Unmanned aerial vehicle inspection identification method and device for power transmission tower

Abstract

The invention provides an unmanned aerial vehicle inspection identification method and device for a power transmission tower, and relates to the technical field of electric power facility health monitoring. Polar coordinates of stay wire endpoints are obtained through unmanned aerial vehicle laser ranging, the length of the stay wire is calculated, and the deformation abnormity degree is evaluated; on the basis of a five-channel hyperspectral imaging system, orthogonal route scanning imaging is utilized, an insulator is positioned in real time in combination with a CNN algorithm, and the state of the insulator is analyzed through a UV wave band radiation value, a discharge gradient and a temperature difference; extracting reflectivity data of a wave band of 550-700nm, and calculating a corrosion area proportion by adopting a wave band ratio method; and evaluating the health state of the iron tower by combining the stay wire deformation, the insulator abnormal state and the corrosion degree. According to the method, multi-source data and an intelligent algorithm are fused, high-precision detection of stay wire deformation, insulator defects and corrosion area is realized, the problems of low efficiency and single parameter of traditional inspection are solved, and automation and reliability of state evaluation of the power transmission tower are remarkably improved.

Description

Technical Field

[0001] The present invention relates to the technical field of power facility health monitoring technology, and specifically to a method and device for unmanned aerial vehicle inspection and identification of transmission towers. Background Art

[0002] With the rapid development of the power industry, transmission towers, as key infrastructure of the power grid system, have a direct impact on the safety and stability of power transmission due to their health. Traditional inspections rely heavily on manual climbing or ground observation, which has drawbacks such as low efficiency, high risk, and single data. In recent years, drone technology has been gradually applied to power inspections, using optical sensors to achieve non-contact detection. However, due to insufficient multi-source data fusion, poor algorithm adaptability, and limited detection accuracy, it is difficult to fully assess the comprehensive conditions of the tower, such as wire deformation, insulator defects, and metal corrosion. This leads to delayed operation and maintenance decisions, becoming a bottleneck restricting the development of intelligent inspections:

[0003] In existing technologies, a typical drone inspection solution typically uses a single sensor, such as a visible light camera or lidar, to acquire local data on the tower. For example, a drone hovers to capture images of the tower, uses threshold segmentation to identify the location of the insulator, and combines manual experience to determine the degree of contamination. Alternatively, a laser rangefinder measures the coordinates of the guy wire endpoints and compares them with historical data to estimate deformation. Furthermore, some methods attempt to use infrared thermal imaging to detect insulator temperature anomalies, but these methods rely on fixed thresholds to determine partial discharge and are susceptible to environmental interference. For metal corrosion detection, existing technologies often use visible light band images to analyze rusted areas and extract the corrosion area based on color features. However, these methods are significantly affected by lighting conditions and background interference, resulting in a high false detection rate.

[0004] The above-mentioned existing technologies have the following technical defects: existing methods mostly process data independently for a single detection target (such as deformation or local temperature), lack multi-dimensional collaborative analysis of the tower structure, electrical components and material status, resulting in insufficient integrity and accuracy of comprehensive health assessments; sensor data is easily affected by changes in lighting, meteorological conditions and background noise, for example, optical detection fails in backlit or shadowed scenes, and thermal imaging is affected by ambient temperature fluctuations; existing solutions mostly rely on static thresholds or offline data analysis, and cannot fuse multimodal data in real time.

[0005] The above information disclosed in this Background section is only for enhancement of understanding of the background of the present disclosure and therefore it may contain information that does not form the prior art that is already known to a person of ordinary skill in the art. Summary of the Invention

[0006] The purpose of the present invention is to provide a method and device for drone inspection and identification of transmission towers to solve the problems raised in the above background technology.

[0007] To achieve the above object, the present invention provides the following technical solutions:

[0008] The UAV inspection and identification method for transmission towers includes the following steps:

[0009] Step 1: Pre-set a reference point on the ground. The drone hovers above the end of the transmission tower's guy wire. Using the laser ranging function, the drone identifies the height difference to the ground reference point and calculates the polar coordinates of the guy wire endpoint. The guy wire length is then calculated based on the polar coordinates of the guy wire endpoint to assess the degree of guy wire deformation anomaly.

[0010] Step 2: The drone is equipped with a 5-channel hyperspectral imaging system, which performs orthogonal flight path scanning and high-precision imaging. It extracts radiation values ​​in the 300-900nm band, uses a CNN algorithm to locate the insulator in real time, and determines the insulator's condition by calculating the average radiation value in the UV band, the gradient of the partial discharge area, and the temperature difference.

[0011] Step 3: Extract the radiation value in the 550-700nm band, shield the insulator area, and convert it into reflectivity. Based on the reflectivity data, use the band ratio method to calculate the corrosion area ratio.

[0012] Step 4: Evaluate the overall health of the transmission tower by comprehensively analyzing the corrosion area ratio, abnormal insulators, and abnormal deformation of guy wires;

[0013] Furthermore, the height difference to the ground reference point is identified by the laser ranging function, and the polar coordinates of the end point of the string are calculated, which includes the following steps:

[0014] A reference point is set on the ground in advance. The coordinates of the reference point are measured by the total station as (x2, y2, z2). The drone hovers above the end of the cable, turns on the laser ranging function, and outputs the height difference h1 from the ranging point to the ground reference point. The measured polar coordinates of the cable end point are Where d1 represents the distance between the end points of the string obtained by laser ranging, θ1 represents the angle of the laser ranging direction in the horizontal plane, Indicates the angle of the ranging direction in the vertical plane.

[0015] Furthermore, the method for calculating the length of the cable based on the polar coordinates of the cable endpoints and evaluating the abnormality of the cable deformation is as follows:

[0016] The measured polar coordinates of the end points of the string are Convert to three-dimensional coordinates:

[0017]

[0018] Calculate the length of the cable:

[0019]

[0020] Where, L represents the length of the wire;

[0021] Construct the deformation anomaly formula:

[0022]

[0023] Where L0 represents the initial length of the wire, L xb Indicates the degree of deformation abnormality.

[0024] Furthermore, the 5-channel hyperspectral imaging system carried by the UAV is based on the spectral range of 300-900nm, divided into bands of 300-400nm, 400-500nm, 550-700nm, 700-800nm ​​and 800-900nm, corresponding to the 1st, 2nd, 3rd, 4th and 5th channels respectively.

