Surface wear detection method and system for extra-high voltage electric power fitting

Abstract

The invention relates to the technical field of power equipment detection, in particular to a surface wear detection method and system for extra-high voltage electric power fittings, and the method comprises the steps: continuously collecting multiple frames of images of each electric power fitting through a sensor carried by an unmanned aerial vehicle, recording a time period between each frame of image and a previous frame as a corresponding interval time period of each frame of image, the wind speed and the wind direction of the environment where the unmanned aerial vehicle is located in the interval time period and the three-axis angular speed and the three-axis acceleration of the unmanned aerial vehicle are obtained; calculating the pose disturbance degree of the interval time period, and determining the first disturbance degree of each frame of image; calculating a second interference degree and an interference coefficient of each frame of image to obtain a reference image; and adjusting a sharpening coefficient of the image sharpening algorithm, and carrying out sharpening enhancement processing and wear detection on the reference image. The detail definition and the visual quality of the image can be improved, and the accurate detection of the real wear characteristics of the surface of the electric power fitting is improved.

Description

Technical Field

[0001] This application relates to the field of power equipment testing technology, specifically to a method and system for detecting surface wear of ultra-high voltage power fittings. Background Technology

[0002] With the rapid development of ultra-high voltage power transmission projects, the requirements for the reliability of key equipment in the power grid are increasing. As the core component connecting conductors, insulators and towers, power fittings are subjected to the combined effects of high voltage, high current, strong electromagnetic fields and complex meteorological environments for a long time. Their surfaces are prone to wear, corrosion and fretting fatigue. Once the mechanical strength of power fittings decreases or the contact performance deteriorates due to wear, it may cause major accidents such as line breakage and string failure, seriously threatening the safe and stable operation of the power grid.

[0003] Power fittings are typically mounted tens of meters in the air. When using drones equipped with high-definition cameras to capture images of these fittings, the intense wind disturbances at high altitudes and the resonance of the drone's own structure cause high-frequency jitter that is difficult to completely suppress when the drone hovers and takes pictures. At the same time, the power fittings themselves are also affected by the interference from the high-altitude environment, resulting in blurred images and loss of details, which seriously affects image quality. Therefore, in order to reduce image blur, traditional image sharpening algorithms generally use a fixed sharpening coefficient, which cannot adapt to the huge quality differences between different images due to different environmental disturbances. This results in insufficient sharpening effect and inadequate detail recovery in strong jitter interference scenarios, while in stable scenarios, it is easy to introduce over-sharpening noise, producing unnatural halo artifacts, which affects the accurate detection of the true wear characteristics of the power fitting surface. Summary of the Invention

[0004] To address the aforementioned technical problems, a method and system for detecting surface wear of ultra-high voltage power fittings are provided to solve the existing issues.

[0005] The solution to the technical problem of this application is to provide a method and system for detecting surface wear of ultra-high voltage power fittings, including the following steps: In a first aspect, embodiments of this application provide a method for detecting surface wear of ultra-high voltage power fittings, the method comprising the following steps: Using the sensors on the drone, multiple frames of images of each power fitting are continuously collected. The time interval between each frame and the previous frame is recorded as the interval time interval corresponding to each frame. The wind speed and wind direction of the environment in which the drone is located during the interval time interval, as well as the three-axis angular velocity and three-axis acceleration of the drone, are obtained. For each frame of image, the energy distribution of the three-axis angular velocity in the frequency domain and the fluctuation of the three-axis acceleration in the horizontal direction are analyzed within the interval. The pose perturbation degree of the interval is calculated. Combined with the influence of wind direction on the attitude stability of the UAV within the interval and the intensity level of wind speed, the first perturbation degree of each frame of image is determined. Based on the differences in the environment represented by wind speed and direction between different time intervals of different images, the images of all power fittings are clustered. By utilizing the texture differences of the same power fitting images between clusters representing different environmental conditions, a texture mapping model is constructed. Combined with the differences in wind speed and direction between adjacent frame images, the second interference degree of each frame image is calculated. Based on the first and second interference degrees, the interference coefficient of each frame image is determined. All images of each power fitting are filtered to obtain a reference image. Based on the average level of the interference coefficient of all images of each power fitting, the sharpening coefficient of the image sharpening algorithm is adjusted. The image sharpening algorithm is then used to sharpen and enhance the reference image and perform wear detection.

[0006] Preferably, the calculation of pose perturbation degree during the time interval includes: Frequency domain analysis is performed on all angular velocities in each axial direction within the interval period to obtain the power spectral density function corresponding to each axial direction; the normalized result of the mean of the integral value of the power spectral density function corresponding to all axial directions within the preset frequency range is calculated as the high-frequency vibration quantity of the interval period. Transform all triaxial accelerations within the interval into the world coordinate system, and positively fuse the dispersion of all accelerations in the X-axis direction and the dispersion of all accelerations in the Y-axis direction within the interval to obtain the horizontal drift degree of the interval. The pose perturbation is the result of a positive fusion of high-frequency vibration and horizontal drift.

