Method and system for detecting strand breakage defect of overhead transmission conductor based on machine vision

Abstract

The invention discloses an overhead transmission conductor strand breakage defect detection method and system based on machine vision. The method comprises the following steps: acquiring an overhead transmission conductor image based on a high-resolution camera; carrying out image enhancement processing based on gray stretching; performing image noise reduction processing based on guide filtering; carrying out edge detection based on an improved Canny algorithm; performing image optimization based on morphological operation; the method can automatically perform strand breakage defect detection on a large number of overhead transmission conductor images, has the characteristic of high automation, can significantly reduce the interference of human factors on the detection result, and meets the requirements of industrial automatic detection.

Description

Technical Field

[0001] This invention belongs to the field of power facility inspection technology, specifically relating to a method and system for detecting broken strand defects in overhead transmission lines based on machine vision. Background Technology

[0002] In power systems, overhead transmission lines are crucial transmission infrastructure, and strand breakage defects can severely impact the safety and reliability of power transmission. Previously, strand breakage detection in overhead transmission lines largely relied on manual inspection, a method that is not only inefficient and labor-intensive but also prone to missed defects. In recent years, with the continuous advancement of machine vision technology, its application in industrial inspection has increased significantly. However, existing machine vision inspection methods still have some shortcomings in identifying strand breakage defects in overhead transmission lines, such as insufficient precision in extracting strand features in complex backgrounds, low recognition accuracy, and the inability to locate defects in real time. Therefore, it is necessary to design a machine vision-based method and system for detecting strand breakage defects in overhead transmission lines to address these issues. Summary of the Invention

[0003] The technical problem this invention aims to solve is to provide a machine vision-based method and system for detecting strand breakage defects in overhead power transmission lines. Based on the grayscale features and edge information of images, this method achieves accurate identification and localization of strand breakage defects through image acquisition, preprocessing, enhancement, edge detection, and morphological optimization. This method directly analyzes overhead power transmission line images using image processing technology, without requiring complex physical modeling or prior knowledge. Through adaptive threshold settings and morphological operations, strand breakage defects can be quickly separated from the background, and false edges and noise can be removed. Once suitable image processing parameters are determined, this method can automatically detect strand breakage defects in a large number of overhead power transmission line images, exhibiting a high degree of automation and significantly reducing the interference of human factors on the detection results, thus meeting the requirements of industrial automated inspection.

[0004] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is as follows: A machine vision-based method and system for detecting strand breakage defects in overhead power transmission lines includes the following steps: S1, based on high-resolution camera to acquire images of overhead power transmission lines; S2, Image enhancement processing based on grayscale stretching; S3, Image noise reduction processing based on guided filtering; S4, edge detection based on the improved Canny algorithm; S5, image optimization based on morphological operations.

[0005] Furthermore, in step S1, an LA-CM-08K08A line scan camera is used as the image acquisition device. This camera has high resolution and fast imaging capabilities, enabling it to clearly capture the detailed features of the conductor surface. The camera is mounted on a movable inspection platform and continuously captures images along the conductor direction with the assistance of a track or drone. During image acquisition, different shooting modes are selected according to the flatness of the conductor surface and the inspection requirements: for conductors with relatively flat surfaces, a continuous shooting mode is used to ensure the continuity and integrity of the images; for conductors with undulating surfaces or complex environments, an intermittent shooting mode is used, automatically acquiring image data at certain intervals based on the platform's movement speed. Simultaneously, a DH-VHS-60mm lens and a DH-COL-180-130-W light source are used to ensure image clarity and contrast, providing high-quality raw data for subsequent image processing and defect detection.

[0006] Preferably, step S2 specifically includes: By scaling the grayscale values ​​of the acquired image, they are uniformly adjusted to the preset grayscale range, thereby improving the contrast of the image; a template is used to scan each pixel in the region of interest and perform linear calculations. The grayscale value of the center pixel is set as the difference between the maximum and minimum values ​​of the template in the region. The specific formula is shown in Equation (1).

[0007] (1) In the formula, This is the output after stretching. The coefficient being multiplied. This is the offset value to be added; By grayscale values Mapped to This significantly enhances image contrast, thus meeting the requirements of the preprocessing stage. To ensure the system's robustness under different input conditions, it is necessary to maintain the output as much as possible. The stability of the parameters was assessed in conjunction with the actual production conditions on site. Preset parameters at this time The value of can be determined by formula (2).