[0025] Furthermore, the method for real-time positioning of insulators using the CNN algorithm is as follows:

[0026] Pre-trained model for locating insulators:

[0027] Collect at least 1,000 hyperspectral images of transmission towers, resize them to 244×244 while maintaining the aspect ratio, and use the LabelImg tool to label the bounding box of the insulator's overall outline: (x min ,y min , x max ,y max ), where x min Indicates the column position on the left side of the bounding box, y min Indicates the row position of the top of the bounding box, x max Indicates the column position on the right side of the bounding box, y max represents the row position at the bottom of the bounding box, and its class label is insulator,

[0028] Afterwards, convert it to normalized center point coordinates:

[0029]

[0030] In the formula, (x c ,y c ,ω,h) represents the coordinates of the center point, x c ,y c represents the center coordinate of the insulator, ω represents the width of the insulator, h represents the height of the insulator, W represents the width of the corresponding image, and H represents the height of the corresponding image;

[0031] The hyperspectral image, the corresponding center point coordinates and the category label are input into the ResNet-18 model. The converted center point coordinates are used as the true label. The output dimension of the last fully connected layer is modified to the corresponding center point coordinates.

[0032] (x c ,y c , ω, h) and confidence score s, outputting a 5-dimensional vector dataset:

[0033] output=[x c ,y c ,ω,h,s]

[0034] Where output represents the output 5-dimensional vector dataset, and the confidence score is calculated as follows:

[0035]

[0036] Where z represents the original confidence score output by the fully connected layer;

[0037] The center point coordinates output by the ResNet-18 model are used as the predicted results, and the error between them and the actual center point coordinates is calculated using the positioning loss function. The positioning loss function is:

[0038]

[0039] Where, L loc represents the positioning loss function, p i represents the predicted value of the i-th parameter, g i Represents the true value of the i-th parameter, where i represents the index of the parameter in the center point coordinates, i = 1, 2, 3, 4, corresponding to the center point horizontal coordinate, center point vertical coordinate, bounding box width and bounding box height, respectively, smooth L1 (p i -g i ) represents the loss function and satisfies:

[0040]

[0041] Construct the classification loss formula:

[0042]

[0043] Where, L cls represents the classification loss;

[0044] Construct the total loss formula:

[0045] L total =L loc +L cls

[0046] Where, L total Indicates total loss;

[0047] Set the training parameters:

[0048] Learning rate: 10 -4 , batch size: 32, training rounds: 50;

[0049] The transmission tower images taken by the drone in real time are unified to a fixed size of 244×244, calibrated into 244×244 hyperspectral images, and input into the pre-trained insulator positioning model to output a 5-dimensional vector dataset:

[0050] output=[x c ,y c ,ω,h,s]

[0051] Set the detection threshold s dw , and 1>s dw ≥0.5, only keep s>s dw The boundary box of the insulator is used to complete the positioning of the insulator.

[0052] Furthermore, the method for judging the insulator status by calculating the average radiation value in the UV band, the gradient of the partial discharge area, and the temperature difference is as follows:

[0053] The pixel coordinates of the output 5-dimensional vector dataset are denormalized. Based on the denormalized pixel coordinates, the insulator area is extracted from the 244×244 hyperspectral image. The radiation values ​​of the five channels are extracted to calculate the average radiation value:

[0054]

[0055] Where N represents the total number of pixels in the insulator area, I aj represents the radiation value of the jth pixel in the ath channel in the insulator area, where a = 1, 2, 3, 4, 5, j = 1, 2, ..., N, and the unit of the radiation value is: μW / cm 2 / nm;

[0056] Calculate the partial discharge regional gradient:

[0057]

[0058] Where G represents the gradient of the local discharge region, R1 represents the average radiation value of the first channel, and R2 represents the average radiation value of the second channel;

[0059] To calculate the temperature difference based on the radiation values ​​of channels 4 and 5, we first construct the Planck's law formula:

[0060]

[0061] Where λ is the central wavelength of the band, T is the temperature, and C1 = 3.7418 × 10 8, C2=1.4388×10 4 , I(λ, T) represents the radiation value at wavelength λ;

[0062] Through inversion, the temperature at a wavelength of 750nm is T5, and the temperature at a wavelength of 850nm is T6;

[0063] Calculate the temperature difference by using the temperature T5 at a wavelength of 750nm and the temperature T6 at a wavelength of 850nm:

[0064] ΔT=T6-T5

[0065] Where ΔT represents the temperature difference;

[0066] Construct anomaly judgment formula:

[0067]

[0068] Wherein, B represents the abnormality judgment formula. When B=1, the insulator is judged to be in an abnormal state. When B=0, the insulator is judged to be in a normal state.

[0069] Furthermore, the shielded insulator area is converted into reflectivity, and the method for calculating the corrosion area ratio based on the reflectivity data using the band ratio method is as follows:

[0070] In the 244×244 hyperspectral image, all insulator areas are shielded and the remaining areas are marked as background areas. The radiation values ​​of the background areas at 550nm and 680nm are converted into reflectance using the radiometric calibration method:

[0071]

[0072] Where, L(λ 550 ) represents the radiation value at a wavelength of 550nm in the background area, L white (λ 550 ) represents the radiation value of the reference white plate at a wavelength of 550nm, ρ white Indicates the standard reflectivity of the whiteboard, the default value is ρ white =99%, L(λ 680 ) represents the radiation value at a wavelength of 680nm in the background area, L white (λ 680 ) represents the radiation value of the reference white plate at a wavelength of 680nm, ρ(λ 550 ) represents the reflectivity at a wavelength of 550 nm in the background area, ρ(λ 680 ) represents the reflectivity at a wavelength of 680 nm in the background area;

[0073] The band ratio method is used to construct the ratio formula:

[0074]

[0075] Where R ratio It represents the ratio formula, that is, the ratio of the reflectivity at a wavelength of 680nm to the reflectivity at a wavelength of 550nm in the background area;

[0076] Set the ratio threshold T yz =1.5, calculate the corrosion area ratio:

[0077]

[0078] Where R ratio，j′ represents the band ratio of the j′th pixel in the background area, where j′=1, 2, …, Z, Z represents the total number of pixels in the background area, Π(R ratio，j′ >T yz ) represents the indicator function, when R ratio，j′ >T yz When ∏(R ratio，j′ >T yz )=1, when R ratio，j′ ≤T yz When ∏(R ratio，j′ >T yz )=0, M represents the corrosion area ratio.

