the direction and depth measurement values. The direction vector should only express the direction and not introduce additional scale.

[0037] After obtaining the normalized viewing direction vector, the camera rotation matrix (describing the rotation relationship between the camera coordinate system and the world coordinate system) and the translation vector (describing the position of the camera optical center in the world coordinate system) are read from the camera pose data corresponding to the peak frame. A rotation transformation is then performed on the normalized viewing direction vector to obtain the world direction vector. The calculation formula is as follows: In the formula, This is the camera rotation matrix corresponding to the peak frame. This is the world direction vector in the world coordinate system. This is the normalized line-of-sight direction vector in the camera coordinate system. The reason for using a rotation matrix transformation is that the normalized line-of-sight direction vector is the direction represented in the camera coordinate system. The world coordinate system is used to unify the representation with the distribution network line polyline data and tower coordinates. By using the camera rotation matrix to rotate the direction from the camera coordinate system to the world coordinate system, the length of the direction vector remains unchanged and only its coordinate expression is changed. This provides a consistent coordinate basis for the subsequent calculation of the three-dimensional spatial coordinates of the fault point by combining the translation vector and depth ranging data.

[0038] Starting from the camera optical center position of the peak frame, extend along the world direction vector. The extension length is the depth ranging value corresponding to the fault orientation point. The resulting endpoint position is used as the three-dimensional spatial coordinates of the fault point. Specifically, after converting the pixel coordinates of the fault orientation point to the world direction vector, the camera optical center position is read from the camera pose data corresponding to the peak frame and used as the starting coordinate in the world coordinate system; the depth ranging value is obtained by sampling the pixel coordinates of the fault orientation point from the depth ranging data corresponding to the peak frame. When the depth ranging data is a depth map, the depth value at the pixel of the fault orientation point is taken and its neighborhood is subjected to median filtering to suppress holes and outliers. The neighborhood window is preferably three times three pixels to reduce ranging noise without significantly smoothing the edges. When the depth ranging data is single-point ranging, the corresponding distance value after registration with the line of sight of the fault orientation point is taken; to avoid ranging abnormalities leading to the endpoint The system then applies a validity constraint to the depth ranging value, limiting it to a feasible range within the common inspection distance between the UAV and insulators. This feasible range is preferably one to fifteen meters to cover the typical close-range inspection range of insulators in power distribution networks and exclude obviously unreasonable ranging outputs. After the depth ranging value falls into this range, the world direction vector is used as the unit direction and multiplied with the depth ranging value to obtain a displacement vector from the camera optical center to the target point. The endpoint position is obtained by adding the starting coordinates to the displacement vector, thus forming the three-dimensional spatial coordinates of the fault point. The three-dimensional coordinates of the endpoint position, together with the peak frame index and the pixel coordinates of the fault orientation point, are recorded as the spatial representation of the fault location result of this inspection.

[0039] S6. Determine the section to which the fault belongs based on the three-dimensional spatial coordinates of the fault point; In an embodiment of the present invention, determining the section to which the fault belongs based on the three-dimensional spatial coordinates of the fault point includes: The three-dimensional spatial coordinates of the fault point are projected onto the polyline data of the distribution network line, which consists of line segments connecting the towers; For each line segment, calculate the vertical projection of the three-dimensional spatial coordinates of the fault point onto the straight line containing the line segment; Specifically, before the inspection task begins, the system reads the polyline data of the distribution network lines from the distribution network ledger or survey results and converts it into geometric objects in the world coordinate system. The polyline data consists of line segments connecting several towers, each segment represented by the coordinates of its endpoints at both ends of the towers. The endpoint coordinates and the three-dimensional spatial coordinates of the fault point use the same coordinate reference and unit to ensure consistency in the dimensions of subsequent distance calculations. After obtaining the three-dimensional spatial coordinates of the fault point, the system traverses each line segment in the polyline data, reads the starting and ending coordinates of the segment, constructs a line segment direction vector, and simultaneously constructs a connection vector pointing from the starting point of the line segment to the fault point. The vertical projection point of the fault point on the line segment is obtained through vector projection. The calculation is as follows: In the formula, The three-dimensional spatial coordinates of the fault point. and These are the starting and ending coordinates of the line segment, respectively. The point is the perpendicular projection of the fault point onto the straight line containing the line segment. For projection coefficients, The length of the line segment direction vector is represented by the projection coefficient. The reason for using the projection coefficient is that the vertical projection point satisfies the condition that the line connecting the projection point to the fault point is orthogonal to the line segment direction vector. Vector projection can directly obtain the magnitude of the component along the line segment direction and thus obtain the three-dimensional position of the projection point. This allows for a unified geometric comparison of the proximity of the fault point to the line among the line segments. To improve computational stability, when the line segment length is too short and may cause the denominator to approach zero, the system sets a minimum effective length threshold for the line segment length. The minimum effective length threshold is preferably 0.5 meters. This is used to exclude numerical instability caused by duplicate points or abnormally short line segments in the ledger, ensuring that each effective line segment can obtain a reliable vertical projection point and providing a basis for subsequent judgments on whether the projection point falls within the line segment range and for calculating the distance from the point to the line segment.

