A power line identification method, system, device and storage medium

Abstract

The application provides a power line identification method, system, device and storage medium, and belongs to the technical field of power system detection. The method comprises the following steps: acquiring a to-be-detected image containing a power line; constructing a phase filter stretching kernel function based on a phase shift kernel function and a convolution theorem; inputting the to-be-detected image into the phase filter stretching kernel function, using the phase filter stretching kernel function to perform stretching transformation on the to-be-detected image, converting the to-be-detected image in the frequency domain into an angle image in the spatial domain, and extracting a contour edge in the angle image; superimposing the to-be-detected image and the angle image containing the contour edge to obtain a superimposed image, using relative total variation (RTV) to perform sharpening enhancement processing on the superimposed image, and identifying the power line. The application can improve the identification accuracy of the power line.

Description

Technical Field

[0001] This invention belongs to the field of power system detection technology, specifically relating to a power line identification method, system, device, and storage medium. Background Technology

[0002] Power line inspection is an important part of routine power grid maintenance and plays a crucial role in ensuring the healthy operation of the power grid. However, many transmission lines are located in mountainous areas with inconvenient transportation, making manual inspection inefficient. While using drones for inspection can overcome these shortcomings, the real-time transmitted images lack clarity and are prone to missing faults. Therefore, deploying fault identification on drones to assist manual judgment is a better solution.

[0003] Segmenting electric field lines from images is fundamental to electric field defect detection. Currently, electric field extraction can be divided into traditional methods and data-based deep learning methods. Some researchers use the Canny operator to obtain edge maps and then use Hough transform to extract electric field lines from the edge maps. However, the Canny algorithm can lead to oversegmentation and inaccurate results when image segmentation is affected by noise. Other researchers have proposed an improved anchor box generation strategy for line detection using the R-CNN (Region-Convolutional Neural Network) algorithm. However, when using R-CNN for image segmentation, it is not sensitive to image edges and results in excessive undersegmentation. Phase stretch transform (PST) is developed based on the time stretch transform (TST) of analog signals. Currently, PST is used in digital image edge feature detection, feature enhancement of visually impaired images, and digital image compression. However, because the phase shift of each spectrum in PST is equal, the detected edges contain a lot of noise. In particular, isolated, fragmented high-frequency noise points are mistakenly identified as high-frequency components in the image and are preserved, which makes subsequent thresholding processing difficult.

[0004] In summary, due to the small size of electric power lines and their weak reflected energy during imaging, UAV aerial images of electric power lines are highly susceptible to interference from environmental linear elements and noise. This results in severe noise in the electric power line images, making them weak, blurry, and difficult to distinguish. Consequently, the aforementioned detection algorithms are not ideal for extracting electric power lines. Summary of the Invention

[0005] To overcome the shortcomings of unsatisfactory power line extraction results, this invention provides a power line identification method, comprising the following steps:

[0006] Acquire the image to be detected, which contains power lines;

[0007] Based on the phase shift kernel function and the convolution theorem, the phase shift kernel function is subjected to phase stretching transformation and convolution operation to construct the phase filter stretching kernel function;

[0008] The image to be detected is input into the phase filter stretching kernel function. The phase filter stretching kernel function is used to stretch the image to be detected, transforming the image in the frequency domain into an angle image in the spatial domain, and extracting the contour edges in the angle image.

[0009] The image to be detected is superimposed with an angle image containing contour edges to obtain a superimposed image. The superimposed image is then sharpened and enhanced using relative total variation (RTV), and electric field lines are identified from the sharpened and enhanced image.

[0010] Preferably, the construction of the phase filter stretching kernel function includes the following steps:

[0011] An angle transformation is performed after convolving the spatial domain complex matrix corresponding to the frequency domain complex stretching kernel of the phase shift kernel function with the image, resulting in a phase-stretched angle image.

[0012] The phase-stretched angle image is simplified, and the nonlinear phase distortion kernel function is transformed and normalized in polar coordinates to obtain a phase filter stretching kernel function with phase stretching intensity parameter S and distortion parameter W.

[0013] Preferably, before inputting the image to be detected into the phase filter stretching kernel function, the method further includes grayscale processing and filtering processing of the image to be detected; the grayscale processing specifically involves taking the weighted average of the R component, G component and B component of the pixel in the image to be detected as the grayscale value of the grayscale image; the filtering processing involves taking a weighted average of all pixel values ​​in the image to be detected, wherein the pixel value of each pixel is obtained by weighting its own value and the values ​​of other pixels in its neighborhood.