[0008] (2) In the formula, The average gray level of the image is given by equation (3): (3) In the formula, Let p be the grayscale value of pixel p within region R.

[0009] Preferably, in step S3, guided filtering is a highly efficient filtering algorithm with constant time complexity. Its computation speed is independent of the window size and it can effectively preserve image edges. Compared with isotropic filters, guided filtering is not only more efficient but also performs better in processing edge textures and details.

[0010] Guided filtering assumes that the guide image and the output image have a linear relationship within a two-dimensional template, mathematically expressed as equation (4): (4) In the formula, It is the output image grayscale value at that location It is the input image The gray values ​​at point k, and a and b are the coefficients of the linear function when the template center is located at point k. This is the template area.

[0011] Preferably, the coefficients of the linear function that minimize the value of equation (5) are calculated using the linear regression method, and this is taken as the optimal solution: (5) In the formula, It is the power transmission line to be filtered. It is an important parameter for adjusting the filtering effect of the filter, and is used to prevent the obtained result from being incorrect. Too large; Equations (6) to (7) are obtained using the least squares method: (6) (7) In equations (6) to (7), The image p to be filtered is in the template region The mean of the middle, Image I is in the template region The average value in It is a template area The number of pixels in the middle, Image I is in the template region The variance in; Substituting equations (6) and (7) into equation (4) yields indivual Value, for all The average of the values ​​yields the final result as shown in equation (8): (8) Using defective transmission lines as a guide diagram, i.e. ,at this time and The solutions are given by equations (9) and (10): (9) (10) When defects exist in a region, i.e., when marginal regions exist, the variance... Much larger ,at this time , This effectively preserves the gradient at the defect edge.

[0012] Preferably, in step S4, the edge detection based on the improved Canny algorithm includes the following steps: Using a 3×3 neighborhood weighted template, the gradient values ​​in the x-direction, 45° direction, y-direction, and 135° direction are calculated and then summed with weights. Let Gx represent the gray value at pixel (i, j) in the filtered image, and Gy represent the gray values ​​detected along the x and y directions at pixel (i, j), respectively. The calculation formulas are shown in equations (11) to (14): (11) (12) (13) (14) The horizontal difference with 45° and 135° characteristics is represented by equation (15): (15) The vertical difference with 45° and 135° features is given by equation (16): (16) Let Amplitude be the gradient magnitude at point (i, j), and θ(i, j) be the gradient direction. Then its calculation formula is as shown in equations (17) and (18): (17) (18) Preferably, for the center pixel (i, j), its gradient intensity is: Based on the gradient intensity in the x-direction gradient intensity in the y direction The sign and magnitude of the gradient can be calculated by linear interpolation along the pixel gradient direction within the connected region to determine the strength of two comparative gradients in the positive and negative gradient directions. and The calculation formulas are as shown in equations (19) and (20), and the value of the linear interpolation t is shown in equation (21): (19) (20) ;(twenty one) like = If the value is 0, then (i, j) is a non-edge point.

[0013] Preferably, the high and low thresholds in the dual-threshold connection are adaptively determined by combining the iterative method and the NIBLACK algorithm: Based on the image after nonmaximum suppression The grayscale histogram is used as an initial threshold, with the median grayscale value Z0 selected. The threshold is then iterated according to the following formula: ;(twenty two) In the formula, L represents the gray level of the image. It is the number of pixels with a gray value of k in the gray-level histogram, and the iteration proceeds to... End, take the iteration threshold Zi at the end.

[0014] Preferably, a grayscale image is used. The size is M×N, the image gray value is f(x,y), the gray mean is m(x,y), and the gray standard deviation is s(x,y); the NIBLACK algorithm realizes dynamic threshold segmentation of the image by calculating the gray mean and standard deviation of a certain pixel in the r×r neighborhood, and there will be different thresholds T in different regions, which has a certain suppression effect on noise; its calculation formula is as shown in equations (23)~(25): ;(twenty three) ;(twenty four) (25) At this time, the image The high threshold within the region is T, and the low threshold is set to 0.5T. The defect edge is extracted by connecting the two thresholds to obtain the detection result Y2.