[0007] Preferably, determining the first interference level for each frame of image includes: Attitude calculations are performed on the three-axis angular velocity and three-axis acceleration to obtain the nose direction of the UAV; the absolute value of the cosine of the angle between the mean of all wind directions and the mean of all nose directions within the time interval is calculated and positively mapped onto it; The mean wind speed is calculated as the average wind speed within the interval period; the ratio between the positive mapping result and the normalized average wind speed is used as the environmental stability of the interval period. The first disturbance degree is positively correlated with the pose disturbance degree, but negatively correlated with the environmental stability degree.

[0008] Preferably, the clustering of images of all power fittings includes: After normalizing all wind speeds and wind directions within the corresponding time interval of each frame image, the normalized wind speeds and wind directions are respectively combined into wind speed sequences and wind direction sequences. All images corresponding to all power fittings are compiled into a dataset, where each frame of the image is considered a sample. The distance between any two samples in the dataset, namely the wind speed sequence and the wind direction sequence, is calculated and denoted as the first distance and the second distance, respectively. The sum of the first distance and the second distance is used as the distance metric to cluster all samples in the dataset.

[0009] Preferably, the process of constructing the texture mapping model is as follows: Extract the gray-level co-occurrence matrix of each frame image; select two samples of the same power fitting from any two clusters, define them as a sample pair, and calculate the difference in gray-level co-occurrence matrix between each sample pair as the texture difference degree; Using the first and second distances of each sample pair as independent variables and the texture difference of each sample pair as the dependent variable, regression analysis was performed on all sample pairs to construct a texture mapping model.

[0010] Preferably, the calculation of the second interference degree of each frame image includes: for each power fitting, taking the first distance and the second distance between each frame image and the previous frame as input to the texture mapping model, and taking the output result as the second interference degree corresponding to each frame image.

[0011] Preferably, determining the interference coefficient of each frame image includes: calculating the difference between each frame image and the first interference degree corresponding to the previous frame, denoted as the relative difference; selecting the maximum value of the first interference degree and the relative difference of each frame image; and multiplying it by the second interference degree as the interference coefficient of each frame image.

[0012] Preferably, the process of acquiring the reference image is as follows: for each power fitting, select the image with the smallest interference coefficient and mark it as the reference image.

[0013] Preferably, the adjusted sharpening coefficient corresponds to the reference image under each power fitting. The calculation formula is: ,in, The normalized result of the mean of the interference coefficients for all images under each power fitting. This is the preset maximum sharpening factor. This is the preset minimum sharpening factor.

[0014] Secondly, embodiments of this application also provide a surface wear detection system for ultra-high voltage power fittings, including a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, it implements the steps of the surface wear detection method for ultra-high voltage power fittings described in any one of the above-mentioned methods.

[0015] This application has at least the following beneficial effects: This application analyzes the frequency domain energy of the angular velocity in different axial directions of a UAV and the fluctuation of acceleration in the horizontal direction to calculate the pose perturbation degree at intervals. Its advantage lies in considering the high-frequency vibration and horizontal drift of the UAV itself, reflecting the overall swaying and jittering intensity exhibited by the UAV, and accurately capturing the instability of the UAV's attitude. It also calculates the first interference degree for each frame, which is beneficial because it considers the influence of wind direction and speed on the UAV's attitude stability, comprehensively representing both the UAV's own jitter and external wind disturbance, and evaluating the impact of UAV jitter on image blur. Furthermore, it clusters images of all power fittings, utilizing the texture differences of images of the same power fitting between clusters representing different environmental conditions to construct a texture mapping model. This is beneficial because it classifies images based on the differences in environmental conditions at different intervals before image acquisition, grouping images acquired under similar environmental conditions into one category, and constructing a texture mapping model based on the texture differences of images under different environmental conditions, thereby quantifying the impact of changes in environmental conditions on image texture features. The impact of environmental disturbance on image acquisition is assessed by calculating the second interference degree of each frame. This is beneficial because it reflects the environmental disturbance affecting the power fittings by analyzing changes in environmental conditions between adjacent frames and the texture differences in the power fittings themselves. Further evaluation of the impact of the power fittings' vibration on image blur is then conducted. The interference coefficient of each frame is determined, which comprehensively assesses the impact of environmental disturbance on image acquisition from two dimensions: the drone's own shaking and the power fittings' own oscillation. All images of each power fitting are filtered to obtain a benchmark image. Based on the average level of the interference coefficients of all images of each power fitting, the sharpening coefficient of the image sharpening algorithm is adjusted. The image sharpening algorithm performs sharpening enhancement and wear detection on the benchmark image. This dynamically adjusts the sharpening coefficient based on the overall environmental conditions during image acquisition, selecting the best-quality image from the acquisition for enhancement and wear detection. This adapts to the blurring differences caused by varying environmental disturbances in different images, significantly and stably improving image detail clarity and visual quality, and enhancing the accurate detection of true wear characteristics on the surface of power fittings. Attached Figure Description

[0016] The following is a detailed description of a surface wear detection method for ultra-high voltage power fittings according to the present application, with reference to the accompanying drawings.