[0014] Preferably, the phase filter stretching kernel function is:

[0015]

[0016] In the formula, S is the phase tensile strength parameter, W is the torsion parameter, and r is the polar radius when transforming to polar coordinates.

[0017] Preferably, the angle image is:

[0018]

[0019] in, The phase filter stretching kernel function is defined by IFFT2, which is the inverse two-dimensional Fourier transform. The result is the inverse Fourier transform of the input image B(x, y).

[0020] This invention also provides a power line identification system, comprising:

[0021] The image acquisition module is used to acquire the image to be detected, which contains power lines;

[0022] The kernel function construction module is used to perform phase stretching transformation and convolution operations on the phase shift kernel function based on the phase shift kernel function and the convolution theorem to construct the phase filter stretching kernel function;

[0023] The contour extraction module is used to input the image to be detected into the phase filter stretching kernel function, use the phase filter stretching kernel function to stretch the image to be detected, transform the image to be detected in the frequency domain into an angle image in the spatial domain, and extract the contour edges in the angle image.

[0024] The power line recognition module is used to overlay the image to be detected with an angle image containing contour edges to obtain an overlay image, and to use relative total variation (RTV) to sharpen and enhance the overlay image, and to identify power lines from the sharpened and enhanced image.

[0025] The present invention also provides a computer device including a memory and a processor; the memory stores a computer program, and the processor is used to run the computer program in the memory to perform the power line identification method.

[0026] The present invention also provides a computer-readable storage medium storing a computer program adapted for loading by a processor to execute the power line identification method.

[0027] The power line identification method, system, device, and storage medium provided by this invention have the following beneficial effects:

[0028] This invention constructs a phase filter stretching kernel function model and uses it to perform stretching transformation on the image to be detected captured by a UAV. This transforms the image to be detected in the frequency domain into an angular image in the spatial domain, which better reflects linear targets in the spatial domain. Furthermore, the superimposed image is sharpened and enhanced using relative total variational RTV, which optimizes the separation of main structures and weak, fragmented textures. This results in a sharpened image that better highlights the main edge contours and structural textures, suppresses weak, fragmented textures, effectively removes noise, and improves the accuracy of power line recognition. Attached Figure Description

[0029] To more clearly illustrate the embodiments and design schemes of the present invention, the accompanying drawings required for this embodiment will be briefly described below. The drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0030] Figure 1 This is a flowchart of an embodiment of the present invention;

[0031] Figure 2 The experimental results correspond to different values ​​of K and tensile strength S; among which, Figure 2 (a) is the original image; Figure 2 (b) is k=2, W=10, S=10 3 The experimental results corresponding to the selected values; Figure 2 (c) is k=2, W=10, S=10 5 The experimental results corresponding to the selected values; Figure 2 The (d) values ​​are k=3, W=10, and S=10. 5 The experimental results corresponding to the selected values; Figure 2 (e) is k=4, W=10, S=10 6 The experimental results corresponding to the selected values; Figure 2 (f) is k=7, W=10, S=10 6 The experimental results corresponding to the selected values;

[0032] Figure 3 The image to be detected; where, Figure 3 (a), (b), (c), and (d) are different original images;

[0033] Figure 4 For S = 1.48 and W = 10.24, various algorithms were used to... Figure 3 (a) Comparison of recognition results; where, Figure 4 (a) is the recognition result when PST is used and S = 1.48 and W = 10.24. Figure 4 (b) shows the recognition results using Canny; Figure 4 (c) shows the recognition result using R-CNN; Figure 4 (d) represents the algorithm of this invention when k=3, W=10, S=10 5 The corresponding recognition result image when the value is obtained;

[0034] Figure 5 To use various algorithms Figure 3 (b) Comparison of recognition results; where, Figure 5 (a) is the recognition result when PST is used and S = 1.48 and W = 10.24. Figure 5 (b) shows the recognition results using Canny; Figure 5 (c) shows the recognition result using R-CNN; Figure 5 (d) represents the algorithm of this invention when k=3, W=10, S=10 5 The corresponding recognition result image when the value is obtained;

