Visual feature data optimization method and system for unmanned aerial vehicle power distribution line inspection

Abstract

The application is suitable for the technical field of unmanned aerial vehicle power distribution line inspection, and provides a visual feature data optimization method and system for unmanned aerial vehicle power distribution line inspection. The application obtains an inspection image photographed by an unmanned aerial vehicle during power distribution line inspection, judges whether there is an adjustable obstruction, when there is an adjustable obstruction, performs influence reconstruction on the obstruction area in the inspection image to generate a reconstructed image, performs line feature recognition, selects the longest straight line segment target, calculates the target width, recognizes the horizontal / vertical deviation angle of the straight line segment target, and performs deviation correction and rotation adjustment on the reconstructed image. After the obstruction reconstruction of the inspection image, the line feature recognition is performed, the longest straight line segment target is selected, the width calculation and comparison are performed, the horizontal / vertical deviation angle is recognized, and the deviation correction and rotation adjustment are performed, so that the image can be rotated to a unified standard direction, the problem of complex and chaotic visual features is reduced, and the difficulty of subsequent manual recognition or automatic recognition is reduced.

Description

Technical Field

[0001] This invention belongs to the field of UAV power line inspection technology, and particularly relates to a method and system for optimizing visual feature data for UAV power line inspection. Background Technology

[0002] Unmanned aerial vehicle (UAV) power line inspection is a power operation and maintenance technology that uses UAVs equipped with visible light cameras, infrared thermal imagers, lidar, multispectral cameras, and other detection equipment to conduct automated and intelligent inspections of power lines and their auxiliary equipment according to preset routes or autonomously planned paths.

[0003] Compared with traditional manual inspection, drone power line inspection has advantages such as high inspection efficiency, wide coverage, low safety risk, high data accuracy and strong environmental adaptability. It is especially suitable for areas that are difficult to reach by humans, such as mountainous areas, forest areas, river crossing areas and complex terrain.

[0004] In actual drone power line inspections, the drone's flight attitude and shooting angle can cause the inspected target to appear at different tilt angles in the image, resulting in complex and confusing visual features, which increases the difficulty of subsequent manual or automatic identification. Summary of the Invention

[0005] The purpose of this invention is to provide a method and system for optimizing visual feature data for unmanned aerial vehicle (UAV) power line inspection, aiming to solve the technical problems existing in the prior art mentioned in the background.

[0006] The embodiments of the present invention are implemented as follows:

[0007] A method for optimizing visual feature data for UAV power line inspection, the method specifically includes the following steps:

[0008] The system acquires inspection images taken by a drone during power line inspection, performs occlusion identification and area comparison on the inspection images, and determines whether there is adjustable occlusion.

[0009] When adjustable occlusion exists, the occlusion area is marked and the affected area is planned. The occlusion area in the inspection image is reconstructed to generate a reconstructed image.

[0010] Line feature recognition is performed on the reconstructed image to identify multiple line segment objects, and the lengths of the multiple line segment objects are compared to select the longest line segment target, and the target width is calculated.

[0011] The target width is compared with a preset width threshold, and the horizontal / vertical deviation angle of the target line segment is identified. The reconstructed image is then rotated and adjusted to correct the deviation.

[0012] As a further limitation of the technical solution of this embodiment of the invention, the step of acquiring inspection images taken by the drone during power line inspection, and performing occlusion identification and area comparison on the inspection images to determine whether there is adjustable occlusion specifically includes the following steps:

[0013] Acquire inspection images taken by drones during power distribution line inspections;

[0014] The inspection image is subjected to occlusion identification to determine whether there is occlusion.

[0015] When there is obstruction, determine the obstruction area;

[0016] Calculate the occlusion ratio based on the occlusion area;

[0017] The occlusion ratio is compared with the preset occlusion invalid ratio and image invalid ratio;

[0018] When the occlusion ratio is less than the invalid occlusion ratio, it is determined that there is invalid occlusion;

[0019] When the occlusion ratio is greater than the invalid image ratio, the inspection image is determined to be invalid.

[0020] When the occlusion ratio is between the occlusion invalid ratio and the image invalid ratio, it is determined that there is adjustable occlusion.

[0021] As a further limitation of the technical solution of this invention embodiment, the step of marking the occlusion area and planning the affected area when adjustable occlusion exists, and reconstructing the occlusion area in the inspection image to generate a reconstructed image specifically includes the following steps:

[0022] When adjustable occlusion exists, mark the occluded area;

[0023] Determine the center point of the occlusion area;

[0024] The affected area is calculated by multiplying the occlusion area by a preset impact factor.

[0025] Based on the center point of the obstruction and the area of ​​influence, plan the affected region;

[0026] Based on the affected area, an inverse distance influence analysis is performed on the occluded area to calculate multiple reconstruction values;

[0027] Based on the multiple reconstruction values, the occluded areas in the inspection image are reconstructed and filled to generate a reconstructed image.