[0017] Figure 1 A flowchart illustrating the steps of a surface wear detection method for ultra-high voltage power fittings provided in this application embodiment; Figure 2 A flowchart illustrating the steps of a method for obtaining the second interference level of each frame of an image, as provided in an embodiment of this application. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description, in conjunction with the accompanying drawings and embodiments, provides a method and system for detecting surface wear of ultra-high voltage power fittings proposed in this application. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0019] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains.

[0020] Please see Figure 1 The diagram illustrates a flowchart of a surface wear detection method for ultra-high voltage power fittings according to an embodiment of this application. The method includes the following steps: Step 1: Using the sensors on the drone, continuously acquire multiple frames of images of each power fitting. Record the time interval between each frame and the previous frame as the interval time interval corresponding to each frame. Obtain the wind speed and direction of the environment in which the drone is located during the interval time interval, as well as the drone's three-axis angular velocity and three-axis acceleration.

[0021] As an important connecting device in the power transmission system, power fittings play a role in bearing mechanical and electrical loads in the transmission lines. During their service life, they are not only affected by complex wind environment and self-weight combined load, but also by the strong electromagnetic environment of power transmission and transformation projects, which leads to severe wear. This manifests as uneven electric field distribution at the wear parts or fatigue fracture, thus causing serious power transmission accidents.

[0022] Based on the above analysis, the wind speed and direction sensors and high-definition cameras carried by the drone were used to scan different power fittings on the UHV transmission line. For each power fitting, when the drone reached the designated altitude, it hovered in the air and continuously collected images of the power fitting using the high-definition camera. The wind speed and direction sensors were used to obtain the wind speed and direction of the environment in which the drone was located in real time. In addition, the drone's three-axis angular velocity and three-axis acceleration are acquired in real time using the drone's built-in inertial measurement unit (IMU); In this embodiment, the image acquisition time interval is 1 second, the continuous acquisition duration is 20 seconds, the acquisition frequency of the wind speed and direction sensor is 10 Hz, and the acquisition frequency of the inertial measurement unit is 200 Hz. As other implementation methods, the implementer can set them according to the actual situation.

[0023] It should be noted that triaxial angular velocity refers to the angular velocity in the X, Y, and Z axes, while triaxial acceleration refers to the acceleration in the X, Y, and Z axes.

[0024] The time interval between each frame of an image and the previous frame is taken as the interval period corresponding to each frame of an image. The collected images, wind speed and direction, as well as the triaxial angular velocity and triaxial acceleration are time-aligned to obtain each frame of image corresponding to different power fittings, and all wind speeds and directions, triaxial angular velocities and triaxial accelerations and the direction of the turbine head within the time interval; The interval period is 1 second long. Since the wind speed and wind direction are collected at a frequency of 10 Hz, and the triaxial angular velocity and triaxial acceleration are collected at a frequency of 200 Hz, there are 10 wind speeds and 10 wind directions, and 200 triaxial angular velocities and triaxial accelerations within the interval period.

[0025] It should be noted that the time alignment process is a well-known technique and will not be described in detail here.

[0026] At this point, we have obtained multiple frames of images corresponding to each power fitting, as well as all wind speeds and directions, three-axis angular velocities, and three-axis accelerations within the time intervals.

[0027] Step 2: For each frame of image, analyze the energy distribution of the three-axis angular velocity in the frequency domain and the fluctuation of the three-axis acceleration in the horizontal direction during the interval period, calculate the pose perturbation degree during the interval period, and determine the first perturbation degree of each frame of image by combining the influence of wind direction on the attitude stability of the UAV during the interval period and the intensity level of wind speed.

[0028] Due to the complexity of the working environment of power fittings, their surfaces may have a lot of dirt, oxidation and other problems, which can interfere with wear detection. Therefore, it is necessary to compare the differences between images of different power fittings under multiple similar working environments. However, the working environment of power fittings is often tens of meters high, which makes drones and power fittings very susceptible to environmental influences. Environmental changes can cause the acquired images to be blurry and lack detail clarity, affecting the accuracy of wear detection of power fittings. Therefore, image enhancement processing is required.