[0035] Figure 6 To use various algorithms Figure 3 (c) Comparison of recognition results; where, Figure 6 (a) is the recognition result when PST is used and S = 1.48 and W = 10.24. Figure 6 (b) shows the recognition results using Canny; Figure 6 (c) shows the recognition result using R-CNN; Figure 6 (d) represents the algorithm of this invention when k=3, W=10, S=10 5 The corresponding recognition result image when the value is obtained;

[0036] Figure 7 To use various algorithms Figure 3 A comparison chart of the recognition results for (d); among which, Figure 7 (a) is the recognition result when PST is used and S = 1.48 and W = 10.24. Figure 7 (b) shows the recognition results using Canny; Figure 7 (c) shows the recognition result using R-CNN; Figure 7 (d) represents the algorithm of this invention when k=3, W=10, S=10 5 The corresponding recognition result image when the value is obtained;

[0037] Figure 8 For the evaluation index P PA Numerical value;

[0038] Figure 9 Evaluation index M MPA Numerical value;

[0039] Figure 10 Evaluation index M MIoU Numerical value. Detailed Implementation

[0040] To enable those skilled in the art to better understand and implement the technical solutions of the present invention, the present invention will be described in detail below with reference to the accompanying drawings and specific embodiments. The following embodiments are only used to more clearly illustrate the technical solutions of the present invention and should not be construed as limiting the scope of protection of the present invention.

[0041] In the description of this invention, it should be understood that the terms "center," "longitudinal," "lateral," "length," "width," "thickness," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," "outer," "axial," "radial," and "circumferential" indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing the technical solution of this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this invention.

[0042] Furthermore, the terms "first," "second," etc., are used for descriptive purposes only and should not be construed as indicating or implying relative importance. In the description of this invention, it should be noted that, unless otherwise explicitly specified or limited, the terms "connected" or "linked" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances. In the description of this invention, unless otherwise stated, "a plurality of" means two or more, which will not be elaborated further here.

[0043] Example

[0044] This invention provides a method for power line identification, specifically as follows: Figure 1 As shown, it includes the following steps:

[0045] Step 1: Obtain the image to be detected, which contains power lines.

[0046] Step 2: Based on the phase shift kernel function and the convolution theorem, perform phase stretching transformation and convolution operations on the phase shift kernel function to construct the phase filter stretching kernel function.

[0047] In 2015, MHAsghari and B. Jalali proposed a digital image transform inspired by physical phenomena, called Phase Stretch Transform (PST). This transform simulates the propagation of electromagnetic waves in a diffractive medium with a distorted dispersive dielectric function. PST uses an all-pass phase filter H(ω) = exp{jβ(ω)} with specific frequency-dependent divergence to simulate the diffraction process, where j is the imaginary part, ω is the frequency, and β(ω) is the group delay of the phase. As an S-shaped linear or sublinear function, it can reshape the complex field of the signal before analog signal sampling and digitization, thereby compressing the analog signal bandwidth without increasing the signal duration in the time domain, i.e., reducing the time-bandwidth product (TBP). This solves two inherent problems of the traditional Nyquist uniform sampling theorem: First, given a sampling rate, traditional Nyquist uniform sampling can only capture twice the information of the signal's maximum frequency component; second, when the analog signal has redundancy, the traditional Nyquist uniform sampling process will result in a much larger number of samples than necessary (because the portion of the signal below the Nyquist frequency is oversampled).

[0048] The Phase Stretch Transform (PST), applied before analog-to-digital converters (ADCs), can also be used in the field of digital signal processing. Its process for digital image edge detection is as follows: First, the original image is smoothed by a local low-pass filter kernel function. Then, a phase operation of a nonlinear frequency function is performed in the frequency domain, which is called the Phase Stretch (Divergence) Transform (PST). Finally, edge detection is achieved through post-processing of thresholding and morphological filtering.

[0049] The mathematical model for the frequency domain phase stretch transform (PST) is as follows:

[0050]

[0051] Where A(m,n) represents the angle image, "∠" represents the angle taking operation, B(m,n) represents the original input image, FFT2 and IFFT2 represent the two-dimensional fast Fourier transform and inverse transform, respectively, and (u,v) represents the frequency variable. It is the frequency response of a locally smoothed low-pass filter. It is a frequency-dependent nonlinear phase-distortion kernel function. It is a nonlinear function of the frequency variable.