[0028] As a further limitation of the technical solution of this embodiment of the invention, the calculation formula for the plurality of reconstructed values ​​is as follows:

[0029] ;

[0030] in, For the first in the occluded area The reconstructed value of each pixel; For the affected area The influence values ​​of each pixel are totaled. 1 pixel; For the first in the occluded area The pixel and the first pixel in the affected area The distance between pixels.

[0031] As a further limitation of the technical solution of this embodiment of the invention, the step of performing line feature recognition on the reconstructed image, determining multiple line segment objects, comparing the lengths of the multiple line segment objects, selecting the longest line segment target, and calculating the target width specifically includes the following steps:

[0032] Line feature recognition is performed on the reconstructed image to identify multiple line segment objects;

[0033] Calculate the lengths of the line segments from multiple line segment objects;

[0034] The lengths of multiple line segments are compared and sorted, and the longest line segment target is selected from the multiple line segment objects;

[0035] Calculate the target width of the straight line segment target.

[0036] As a further limitation of the technical solution of this embodiment of the invention, the step of comparing the target width with a preset width threshold and identifying the horizontal / vertical deviation angle of the straight line segment target, and performing correction and rotation adjustment on the reconstructed image specifically includes the following steps:

[0037] The target width is compared with a preset width threshold;

[0038] When the target width is less than a width threshold, the horizontal deviation angle of the straight line segment target is identified;

[0039] The reconstructed image is horizontally rotated to correct the deviation according to the stated horizontal deviation angle.

[0040] When the target width is not less than the width threshold, identify the vertical deviation angle of the target line segment;

[0041] The reconstructed image is vertically rotated to correct the deviation according to the stated vertical deviation angle.

[0042] The invalid borders of the image after horizontal or vertical skew rotation are cropped to obtain the optimized image.

[0043] A visual feature data optimization system for UAV power line inspection includes an image occlusion judgment module, an occlusion region reconstruction module, a straight line segment recognition and comparison module, and a deviation correction and rotation adjustment module, wherein:

[0044] The image occlusion judgment module is used to acquire inspection images taken by the drone during power line inspection, perform occlusion recognition and area comparison on the inspection images, and determine whether there is adjustable occlusion.

[0045] The occlusion region reconstruction module is used to mark the occlusion region and plan the affected region when there is adjustable occlusion, and to reconstruct the occlusion region in the inspection image to generate a reconstructed image.

[0046] The line segment recognition and comparison module is used to perform line feature recognition on the reconstructed image, identify multiple line segment objects, compare the lengths of the multiple line segment objects, select the longest line segment target, and calculate the target width.

[0047] The skew correction and rotation adjustment module is used to compare the target width with a preset width threshold, identify the horizontal / vertical deviation angle of the straight line segment target, and perform skew correction and rotation adjustment on the reconstructed image.

[0048] As a further limitation of the technical solution of this embodiment of the invention, the image occlusion determination module specifically includes:

[0049] The inspection image acquisition unit is used to acquire inspection images taken by the drone during the inspection of power distribution lines;

[0050] An occlusion recognition unit is used to perform occlusion recognition on the inspection image and determine whether there is an occlusion.

[0051] The occlusion area determination unit is used to determine the occlusion area when occlusion exists.

[0052] An occlusion ratio calculation unit is used to calculate the occlusion ratio based on the occlusion area;

[0053] A ratio comparison unit is used to compare the occlusion ratio with a preset occlusion invalid ratio and an image invalid ratio;

[0054] An invalid occlusion determination unit is used to determine that there is invalid occlusion when the occlusion ratio is less than the invalid occlusion ratio;

[0055] An image invalidity determination unit is used to determine that the inspection image is invalid when the occlusion ratio is greater than the image invalidity ratio;

[0056] An adjustable occlusion determination unit is used to determine that adjustable occlusion exists when the occlusion ratio is between the invalid occlusion ratio and the invalid image ratio.

[0057] As a further limitation of the technical solution of this embodiment of the invention, the occlusion area reconstruction module specifically includes:

[0058] An occlusion area marking unit is used to mark the occlusion area when adjustable occlusion exists;

[0059] A center point determination unit is used to determine the occlusion center point of the occlusion area;

[0060] An impact area calculation unit is used to multiply the shading area by a preset impact factor to calculate the impact area.

[0061] An impact area planning unit is used to plan the impact area based on the shading center point and the impact area;

[0062] The reconstruction value calculation unit is used to perform inverse distance influence analysis on the occluded area based on the influence area and calculate multiple reconstruction values;

[0063] The reconstruction filling unit is used to perform relevant reconstruction filling on the occluded areas in the inspection image according to multiple reconstruction values ​​to generate a reconstructed image.

