A method and device for detecting foreign objects in a transmission line

Abstract

The present invention discloses a method and device for detecting foreign objects on power transmission lines. The method comprises: preprocessing a captured power transmission line image to obtain a detection image; acquiring edge features of the detection image to form an edge feature map; identifying and extracting power transmission line information from the edge feature map using a straight line detection mechanism; identifying foreign object information in the edge feature map based on region screening; and locating the identified foreign objects using a region expansion algorithm to complete the detection of foreign objects on the power transmission line. The power transmission line foreign object detection technology provided by the present invention has a higher accuracy rate for the extracted power transmission line and foreign object detection, significantly improving the safety and efficiency of foreign object detection.

Description

Technical Field

[0001] The present invention belongs to the technical field of power transmission line detection, and in particular relates to a method and device for detecting foreign objects in a power transmission line. Background Art

[0002] Transmission lines often serve as a vehicle for the transmission of electric energy. Their safe and stable operation plays a crucial role in the overall operation of power systems. Transmission line foreign object detection systems based on drone-generated imagery have emerged as a necessary response. Numerous scholars, both domestically and internationally, have conducted research on foreign object detection in transmission line imagery.

[0003] Common methods for power line feature extraction include Hough transform and phase grouping. Numerous research results exist in foreign object detection algorithms, but these are generally categorized into two types: those based on image processing and those based on deep learning.

[0004] Patent CN106960438A proposes a method for identifying foreign objects on power transmission lines based on the Hough linear transform. The method grayscales the captured transmission line image, then applies a median filter to the grayscaled transmission line image. The Otsu method is used to segment the image to obtain a binary image to be identified, followed by Canny edge detection. After performing the Hough linear transform, the method proposes a convolution operation on the transmission line, leveraging the characteristic that foreign objects are attached to the periphery of the transmission line. This method, based on the Hough linear transform, provides excellent foreign object identification results and meets the requirements for foreign object detection and identification in basic inspection images. Its application in power transmission line inspections can effectively reduce inspection labor costs and improve inspection efficiency and accuracy, possessing significant engineering value and providing a new approach to identifying foreign objects on transmission lines.

[0005] For example, a RetinaNet-based detection method was established for bird nests and other foreign objects on transmission lines. The detection accuracy of this method can reach about 94%, but the actual recognition speed is relatively slow. A local context feature pyramid network (LCFPN) is used to refine local context information and perform multi-scale learning to achieve defect detection based on transmission line images acquired by drones.

[0006] For example, to address the poor accuracy of foreign object recognition in transmission line image detection, an improved YOLOX technology is proposed to detect foreign objects in transmission lines. This solution improves the YOLOX object detection network, uses Atrous spatial pyramid pooling to increase sensitivity to foreign objects of different scales, and embeds a convolutional block attention module to improve model recognition accuracy.

[0007] However, the above methods have low detection accuracy or low detection efficiency for foreign object detection in actual power transmission lines, resulting in unsatisfactory results in the actual process.

[0008] Therefore, how to design a detection solution for foreign objects in power transmission lines to improve the efficiency and accuracy of detection is an urgent problem to be solved by those skilled in the art. Summary of the Invention

[0009] To address the shortcomings of the aforementioned prior art, the present invention provides a method and device for detecting foreign objects on power transmission lines. This method uses a similar mechanism for recognizing linear features to extract power transmission lines. Furthermore, an algorithm based on rectangular region expansion is proposed to precisely locate foreign objects on power transmission lines. This method achieves higher accuracy in detecting both the extracted power lines and foreign objects, significantly improving the safety and efficiency of foreign object detection.

[0010] In a first aspect, the present invention provides a method for detecting foreign objects in a power transmission line, comprising the following steps:

[0011] Preprocessing the collected transmission line image to obtain a detection image;

[0012] Obtain edge features of the detection image and form an edge feature map;

[0013] Combined with the straight line detection mechanism, the transmission line information in the edge feature map is identified and extracted;

[0014] Based on region screening, identify foreign matter information in edge feature maps;

[0015] Combined with the regional expansion algorithm, the identified foreign objects are located to complete the detection of foreign objects on the transmission line.

[0016] Furthermore, preprocessing includes de-illumination processing, image grayscale processing, image compression, image enhancement and image denoising.