[0079] 8. The method for inspection and identification of transmission towers by drones according to claim 1, wherein the method for evaluating the comprehensive health status of the transmission tower is:

[0080] Set the corrosion area ratio threshold to M threshold =0.05, the wire deformation threshold is L threshold =0.02;

[0081] When the corrosion area ratio M≤M threshold 、Number of abnormal insulators abnormal = 0 and the abnormality of wire deformation L xb ≤L threshold When , the evaluation result of the transmission tower is normal;

[0082] When the corrosion area ratio M>M threshold 、Number of abnormal insulators abnormal ≠0 or abnormal deformation of the wire L xb >L threshold If any of these conditions occur, the evaluation result of the transmission tower will be a general alarm state;

[0083] When the corrosion area ratio M>M threshold 、Number of abnormal insulators abnormal ≠0 or abnormal deformation of the wire L xb >L thresholdWhen any two or more of the above conditions occur, the evaluation result of the transmission tower will be an emergency alarm state.

[0084] The present invention further provides a UAV inspection and identification device for transmission towers, wherein the device is used to execute the above-mentioned UAV inspection and identification method for transmission towers, comprising:

[0085] The guy wire deformation monitoring module is used to pre-set reference points on the ground. The drone hovers above the end of the guy wire of the transmission tower and uses the laser ranging function to identify the height difference to the ground reference point. The polar coordinates of the guy wire endpoint are then calculated. The guy wire length is then calculated based on the polar coordinates of the guy wire endpoint to assess the degree of guy wire deformation anomaly.

[0086] The insulator detection module is used on drones equipped with a 5-channel hyperspectral imaging system. It performs orthogonal flight path scanning and high-precision imaging, extracts radiation values ​​in the 300-900nm band, locates insulators in real time using a CNN algorithm, and determines the insulator status by calculating the average radiation value in the UV band, the gradient of the partial discharge area, and the temperature difference.

[0087] The corrosion assessment module extracts radiation values ​​in the 550-700nm band, converts them into reflectivity after shielding the insulator area, and calculates the corrosion area ratio based on the reflectivity data using the band ratio method.

[0088] The health assessment module is used to evaluate the overall health status of transmission towers by comprehensively considering the corrosion area ratio, abnormal insulators, and abnormal deformation of guy wires.

[0089] Compared with the prior art, the present invention has the following beneficial effects:

[0090] The present invention integrates laser ranging to evaluate wire deformation, hyperspectral imaging to monitor insulator status, and detects corrosion area to construct a multi-data fusion model, thus overcoming the limitations of single-parameter evaluation in traditional methods. The improved CNN algorithm is used to achieve real-time positioning of insulators with pixel-level positioning accuracy. The insulator status is judged by the average radiation value in the UV band, the gradient of the local discharge area, and the temperature difference, thus avoiding manual reliance on experience to identify insulator quality and achieving higher accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0091] Figure 1 Schematic diagram of the overall method flow of the present invention;

[0092] Figure 2 This is a statistical diagram of the detection threshold and summon rate of the present invention;

[0093] Figure 3 This is a statistical diagram of the detection threshold and accuracy of the present invention;

[0094] Figure 4 This is a statistical diagram of the detection threshold and accuracy rate of the present invention;

[0095] Figure 5 This is a statistical diagram of the detection threshold and average accuracy of the present invention;

[0096] Figure 6 This is a statistical diagram of the corrosion area ratio of the present invention;

[0097] Figure 7 This is a statistical diagram of the number of abnormal insulators in the present invention;

[0098] Figure 8 This is a statistical diagram of the abnormality of wire deformation in the present invention;

[0099] Figure 9 Schematic diagram of the overall device module of the present invention. DETAILED DESCRIPTION

[0100] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below with reference to specific embodiments.

[0101] It should be noted that, unless otherwise defined, the technical or scientific terms used in the present invention should have the usual meanings understood by people with ordinary skills in the field to which the present invention belongs. The "first", "second" and similar words used in the present invention do not indicate any order, quantity or importance, but are only used to distinguish different components. "Include" or "comprise" and similar words mean that the elements or objects appearing before the word include the elements or objects listed after the word and their equivalents, without excluding other elements or objects. "Connect" or "connected" and similar words are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. "Up", "down", "left", "right" and the like are only used to indicate relative position relationships. When the absolute position of the object being described changes, the relative position relationship may also change accordingly.

[0102] Example:

[0103] See also Figures 1 to 8 , the present invention provides a technical solution:

[0104] The UAV inspection and identification method for transmission towers includes the following steps:

[0105] Step 1: Pre-set a reference point on the ground. The drone hovers above the end of the transmission tower's guy wire. Using the laser ranging function, the drone identifies the height difference to the ground reference point and calculates the polar coordinates of the guy wire endpoint. The guy wire length is then calculated based on the polar coordinates of the guy wire endpoint to assess the degree of guy wire deformation anomaly.

[0106] A reference point is set on the ground in advance and a crosshair is engraved on it. This is because the ground reference point is the conversion benchmark between the drone laser coordinates and the geodetic coordinate system, ensuring that the length of the cable can be calculated using three-dimensional coordinates. Therefore, the coordinates of the ground reference point are measured by a total station, such as Leica TS60, as (x2, y2, z2). This provides accurate ground reference point coordinate data. The drone hovers over the top connection point of the cable, and the visible light camera is zoomed to confirm that the crosshairs are aligned with the center of the connection point. The laser ranging function is turned on, and the angle deviation caused by the tilt of the laser beam is avoided. The height difference h1 from the ranging point to the ground reference point is output, and the polar coordinates of the measured cable endpoint are Among them, d1 represents the distance between the end points of the string obtained by laser ranging, θ1 represents the angle of the laser ranging direction in the horizontal plane, and θ1 represents the angle of the ranging direction in the vertical plane.