[0029] Secondly, drones are subject to significant wind interference during high-altitude flight. Existing drone systems typically utilize IMUs and flight control systems to achieve dynamic compensation control, aiming to ensure the stability of the drone's attitude as much as possible. However, the purpose of dynamic compensation by the flight control system is to ensure the overall position and attitude of the drone remain within a safe range and suppress large-scale disturbances. High-frequency jitter that does not affect flight safety cannot be completely eliminated. Furthermore, due to the mechanical delays inherent in the drone's motors, propellers, and fuselage, it is difficult to eliminate these jitters. Therefore, when acquiring images, the drone needs to hover near power fittings. Even minor disturbances caused by wind changes will still result in some jitter during hovering. This jitter does not affect the drone's flight safety but reduces the clarity of the acquired images, blurring abnormal wear features and affecting the accuracy of wear detection.

[0030] Based on the above analysis, for each frame of image corresponding to a time interval, the high-frequency vibration is calculated based on the frequency domain variation of the triaxial angular velocity within that time interval. Specifically: Frequency domain analysis was performed on all angular velocities in each axial direction within the time interval to obtain the power spectral density function corresponding to each axial direction. In this embodiment, the power spectral density functions in the X-axis, Y-axis and Z-axis directions are obtained by fast Fourier transform. The fast Fourier transform and the power spectral density function are well-known techniques and will not be described in detail here.

[0031] The normalized result of the mean of the integral values ​​of the power spectral density function corresponding to all axial directions within the preset frequency range during the interval period is used as the high-frequency vibration quantity during the interval period. In this embodiment, 2Hz~20Hz is a common high-frequency vibration band in many mechanical systems, corresponding to the natural frequency of components or the vibration mode caused by operation. Therefore, the preset frequency range is 2Hz~20Hz. As other implementations, implementers can set it according to the actual situation. Secondly, the sigmoid function is used for normalization processing. The sigmoid function is a well-known technology and will not be described in detail here. As other implementations, implementers can use other methods of the prior art, such as the tanh function, etc. This embodiment does not impose any special restrictions on this.

[0032] It should be noted that the greater the amount of high-frequency vibration, the more severe the high-frequency shaking of the drone's body when it is hovering.

[0033] Secondly, based on the triaxial acceleration within the time interval, the horizontal drift of the UAV is analyzed, and the horizontal drift degree is calculated, specifically: Transform all triaxial accelerations within the interval into the world coordinate system, and positively fuse the dispersion of all accelerations in the X-axis direction and the dispersion of all accelerations in the Y-axis direction within the interval to obtain the horizontal drift degree of the interval. In this embodiment, the degree of dispersion is measured by calculating the variance of all accelerations in the X-axis direction and the variance of all accelerations in the Y-axis direction within the time interval. As another implementation, the implementer can use other methods of the prior art, such as the coefficient of variation, etc., and there are no special restrictions on this. The specific process of forward fusion is as follows: the square root of the sum of the degree of dispersion between the X-axis direction and the Y-axis direction is normalized and used as the horizontal drift degree of the time interval. The normalization process is as follows: the maximum acceleration in the X-axis direction and the Y-axis direction is obtained, and the ratio of the square root to the maximum acceleration is used as the horizontal drift degree. As another implementation, the forward fusion process can also be carried out by calculating the product of the degree of dispersion between the X-axis direction and the Y-axis direction.

[0034] It should be noted that the greater the degree of dispersion, the more violent the acceleration fluctuation of the UAV in the X-axis or Y-axis direction, reflecting the more unstable the movement of the UAV in that axis and the more frequent and violent the shaking. The greater the degree of horizontal drift, the worse the overall stability of the UAV on the entire horizontal plane and the weaker its position-keeping ability, reflecting the displacement of the UAV on the horizontal plane.

[0035] Furthermore, based on the high-frequency vibration and horizontal drift, the pose perturbation is determined, specifically as follows: The high-frequency vibration quantity and the horizontal drift degree are positively fused to obtain the pose perturbation degree at intervals. In this embodiment, the specific process of forward fusion is as follows: based on a preset first weight and a preset second weight, the high-frequency vibration amount and the horizontal drift degree are weighted and summed to obtain the pose perturbation degree of the interval period; wherein, the preset first weight is greater than the preset second weight, and the sum of the preset first weight and the preset second weight is 1, wherein the preset first weight is 0.7 and the preset second weight is 0.3. As other implementation methods, the implementer can set it according to the actual situation. As other implementation methods, the forward fusion process can also use the product of the high-frequency vibration amount and the horizontal drift degree as the pose perturbation degree of the interval period.

[0036] It should be noted that the greater the pose disturbance, the more severe the overall shaking and trembling of the UAV before it hovers to collect images.