[0052] The PST kernel function is described below:

[0053] Although arbitrary phase kernel functions can be considered in the Phase Stretch Transform (PST), research results show that the kernel phase function... The derivative of the group delay is a linear or sublinear function of the frequency variable. A simple example of such a phase kernel function is the S-shaped arctangent function. For simplicity, if we further require that this phase warping operation is isotropic in the frequency plane, and that its degree of warping depends only on the polar radius r in the o-uv frequency plane polar coordinate system, and is independent of the polar angle θ, that is, assuming that the prototype of the PST phase stretching kernel function is circularly symmetric about the frequency variable, we have:

[0054]

[0055] Where r is the polar radius in the frequency plane o-uv polar coordinate system, θ is the polar angle, and its relationship with the uv frequency variables is as follows: u = r * cosθ v = r * sinθ

[0056] ,, If required If the derivative of r is an sigmoid arctangent function, then:

[0057]

[0058] Note that the uv frequency plane after the Fourier transform of the image is a finite region, so it can be solved according to equation (3).

[0059]

[0060] In the formula, x is the Fourier transform variable.

[0061] Normalizing the phase function in equation (4), we get

[0062]

[0063] By adding the phase stretching strength S and warped W from the nonlinear twist-stretch transform to the phase function in equation (5), the final PST transform phase translation kernel function with phase stretching strength parameter S and warped parameter W is obtained.

[0064]

[0065] Among them, tan -1 (·) denotes the arctangent function, ln(·) is the natural logarithm, and r max This represents the maximum frequency radius of the uv frequency plane.

[0066] The derivation of the improved phase filter kernel function is as follows:

[0067] Let t = W·r and C = S / {W·r} in formula (6) max ·tan -1 (W·r max )-0.5ln[1+(W·r max ) 2 We can obtain:

[0068]

[0069] but

[0070]

[0071] set up:

[0072]

[0073] get

[0074] A(m,n)=∠IFFT2{D(m,n)·FFT2[B(m,n)]} (11)

[0075] A(m,n)=∠IFFT2{FFT2[E(m,n)]·FFT2[B(m,n)]} (12)

[0076] According to the convolution theorem, it is obvious that:

[0077]

[0078] Therefore, we get:

[0079] E(x,y)*B(x,y)=IFFT2{FFT2[E(m,n)]·FFT2[B(m,n)]} (14)

[0080] A(m,n)=∠E(x,y)*B(x,y) (15)

[0081] Where B(x,y) is the input image, E(x,y) is the spatial matrix corresponding to the phase stretching kernel after calculation, and E(m,n) is the result of the inverse Fourier transform of D(m,n).

[0082] The above results indicate that taking an angle after phase stretching transformation is equivalent to convolving the spatial domain "complex matrix" corresponding to the "complex stretching kernel" in the frequency domain with the image and then "taking" the angle. This is equivalent to the gradient of the input image being "reflected" into the angle image formed after "taking" the angle through convolution. Therefore, the theoretical basis for the edge extraction function of the "angle image" after phase stretching may lie in this. Different selections of parameters W and S may result in different proportions of the image gradient reflected in the angle image, thus affecting the edge detection effect.

[0083] From formula (1) and the principle that edge extraction can be performed using the angle information of the PST angle image A(m,n), we can know that:

[0084]

[0085] Simplified to:

[0086]

[0087] set up

[0088] In the formula, A(x) is the simplified angle image, and B(x) is the simplified form of the original input image. The result is the inverse Fourier transform of B(x). It is a frequency-dependent nonlinear phase-distortion kernel function, where j is the imaginary part, u is the function variable, and x is the Fourier transform variable.

[0089] In a two-dimensional image, we have: If expressed in polar coordinates:

[0090]

[0091] Normalization:

[0092]

[0093] The improved phase filter stretching kernel function with phase stretching intensity parameter S and torsion parameter W is obtained:

[0094]

[0095] Corresponding angle image:

[0096]

[0097] Among them, v 2 The y-axis variable in a planar coordinate system. The result is the inverse Fourier transform of the input image B(x, y).

[0098] Step 3: Input the image to be detected into the phase filter stretching kernel function, use the phase filter stretching kernel function to stretch the image to be detected, transform the image to be detected in the frequency domain into an angle image in the spatial domain, and extract the contour edges in the angle image.