[0040] Determine whether the vertical projection point is located between the two ends of the line segment. If it is located between the two ends of the line segment, then take the vertical projection point as the corrected projection point; otherwise, take the line segment endpoint closest to the fault point as the corrected projection point. Calculate the Euclidean distance between the fault point and the corrected projection point; Compare the Euclidean distances of all line segments and identify the line segment with the smallest Euclidean distance as the faulty segment. Specifically, after obtaining the vertical projection point of the fault point on the straight line of each line segment, the projection coefficient is used to characterize the relative position of the projection point in the direction of the line segment. The system determines whether the vertical projection point is located between the two endpoints of the line segment by checking if the projection coefficient falls within a closed interval of zero to one. If the projection coefficient falls between zero and one, it means the vertical projection point is located within the line segment connecting the start and end points. The system directly uses this vertical projection point as the correction projection point to ensure the shortest distance from the corresponding point to the line segment is calculated. When the projection coefficient is less than zero or greater than one, it means the vertical projection point is located on the extension line of the line segment and exceeds the line segment's range. In this case, the shortest distance from the point to the line segment must appear at one of the endpoints. The Euclidean distances from the fault point to the start point and the fault point to the end point of the line segment are calculated separately, and the endpoint corresponding to the smaller distance is selected as the correction projection point, thus correcting the projection point to the closest point within the line segment's range. After obtaining the correction projection point, the Euclidean distance between the fault point and the correction projection point is calculated as the point-to-line-segment distance for that line segment. The Euclidean distance is obtained by taking the square root of the sum of the squares of the differences in three-dimensional coordinates, which can directly reflect the spatial proximity of the fault point to the geometric location of the line. Considering that there may be local coordinate errors in the distribution network ledger or slight deviations caused by the approximation of the broken lines between towers, in order to avoid the section jump caused by random errors between extremely close line segments, a minimum distance difference margin is introduced as a stability condition when comparing the consistency of the Euclidean distance. The minimum distance difference margin is preferably 0.2 meters. When the difference between the minimum distance and the second smallest distance is less than this margin, the section to which the fault belongs is determined as the line segment that is closer to the fault point in the axial projection, or the determination result of the previous frame is maintained, thereby improving the stability of the section determination. After traversing all line segments and obtaining the corresponding Euclidean distances, the line segment corresponding to the minimum value in all Euclidean distances is selected as the section to which the fault belongs, and the index of the line segment in the broken line data of the distribution network line and the tower identifiers corresponding to its two ends are output, realizing the mapping of the three-dimensional spatial coordinates of the fault point to the landing point of the tower section.

[0041] like Figure 2 The diagram shown is a functional block diagram of a power distribution network fault location system based on UAV inspection, provided in an embodiment of the present invention.