[0017] Furthermore, edge features of the detected image are obtained to form an edge feature map, which specifically includes the following steps:

[0018] The edge features of the image are extracted using the pre-built edge detection algorithm in the horizontal, vertical and diagonal directions to form an edge feature map.

[0019] The pre-built edge detection algorithm is specifically expressed as:

[0020]

[0021] Among them, g(x, y) is the edge feature map, is the Laplace operator, f(x, y) is the detection image, G(x, y) is the Gaussian two-dimensional function, and (x, y) is the image pixel coordinate.

[0022] Furthermore, combined with the straight line detection mechanism, the transmission line information in the edge feature map is identified and extracted, which specifically includes the following steps:

[0023] Optimize the edge feature map and perform preliminary identification of transmission lines based on their topological structure and geometric distribution characteristics;

[0024] Combined with the straight line detection mechanism, the slope and intercept of the transmission line are analyzed to verify the initially identified transmission line and complete the identification and extraction of the transmission line information.

[0025] Furthermore, the edge feature map is optimized, and the transmission lines are preliminarily identified based on their topological structure and geometric distribution characteristics. Specifically, the following steps are included:

[0026] The edge feature map is smoothed and blurred by using the hierarchical recursion of the average gray value to obtain a low-resolution edge feature map, which is specifically expressed as:

[0027] g n (x, y) = (σ x σ y ) -1 ×{g n-1 (2x-1,2y-1)+g n-1 (2x-1, 2y)

[0028] +g n-1 (2x,2y-1)+g n-1 (2x, 2y)

[0029] Among them, σ x , σ y Represents the multiple of the size reduction in the x and y directions of the image, g n (x, y) is the edge feature of (x, y) after n times of hierarchical recursion, g n-1 (2x-1, 2y-1) is the edge feature of (2x-1, 2y-1) after n-1 hierarchical recursions, g n-1 (2x-1, 2y) is the edge feature of (2x-1, 2y) after n-1 hierarchical recursions, g n-1 (2x, 2y-1) is the edge feature of (2x, 2y-1) after n-1 hierarchical recursions, g n-1 (2x, 2y) is the edge feature of (2x, 2y) after n-1 hierarchical recursions;

[0030] Perform region segmentation on the low-resolution edge feature map and determine the image area for searching the transmission line. The specific expression is:

[0031]

[0032] Among them, H is the height of the low-resolution edge feature map, h1 is the height of region a, h2 is the height of region b, and h3 is the height of region c;

[0033] The eigenvectors and eigenvalues ​​of the pixels in the image area are analyzed, and the preliminary identification of the transmission lines is completed based on the topological structure and geometric distribution characteristics of the transmission lines.

[0034] Furthermore, the eigenvectors and eigenvalues ​​of the pixels in the image area are analyzed to complete the preliminary identification of the transmission line, which is specifically expressed as:

[0035] By statistically analyzing the pixels in the image area, the covariance matrix of the image area is obtained, which is specifically expressed as:

[0036]

[0037] Among them, A is the covariance matrix of the image area, n is the total number of pixels in the image area, (x i ,y i ) is the coordinate of the i-th pixel;

[0038] The eigenvalues ​​of the corresponding covariance matrix are given, and the angles of the pixels in the image area are determined, which can be expressed as:

[0039]

[0040]

[0041] Among them, λ is the characteristic root of the covariance matrix A, α1 is the angle between the principal component directions of the pixels in the image area;

[0042] Traverse the covariance matrix and characteristic roots of each pixel in the image area to identify the main direction angle of the transmission line. Combined with the geometric distribution characteristics of the transmission line, the preliminary detection of the transmission line is completed. The main direction angle of the transmission line is specifically expressed as:

[0043]

[0044] Where α is the main direction angle of the transmission line in the image area.

[0045] Furthermore, combined with the straight line detection mechanism, the slope and intercept of the transmission line are analyzed to verify the initially identified transmission line and complete the identification and extraction of the transmission line information. Specifically, the following steps are included:

[0046] Obtain the initially identified adjacent transmission lines and analyze the distance converted into polar coordinates in the parameter space, which can be expressed as:

[0047]

[0048] Where s is the distance between the pixel points of transmission lines a and b converted into polar coordinates in the parameter space, (β a , α a ) is the polar coordinate of the pixel point on the transmission line a, (β b , α b ) is the polar coordinate of the pixel point on the transmission line b;

[0049] like The adjacent transmission lines initially identified are merged to complete the identification and extraction of transmission line information.