[0037] Furthermore, the consistency between wind direction and the drone's nose direction during the time intervals, as well as changes in wind speed, are analyzed to calculate environmental stability, specifically: Attitude calculations are performed on the three-axis angular velocity and three-axis acceleration to obtain the nose direction of the UAV; It should be noted that the attitude calculation process is a well-known technique and will not be described in detail here. The yaw angle can be obtained after attitude calculation, and the direction of the nose represents the yaw angle.

[0038] Calculate the absolute value of the cosine of the angle between the mean of all wind directions and the mean of all nose directions within the time interval, and then perform a positive mapping on it. In this embodiment, the specific process of positive mapping is as follows: the sum of the absolute value and a preset value greater than 0 is taken as the result of positive mapping, where the preset value greater than 0 is 1. Through the process of positive mapping, the result of positive mapping is made greater than 0. As another implementation, positive mapping can be performed using an exponential function. Let's assume the absolute value of the cosine is denoted as... ,but The result is taken as the result of the positive mapping, where, It is an exponential function with the natural constant as the base.

[0039] The mean wind speed is calculated as the average wind speed within the interval period; the ratio between the positive mapping result and the normalized average wind speed is used as the environmental stability of the interval period. In this embodiment, the normalization process is as follows: the ratio between the average wind speed and the maximum value of all collected wind speeds is used as the result of normalization. As for other implementation methods, the implementer may use other methods of existing technology, such as Z-score normalization, etc. This embodiment does not impose any special restrictions on this.

[0040] It should be noted that the larger the positive mapping result, the closer the wind direction is to the direction of the drone's nose, reflecting that the wind direction has less disturbance to the drone's attitude. The smaller the average wind speed, the weaker the wind, the less lateral disturbance to the drone. The greater the environmental stability, the better the overall stability of the wind disturbance environment in which the drone is located, which is more conducive to image acquisition.

[0041] Therefore, based on the pose perturbation degree and the environmental stability, the first perturbation degree is determined as follows: The first degree of disturbance in each frame is positively correlated with the pose perturbation degree, but negatively correlated with the environmental stability. It should be noted that a positive correlation means that the dependent variable increases as the independent variable increases and decreases as the independent variable decreases; a negative correlation means that the dependent variable decreases as the independent variable increases and increases as the independent variable decreases.

[0042] In this embodiment, the ratio of pose perturbation degree to environmental stability for each frame of image at corresponding time intervals is used as the first perturbation degree for each frame of image. In other embodiments, the implementer can calculate the difference between pose perturbation degree and environmental stability. To avoid the difference being less than or equal to 0, the result of an exponential function with the natural constant as the base and the difference as the exponent is used as the first perturbation degree for each frame of image.

[0043] It should be noted that during drone hovering, vibrations caused by environmental factors can lead to slight displacements in the three axes of the drone and changes in its tilt angles along those axes, resulting in greater attitude disturbance. Secondly, the impact of wind on drone vibration is mainly due to changes in wind speed and direction. When the angle between the wind direction and the drone's nose is larger, and the wind speed is higher, the tilt angle fluctuations of the drone will be more severe, resulting in lower environmental stability. This increases the likelihood of blurred images. Therefore, a higher degree of first-order disturbance indicates more severe drone vibrations caused by wind changes, and a greater probability of blurred images.

[0044] At this point, the first level of interference for each frame of the image is obtained.

[0045] Step 3: Based on the differences in the environment represented by wind speed and direction between different time intervals of different images, cluster all images of power fittings. Utilize the texture differences of the same power fitting images between clusters representing different environmental conditions to construct a texture mapping model. Combine the differences in wind speed and direction between adjacent frame images to calculate the second interference degree of each frame image.

[0046] Furthermore, in high-altitude environments, not only are drones affected by environmental interference, but the power fittings under test are also subject to significant environmental influences. Under the influence of wind, the position and attitude of the power fittings also experience some shaking, which in turn affects image acquisition, resulting in significant differences in the texture and details of the power fittings in multiple consecutively acquired images.

[0047] Based on the above analysis, by analyzing the differences in wind speed and direction within the corresponding time intervals of different images, all images corresponding to all power fittings are classified as follows: After normalizing all wind speeds and wind directions within the corresponding time interval of each frame image, the normalized wind speeds and wind directions are respectively combined into wind speed sequences and wind direction sequences. In this embodiment, the maximum-minimum normalization method is used for normalization processing. The maximum-minimum normalization method is a well-known technique and will not be described in detail here.