[0107] The measured polar coordinates of the end points of the string are Convert to three-dimensional coordinates:

[0108]

[0109] Calculate the length of the cable:

[0110]

[0111] Where, L represents the length of the cable;

[0112] Construct the deformation anomaly formula:

[0113]

[0114] Where, L0 represents the initial length of the guy wire, which takes precedence over the guy wire length marked on the tower drawing, or the actual measured value during operation as the initial length of the guy wire, L xb Indicates the degree of deformation abnormality. When the wire undergoes plastic deformation due to corrosion, fatigue, overload, etc., its actual length will deviate from the initial design value. By monitoring L xb , can detect the degradation of the mechanical properties of the guy wire in advance, and avoid the tilt or collapse of the tower due to insufficient tension. According to industry standards, the maximum allowable deviation of the guy wire length is only 3%, so in the subsequent guy wire deformation threshold L threshold The setting is L threshold =0.02.

[0115] Step 2: The drone is equipped with a 5-channel hyperspectral imaging system, which performs orthogonal flight path scanning and high-precision imaging. It extracts radiation values ​​in the 300-900nm band, uses a CNN algorithm to locate the insulator in real time, and determines the insulator's condition by calculating the average radiation value in the UV band, the gradient of the partial discharge area, and the temperature difference.

[0116] Insulators are critical components in transmission tower inspections, and their condition directly impacts the safe operation of transmission lines. However, in complex transmission tower scenarios, insulators may be obscured or blend in with the background. A localization model can quickly and accurately locate insulators within an image, providing a foundation for subsequent condition assessment. Therefore, pre-training the insulator localization model improves inspection efficiency and reduces the workload of manual insulator identification.

[0117] Collect at least 1,000 hyperspectral images of transmission towers. Hyperspectral images contain rich spectral information. Different substances have unique spectral characteristics in different bands. Channels 1-5 correspond to 300-400nm, 400-500nm, 550-700nm, 700-800nm, and 800-900nm, respectively. These features can be used to more accurately identify the material and status of insulators. The included RGB images are common color images that can provide intuitive visual information. The combination of the two can fully utilize the advantages of different types of images to improve the generalization ability and recognition accuracy of the model. While maintaining the aspect ratio, the image size is uniformly modified to 244×244. The unified image size ensures that the model can process the input data normally and better adapt to the deep learning model. Use the LabelImg tool and the rectangular box drawing tool provided by the tool to select the overall outline of the insulator. The bounding box of the overall outline of the insulator is marked as: (x min ,y min , x max ,y max ), where x min Indicates the column position on the left side of the bounding box, y min Indicates the row position of the top of the bounding box, x max Indicates the column position on the right side of the bounding box, y max Indicates the row position at the bottom of the bounding box and its category label is insulator, saved in XML format;

[0118] Afterwards, convert it to normalized center point coordinates:

[0119]

[0120] In the formula, (x c ,y c ,ω,h) represents the coordinates of the center point, x c ,y crepresents the center coordinates of the insulator, ω represents the width of the insulator, h represents the height of the insulator, W represents the width of the corresponding image, and H represents the height of the corresponding image. Normalized center coordinates eliminate the influence of image size, enabling the model to perform unified learning and prediction on images of different sizes. Furthermore, center coordinates and width and height information are more consistent with the model's input and processing methods, facilitating model calculation and learning.

[0121] The processed hyperspectral image is used as input data, the corresponding center point coordinates are used as the true label, and the category label is used for the classification task. These data are input into the ResNet-18 model in batches according to a certain batch size, such as 32. For the first convolutional layer, the number of input channels is changed from 3 to 5 to match the 5 channels of the hyperspectral image, and the weights are randomly initialized. The output dimension of the last fully connected layer is modified to the corresponding center point coordinates (x c ,y c , ω, h) and confidence score s, outputting a 5-dimensional vector dataset:

[0122] output=[x c ,y c ,ω,h,s]

[0123] In the formula, "output" represents the output 5-dimensional vector dataset. ResNet-18 is a pre-trained deep convolutional neural network with powerful feature extraction capabilities. By inputting data into this model for training, its pre-trained weights and network structure can be used to quickly learn the characteristics of insulators, improving the model's training efficiency and accuracy. The confidence score indicates the model's confidence in the prediction results, which helps in subsequent screening and judgment of the prediction results.

[0124] The confidence score is calculated as follows:

[0125]

[0126] Where z represents the original confidence score output by the fully connected layer;

[0127] The center point coordinates output by the ResNet-18 model are used as the predicted results, and the error between them and the actual center point coordinates is calculated using the positioning loss function. The positioning loss function is:

[0128]

[0129] Where, L koc represents the positioning loss function, p i represents the predicted value of the i-th parameter, g iRepresents the true value of the i-th parameter, where i represents the index of the parameter in the center point coordinates, i = 1, 2, 3, 4, corresponding to the center point horizontal coordinate, center point vertical coordinate, bounding box width and bounding box height, respectively, used to calculate the error between the center point coordinates predicted by the model and the true center point coordinates. For example, p1 represents the predicted x coordinate, g1 represents the true x coordinate, and smoothing is used. L1 The loss function can use square loss when the error is small to make the model converge faster; when the error is large, use linear loss to avoid gradient explosion and smooth L1 (p i -g i ) represents the loss function and satisfies:

[0130]

[0131] Construct the classification loss formula:

[0132]

[0133] Where, L cls represents the classification loss, This is the sigmoid function, which is used to map the output confidence score to the probability interval [0, 1]. This term further takes the logarithm of the probability. When the confidence score is higher, the classification loss value is closer to 1, indicating a smaller loss. Conversely, when the confidence score is lower, the classification loss value is closer to 0, indicating a larger loss. A large number of experiments and studies have shown that this method is suitable for binary classification problems, can reduce misclassification, promote rapid model correction, and is used to calculate the model's classification error for the insulator category. It can measure the difference between the model's predicted category probability and the true category.

[0134] Construct the total loss formula:

[0135] L total =L loc +L cls

[0136] Where, L total Represents the total loss, adding the positioning loss and the classification loss to achieve linear addition of the positioning loss and the classification loss, so that the model can optimize the positioning and classification tasks simultaneously during the training process, ensuring the accuracy of classification and positioning and improving the overall performance of the model;

[0137] Set the training parameters:

[0138] Learning rate: 10 -4 , making the model more stable during training and avoiding skipping the optimal solution due to excessive step size;

[0139] Batch size: 32. The batch size determines the number of data samples input into the model during each training. It can improve training efficiency and prevent the model from falling into a local optimal solution.