[0050] Where Δx a is the x-axis coordinate difference of the pixel point of transmission line a, Δy a is the y-axis coordinate difference of the pixel point of transmission line a, Δx b is the x-axis coordinate difference of the pixel point of transmission line b, Δy b is the y-axis coordinate difference of the pixel point of transmission line b.

[0051] Furthermore, based on region screening, foreign body information in the edge feature map is identified, which specifically includes the following steps:

[0052] Taking the matrix box as the region of interest, the rectangular box is horizontally translated from the left side to the right side of the image area to traverse each pixel point, where the length d1 of the rectangular box is the horizontal spacing of the pixel points on the transmission line;

[0053] If there is a pixel in a rectangular box during translation whose horizontal spacing of pixels exceeds the first threshold with respect to d1, it is determined to be an outlier, and the number of outliers in the rectangular box is counted. j ;

[0054] If n j If the value exceeds the second threshold, it is determined that the rectangular frame contains the target rectangular frame of the foreign object, and the coordinates of the center point of the foreign object area are given.

[0055] Furthermore, the identified foreign objects are located by combining the region expansion algorithm, which specifically includes the following steps:

[0056] Take the upper boundary, lower boundary, left boundary and right boundary of the target rectangle as the reference line respectively, with a step size of O u to expand;

[0057] Determine the number of abnormal points in the target rectangular box before and after expansion respectively;

[0058] If the number of outliers before and after the expansion does not change, the expansion of the target rectangular box is stopped. If the number of outliers before and after the expansion increases, the expansion of the target rectangular box is continued until the number of outliers does not change.

[0059] Based on the determined size of the expanded target rectangle, the location of the foreign object is given.

[0060] In a second aspect, the present invention further provides a transmission line foreign object detection device, which adopts the above-mentioned transmission line foreign object detection method, comprising:

[0061] A preprocessing module preprocesses the collected transmission line image to obtain a detection image;

[0062] The feature extraction module extracts edge features of the detection image based on the line detection mechanism, and identifies and extracts transmission line information;

[0063] The analysis and detection module uses a regional expansion algorithm to identify and analyze the information of foreign objects on the transmission lines in the detection image, locate the foreign objects, and complete the detection of foreign objects on the transmission lines.

[0064] The present invention provides a method and device for detecting foreign matter in a transmission line, which at least includes the following:

[0065] Beneficial effects:

[0066] (1) The power transmission line foreign body detection technology provided by the present invention has a higher accuracy rate in extracting power transmission lines and foreign body detection, which can greatly improve the safety and efficiency of foreign body detection.

[0067] (2) The present invention is based on a similar mechanism of recognizing straight line features, which is beneficial for preserving the rich information of the image itself, reducing the amount of processing calculations, and eliminating interference around the transmission lines and possible line overlap or intersection problems.

[0068] (3) The present invention is based on an algorithm for rectangular area expansion, which can further accurately select the location of foreign objects without increasing the complexity of detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0069] Figure 1 A schematic diagram of a power transmission line image acquisition system provided by the present invention;

[0070] Figure 2 A schematic flow chart of a method for detecting foreign matter in a power transmission line provided by the present invention;

[0071] Figure 3 A schematic diagram of the foreign body detection principle provided by the present invention;

[0072] Figure 4 A schematic diagram of the foreign body positioning principle of the rectangular area expansion method provided by the present invention;

[0073] Figure 5 This is an edge feature rendering of an embodiment provided by the present invention;

[0074] Figure 6A foreign body detection effect diagram of a certain embodiment provided by the present invention;

[0075] Figure 7 A schematic diagram of the accuracy of different foreign body detections provided by an embodiment of the present invention;

[0076] Figure 8 This is a schematic diagram of foreign body detection time under different backgrounds of an embodiment provided by the present invention. DETAILED DESCRIPTION

[0077] In order to better understand the above technical solution, the above technical solution will be described in detail below in conjunction with the accompanying drawings and specific implementation methods. Obviously, the embodiments described are only part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0078] The terms used in the embodiments of the present invention are for the purpose of describing specific embodiments only and are not intended to limit the present invention. The singular forms "a," "an," "the," and "the" used in the embodiments of the present invention and the appended claims are also intended to include plural forms, and unless the context clearly indicates otherwise, "a plurality" generally includes at least two.