[0048] All images corresponding to all power fittings are compiled into a dataset, where each frame of the image is considered a sample. Calculate the distance between any two wind speed sequences within the dataset, denoted as the first distance; Calculate the distance between the wind direction sequences of any two samples within the dataset, and denote it as the second distance; In this embodiment, the distance is measured by calculating the Manhattan distance between the wind speed sequences of any two samples and the Manhattan distance between the wind direction sequences of any two samples. The calculation of the Manhattan distance is a well-known technique and will not be described in detail here. As other implementation methods, implementers may use other methods of the prior art, such as Euclidean distance, etc. This embodiment does not impose any special restrictions on this.

[0049] The sum of the first distance and the second distance is used as the distance metric to cluster all samples in the dataset; In this embodiment, the K-means clustering algorithm is used for clustering. The K-means clustering algorithm is a well-known technology and will not be described in detail here. As other implementation methods, implementers may use other methods of existing technology, such as hierarchical clustering algorithms. This embodiment does not impose any special restrictions on this.

[0050] It should be noted that the same cluster represents similar environmental conditions.

[0051] Secondly, based on the texture differences between images of the same power fittings in clusters representing different environmental conditions, a texture mapping model is constructed. This model, combined with environmental changes during the acquisition of adjacent frames of images for the same power fitting, calculates the second interference degree. The flowchart of the method for obtaining the second interference degree for each frame provided in this embodiment is shown below. Figure 2 As shown, it specifically includes: Each frame of the image is converted to grayscale to obtain a grayscale image, and the grayscale co-occurrence matrix is ​​extracted. In this embodiment, a weighted average method is used for grayscale processing. The weighted average method and the grayscale co-occurrence matrix are well-known techniques and will not be described in detail here.

[0052] From any two clusters, select two samples of the same power fitting, define them as a sample pair, and calculate the difference in the gray-level co-occurrence matrix between each sample pair as the texture difference degree. In this embodiment, the KL divergence of the gray-level co-occurrence matrix between two samples is calculated as the texture difference. KL divergence is a well-known technique and will not be described in detail here. As other implementation methods, implementers may use other methods of the prior art, such as JS divergence, etc. This embodiment does not impose any special restrictions on this.

[0053] Using the first and second distances of each sample pair as independent variables and the texture difference of each sample pair as the dependent variable, a multivariate regression analysis was performed on all sample pairs to construct a texture mapping model. It should be noted that multiple regression analysis is a well-known technique and will not be elaborated upon here. The multiple regression function obtained through multiple regression analysis is the texture mapping model.

[0054] For each power fitting, the first distance and the second distance between each frame of image and the previous frame are used as inputs to the texture mapping model, and the output is used as the second interference degree corresponding to each frame of image. It should be noted that the greater the second interference degree, the more likely the power fitting itself will experience violent swinging and posture changes between two adjacent frames, resulting in significant changes in image texture and reflecting the more significant the vibration impact of environmental wind disturbance on the power fitting.

[0055] Thus, the second interference degree corresponding to each frame of the image is obtained.

[0056] Step 4: Based on the first interference degree and the second interference degree, determine the interference coefficient of each frame image, filter all images of each power fitting to obtain the reference image; adjust the sharpening coefficient of the image sharpening algorithm according to the average level of the interference coefficient of all images of each power fitting, and use the image sharpening algorithm to perform sharpening enhancement processing and wear detection on the reference image.

[0057] Furthermore, based on the first interference degree and the second interference degree, the interference coefficient is determined as follows: Calculate the difference between the first interference degree of each frame and the corresponding first interference degree of the previous frame, and record it as the relative difference. Select the maximum value of the first interference degree and the relative difference of each frame, and multiply it by the second interference degree as the interference coefficient of each frame. In this embodiment, the absolute value of the difference between each frame of image and the first interference degree corresponding to the previous frame is calculated and denoted as the relative difference.

[0058] It should be noted that the greater the relative difference, the more abrupt the drone's state has changed; the greater the first interference degree, the more the drone is in a state of continuous and severe shaking. By selecting the maximum value of the two, the more severe environmental disturbance factors in the two unstable modes of sudden and continuous states can be captured. The greater the maximum value, the more unstable the drone is in. The greater the second interference degree, the more violently the power fitting itself is swinging. The greater the obtained interference coefficient, the worse the stability of the drone is when the image frame is acquired, and the more violent the vibration of the power fitting itself is, resulting in a higher blurriness of the image, making it the least suitable for subsequent wear detection and comparative analysis.

[0059] Based on the interference coefficient, all images corresponding to each power fitting are filtered, specifically as follows: For each power fitting, select the image with the smallest interference coefficient and mark it as the reference image; Because the shaking of drones and high-altitude hardware causes image blurring, image sharpening is necessary to make the blurred images clearer and improve image quality. Among these methods, the reference image is least affected by environmental interference and contains the richest detail information. Sharpening on this high-quality basis can most effectively and safely enhance the edge and texture features that characterize wear, fundamentally improving the accuracy and reliability of wear detection.