[0140] Training epochs: 50, the model fully learns the characteristics of the data and avoids overfitting;

[0141] The transmission tower images taken by the drone in real time are unified to a fixed size of 244×244, calibrated into 244×244 hyperspectral images, and input into the pre-trained insulator positioning model to output a 5-dimensional vector dataset:

[0142] output=[x c ,y c ,ω,h,s]

[0143] Set the detection threshold s dw , and 1>s dw ≥0.5, only keep s>s dw The bounding box of the insulator is used to complete the positioning of the insulator. Since the real-time inspection of the drone needs to ensure the detection speed and result continuity, a too high detection threshold may cause the detection frame to lose the target and affect the positioning stability. According to the tolerance and consequences of missed detection, the detection threshold is set independently, such as setting the detection threshold s. dw =0.5, missed inspection may lead to serious accidents.

[0144] The normalized center point coordinates need to be converted back to the pixel coordinates of the original image in order to accurately extract the insulator area from the image. Therefore, the pixel coordinates of the output 5-dimensional vector dataset output are denormalized and the output (x min ,y min , x max ,y max ), extract the insulator area from the 244×244 hyperspectral image based on the denormalized pixel coordinates, extract the radiation values ​​of the five channels and calculate the average radiation value:

[0145]

[0146] Where N represents the total number of pixels in the insulator area, I aj represents the radiation value of the jth pixel in the ath channel in the insulator area, where a = 1, 2, 3, 4, 5, j = 1, 2, ..., N;

[0147] The radiation value of corona discharge in the UV band is several times that of blue light. Therefore, the radiation difference between UV (300-400nm) and blue light (400-500nm) is extracted to calculate the gradient of the partial discharge area:

[0148]

[0149] Where G represents the gradient of the local discharge region, R1 represents the average radiation value of the first channel, R2 represents the average radiation value of the second channel, and 100 is the difference between the center wavelengths of the first and second channels. Eliminating the dimension effect is closer to the physical characteristics of the discharge.

[0150] The temperature difference is calculated based on the radiation values ​​of channels 4 and 5. These two channels correspond to wavelengths of 750nm and 850nm, which is an ideal combination for dual-band infrared temperature measurement. Planck's law shows that the monochromatic radiation value has a nonlinear relationship with temperature, but a single wavelength cannot eliminate the influence of emissivity. Selecting two adjacent wavelengths to jointly invert the temperature can eliminate interference. First, construct the Planck's law formula:

[0151]

[0152] Where λ is the central wavelength of the band, T is the temperature, and C1 = 3.7418 × 10 8 , C2=1.4388×10 4 , I(λ, T) represents the radiation value at wavelength λ;

[0153] Through inversion, the temperature at a wavelength of 750nm is T5, and the temperature at a wavelength of 850nm is T6;

[0154] Calculate the temperature difference by using the temperature T5 at a wavelength of 750nm and the temperature T6 at a wavelength of 850nm:

[0155] ΔT=T6-T5

[0156] Where ΔT represents the temperature difference. The combination of temperature T5 and temperature T6 can capture the small temperature gradient on the surface of the insulator, that is, the temperature difference between the partial discharge point and the normal area.

[0157] Construct anomaly judgment formula:

[0158]

[0159] Wherein, B represents the abnormality judgment formula. When B=1, the insulator is judged to be in an abnormal state. When B=0, the insulator is judged to be in a normal state.

[0160] According to the statistics of 1,000 transmission towers in the past three years, the average UV band radiation value R1 of normal insulators is 130. Therefore, R1>150 is considered to be abnormal. According to experiments, during artificial corona discharge, the local discharge area gradient G∈[25,50]. According to the recommended standards of the power industry, an insulator temperature difference higher than 10°C is a serious defect, and higher than 5°C is an early warning threshold. For example, the reflection of morning dew may cause R1>150. The presence of dirt on the insulator surface may reach G>25. Areas with large day and night temperature differences may cause Δ T>5, so when R1>150∧G>25∧ΔT>5 is satisfied, the output is 1 to judge that the insulator is abnormal, otherwise the output is 0 to judge that the insulator is in normal state. Table 1 shows that after collecting 40 sets of hyperspectral images of different transmission tower images and processing them with the above calculation formula to obtain UV radiation value, partial discharge area gradient and temperature difference, the insulator status judgment result is output by comparing with each set threshold value. Finally, the UV radiation value, partial discharge area gradient, temperature difference and insulator status judgment results are summarized to form a data report.

[0161]

[0162]

[0163] Table 1 Insulator status determination data report

[0164] like Figure 2-Figure 5 As shown in the figure, the detection threshold setting of the insulator positioning model is experimentally verified for a sample set containing 100 real insulator hyperspectral images and 100 non-insulator hyperspectral images. The recall rate, precision rate, accuracy rate and average precision under different thresholds are output, and the number of insulators correctly identified, the number of insulators not detected, the number of non-insulators not misdetected, and the number of non-insulators misidentified as insulators are recorded.

[0165] Calculate the recall of the model:

[0166]

[0167] Where TP represents the number of correctly identified insulators, FN represents the number of undetected insulators, and Recall represents the recall rate. The recall rate measures the model's ability to correctly detect all insulators. A higher value indicates fewer missed insulators. As the detection threshold gradually increases, the model only retains detection boxes with higher confidence, so the recall rate continues to decrease.

[0168] Calculate the accuracy of the model:

[0169]

[0170] Where TN represents the number of non-insulators that are not falsely detected, FP represents the number of non-insulators that are falsely identified as insulators, and Accuracy represents the accuracy rate. The accuracy rate measures the overall classification ability of the model for all samples, that is, the proportion of correct predictions.

[0171] Calculate the accuracy of the model:

[0172]

[0173] Where Precision represents the accuracy rate, which reflects how many of the insulator detection results from the model are truly correct. It is used to evaluate the reliability of the model detection results. A high accuracy rate can reduce the workload of manual review and avoid dealing with a large number of false insulators that are falsely detected.