[0079] It should also be noted that the terms "include," "comprises," or any other variations thereof are intended to encompass non-exclusive inclusion, such that a product or device comprising a series of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such product or device. In the absence of further limitations, an element defined by the phrase "comprises a..." does not exclude the presence of other identical elements in the product or device comprising the element.

[0080] In order to solve the existing problem of abnormality detection of power transmission lines, a 5G wireless signal transmitter is used to transmit the collected abnormal images to the ground station monitoring room, and then the abnormal situation is automatically detected and judged by software. The entire power transmission line foreign body detection system mainly includes data image acquisition system, drone control and 5G transmission system. Figure 1 As shown in the figure, the system can be used to realize the aerial photography of transmission line images by UAVs, extract effective information and realize transmission line anomaly detection.

[0081] like Figure 2 As shown, the present invention provides a method for detecting foreign objects in a transmission line, which specifically includes the following steps:

[0082] Preprocessing the collected transmission line image to obtain a detection image;

[0083] Obtain edge features of the detection image and form an edge feature map;

[0084] Combined with the straight line detection mechanism, the transmission line information in the edge feature map is identified and extracted;

[0085] Based on region screening, identify foreign matter information in edge feature maps;

[0086] Combined with the regional expansion algorithm, the identified foreign objects are located to complete the detection of foreign objects on the transmission line.

[0087] Preprocessing includes de-illumination, image grayscale processing, image compression, image enhancement and image denoising.

[0088] The collected transmission line images cannot be directly used for transmission line anomaly detection, mainly because the transmission line inspection environment is relatively complex and the image quality collected by drones is also uneven. Therefore, the collected images must be preprocessed before anomaly detection.

[0089] Based on the characteristics of drone-captured images, it's worth noting that this invention first removes illumination to enhance image contrast before proceeding with subsequent processing. Furthermore, image grayscale conversion utilizes a weighted average method, image compression employs a pyramid compression method, and image enhancement and denoising utilize linear regression to calibrate the image and Gaussian filtering for denoising, respectively.

[0090] The edges of transmission line images contain a wealth of important information, so edge detection on captured images helps preserve the rich information inherent in the image. Traditional image edge detection algorithms primarily include the Canny operator, the Sobel operator, and the Log operator. However, these algorithms often suffer from high algorithmic complexity, heavy computational effort, and low detection efficiency in actual edge feature detection.

[0091] Obtaining the edge features of the detection image and forming an edge feature map specifically includes the following steps:

[0092] The edge features of the image are extracted using the pre-built edge detection algorithm in the horizontal, vertical and diagonal directions to form an edge feature map.

[0093] The pre-built edge detection algorithm is specifically expressed as:

[0094]

[0095] Among them, g(x, y) is the edge feature map, is the Laplace operator, f(x, y) is the detection image, G(x, y) is the Gaussian two-dimensional function, and (x, y) is the image pixel coordinate.

[0096] The pre-built edge detection algorithm integrates the Sobel operator using templates in four directions: horizontal, vertical, and diagonal. It also introduces the Log operator for feature detection, and combines the Gaussian and Laplacian operators, first using Gaussian filtering for smoothing and then applying the Laplacian operator. This reduces computational complexity and improves detection efficiency.

[0097] After obtaining smooth, continuous edge features, interference around the transmission lines needs to be eliminated to increase the accuracy of transmission line extraction. After obtaining smooth, continuous edges, vertical and horizontal interference around the transmission lines needs to be eliminated to ensure accurate transmission line extraction. Based on the topological structure (linear structure, low grayscale value) and geometric distribution characteristics of the transmission lines, a method for determining the coarseness of the line direction angle is proposed.

[0098] Combined with the straight line detection mechanism, the power transmission line information in the edge feature map is identified and extracted, which specifically includes the following steps:

[0099] Optimize the edge feature map and perform preliminary identification of transmission lines based on their topological structure and geometric distribution characteristics;

[0100] Combined with the straight line detection mechanism, the slope and intercept of the transmission line are analyzed to verify the initially identified transmission line and complete the identification and extraction of the transmission line information.