[0042] In this embodiment, the functions of each module / unit are as follows: The data acquisition module is used to control the UAV to hover and inspect the insulator targets of the power distribution network, and simultaneously collect infrared apparent temperature frame sequences, electrical voltage sequences, camera pose data and depth ranging data. The radiation correction module is used to perform radiation correction on the infrared apparent temperature frame sequence and remove the background to obtain the corresponding temperature rise map. The ring extraction module is used to extract the stationary ring at the root of the insulator skirt in the temperature rise map and generate a ring sampling point set on the outline of the stationary ring; The electrical orientation module is used to calculate the electrical energy of the electrical voltage sequence, determine the peak frame corresponding to the time window when the electrical energy reaches its peak, and lock the fault orientation point in the ring sampling points by combining the depth ranging data. The three-dimensional positioning module is used to determine the three-dimensional spatial coordinates of the fault point based on the image pixel coordinates of the fault orientation point, the camera pose data corresponding to the peak frame, and the depth ranging data. The section point module is used to determine the section to which the fault belongs based on the three-dimensional spatial coordinates of the fault point.

[0043] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A power distribution network fault location method based on unmanned aerial vehicle inspection, characterized in that, The method comprises the following steps: S1. Controlling the unmanned aerial vehicle to hover and patrol the insulator target of the power distribution network, and synchronously collecting an infrared apparent temperature frame sequence, an electric measurement voltage sequence, camera pose data, and depth ranging data; S2. Radiometrically correcting the infrared apparent temperature frame sequence and performing background elimination to obtain a corresponding temperature rise map; S3. In the temperature rise map, a standing ring belt at the root of the insulator shed is extracted, and a ring belt sample point set is generated on the contour of the standing ring belt; S4. The electric measurement energy of the electric measurement voltage sequence is calculated, the peak value frame corresponding to the time window in which the electric measurement energy reaches the peak value is determined, and the fault orientation point is locked in the ring belt sample point set in combination with the depth ranging data; S5. The three-dimensional space coordinates of the fault point are determined according to the image pixel coordinates of the fault orientation point, the camera pose data corresponding to the peak value frame, and the depth ranging data; S6. The fault section to which the fault belongs is determined based on the three-dimensional space coordinates of the fault point. 2.The power distribution network fault locating method based on unmanned aerial vehicle inspection of claim 1, wherein, The method for radiometrically correcting the infrared apparent temperature frame sequence and performing background elimination to obtain a corresponding temperature rise map comprises the following steps: The target emissivity of the insulator material and the reflection equivalent temperature of the environment are obtained; According to the law of thermal radiation, the reflection component determined by the reflection equivalent temperature is deducted from the radiation energy of the infrared apparent temperature frame, and the corrected temperature is obtained by normalization based on the target emissivity; The average temperature in the neighborhood background area of the corrected temperature is taken as the background temperature, and the difference between the corrected temperature and the background temperature is calculated to obtain the temperature rise map. 3.The power distribution network fault locating method based on unmanned aerial vehicle inspection of claim 1, wherein, In the temperature rise map, the standing ring belt at the root of the insulator shed is extracted, and a ring belt sample point set is generated on the contour of the standing ring belt, which comprises the following steps: A pixel displacement of one to three pixels is set in the axial direction of the insulator; Different candidate axial positions are traversed in the temperature rise map to generate corresponding candidate ring belt contours; For each candidate ring belt contour, the average value of the absolute difference in temperature rise between the upstream and downstream pixels of the pixel point on the contour at the pixel displacement is calculated as the gradient score of the candidate axial position; The candidate axial position with the maximum gradient score is selected as the position of the standing ring belt, and a preset number of pixel points are selected on the contour of the standing ring belt at equal arc length intervals to form a ring belt sample point set. 4.The power distribution network fault locating method based on unmanned aerial vehicle inspection of claim 1, wherein, The method for calculating the electric measurement energy of the electric measurement voltage sequence comprises the following steps: The electric measurement voltage sequence is segmented based on the frame time window corresponding to the infrared apparent temperature frame; The noise mean value and the noise standard deviation are calculated by counting the electric measurement voltage sequence in each frame time window; The noise mean value plus a preset multiple of the noise standard deviation is taken as the adaptive threshold; The effective signal segment in which the absolute value of the electric measurement voltage sequence exceeds the adaptive threshold is identified; The receiving end equivalent resistance is obtained, and the electric measurement energy is calculated based on the voltage in the effective signal segment and the receiving end equivalent resistance.