[0174] The average precision is the area under the precision-recall curve, which combines the recall rate and precision rate. It is the core indicator for measuring the performance of the model in detecting a single insulator in the target detection task. In the actual engineering scenario of drone inspection, the dynamic adjustment of the detection threshold needs to be closely combined with business needs to achieve a balance between the risk control of missed detection and the optimization of false detection costs. Selecting a low detection threshold between 0.5-0.6 can ensure that insulators can be detected, but the number of non-insulators that are not falsely detected increases, and secondary verification is required. For example, to avoid the possibility that low recall rate may lead to the detection of insulators. If the detection is missed, it will cause safety hazards. If hidden dangers are required to be checked and missed, a lower detection threshold, such as 0.5, can be selected. If a high detection threshold is selected, such as between 0.8-0.95, the number of non-insulators misidentified as insulators will be greatly reduced, avoiding the wrong dispatch of maintenance personnel and increasing costs, but some insulators may be missed. It is recommended to select a detection threshold based on the peak value of the average accuracy so that it can detect most real insulators and control the number of non-insulators misidentified as insulators within an acceptable range. When the average accuracy is between 0.7-0.75, such as Figure 2 The recall shown is between 75% and 80%, as shown in Figure 3 The accuracy shown is between 90%-93%.

[0175] Step 3: Extract the radiation value in the 550-700nm band, shield the insulator area, and convert it into reflectivity. Based on the reflectivity data, use the band ratio method to calculate the corrosion area ratio.

[0176] In the 244×244 hyperspectral image, all insulator areas are masked, and the remaining area is marked as background. This is because corrosion primarily occurs in metal components such as the tower main material and the guy wire anchors. Insulators themselves are non-metallic, and their spectral characteristics are unrelated to metal corrosion. This masking allows for 100% focus on the background metal structure, eliminating interference from components such as the insulator sheds and steel legs.

[0177] The radiation values ​​of the background area at 550nm and 680nm are converted into reflectivity by the radiation calibration method:

[0178]

[0179] Where, L(λ 550 ) represents the radiation value at a wavelength of 550nm in the background area, L white (λ 550 ) represents the radiation value of the reference white plate at a wavelength of 550nm, ρ white Indicates the standard reflectivity of the whiteboard, the default value is ρ white =99%, L(λ 680 ) represents the radiation value at a wavelength of 680nm in the background area, L white (λ 680 ) represents the radiation value of the reference white plate at a wavelength of 680nm, ρ(λ 550 ) represents the reflectivity at a wavelength of 550 nm in the background area, ρ(λ 680 ) represents the reflectivity at a wavelength of 680nm in the background area. Iron oxide has strong absorption in the 550nm band (green light region), and the reflectivity is significantly reduced. The reflectivity is significantly increased at 680nm (red light region). The characteristics of low reflection of green light and high reflection of red light make the ratio of the reflectivity at 680nm to 550nm in the corroded area significantly higher than that in the uncorroded area, and the ratio is between 1.5-2.

[0180] Therefore, the band ratio method is used to construct the ratio formula:

[0181]

[0182] Where R ratio It represents the ratio formula, that is, the ratio of the reflectivity at a wavelength of 680nm to the reflectivity at a wavelength of 550nm in the background area;

[0183] Set the ratio threshold T yz =1.5, when R ratio =1.5, indicating that iron oxide begins to accumulate. Calculate the percentage of corrosion area:

[0184]

[0185] Where R ratio，j′ represents the band ratio of the j′th pixel in the background area, where j′=1, 2, …, z, Z represents the total number of pixels in the background area, Π(R ratio，j′ >T yz ) represents the indicator function, when R ratio，j′ >T yz When ∏(R ratio，j′ >T yz 0=1, when Rratio，j′ ≤T yz When ∏(R ratio，j′ >T yz )=0. This conversion into a 0 / 1 binary variable shields noise fluctuations and avoids misjudging a certain area as corroded. M represents the proportion of the corroded area.

[0186] Step 4: Evaluate the overall health of the transmission tower by comprehensively analyzing the corrosion area ratio, abnormal insulators, and abnormal deformation of guy wires;

[0187] Set the corrosion area ratio threshold to M threshold =0.05. According to the recommended standards of the power industry, if the corrosion depth of the main material of the tower is greater than 10%, the strength of the steel will drop by more than 15%. Therefore, 0.05 (5%) is set as the threshold, leaving a 50% safety margin.

[0188] The wire deformation threshold is L threshold =0.02, or 2%, which is set according to the industry standard that the deviation of the wire length is ±2%. At this time, tightening can restore the structure. Otherwise, the risk of tower tilt will increase significantly.

[0189] When the corrosion area ratio M≤M threshold 、Number of abnormal insulators abnormal = 0 and the abnormality of wire deformation L xb ≤L threshold When , the evaluation result of the transmission tower is normal;

[0190] When the corrosion area ratio M>M threshold 、Number of abnormal insulators abnormal ≠0 or abnormal deformation of the wire L xb >L threshold If any of these conditions occur, the transmission tower will be assessed as in a general alarm state, and it will be included in the regular maintenance plan to avoid overreaction.

[0191] When the corrosion area ratio M>M threshold 、Number of abnormal insulators abnormal ≠0 or abnormal deformation of the wire L xb >L threshold If any two or more of these conditions occur, it indicates a high-risk scenario for fault coupling, and the transmission tower assessment results in an emergency alert state to prevent serious consequences. Table 2 shows that after collecting hyperspectral images of 40 different transmission towers, the aforementioned calculation formula was used to calculate the corrosion area percentage, the number of abnormal insulators, and the degree of abnormal deformation of the guy wires. The assessment results were then compared with the set thresholds to output the results and summarize them in a data report.

[0192]

[0193]

[0194] Table 2 Transmission tower evaluation results

[0195] like Figure 6-Figure 8 As shown, the number of abnormal insulators, abnormal deformation degree of guy wires and proportion of corrosion area of ​​each sample are displayed. By observing whether each indicator exceeds the threshold, the status of the tower can be quickly judged. The number of abnormal insulators in most samples is 0, and some samples have a certain number of abnormal insulators. Transmission towers with abnormal insulators need to be paid attention to. The abnormal deformation degree of guy wires reflects the stability of the guy wires. For some samples that exceed the threshold, the risk of tower tilt is significantly increased, and timely inspection and repair are required to avoid unstable tower structure due to guy wire problems and cause safety accidents. The proportion of corrosion area directly reflects the corrosion of metal parts. When it exceeds the threshold, the strength of steel decreases, affecting the safety of the tower structure. By integrating the indicators of the three dimensions, the output evaluation results can be directly observed. According to the evaluation results of different levels, it is chosen to be included in the regular maintenance plan or immediately handled to avoid serious consequences.