[0099] Before inputting the image to be detected into the phase filter stretching kernel function, the image needs to be grayscaled and filtered. Grayscaled processing specifically involves taking the weighted average of the R, G, and B components of each pixel in the image as the grayscale value. In the RGB model, the color of a pixel at spatial location (x, y) is represented by its R (R(x, y)), G (G(x, y)), and B (B(x, y)) components, as shown in the formula:

[0100] Gray(x,y)=0.299*R(x,y)+0.578*G(x,y)+0.114*B(x,y)(26)

[0101] After grayscale processing, the matrix dimension of the image to be detected decreases, the processing speed is greatly improved, and the gradient information is still preserved.

[0102] The filtering process used in this invention is low-pass Gaussian filtering. Gaussian filtering is a linear smoothing filter suitable for eliminating Gaussian noise and widely used in image denoising. Simply put, Gaussian filtering denoising involves weighted averaging of the pixel values ​​across the entire image. The value of each pixel is obtained by weighted averaging of its own value and the values ​​of its neighboring pixels. The one-dimensional and two-dimensional Gaussian distribution functions are as follows:

[0103]

[0104] In the formula, σ is the standard deviation of the Gaussian function, and x and y are pixel coordinates.

[0105] High-frequency noise is often contained in the high-frequency components of a signal. Low-pass Gaussian filtering can remove this noise and improve the quality and reliability of the signal.

[0106] Step 4: Overlay the image to be detected with the angle image containing the contour edges to obtain the overlay image. Use the relative total variation RTV to sharpen and enhance the overlay image, and identify the electric field lines from the sharpened and enhanced image.

[0107] The purpose of image sharpening is to highlight meaningful, large-scale structural features, such as contour edges and structural details, while suppressing unimportant fine textures, such as irregular, chaotic, and recurring patterns. To achieve the desired image enhancement, the extracted enhanced edges and important structural images are superimposed on the original image, making the resulting image emphasize edges and important structural features, which is beneficial for subsequent specific applications. However, the superimposed image may exhibit noise and edge jaggedness due to insufficient suppression of unimportant fine textures. Therefore, this invention utilizes Relative Total Variation (RTV) to post-process the superimposed image, achieving good results.

[0108] The relative total variation (RTV) measure and clearly distinguishes important structures from fine textures well, thanks to its metric for relative total variation, which is based on a pixel-by-pixel windowed total variation measure D. x (p), D y (p) and windowing intrinsic variational measure L x (p), L y (p) Two sets of indicators determine:

[0109]

[0110] Here, S can be temporarily considered as the input image, R(p) is a rectangular local neighborhood window centered at pixel p, q is any pixel in R(p), and g p，q The weighting factor is defined based on spatial similarity. Let S be the partial derivative of the input image S in the x-direction. Let S be the partial derivative of the input image S in the y-direction.

[0111] Drawing on the advantages of Relative Total Variation (RTV) optimization in separating major structures from subtle textures, the improved phase filter stretch transform image feature extraction algorithm in this invention incorporates an RTV step. This allows the sharpened image to better highlight major edge contours and structural textures while suppressing subtle textures.

[0112] This invention also provides a power line recognition system, including an image acquisition module, a kernel function construction module, a contour extraction module, and a power line recognition module. The image acquisition module acquires an image to be detected containing power lines; the kernel function construction module constructs a phase filter stretching kernel function based on a phase translation kernel function and the convolution theorem; the contour extraction module inputs the image to be detected into the phase filter stretching kernel function, performs a stretching transformation on the image to be detected using the phase filter stretching kernel function, converts the frequency domain image to be detected into a spatial domain angle image, and extracts the contour edges in the angle image; the power line recognition module superimposes the image to be detected with the angle image containing contour edges to obtain a superimposed image, and uses relative total variation (RTV) to sharpen and enhance the superimposed image to identify the power lines.

[0113] The present invention also provides a computer device including a memory and a processor; the memory stores a computer program, and the processor is used to run the computer program in the memory to perform a power line identification method.

[0114] The present invention also provides a computer-readable storage medium storing a computer program adapted for loading by a processor to execute a power line identification method.