5. The power distribution network fault locating method based on unmanned aerial vehicle inspection according to claim 4, characterized in that, The time window corresponding to the peak value frame in which the electric measurement energy reaches the peak value is determined, and the fault orientation point is locked in the ring belt sample point set in combination with the depth ranging data, which comprises the following steps: The electric measurement energy calculated for all frame time windows is compared, and the infrared apparent temperature frame in which the frame time window with the maximum electric measurement energy is located is determined as the peak value frame; The depth ranging data corresponding to the peak value frame is obtained; In the ring belt sample point set, the depth ranging data of each sample point is compared, and the sample point with the smallest depth ranging data is taken as the forward point. The forward point is determined as a fault orientation point of the current inspection lock.

6. The power distribution network fault locating method based on unmanned aerial vehicle inspection according to claim 1, characterized in that, According to the image pixel coordinates of the fault orientation point, the camera pose data corresponding to the peak frame, and the depth ranging data, three-dimensional space coordinates of the fault point are determined, including: The pixel coordinates of the fault orientation point are converted into a normalized line-of-sight direction vector in the camera coordinate system based on the camera intrinsic matrix; The normalized line-of-sight direction vector is transformed into a world direction vector in the world coordinate system according to the camera rotation matrix and the translation vector corresponding to the peak frame; The camera optical center position of the peak frame is taken as a starting point, and the world direction vector is extended, and the extension length is the depth ranging value corresponding to the fault orientation point, and the terminal position obtained is taken as the three-dimensional space coordinates of the fault point.

7. The power distribution network fault locating method based on unmanned aerial vehicle inspection according to claim 1, characterized in that, Based on the three-dimensional space coordinates of the fault point, the section to which the fault belongs is determined, including: The three-dimensional space coordinates of the fault point are projected onto the power distribution network line polyline data composed of line segments connecting the towers; For each line segment, the vertical projection point of the three-dimensional space coordinates of the fault point on the straight line of the line segment is calculated; It is judged whether the vertical projection point is located between the two end points of the line segment, if it is located between the two end points of the line segment, the vertical projection point is taken as the corrected projection point, otherwise the end point closest to the fault point is taken as the corrected projection point; The Euclidean distance between the fault point and the corrected projection point is calculated; The Euclidean distances corresponding to all line segments are compared, and the line segment with the smallest Euclidean distance is taken as the section to which the fault belongs.

8. A power distribution network fault positioning system based on unmanned aerial vehicle inspection, applied to the power distribution network fault positioning method based on unmanned aerial vehicle inspection in any one of claims 1-7, the system comprising: A data acquisition module for controlling the unmanned aerial vehicle to hover and inspect the target insulator of the power distribution network, and synchronously acquiring an infrared apparent temperature frame sequence, an electric measurement voltage sequence, camera pose data, and depth ranging data; A radiation correction module for performing radiation correction on the infrared apparent temperature frame sequence and background removal to obtain a corresponding temperature rise map; A ring band extraction module for extracting the standing ring band at the root of the insulator shed in the temperature rise map, and generating a ring band sample point set on the contour of the standing ring band; An electric measurement orientation module for calculating the electric measurement energy of the electric measurement voltage sequence, determining the peak frame corresponding to the time window in which the electric measurement energy reaches the peak, and locking the fault orientation point in the ring band sample point set in combination with the depth ranging data; A three-dimensional positioning module for determining the three-dimensional space coordinates of the fault point according to the image pixel coordinates of the fault orientation point, the camera pose data corresponding to the peak frame, and the depth ranging data; A section landing point module for determining the section to which the fault belongs based on the three-dimensional space coordinates of the fault point.

Images