[0196] See also Figure 9 The present invention further provides a UAV inspection and identification device for transmission towers, wherein the device is used to execute the above-mentioned UAV inspection and identification method for transmission towers, comprising:

[0197] The guy wire deformation monitoring module is used to pre-set reference points on the ground. The drone hovers above the end of the guy wire of the transmission tower and uses the laser ranging function to identify the height difference to the ground reference point. The polar coordinates of the guy wire endpoint are then calculated. The guy wire length is then calculated based on the polar coordinates of the guy wire endpoint to assess the degree of guy wire deformation anomaly.

[0198] The insulator detection module is used on drones equipped with a 5-channel hyperspectral imaging system. It performs orthogonal flight path scanning and high-precision imaging, extracts radiation values ​​in the 300-900nm band, locates insulators in real time using a CNN algorithm, and determines the insulator status by calculating the average radiation value in the UV band, the gradient of the partial discharge area, and the temperature difference.

[0199] The corrosion assessment module extracts radiation values ​​in the 550-700nm band, converts them into reflectivity after shielding the insulator area, and calculates the corrosion area ratio based on the reflectivity data using the band ratio method.

[0200] The health assessment module is used to evaluate the overall health status of transmission towers by comprehensively considering the corrosion area ratio, abnormal insulators, and abnormal deformation of guy wires.

[0201] The above formulas are all dimensionless and numerical calculations. The formulas are obtained by collecting a large amount of data and performing software simulation to obtain the most recent real situation. The preset parameters in the formulas are set by technicians in this field according to actual conditions.

[0202] The above embodiments can be implemented in whole or in part by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product. Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed by hardware or software depends on the specific application and design constraints of the technical solution.

[0203] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, and may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment as needed.

[0204] The above is only a specific implementation method of the present application, but the scope of protection of the present application is not limited thereto. Any technician familiar with this technical field can easily think of changes or replacements within the technical scope disclosed in this application, which should be covered by the scope of protection of the present application.

Claims

1. A method for inspection and identification of transmission towers by drones, characterized in that: The specific steps include: Step 1: Pre-set a reference point on the ground. The drone hovers above the end of the transmission tower's guy wire. Using the laser ranging function, the drone identifies the height difference to the ground reference point and calculates the polar coordinates of the guy wire endpoint. The guy wire length is then calculated based on the polar coordinates of the guy wire endpoint to assess the degree of guy wire deformation anomaly. Step 2: The drone is equipped with a 5-channel hyperspectral imaging system, which performs orthogonal flight path scanning and high-precision imaging. It extracts radiation values ​​in the 300-900nm band, uses a CNN algorithm to locate the insulator in real time, and determines the insulator's condition by calculating the average radiation value in the UV band, the gradient of the partial discharge area, and the temperature difference. Step 3: Extract the radiation value of the 550-700nm band, shield the insulator area, and convert it into reflectivity. Based on the reflectivity data, use the band ratio method to calculate the corrosion area ratio; Step 4: Evaluate the overall health of the transmission tower by comprehensively considering the corrosion area ratio, abnormal insulators, and abnormal deformation of the guy wires.

2. The method for inspection and identification of transmission towers by unmanned aerial vehicles according to claim 1, wherein: The laser ranging function is used to identify the height difference to the ground reference point, and the polar coordinates of the end point of the cable are calculated. The following steps are included: A reference point is set on the ground in advance. The coordinates of the reference point are measured by the total station as (x2, y2, z2). The drone hovers above the end of the cable, turns on the laser ranging function, and outputs the height difference h1 from the ranging point to the ground reference point. The measured polar coordinates of the cable end point are Where d1 represents the distance between the end points of the string obtained by laser ranging, θ1 represents the angle of the laser ranging direction in the horizontal plane, Indicates the angle of the ranging direction in the vertical plane.

3. The method for inspection and identification of transmission towers by unmanned aerial vehicles according to claim 2, wherein: The method for calculating the length of the cable based on the polar coordinates of the cable endpoints and evaluating the abnormality of the cable deformation is as follows: The measured polar coordinates of the end points of the string are Convert to three-dimensional coordinates: Calculate the length of the cable: Where, L represents the length of the wire; Construct the deformation anomaly formula: Where L0 represents the initial length of the wire, L xb Indicates the degree of deformation abnormality.

4. The method for inspection and identification of transmission towers by unmanned aerial vehicles according to claim 1, wherein: The 5-channel hyperspectral imaging system carried by the UAV is based on the spectral range of 300-900nm, divided into bands of 300-400nm, 400-500nm, 550-700nm, 700-800nm ​​and 800-900nm, corresponding to the 1st, 2nd, 3rd, 4th and 5th channels respectively.

5. The method for inspection and identification of transmission towers by unmanned aerial vehicles according to claim 1, wherein: The method for real-time positioning of insulators using CNN algorithm is as follows: Pre-trained model for locating insulators: Collect at least 1,000 hyperspectral images of transmission towers, resize them to 244×244 while maintaining the aspect ratio, and use the LabelImg tool to label the bounding box of the insulator's overall outline: (x min ,y min , x max ,y max ), where x min Indicates the column position on the left side of the bounding box, y min Indicates the row position of the top of the bounding box, x max Indicates the column position on the right side of the bounding box, y max represents the row position at the bottom of the bounding box, and its class label is insulator, Afterwards, convert it to normalized center point coordinates: In the formula, (x c ,y c ,ω,h) represents the coordinates of the center point, x c ,y c represents the center coordinate of the insulator, ω represents the width of the insulator, h represents the height of the insulator, W represents the width of the corresponding image, and H represents the height of the corresponding image; The hyperspectral image, the corresponding center point coordinates and the category label are input into the ResNet-18 model, and the converted center point coordinates are used as the true label. The output dimension of the last fully connected layer is modified to the corresponding center point coordinates (x c ,y c , ω, h) and confidence score s, outputting a 5-dimensional vector dataset: output=[x c ,y c ,ω,h,s] Where output represents the output 5-dimensional vector dataset, and the confidence score is calculated as follows: Where z represents the original confidence score output by the fully connected layer; The center point coordinates output by the ResNet-18 model are used as the predicted results, and the error between them and the actual center point coordinates is calculated using the positioning loss function. The positioning loss function is: Where, L loc represents the positioning loss function, p i represents the predicted value of the i-th parameter, g i Represents the true value of the i-th parameter, where i represents the index of the parameter in the center point coordinates, i = 1, 2, 3, 4, corresponding to the center point horizontal coordinate, center point vertical coordinate, bounding box width and bounding box height, respectively, smooth L1 (p i -g i ) represents the loss function and satisfies: Construct the classification loss formula: Where, L cls represents the classification loss; Construct the total loss formula: L total =L loc +L cls Where, L total Indicates total loss; Set the training parameters: Learning rate: 10 -4 , batch size: 32, training rounds: 50; The transmission tower images taken by the drone in real time are unified to a fixed size of 244×244, calibrated into 244×244 hyperspectral images, and input into the pre-trained insulator positioning model to output a 5-dimensional vector dataset: output=[x c ,y c ,ω,h,s] Set the detection threshold s dw , and 1>s dw ≥0.5, only keep s>s dw The boundary box of the insulator is used to complete the positioning of the insulator.