[0101] Furthermore, the edge feature map is optimized, and the transmission lines are preliminarily identified based on their topological structure and geometric distribution characteristics. Specifically, the following steps are included:

[0102] The edge feature map is smoothed and blurred by using the hierarchical recursion of the average gray value to obtain a low-resolution edge feature map, which is specifically expressed as:

[0103] g n (x, y) = (σ x σ y ) -1 ×{g n-1 (2x-1,2y-1)+g n-1 (2x-1, 2y)+g n-1 (2x,2y-1)+g n-1 (2x, 2y)}

[0104] Among them, σ x , σ y Represents the multiple of the size reduction in the x and y directions of the image, g n (x, y) is the edge feature of (x, y) after n times of hierarchical recursion, g n-1 (2x-1, 2y-1) is the edge feature of (2x-1, 2y-1) after n-1 hierarchical recursions, gn-1 (2x-1, 2y) is the edge feature of (2x-1, 2y) after n-1 hierarchical recursions, g n-1 (2x, 2y-1) is the edge feature of (2x, 2y-1) after n-1 hierarchical recursions, g n-i1 (2x, 2y) is the edge feature of (2x, 2y) after n-1 hierarchical recursions;

[0105] Perform region segmentation on the low-resolution edge feature map and determine the image area for searching the transmission line. The specific expression is:

[0106]

[0107] Among them, H is the height of the low-resolution edge feature map, h1 is the height of region a, h2 is the height of region b, and h3 is the height of region c;

[0108] The eigenvectors and eigenvalues ​​of the pixels in the image area are analyzed, and the preliminary identification of the transmission lines is completed based on the topological structure and geometric distribution characteristics of the transmission lines.

[0109] Generally speaking, according to the characteristics of drone-collected images, the power lines are located in the middle of the image area. In order to improve the detection speed, the search is performed in area b.

[0110] The eigenvectors and eigenvalues ​​of the pixels in the image area are analyzed to complete the preliminary identification of the transmission line, which is specifically expressed as:

[0111] By statistically analyzing the pixels in the image area, the covariance matrix of the image area is obtained, which is specifically expressed as:

[0112]

[0113] Among them, A is the covariance matrix of the image area, n is the total number of pixels in the image area, (x i ,y i ) is the coordinate of the i-th pixel;

[0114] The eigenvalues ​​of the corresponding covariance matrix are given, and the angles of the pixels in the image area are determined, which can be expressed as:

[0115]

[0116] Among them, λ is the characteristic root of the covariance matrix A, α1 is the principal component angle of the pixel points in the image area;

[0117] Traverse the covariance matrix and characteristic roots of each pixel in the image area to identify the main direction angle of the transmission line. Combined with the geometric distribution characteristics of the transmission line, the preliminary detection of the transmission line is completed. The main direction angle of the transmission line is specifically expressed as:

[0118]

[0119] Where α is the main direction angle of the transmission line in the image area.

[0120] Transmission line identification not only involves background images like trees and utility poles, but also involves overlapping or intersecting lines, necessitating the merging of adjacent lines. Since transmission lines have a certain width, this paper proposes a similarity mechanism based on line feature recognition to find the slope and intercept of a line, enabling transmission line identification and extraction. Adjacent lines with roughly the same slope and spacing within a pixel are merged into a cluster for tracking.

[0121] Combined with the straight line detection mechanism, the slope and intercept of the transmission line are analyzed to verify the initially identified transmission line and complete the identification and extraction of transmission line information. The specific steps include:

[0122] Obtain the initially identified adjacent transmission lines and analyze the distance converted into polar coordinates in the parameter space, which can be expressed as:

[0123]

[0124] Where s is the distance between the pixel points of transmission lines a and b converted into polar coordinates in the parameter space, (β a , α a ) is the polar coordinate of the pixel point on the transmission line a, (β b , α b ) is the polar coordinate of the pixel point on the transmission line b;

[0125] The similarity of lines is identified by comparing the distances between the polar coordinates of adjacent lines in the parameter space. The smaller the s value, the closer the two point elements are.

[0126] like The adjacent transmission lines initially identified are merged to complete the identification and extraction of transmission line information.