[0015] Preferably, the advantages of the iterative method in rapidly segmenting the target and background are combined with the ability of the NIBLACK algorithm to accurately identify the edges of the region of interest. The result Y1 processed by the iterative method is used as the background, and a logical AND operation is performed with the detection result Y2 of the NIBLACK algorithm. Only when the detection results of both algorithms at a certain point are true, the point is marked as an edge point; otherwise, if the detection result of either algorithm is false, the point is marked as a non-edge point. Subsequently, the detection results are subjected to dilation operation to connect the disconnected connected components, and the final edge result is obtained. The relevant operation expressions are shown in equations (26) and (27): (26) (27) The morphological operation element of P is a 3×3 template as shown in equation (28): (28) In the formula, Y3 is the result of the logical operation. Y4 is the output edge contour. ⊙ represents the expansion operation, and ⊙ represents the erosion operation.

[0016] Preferably, in step S5, the image optimization based on morphological operations includes: Based on the geometry of the transmission conductor and the characteristics of strand breakage, select appropriate morphological structural elements; Perform an erosion operation on the pre-processed image; After the erosion operation, the image is dilated. The image is further optimized by combining opening and closing operations; The edge features of the broken strand defect are highlighted by calculating the morphological gradient.

[0017] The beneficial effects of this invention are as follows: 1. Based on high-resolution camera image acquisition of overhead power transmission lines, the high-resolution camera can clearly capture the details of the lines, adapt to complex environments, provide a high-quality foundation for subsequent processing, and the detection process is non-contact, safe and reliable.

[0018] 2. Image enhancement processing based on grayscale stretching: Image contrast is enhanced by grayscale stretching to adapt to different lighting and backgrounds, highlight defect features, reduce the risk of missed detection, and provide clear image data for subsequent processing.

[0019] 3. Guided filtering-based image denoising: Guided filtering denoising can preserve the details of defect edges, avoid excessive smoothing, improve the image signal-to-noise ratio, enhance visual effects, and improve the accuracy and reliability of subsequent edge detection.

[0020] 4. Edge detection based on the improved Canny algorithm: The improved Canny algorithm provides multi-angle edge information, combined with adaptive threshold acquisition, which improves detection robustness and accuracy, reduces false judgments, and provides high-quality edge information for defect localization.

[0021] 5. Image optimization based on morphological operations: Morphological operations optimize images by removing noise, filling breaks, connecting connected components, preserving real defect features, making defect edges complete and clear, and improving detection accuracy and reliability. Attached Figure Description

[0022] Figure 1This is an image of a defective power transmission line in an embodiment of the present invention; Figure 2 This is a block diagram illustrating the improved Canny algorithm in an embodiment of the present invention; Figure 3 This is a gradient template diagram in an embodiment of the present invention; Figure 4 These are non-maximum suppression redundancy points in the embodiments of the present invention; Figure 5 This is the original image of the edge detection results in an embodiment of the present invention; Figure 6 This is a result image of edge detection in an embodiment of the present invention. Detailed Implementation

[0023] A method and system for detecting strand breakage defects in overhead power transmission lines based on machine vision, with specific embodiments as follows: Example 1: Acquiring images of overhead power transmission lines using a high-resolution camera; Example 2: Image enhancement processing based on grayscale stretching; Example 3: Image noise reduction processing based on guided filtering; Example 4: Edge detection based on the improved Canny algorithm; Example 5: Image optimization based on morphological operations.

[0024] Furthermore, in Example 1, an LA-CM-08K08A line scan camera is used as the image acquisition device. This camera has high resolution and fast imaging capabilities, enabling it to clearly capture the detailed features of the conductor surface. The camera is mounted on a movable inspection platform and continuously captures images along the conductor direction with the assistance of a track or drone. During image acquisition, different shooting modes are selected according to the flatness of the conductor surface and the inspection requirements: for conductors with relatively flat surfaces, a continuous shooting mode is used to ensure the continuity and integrity of the images; for conductors with undulating surfaces or complex environments, an intermittent shooting mode is used, automatically acquiring image data at certain intervals based on the platform's movement speed. Simultaneously, a DH-VHS-60mm lens and a DH-COL-180-130-W light source are used to ensure image clarity and contrast. This provides high-quality raw data for subsequent image processing and defect detection; such as... Figure 1 Image of a defective power transmission line.

[0025] Preferably, Example 2 specifically includes: By scaling the grayscale values ​​of the acquired image, they are uniformly adjusted to the preset grayscale range, thereby improving the contrast of the image; a template is used to scan each pixel in the region of interest and perform linear calculations. The grayscale value of the center pixel is set as the difference between the maximum and minimum values ​​of the template in the region. The specific formula is shown in Equation (1).