[0060] However, traditional image sharpening algorithms generally use a fixed sharpening coefficient. The larger the sharpening coefficient, the stronger the sharpening effect, the more obvious the image edges are enhanced, and the more prominent the details will be. However, an excessively high sharpening coefficient may cause overshooting, that is, unnatural halos or bright edges appear around the edges, and at the same time, it will amplify image noise. If the sharpening coefficient is too small, the enhancement of image edges and details is not obvious, and it cannot effectively improve blur or low contrast problems, resulting in limited improvement in overall image clarity.

[0061] Based on the above analysis, the sharpening coefficient of the image sharpening algorithm is adjusted using the interference coefficient, specifically as follows:

[0062] in, The adjusted sharpening factor corresponds to the reference image for each power fitting. The normalized result of the mean of the interference coefficients for all images under each power fitting. This is the preset maximum sharpening factor. This is the preset minimum sharpening factor; In this embodiment, the preset minimum sharpening coefficient is 0.3 and the preset maximum sharpening coefficient is 1.2. This is to avoid excessive amplification of noise when the adjusted sharpening intensity is too high, which would result in a worse image quality. The sharpening coefficient range is 0.3 to 1.2. As for other implementation methods, the implementer can set it according to the actual situation. Secondly, the sigmoid function is used for normalization processing. The sigmoid function is a well-known technology and will not be described in detail here.

[0063] It should be noted that the greater the environmental interference to the power fittings, the higher the interference coefficient and the larger the average value, reflecting a very harsh environment. In this case, the image will be more blurred, and a larger sharpening factor is needed to improve the clarity of details.

[0064] Furthermore, based on the adjusted sharpening coefficient, the reference image is sharpened to detect wear on power fittings, specifically as follows: Based on the adjusted sharpening coefficient, the reference image of each power fitting is sharpened and enhanced using an image sharpening algorithm to obtain the enhanced reference image. In this embodiment, the Laplacian sharpening algorithm is used for sharpening enhancement. The Laplacian sharpening algorithm is a well-known technology and will not be described in detail here.

[0065] Image difference is performed between the grayscale images corresponding to the enhanced reference images of each power fitting and the rest of the power fittings to obtain the difference image; Collect all images of each power fitting that were captured multiple times during a historical period to form a historical set. For each image in the historical set corresponding to each power fitting, perform image difference between the grayscale image and the enhanced reference image to obtain the difference image. It should be noted that the frame difference method is used for image differencing. The frame difference method is a well-known technique and will not be described in detail here.

[0066] Pixels with gray values ​​greater than a preset threshold in the differential image are marked, and the area where pixels that are marked simultaneously in all differential images are located is the wear area. In this embodiment, the process of determining the preset threshold is as follows: the Otsu thresholding method is used to obtain the segmentation threshold of the gray values ​​of all pixels in the difference image, which is used as the preset threshold. The Otsu thresholding method is a well-known technique and will not be described in detail here.

[0067] It should be noted that, due to the irreversibility and uniqueness of wear on power fittings, if a pixel in the differential image has a gray value greater than a preset threshold, it indicates that the gray value difference of the same pixel between the two frames that generated the differential image is large, and the likelihood of wear at that location is greater.

[0068] Based on the same inventive concept as the above method, this application embodiment also provides a surface wear detection system for ultra-high voltage power fittings, including a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, it implements the steps of any one of the above-described surface wear detection methods for ultra-high voltage power fittings.

[0069] It should be understood that, although Figure 1 The steps in the flowchart are shown sequentially as indicated by the arrows, but these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order in which these steps are executed, and they can be performed in other orders. Figure 1 At least some of the steps in the process may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be executed in turn or alternately with other steps or at least some of the sub-steps or stages of other steps.

[0070] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0071] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application. Therefore, any simple modifications, equivalent changes, and alterations made to the above embodiments based on the technical essence of this application, without departing from the content of the technical solution of this application, shall fall within the protection scope of the technical solution of this application.