[0127] Where Δx a is the x-axis coordinate difference of the pixel point of transmission line a, Δy a is the y-axis coordinate difference of the pixel point of transmission line a, Δx b is the x-axis coordinate difference of the pixel point of transmission line b, Δy b is the y-axis coordinate difference of the pixel point of transmission line b.

[0128] The parameters of the transmission line itself usually occupy about two pixels, so when the above situation occurs, the two transmission lines are considered to be the same straight line.

[0129] M iis a set of points on the fitted line, and its coordinates are set as (x i ,y i ), the horizontal spacing between these points is d1. i The point is used as the core to establish a rectangular screening area, and the length and width of the rectangle are set to d1 and h1 respectively. Each black foreground pixel in the intermediate image is traversed with the rectangular area as the region of interest. Assuming that the set threshold is d2, when the difference between the distance from the point to the fitted straight line and d1 is greater than d2, it means that the point is not a point on the transmission line, and the total number of such points in the rectangular area n is calculated. j .

[0130] like Figure 3 As shown in the figure, based on region screening, foreign body information in the edge feature map is identified, which specifically includes the following steps:

[0131] Taking the matrix box as the region of interest, the rectangular box is horizontally translated from the left side to the right side of the image area to traverse each pixel point, where the length d1 of the rectangular box is the horizontal spacing of the pixel points on the transmission line;

[0132] Assuming that there is a foreign object on the transmission line Q, it can be seen from the fitted equation extracted from the transmission line that it satisfies the straight line Ax+By+c=0. In order to facilitate calculation and better explain the principle of foreign object detection, it can be expressed as a two-point formula:

[0133]

[0134] M i (x i ,y i ) point as the core, create a rectangular frame filter starting from the leftmost side of the image area. If the rectangular frame exceeds the right boundary of the image area at this time, it is necessary to end the creation of the rectangular boundary.

[0135] If there is a pixel in a rectangular box during translation whose horizontal spacing of pixels exceeds the first threshold with respect to d1, it is determined to be an outlier, and the number of outliers in the rectangular box is counted. j ;

[0136] Translate to find the next core point M i+1 point, you can pass

[0137] If n j If the second threshold is exceeded, the target rectangular box is determined to contain the foreign body, and the center coordinates of the foreign body area are given. The total number of abnormal points n in the rectangular box is selected. j Calculate and determine whether it is greater than the set second threshold m. j >m, it means that the rectangular frame contains foreign matter.

[0138] In order to further accurately select the location of the foreign object, a frame selection area expansion algorithm is proposed to solve the problem based on the intersection of the foreign object and the transmission line.

[0139] like Figure 4 As shown, assuming that the image size for foreign body detection is L E ×H E , with M i Create a rectangular search box of size f × g with the point as the core. Expand the four sides of the rectangular search box outward. When a side encounters two consecutive outliers or reaches the image boundary, expansion in that direction stops. Continue expansion in all directions until the location of the foreign object is accurately selected.

[0140] Combined with the region expansion algorithm, the identified foreign objects are located, which specifically includes the following steps:

[0141] Take the upper boundary, lower boundary, left boundary and right boundary of the target rectangle as the reference line, and u Indicates the step size of outward expansion in each direction, P u Represents the total number of outward expansions, q u Indicates the total number of times that the singular point does not grow and r u Indicates the distance between the intersection point Mz and the surrounding rectangular positioning frame;

[0142] Determine the number of abnormal points in the target rectangular box before and after expansion respectively;

[0143] If the number of outliers before and after expansion does not change, then stop expanding the target rectangle. If the number of outliers before and after expansion increases, then the parameter P u =P u +1, then continue to expand the target rectangle until the number of outliers remains unchanged;

[0144] The total number of outliers has not increased, so q u =q u +1, based on the size of the determined expanded target rectangle, output r u , or Provide the location of the foreign body.

[0145] In a second aspect, the present invention further provides a transmission line foreign object detection device, which adopts the above-mentioned transmission line foreign object detection method, comprising:

[0146] A preprocessing module preprocesses the collected transmission line image to obtain a detection image;

[0147] The feature extraction module extracts edge features of the detection image based on the line detection mechanism, and identifies and extracts transmission line information;

[0148] The analysis and detection module uses a regional expansion algorithm to identify and analyze the information of foreign objects on the transmission lines in the detection image, locate the foreign objects, and complete the detection of foreign objects on the transmission lines.