[0026] (1) In the formula, This is the output after stretching. The coefficient being multiplied. This is the offset value to be added; By grayscale values Mapped to This significantly enhances image contrast, thus meeting the requirements of the preprocessing stage. To ensure the system's robustness under different input conditions, it is necessary to maintain the output as much as possible. The stability of the parameters was assessed in conjunction with the actual production conditions on site. Preset parameters at this time The value of can be determined by formula (2).

[0027] (2) In the formula, The average gray level of the image is given by equation (3): (3) In the formula, Let p be the grayscale value of pixel p within region R.

[0028] Preferably, in Embodiment 3, guided filtering is a highly efficient filtering algorithm with constant time complexity. Its computation speed is independent of window size and it can effectively preserve image edges. Compared with isotropic filters, guided filtering is not only more efficient but also performs better in processing edge textures and details.

[0029] Guided filtering assumes that the guide image and the output image have a linear relationship within a two-dimensional template, mathematically expressed as equation (4): (4) In the formula, It is the output image grayscale value at that location It is the input image The gray values ​​at point k, and a and b are the coefficients of the linear function when the template center is located at point k. This is the template area.

[0030] Preferably, the coefficients of the linear function that minimize the value of equation (5) are calculated using the linear regression method, and this is taken as the optimal solution: (5) In the formula, It is the power transmission line to be filtered. It is an important parameter for adjusting the filtering effect of the filter, and is used to prevent the obtained result from being incorrect. Too large; Equations (6) to (7) are obtained using the least squares method: (6) (7) In equations (6) to (7), The image p to be filtered is in the template region The mean of the middle, Image I is in the template region The average value in It is a template area The number of pixels in the middle, Image I is in the template region The variance in; Substituting equations (6) and (7) into equation (4) yields indivual Value, for all The average of the values ​​yields the final result as shown in equation (8): (8) Using defective transmission lines as a guide diagram, i.e. ,at this time and The solutions are given by equations (9) and (10): (9) (10) When defects exist in a region, i.e., when marginal regions exist, the variance... Much larger ,at this time , This effectively preserves the gradient at the defect edge.

[0031] Preferably, in Example 4, the Canny algorithm for edge detection primarily relies on maximizing the gradient magnitude and direction of the region of interest to approximate the target edge as closely as possible. Compared to other common algorithms, it exhibits a higher signal-to-noise ratio. This paper optimizes the traditional Canny algorithm in two aspects: image denoising and dual-threshold connection, effectively compensating for its original shortcomings and thus achieving rapid extraction of wire defect edges. The block diagram of the improved algorithm is shown in Figure 2.

[0032] Edge detection based on the improved Canny algorithm includes the following steps: In a four-connected domain, the gradient information obtained by calculating finite differences using a 2×2 template fails to cover the diagonal direction. To improve the gradient calculation method, a scheme based on the Sobel algorithm is introduced. This scheme uses a weighted template of a 3×3 neighborhood to calculate the gradient values ​​in the x-direction, 45° direction, y-direction, and 135° direction respectively, and then performs a weighted summation. Compared with the traditional method, it increases the weights in the horizontal and vertical directions, thereby improving the accuracy of wire defect edge location and having a certain noise suppression effect. The gradient template of this algorithm is shown in Figure 3.

[0033] set up Let Gx represent the gray value at pixel (i, j) in the filtered image, and Gy represent the gray values ​​detected along the x and y directions at pixel (i, j), respectively. The calculation formulas are shown in equations (11) to (14): (11) (12) (13) (14) The horizontal difference with 45° and 135° characteristics is represented by equation (15): (15) The vertical difference with 45° and 135° features is given by equation (16): (16) Let Amplitude be the gradient magnitude at point (i, j), and θ(i, j) be the gradient direction. Then its calculation formula is as shown in equations (17) and (18): (17) (18) Preferably, non-maximum gradient suppression can eliminate redundant points at the defect edges, effectively reduce the error caused by global gradient extraction of edges, and reduce stray responses generated during edge detection.