[0115] Example 2

[0116] The effect of using phase filter stretch transform kernel function to recognize images is analyzed, equation (21) u can be transformed into u raised to the power of k, i.e.: The experimental results differ for different values ​​of k, different phase tensile strength parameters S, and different torsion parameters W. Specific effects are as follows: Figure 2 As shown. Figure 2 (a) is the original image; Figure 2 (b) is k=2, W=10, S=10 3The experimental results corresponding to the selected values; Figure 2 (c) is k=2, W=10, S=10 5 The experimental results corresponding to the selected values; Figure 2 The (d) values ​​are k=3, W=10, and S=10. 5 The experimental results corresponding to the selected values; Figure 2 (e) is k=4, W=10, S=10 6 The experimental results corresponding to the selected values; Figure 2 (f) is k=7, W=10, S=10 6 The experimental result corresponding to the value. For example... Figure 2 The stretching results of (f) also show that when k=7, the higher the stretching power, the better the edge detection for "more curved" edges. Fine, delicate edges such as hair, hats, and shoulders are extracted very delicately and perfectly. This is because a higher power better characterizes the higher-order derivatives in the image. The higher-order derivatives of a function represent the curvature, singularity, and higher-order non-stationary information within the function, but they also generate more subtle noise. Therefore, improving the phase filter stretching transform requires selecting appropriate power exponent k and phase stretching intensity parameter S in specific target segmentation algorithms.

[0117] The image segmentation results obtained by the algorithm of this invention are compared and analyzed with image segmentation based on Canny, image segmentation based on traditional isotropic PST, and image segmentation based on R-CNN algorithm. Figure 3 (a), (b), (c), and (d) are different images to be detected; Figures 4-7 The comparison images show the results, where... Figures 4-7 (a) is the result of using PST to respectively Figure 3 The recognition results corresponding to (a), (b), (c) and (d) when S = 1.48 and W = 10.24; Figures 4-7 (b) Using Canny to respectively Figure 3 The recognition results for (a), (b), (c), and (d) are shown. Figures 4-7 (c) Using R-CNN to respectively Figure 3 The recognition results for (a), (b), (c), and (d) are shown. Figures 4-7 (d) represents the results of using the algorithm of this invention to respectively... Figure 3 (a), (b), (c), and (d) are given at k=3, W=10, and S= 10 5 The corresponding recognition effect diagram when the value is selected.

[0118] Analysis of electron wire segmentation effect

[0119] To effectively evaluate the image segmentation performance of the algorithm, three commonly used evaluation metrics in image segmentation are selected: image pixel accuracy, average pixel accuracy, and average intersection-over-union ratio.

[0120] Image pixel accuracy reflects the ratio of accurately predicted pixels to the total number of pixels, and its calculation formula is as follows:

[0121]

[0122] In the formula, n ii Let n be the number of real pixels of class i. ij The number of pixels that are misidentified as class j for class i.

[0123] Average pixel precision reflects the proportion of pixels that are correctly predicted for a category. It is calculated by averaging these proportions using the following formula:

[0124]

[0125] In the formula, N is the number of pixels.

[0126] The average intersection-union ratio (AUC) reflects the ratio of correctly predicted pixel regions to all predicted pixel regions, and its calculation formula is as follows:

[0127]

[0128] against Figures 4-7 The four image segmentation methods—traditional PST, Canny, R-CNN, and the algorithm of this invention—are compared. PA M MPA and M MIoU The values ​​are shown in Tables 1, 2 and 3.

[0129] Table 1. P values ​​for four different segmentation algorithms PA value

[0130]

[0131] To further illustrate the advantages of the algorithm of this invention, 60 power line detection images were captured by drone. These 60 images were then processed using four image segmentation algorithms, including traditional PST and R-CNN, to obtain the evaluation index P. PA M MPA and M MIoU Values ​​such as Figure 8 , Figure 9 and Figure 10 As shown.

[0132] Table 2. M values ​​for four different segmentation algorithms MPA value

[0133]

[0134] Table 3. M values ​​for four different segmentation algorithms MIoU value

[0135]

[0136] From Tables 1 to 3, Figure 8 , Figure 9 and Figure 10 The results show that the image segmentation result evaluation index P of the algorithm of this invention is... PA M MPA and M MIoU The values ​​are all higher than those of traditional image segmentation algorithms such as PST and R-CNN, which fully demonstrates that the algorithm of this invention has a better noise suppression effect. When segmenting aerial power line images containing noise and with unclear edges, the probability of over-segmentation, under-segmentation, and incomplete segmentation is reduced. Moreover, due to the enhancement of image edges, the segmentation results are closer to the ideal contour, achieving better segmentation results. Image segmentation based on traditional PST is good at handling image edges, but its effect on suppressing noise at the edges of power lines is poor; image segmentation based on Canny may result in inaccurate results such as excessive over-segmentation under the influence of noise; image segmentation based on R-CNN is insensitive to image edges and may result in excessive under-segmentation.