[0149] like Figure 5 As described above, the detection method of the present invention is used to detect and extract power transmission lines in images with backgrounds including cloudy days, sunny days, trees, etc. Under different backgrounds, the power transmission lines and foreign objects can be clearly and accurately extracted.

[0150] like Figure 6 、 7 As shown, the system can fully detect foreign objects on power lines, and the detection performance for large foreign objects is generally better than for small ones. For example, the detection accuracy for large objects like plastic bags reached 99.32%, and for small objects like bird nests, the accuracy reached 92.16%. However, it is worth noting that the detection of suspended objects, whether large or small, is not particularly ideal. Because suspended objects are not considered foreign objects on power lines, their intersection with the power lines cannot be calculated, making it impossible to accurately locate them.

[0151] In the actual application of foreign body detection on power transmission lines, another indicator worth noting is the speed of foreign body detection on power transmission lines. The speed of detecting large and small foreign body targets under different backgrounds is as follows: Figure 8 As shown in the figure, it is clear that the foreign object detection speed is relatively fast under good lighting and simple image backgrounds, but is slower under cloudy conditions and complex backgrounds. Furthermore, under the same environmental conditions, the detection speed for large foreign objects is slower than that for small foreign objects. The foreign object detection speed proposed by the present invention can reach as fast as 70ms, and can be achieved within 200ms even under complex backgrounds.

[0152] Although preferred embodiments of the present invention have been described, those skilled in the art may make additional changes and modifications to these embodiments once they are aware of the basic inventive concepts. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the invention. Obviously, those skilled in the art may make various changes and modifications to the present invention without departing from the spirit and scope of the invention. Thus, the present invention is intended to include such changes and modifications as fall within the scope of the claims and their equivalents.

Claims

1. A method for detecting foreign objects in a power transmission line, characterized in that: The specific steps include: Preprocessing the collected transmission line image to obtain a detection image; Obtain edge features of the detection image and form an edge feature map; The edge feature map is smoothed and blurred by using the hierarchical recursion of the average gray value to obtain a low-resolution edge feature map, which is specifically expressed as: g n (x,y)=(σ x σ y ) -1 ×{g n-1 (2x-1,2y-1)+g n-1 (2x-1,2y)+g n-1 (2x,2y-1)+g n-1 (2x,2y)} Among them, σ x , σ y Represents the multiple of the size reduction in the x and y directions of the image, g n (x, y) is the edge feature of (x, y) after n times of hierarchical recursion, g n-1 (2x-1, 2y-1) is the edge feature of (2x-1, 2y-1) after n-1 hierarchical recursions, g n-1 (2x-1, 2y) is the edge feature of (2x-1, 2y) after n-1 hierarchical recursions, g n-1 (2x, 2y-1) is the edge feature of (2x, 2y-1) after n-1 hierarchical recursions, g n-1 (2x, 2y) is the edge feature of (2x, 2y) after n-1 hierarchical recursions; Perform region segmentation on the low-resolution edge feature map and determine the image area for searching the transmission line. The specific expression is: Among them, H is the height of the low-resolution edge feature map, h1 is the height of area a, h2 is the height of area b, and h3 is the height of area c. The transmission line is located in the middle of the image area and is searched in area b; Analyze the eigenvectors and eigenvalues ​​of the pixels in the image area, and complete the preliminary identification of the transmission line based on the topological structure and geometric distribution characteristics of the transmission line; Combined with the straight line detection mechanism, the slope and intercept of the transmission line are analyzed to verify the initially identified transmission line and complete the identification and extraction of transmission line information; Based on region screening, identify foreign matter information in edge feature maps; Combined with the regional expansion algorithm, the identified foreign objects are located to complete the detection of foreign objects on the transmission line.

2. The method for detecting foreign matter in a power transmission line according to claim 1, wherein: Preprocessing includes de-illumination, image grayscale processing, image compression, image enhancement and image denoising.