[0034] like Figure 4 As shown, for the center pixel (i, j), its gradient intensity is Based on the gradient intensity in the x-direction gradient intensity in the y direction The sign and magnitude of the gradient can be calculated by linear interpolation along the pixel gradient direction within the connected region to determine the strength of two comparative gradients in the positive and negative gradient directions. and The calculation formulas are as shown in equations (19) and (20), and the value of the linear interpolation t is shown in equation (21): (19) (20) ;(twenty one) like = If the value is 0, then (i, j) is a non-edge point.

[0035] Preferably, the iterative method and the Niblack algorithm are combined to adaptively determine the high and low thresholds in the dual-threshold connection. The iterative method has a simple calculation process and can quickly distinguish the target from the background; the Niblack algorithm performs well in identifying the edges of broken strand defects in transmission lines. By fusing these two algorithms, the detection of defect edges in transmission lines is realized.

[0036] Based on the image after nonmaximum suppression The grayscale histogram is used as an initial threshold, with the median grayscale value Z0 selected. The threshold is then iterated according to the following formula: ;(twenty two) In the formula, L represents the gray level of the image. It is the number of pixels with a gray value of k in the gray-level histogram, and the iteration proceeds to... End, take the iteration threshold Zi at the end.

[0037] Preferably, a grayscale image is used. The size is M×N, the image gray value is f(x,y), the gray mean is m(x,y), and the gray standard deviation is s(x,y); the NIBLACK algorithm realizes dynamic threshold segmentation of the image by calculating the gray mean and standard deviation of a certain pixel in the r×r neighborhood, and there will be different thresholds T in different regions, which has a certain suppression effect on noise; its calculation formula is as shown in equations (23)~(25): ;(twenty three) ;(twenty four) (25) At this time, the image The high threshold within the region is T, and the low threshold is set to 0.5T. The defect edge is extracted by connecting the two thresholds to obtain the detection result Y2.

[0038] Preferably, the iterative method's advantage in rapidly segmenting the target and background is combined with the NIBLACK algorithm's ability to accurately identify the edges of regions of interest. The result Y1 processed by the iterative method is used as the background, and a logical AND operation is performed with the detection result Y2 from the NIBLACK algorithm. A point is marked as an edge only if both algorithms detect a true value at a given point; otherwise, if either algorithm detects a false value, the point is marked as a non-edge. Subsequently, dilation is performed on the detection results to connect disconnected connected components, resulting in the final edge result as shown below. Figure 5 and Figure 6 As shown; the relevant operational expressions are shown in equations (26) and (27): (26) (27) The morphological operation element of P is a 3×3 template as shown in equation (28): (28) In the formula, Y3 is the result of the logical operation. Y4 is the output edge contour. ⊙ represents the expansion operation, and ⊙ represents the erosion operation.

[0039] Preferably, in embodiment 5, the image optimization based on morphological operations includes: Based on the geometry of the transmission conductor and the characteristics of strand breakage, select an appropriate morphological structural element. The structural element is typically linear or circular, and its dimensions should be adjusted according to the conductor's diameter and the typical width of the strand breakage. For example, for common overhead transmission conductors, a linear structural element with a length of 3-5 pixels and a width of 1 pixel can be selected to capture the conductor's longitudinal features and subtle variations in strand breakage.

[0040] Erosion is performed on the image after preprocessing (such as binarized images after edge detection). The purpose of erosion is to eliminate small noise points, burrs, and redundant connections at the edges of broken strand defects. Erosion makes the edges of broken strand defects clearer, reducing misjudgments caused by noise interference. The number of iterations for erosion should be adjusted according to the noise level of the actual image and the characteristics of the broken strand defect, generally 1-2 times.

[0041] After the erosion operation, a dilation operation is performed on the image. The purpose of dilation is to restore the broken strand defect areas that may have been over-cut during the erosion operation, while also connecting small defect segments that were broken off by the erosion operation. Dilation enhances the connectivity of the broken strand defects, making the defect areas more complete and facilitating subsequent feature extraction and recognition. The number of iterations for the dilation operation should also be adjusted according to the actual image conditions, generally the same as or slightly more than the number of iterations for the erosion operation.

[0042] The image is further optimized by combining opening and closing operations. Opening is an operation that involves erosion followed by dilation, primarily used to remove small objects and connecting parts in the image, while smoothing the boundaries of larger objects. Closing is an operation that involves dilation followed by erosion, primarily used to fill small holes and broken parts in the image, and to connect adjacent objects. In this step, the appropriate order and number of opening and closing operations can be selected based on the characteristics of the transmission line image and the morphological features of the broken strand defects to achieve the best image optimization effect. For example, for images with many small breaks and noise interference, opening operations can be performed first to remove noise and small breaks, followed by closing operations to connect the main body of the broken strand defects.