6. The method for inspection and identification of transmission towers by unmanned aerial vehicles according to claim 5, characterized in that: The method for judging the insulator status by calculating the average radiation value in the UV band, the gradient of the partial discharge area and the temperature difference is as follows: The pixel coordinates of the output 5-dimensional vector dataset are denormalized. Based on the denormalized pixel coordinates, the insulator area is extracted from the 244×244 hyperspectral image. The radiation values ​​of the five channels are extracted to calculate the average radiation value: Where N represents the total number of pixels in the insulator area, I aj represents the radiation value of the jth pixel in the ath channel in the insulator area, where a = 1, 2, 3, 4, 5, j = 1, 2, ..., N, and the unit of the radiation value is: μW / cm 2 / nm; Calculate the partial discharge regional gradient: Where G represents the gradient of the local discharge region, R1 represents the average radiation value of the first channel, and R2 represents the average radiation value of the second channel; To calculate the temperature difference based on the radiation values ​​of channels 4 and 5, we first construct the Planck's law formula: Where λ is the central wavelength of the band, T is the temperature, and C1 = 3.7418 × 10 8 , C2=1.4388×10 4 , I(λ, T) represents the radiation value at wavelength λ; Through inversion, the temperature at a wavelength of 750nm is T5, and the temperature at a wavelength of 850nm is T6; Calculate the temperature difference by using the temperature T5 at a wavelength of 750nm and the temperature T6 at a wavelength of 850nm: ΔT=T6-T5 Where ΔT represents the temperature difference; Construct anomaly judgment formula: Wherein, B represents the abnormality judgment formula. When B=1, the insulator is judged to be in an abnormal state. When B=0, the insulator is judged to be in a normal state.

7. The method for inspection and identification of transmission towers by drones according to claim 6, characterized in that: After the shielding insulator area is converted into reflectivity, the method of calculating the corrosion area ratio based on the reflectivity data using the band ratio method is as follows: In the 244×244 hyperspectral image, all insulator areas are shielded and the remaining areas are marked as background areas. The radiation values ​​of the background areas at 550nm and 680nm are converted into reflectance using the radiometric calibration method: Where, L(λ 550 ) represents the radiation value at a wavelength of 550nm in the background area, L white (λ 550 ) represents the radiation value of the reference white plate at a wavelength of 550nm, ρ white Indicates the standard reflectivity of the whiteboard, the default value is ρ white =99%, L(λ 680 ) represents the radiation value at a wavelength of 680nm in the background area, L white (λ 680 ) represents the radiation value of the reference white plate at a wavelength of 680nm, ρ(λ 550 ) represents the reflectivity at a wavelength of 550 nm in the background area, ρ(λ 680 ) represents the reflectivity at a wavelength of 680 nm in the background area; The band ratio method is used to construct the ratio formula: Where R ratio It represents the ratio formula, that is, the ratio of the reflectivity at a wavelength of 680nm to the reflectivity at a wavelength of 550nm in the background area; Set the ratio threshold T yz =1.5, calculate the corrosion area ratio: Where R ratio，j′ represents the band ratio of the j′th pixel in the background area, where j′=1, 2, …, Z, Z represents the total number of pixels in the background area, Π(R ratio，j′ >T yz ) represents the indicator function, when R ratio，j′ >T yz When ∏(R ratio，j′ >T yz )=1, when R ratio，j′ ≤T yz When ∏(R ratio，j′ >T yz )=0, M represents the corrosion area ratio.

8. The method for inspection and identification of transmission towers by drones according to claim 1, characterized in that: The method for evaluating the comprehensive health status of transmission towers is: Set the corrosion area ratio threshold to M threshold =0.05, the wire deformation threshold is L threshold =0.02; When the corrosion area ratio M≤M threshold , Number of insulators in abnormal state n abbormal = 0 and the abnormality of wire deformation L xb ≤L threshold When , the evaluation result of the transmission tower is normal; When the corrosion area ratio M>M threshold 、Number of abnormal insulators abnormal ≠0 or abnormal deformation of the wire L xb >L threshold If any of these conditions occur, the evaluation result of the transmission tower will be a general alarm state; When the corrosion area ratio M>M threshold 、Number of abnormal insulators abnormal ≠0 or abnormal deformation of the wire L xb >L threshold When any two or more of the above conditions occur, the evaluation result of the transmission tower will be an emergency alarm state.

9. UAV inspection and identification device for transmission towers, characterized by: The device is used in the UAV inspection and identification method for transmission towers according to any one of claims 1 to 8: The guy wire deformation monitoring module is used to pre-set reference points on the ground. The drone hovers above the end of the guy wire of the transmission tower and uses the laser ranging function to identify the height difference to the ground reference point. The polar coordinates of the guy wire endpoint are then calculated. The guy wire length is then calculated based on the polar coordinates of the guy wire endpoint to assess the degree of guy wire deformation anomaly. The insulator detection module is used on drones equipped with a 5-channel hyperspectral imaging system. It performs orthogonal flight path scanning and high-precision imaging, extracts radiation values ​​in the 300-900nm band, locates insulators in real time using a CNN algorithm, and determines the insulator status by calculating the average radiation value in the UV band, the gradient of the partial discharge area, and the temperature difference. The corrosion assessment module extracts radiation values ​​in the 550-700nm band, converts them into reflectivity after shielding the insulator area, and calculates the corrosion area ratio based on the reflectivity data using the band ratio method. The health assessment module is used to evaluate the overall health status of transmission towers by comprehensively considering the corrosion area ratio, abnormal insulators, and abnormal deformation of guy wires.

Images