3. The method for detecting foreign matter in a power transmission line according to claim 1, wherein: Obtaining the edge features of the detection image and forming an edge feature map specifically includes the following steps: The edge features of the image are extracted using the pre-built edge detection algorithm in the horizontal, vertical and diagonal directions to form an edge feature map. The pre-built edge detection algorithm is specifically expressed as: Among them, g(x,y) is the edge feature map, is the Laplace operator, f(x,y) is the detection image, G(x,y) is the Gaussian two-dimensional function, and (x,y) is the image pixel coordinate.

4. The method for detecting foreign matter in a power transmission line according to claim 1, wherein: The eigenvectors and eigenvalues ​​of the pixels in the image area are analyzed to complete the preliminary identification of the transmission line, which is specifically expressed as: By statistically analyzing the pixels in the image area, the covariance matrix of the image area is obtained, which is specifically expressed as: Among them, A is the covariance matrix of the image area, n is the total number of pixels in the image area, (x i ,y i ) is the coordinate of the i-th pixel; The eigenvalues ​​of the corresponding covariance matrix are given, and the angles of the pixels in the image area are determined, which can be expressed as: Among them, λ is the characteristic root of the covariance matrix A, α1 is the angle between the principal component directions of the pixels in the image area; Traverse the covariance matrix and characteristic roots of each pixel in the image area to identify the main direction angle of the transmission line. Combined with the geometric distribution characteristics of the transmission line, the preliminary detection of the transmission line is completed. The main direction angle of the transmission line is specifically expressed as: Where α is the main direction angle of the transmission line in the image area.

5. The method for detecting foreign matter in a power transmission line according to claim 1, wherein: Combined with the straight line detection mechanism, the slope and intercept of the transmission line are analyzed to verify the initially identified transmission line and complete the identification and extraction of transmission line information. The specific steps include: Obtain the initially identified adjacent transmission lines and analyze the distance converted into polar coordinates in the parameter space, which can be expressed as: Where s is the distance between the pixel points of transmission lines a and b converted into polar coordinates in the parameter space, (β a , α a ) is the polar coordinate of the pixel point on the transmission line a, (β b , α b ) is the polar coordinate of the pixel point on the transmission line b; like The adjacent transmission lines initially identified are merged to complete the identification and extraction of transmission line information. Where Δx a is the x-axis coordinate difference of the pixel point of transmission line a, Δy a is the y-axis coordinate difference of the pixel point of transmission line a, Δx b is the x-axis coordinate difference of the pixel point of transmission line b, Δy b is the y-axis coordinate difference of the pixel point of transmission line b.

6. The method for detecting foreign matter in a power transmission line according to claim 1, wherein: Based on region screening, foreign body information in the edge feature map is identified, which specifically includes the following steps: Taking the matrix box as the region of interest, the rectangular box is horizontally translated from the left side to the right side of the image area to traverse each pixel point, where the length d1 of the rectangular box is the horizontal spacing of the pixel points on the transmission line; If there is a pixel in a rectangular box during translation whose horizontal spacing of pixels exceeds the first threshold with respect to d1, it is determined to be an outlier, and the number of outliers in the rectangular box is counted. j ; If n j If the value exceeds the second threshold, it is determined that the rectangular frame contains the target rectangular frame of the foreign object, and the coordinates of the center point of the foreign object area are given.

7. The method for detecting foreign matter in a power transmission line according to claim 6, wherein: Combined with the region expansion algorithm, the identified foreign objects are located, which specifically includes the following steps: Take the upper boundary, lower boundary, left boundary and right boundary of the target rectangle as the reference line respectively, with a step size of O u to expand; Determine the number of abnormal points in the target rectangular box before and after expansion respectively; If the number of outliers before and after the expansion does not change, the expansion of the target rectangular box is stopped. If the number of outliers before and after the expansion increases, the expansion of the target rectangular box is continued until the number of outliers does not change. Based on the determined size of the expanded target rectangle, the location of the foreign object is given.

8. A foreign body detection device for a power transmission line, characterized in that: The method for detecting foreign objects in a power transmission line according to any one of claims 1 to 7 comprises: A preprocessing module preprocesses the collected transmission line image to obtain a detection image; The feature extraction module extracts edge features of the detection image based on the line detection mechanism, and identifies and extracts transmission line information; The analysis and detection module uses a regional expansion algorithm to identify and analyze the information of foreign objects on the transmission lines in the detection image, locate the foreign objects, and complete the detection of foreign objects on the transmission lines.

Images