[0043] Morphological gradients are used to highlight the edge features of the broken strand defect. The morphological gradient is obtained by subtracting the dilated image from the eroded image, effectively highlighting the edge information of objects in the image. In this step, morphological gradient calculation further enhances the contrast between the broken strand defect and the background, making the edges of the defect more prominent and facilitating accurate location and identification of the defect by subsequent feature extraction and recognition algorithms.

Claims

1. A machine vision based method for detection of broken strand defects in overhead power transmission conductors, characterized by, The method comprises the following steps: S1, collecting overhead power transmission conductor image based on high-resolution camera; S2, image enhancement processing based on gray scale stretching; S3, image noise reduction processing based on guided filtering; S4, edge detection based on improved Canny algorithm; S5, image optimization based on morphological operation.

2. The machine vision-based method for detecting broken conductor defects of overhead power transmission lines according to claim 1, wherein, Step S1 specifically comprises: A line scan camera is used as an image acquisition device, and is matched with a lens and a light source; the camera is installed on a movable detection platform, and continuous shooting mode or intermittent shooting mode is selected according to the flatness of the conductor surface. 3.The machine vision-based detection method for broken conductor defect of overhead transmission line according to claim 1, characterized in that, Step S2 specifically comprises: The gray scale value of the collected image is scaled to adjust it to a preset gray scale range, so as to improve the contrast of the image; a template is used to scan each pixel in the region of interest, and linear calculation is performed, and the gray scale value of the center pixel is set as the difference between the maximum value and the minimum value of the template in the region, and the specific formula is shown as formula (1): ；（1） In the formula, is the output after stretching, is the coefficient to be multiplied, is the offset value to be added.

4. The machine vision-based method for detecting broken conductor defects of overhead power transmission lines according to claim 3, wherein, In the step S2: by mapping the gray value to , combining the actual situation of field production to preset the parameter , at this time the value of the parameter can be determined by formula (2): ；（2） In the formula, is the average gray level of the image, which has a value as in equation (3): ；（3） In the formula, is the gray value of the pixel point p in the region R.

5. The machine vision-based method for detecting broken conductor defects of overhead power transmission lines according to claim 1, wherein, In step S3, the guided filtering includes assuming that the guided image and the output image are in linear relationship in a two-dimensional template, and the mathematical expression is shown as formula (4): ；（4） wherein is the gray value at the output image wherein is the gray value at the input image wherein a and b are coefficients of the linear function when the template center is located at k, is the template region.

6. The machine vision-based method for detecting broken conductor defects of overhead power transmission lines according to claim 5, wherein, The linear function coefficients that make the value of formula (5) reach the minimum are calculated by using the linear regression method as the optimal solution: ； （5） In the formula, is the power transmission line to be filtered, is an important parameter for adjusting the filtering effect of the filter, and is used to prevent the obtained from being too large; Formula (6)~(7) are obtained by the least square method: ；（6） ；（7） in formulas (6) to (7), is the mean value of the image p to be filtered in the template region , is the mean value of the image I in the template region , is the number of pixels in the template region , is the variance of the image I in the template region ; Substituting equations (6) and (7) into equation (4) gives one values, and averaging over all values gives the final result as equation (8): ；（8） The defective power transmission line is used as a guide map, i.e. at this time and the solution of formula (9) and formula (10): ；（9） ； （10） When the region is defective, i.e. there is an edge region, the variance is much larger than , , effectively preserving the gradient of the defect edge.

7. The machine vision-based method for detecting broken conductor defects of overhead power transmission lines according to claim 1, wherein, In the step S4, the edge detection based on the improved Canny algorithm comprises the following steps: A 3*3 neighborhood weighted template is used to calculate the gradient values in x direction, 45° direction, y direction and 135° direction respectively, and weighted summation is performed.