[0137] In summary, both the overall segmentation effect and the average value of the evaluation index further verify the effectiveness and stability of the algorithm of this invention.

[0138] According to objective evaluation index P PA M MPA and M MIoU The values ​​indicate that the electric field line segmentation results of the algorithm of this invention are significantly better than those of traditional PST, Canny, and R-CNN algorithms. Compared with image segmentation algorithms based on traditional PST, the evaluation index P of the electric field line results is significantly better. PA M MPA and M MIoU The values ​​all increased by about 15%, and the research results can be applied to aerial power line inspection and detection.

[0139] The above embodiments are merely preferred embodiments of the present invention, and the scope of protection of the present invention is not limited thereto. Any simple changes or equivalent substitutions of the technical solutions that can be obviously obtained by those skilled in the art within the scope of the technology disclosed in the present invention shall fall within the scope of protection of the present invention.

Claims

1. A method for identifying power lines, characterized in that, Includes the following steps: Acquire the image to be detected, which contains power lines; Based on the phase shift kernel function and the convolution theorem, the spatial domain complex matrix corresponding to the complex stretching kernel in the frequency domain of the phase shift kernel function is convolved with the image, and then an angle transformation is performed to obtain the phase-stretched angle image. The phase-stretched angle image is simplified, and the nonlinear phase distortion kernel function is subjected to polar coordinate transformation and normalization to obtain an image with phase stretching intensity parameters. S and twist parameters W The phase filter stretching kernel function, wherein the phase filter stretching kernel function is: In the formula, S For phase tensile strength parameters, W For the distortion parameters, r This refers to the polar radius when transforming to polar coordinates; The image to be detected is input into the phase filter stretching kernel function. The phase filter stretching kernel function is used to stretch the image to be detected, transforming the image in the frequency domain into an angle image in the spatial domain, and extracting the contour edges in the angle image. The image to be detected is superimposed with an angle image containing contour edges to obtain a superimposed image. The superimposed image is then sharpened and enhanced using relative total variation (RTV), and electric field lines are identified from the sharpened and enhanced image.

2. The power line identification method according to claim 1, characterized in that, Before inputting the image to be detected into the phase filter stretching kernel function, the process further includes grayscale conversion and filtering of the image to be detected; the grayscale conversion specifically involves: converting the pixels in the image to be detected... R Quantity, G Components and B The weighted average of the three component values ​​is used as the gray value of the grayscale image; the filtering process is to perform a weighted average of all pixel values ​​of the image to be detected, wherein the pixel value of each pixel is obtained by weighting its own value and the other pixel values ​​in its neighborhood.

3. The power line identification method according to claim 1, characterized in that, The angle image is: in, Stretch the kernel function for the phase filter. IFFT 2 represents the inverse two-dimensional Fourier transform. For the input image B ( x , y The result after inverse Fourier transform processing.

4. A power line identification system, characterized in that, include: The image acquisition module is used to acquire the image to be detected, which contains power lines; The kernel function construction module is used to perform angular transformation on the image after convolving the spatial domain complex matrix corresponding to the complex stretching kernel in the frequency domain of the phase shift kernel function with the image, based on the phase shift kernel function and the convolution theorem, to obtain the phase-stretched angular image. The phase-stretched angular image is then simplified, and the nonlinear phase distortion kernel function is subjected to polar coordinate transformation and normalization to obtain an image with phase stretching intensity parameters. S and twist parameters W The phase filter stretching kernel function, wherein the phase filter stretching kernel function is: In the formula, S For phase tensile strength parameters, W For the distortion parameters, r This refers to the polar radius when transforming to polar coordinates; The contour extraction module is used to input the image to be detected into the phase filter stretching kernel function, use the phase filter stretching kernel function to stretch the image to be detected, transform the image to be detected in the frequency domain into an angle image in the spatial domain, and extract the contour edges in the angle image. The power line recognition module is used to overlay the image to be detected with an angle image containing contour edges to obtain an overlay image, and to use relative total variation (RTV) to sharpen and enhance the overlay image, and to identify power lines from the sharpened and enhanced image.

5. A computer device, characterized in that, It includes a memory and a processor; the memory stores a computer program, and the processor is used to run the computer program in the memory to perform the power line identification method according to any one of claims 1-3.

6. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program adapted for loading by a processor to execute the power line identification method according to any one of claims 1-3.

Images