Claims

1. A method for detecting surface wear of ultra-high voltage power fittings, characterized in that, The method includes the following steps: Using the sensors on the drone, multiple frames of images of each power fitting are continuously collected. The time interval between each frame and the previous frame is recorded as the interval time interval corresponding to each frame. The wind speed and wind direction of the environment in which the drone is located during the interval time interval, as well as the three-axis angular velocity and three-axis acceleration of the drone, are obtained. For each frame of image, the energy distribution of the three-axis angular velocity in the frequency domain and the fluctuation of the three-axis acceleration in the horizontal direction are analyzed within the interval. The pose perturbation degree of the interval is calculated. Combined with the influence of wind direction on the attitude stability of the UAV within the interval and the intensity level of wind speed, the first perturbation degree of each frame of image is determined. Based on the differences in the environment represented by wind speed and direction between different time intervals of different images, the images of all power fittings are clustered. By utilizing the texture differences of the same power fitting images between clusters representing different environmental conditions, a texture mapping model is constructed. Combined with the differences in wind speed and direction between adjacent frame images, the second interference degree of each frame image is calculated. Based on the first and second interference degrees, the interference coefficient of each frame image is determined. All images of each power fitting are filtered to obtain a reference image. Based on the average level of the interference coefficient of all images of each power fitting, the sharpening coefficient of the image sharpening algorithm is adjusted. The image sharpening algorithm is then used to sharpen and enhance the reference image and perform wear detection.

2. The surface wear detection method for ultra-high voltage power fittings as described in claim 1, characterized in that, The pose perturbation degree calculated over the time interval includes: Frequency domain analysis is performed on all angular velocities in each axial direction within the interval period to obtain the power spectral density function corresponding to each axial direction; the normalized result of the mean of the integral value of the power spectral density function corresponding to all axial directions within the preset frequency range is calculated as the high-frequency vibration quantity of the interval period. Transform all triaxial accelerations within the interval into the world coordinate system, and positively fuse the dispersion of all accelerations in the X-axis direction and the dispersion of all accelerations in the Y-axis direction within the interval to obtain the horizontal drift degree of the interval. The pose perturbation is the result of a positive fusion of high-frequency vibration and horizontal drift.

3. The surface wear detection method for ultra-high voltage power fittings as described in claim 1, characterized in that, Determining the first interference level for each frame of image includes: Attitude calculations are performed on the three-axis angular velocity and three-axis acceleration to obtain the nose direction of the UAV; the absolute value of the cosine of the angle between the mean of all wind directions and the mean of all nose directions within the time interval is calculated and positively mapped onto it; The mean wind speed is calculated as the average wind speed within the interval period; the ratio between the positive mapping result and the normalized average wind speed is used as the environmental stability of the interval period. The first disturbance degree is positively correlated with the pose disturbance degree, but negatively correlated with the environmental stability degree.

4. The surface wear detection method for ultra-high voltage power fittings as described in claim 1, characterized in that, The clustering of images of all power fittings includes: After normalizing all wind speeds and wind directions within the corresponding time interval of each frame image, the normalized wind speeds and wind directions are respectively combined into wind speed sequences and wind direction sequences. All images corresponding to all power fittings are compiled into a dataset, where each frame of the image is considered a sample. The distance between any two samples in the dataset, namely the wind speed sequence and the wind direction sequence, is calculated and denoted as the first distance and the second distance, respectively. The sum of the first distance and the second distance is used as the distance metric to cluster all samples in the dataset.

5. The surface wear detection method for ultra-high voltage power fittings as described in claim 4, characterized in that, The process of constructing the texture mapping model is as follows: Extract the gray-level co-occurrence matrix of each frame image; select two samples of the same power fitting from any two clusters, define them as a sample pair, and calculate the difference in gray-level co-occurrence matrix between each sample pair as the texture difference degree; Using the first and second distances of each sample pair as independent variables and the texture difference of each sample pair as the dependent variable, regression analysis was performed on all sample pairs to construct a texture mapping model.

6. The surface wear detection method for ultra-high voltage power fittings as described in claim 5, characterized in that, The calculation of the second interference degree of each frame image includes: for each power fitting, taking the first distance and the second distance between each frame image and the previous frame as the input of the texture mapping model, and taking the output result as the second interference degree corresponding to each frame image.

7. The surface wear detection method for ultra-high voltage power fittings as described in claim 1, characterized in that, Determining the interference coefficient of each frame image includes: calculating the difference between the first interference degree of each frame image and the corresponding first interference degree of the previous frame, denoted as the relative difference; selecting the maximum value of the first interference degree and the relative difference of each frame image; and multiplying it by the second interference degree as the interference coefficient of each frame image.

8. The surface wear detection method for ultra-high voltage power fittings as described in claim 1, characterized in that, The process of acquiring the reference image is as follows: for each power fitting, select the image with the smallest interference coefficient and mark it as the reference image.

9. The surface wear detection method for ultra-high voltage power fittings as described in claim 1, characterized in that, The adjusted sharpening factor corresponds to the reference image for each power fitting. The calculation formula is: ,in, The normalized result of the mean of the interference coefficients for all images under each power fitting. This is the preset maximum sharpening factor. This is the preset minimum sharpening factor.

10. A surface wear detection system for ultra-high voltage power fittings, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the surface wear detection method for ultra-high voltage power fittings as described in any one of claims 1-9.

Images