8. The machine vision-based method for detecting broken conductor defects of overhead power transmission lines according to claim 7, wherein, In the step S4: Set G(i, j) represents the gray value at pixel point (i, j) in the filtered image, Gx and Gy represent the gray values of edge detection in x direction and y direction at pixel point (i, j) respectively, and the calculation formula is as formula (11)~formula (14): ；（11） ；（12） ；（13） ； （14） The horizontal difference with 45° and 135° characteristics is formula (15): ；（15） The vertical difference with 45° and 135° characteristics is formula (16): ；（16） Assuming that Amplitude is the gradient amplitude at point (i, j), the calculation formula is shown as formula (17) and formula (18): ；（17） ；（18） wherein is the gradient direction. 9.The machine vision based overhead power transmission conductor broken strand defect detection method of claim 8, wherein, For the center pixel point (i, j), its gradient strength is ; according to the positive and size of the gradient strength in x direction and the gradient strength in y direction , two comparative gradient strengths in the positive and negative gradient directions and can be calculated by linear interpolation along the pixel gradient direction in the connected domain, and the calculation formula is as formula (19) and formula (20), and the value of linear interpolation t is as formula (21): ；（19） ；（20） ； （21） like = If the value is 0, then (i, j) is a non-edge point. 10.The machine vision-based detection method for broken conductor defect of overhead power transmission line according to claim 9, characterized in that, In the step S4, the high and low threshold values in the double threshold connection are adaptively determined by combining the iterative method and the NIBLACK algorithm: According to the gray scale histogram of the image after non-maximum suppression , the gray scale median Z0 is selected as the initial threshold, and threshold iteration is performed according to the following formula: ；（22） where L is the gray level of the image, is the number of pixels with gray value k in the gray histogram, and the iteration is performed until Zi at the end is the iteration threshold.

11. The machine vision-based method for detecting broken conductor defects of overhead power transmission lines according to claim 10, wherein, Set the gray image The size of the image is M x N, the image gray value is f(x, y), the gray mean value is m(x, y), and the gray standard deviation is s(x, y); the NIBLACK algorithm realizes dynamic threshold segmentation of the image by calculating the gray mean value and standard deviation of a certain pixel point in a r x r field, and different threshold values T will be obtained in different regions, which has a certain inhibitory effect on noise. The calculation formula is shown as formula (23)~(25): ；（23） ；（24） ；（25） At this time, the image The high threshold value corresponding to the region is T, and the low threshold value is set to 0.5T; the defect edge is extracted by double-threshold connection to obtain a detection result Y2.

12. The machine vision-based method for detecting broken conductor defects of overhead power transmission lines according to claim 11, wherein, In the step S4, the advantages of the iterative method in fast segmentation of targets and backgrounds and the ability of the NIBLACK algorithm to accurately identify the edges of the region of interest are combined, the result Y1 processed by the iterative method is taken as the background, and the logical AND operation is performed with the detection result Y2 of the NIBLACK algorithm; only when the detection results of the two algorithms at a certain point are both true, the point is marked as an edge point; Otherwise, if the detection result of any algorithm is false, the point is marked as a non-edge point; then, the detection result is dilated to connect the disconnected connected domains, and finally the edge result is obtained; the related operation expressions are shown as formula (26) and formula (27): ；（26） ；（27） The P morphological operation element is a 3*3 template as shown in formula (28): ；（28） In the formula, Y3 is the result of logical operation; Y4 is the output edge profile, is the dilatation operation, and is the erosion operation.

13. The machine vision-based method for detecting broken conductor defects of overhead power transmission lines according to claim 1, wherein, In the step S5, the image optimization based on morphological operation comprises: According to the geometric shape of the power transmission conductor and the characteristics of the broken strand defect, a suitable morphological structure element is selected; An erosion operation is performed on the image processed in the previous stage; After the erosion operation, an expansion operation is performed on the image; Further optimization is performed on the image by combining the open operation and the close operation; Edge features of the broken strand defects are highlighted by calculating morphological gradients.

14. The machine vision-based method for detecting broken conductor defects of overhead power transmission lines according to claim 13, wherein, In the step S5, edge features of the broken strand defects are highlighted by calculating morphological gradients, which are obtained by subtracting the eroded image from the dilated image.

15. A machine vision based aerial power conductor strand breakage defect detection system characterized in that, The system is used for implementing the method according to any one of claims 1-14, and comprises: An image acquisition module configured to acquire images of overhead transmission lines based on a high-resolution camera; An image processing module configured to sequentially perform operations of gray stretch enhancement, guided filter denoising, improved Canny algorithm edge detection and morphological optimization on the images; A defect identification and output module configured to identify and locate the broken strand defects based on the processed images.

Images