[0064] As a further limitation of the technical solution of this embodiment of the invention, the straight line segment recognition and comparison module specifically includes:

[0065] A line feature recognition unit is used to perform line feature recognition on the reconstructed image to determine multiple line segment objects;

[0066] A line segment length calculation unit is used to calculate the line segment length of multiple line segment objects;

[0067] A line segment length comparison unit is used to compare and arrange the lengths of multiple line segments, and select the longest line segment target from the multiple line segment objects;

[0068] The target width calculation unit is used to calculate the target width of the straight line segment target.

[0069] Compared with the prior art, the beneficial effects of the present invention are:

[0070] This invention acquires inspection images taken by a drone during power line inspections, determines whether adjustable occlusion exists, and if so, reconstructs the occluded area in the inspection image to generate a reconstructed image. Line feature recognition is performed, the longest straight line segment is selected, and the target width is calculated. The horizontal / vertical deviation angle of the straight line segment is identified, and the reconstructed image is adjusted by rotation to correct this deviation. By reconstructing the inspection image after occlusion, performing line feature recognition, selecting the longest straight line segment, calculating and comparing its width, identifying the horizontal / vertical deviation angle, and performing rotation adjustment, the image can be rotated to a uniform standard direction, reducing the complexity and confusion of visual features and lowering the difficulty of subsequent manual or automatic recognition. Attached Figure Description

[0071] Figure 1 A flowchart of a visual feature data optimization method for UAV power line inspection provided by an embodiment of the present invention is shown;

[0072] Figure 2 The diagram illustrates the application architecture of a visual feature data optimization system for UAV power line inspection provided by an embodiment of the present invention. Detailed Implementation

[0073] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0074] Understandably, in the current process of actual drone power line inspection, the changes in the drone's flight attitude and shooting angle will cause the inspected target to appear at different tilt angles in the image, resulting in complex and confusing visual features, which increases the difficulty of subsequent manual or automatic identification.

[0075] To address the aforementioned issues, this invention discloses a visual feature data optimization method and system for UAV power line inspection. This method acquires inspection images captured by the UAV during power line inspection, performs occlusion identification and area comparison on the images to determine if adjustable occlusion exists. If adjustable occlusion exists, the occlusion area is marked, and the affected area is planned. The occlusion area in the inspection image is then reconstructed to generate a reconstructed image. Line feature recognition is performed on the reconstructed image to identify multiple straight line segments. The lengths of these segments are compared, and the longest segment is selected. The target width is calculated. The target width is compared with a preset width threshold, and the horizontal / vertical deviation angle of the straight line segment is identified. The reconstructed image is then adjusted by rotation to correct this deviation. This method enables the image to be rotated to a uniform standard direction after occlusion reconstruction, reducing the complexity and confusion of visual features and lowering the difficulty of subsequent manual or automatic identification.

[0076] Specifically, Figure 1 A flowchart of a visual feature data optimization method for UAV power line inspection provided by an embodiment of the present invention is shown.

[0077] In a preferred embodiment of the present invention, a method for optimizing visual feature data for UAV power line inspection specifically includes the following steps:

[0078] Step S101: Obtain inspection images taken by the drone during power line inspection, perform occlusion identification and area comparison on the inspection images, and determine whether there is adjustable occlusion.

[0079] In this embodiment of the invention, when a drone inspects power distribution lines, it takes inspection photos and transmits images. By acquiring the inspection images transmitted by the drone, it performs occlusion identification on the images to determine if occlusion exists. If occlusion exists, it determines the occlusion area in the inspection image. It then calculates the occlusion ratio by dividing the occlusion area by the total area of ​​the inspection image. This ratio is then compared with preset invalid occlusion ratios and invalid image ratios to determine the nature of the occlusion. Specifically, if the occlusion ratio is less than the invalid occlusion ratio, invalid occlusion is determined to exist. In this case, the occlusion in the inspection image can be considered as minor noise and ignored, requiring no further reconstruction processing. If the occlusion ratio is greater than the invalid image ratio, the inspection image is determined to be invalid. In this case, the occlusion in the inspection image is too large, severely affecting the image's validity. If the occlusion ratio is between the invalid occlusion ratio and the invalid image ratio, adjustable occlusion is determined to exist, allowing for subsequent reconstruction processing.

[0080] It is understood that the occlusion invalidity ratio is much smaller than the image invalidity ratio. In this embodiment of the invention, the occlusion invalidity ratio can be set to 0.005; the image invalidity ratio can be set to 0.05.

[0081] Specifically, in another preferred embodiment provided by the present invention, the step of acquiring inspection images taken by the UAV during power line inspection, and performing occlusion identification and area comparison on the inspection images to determine whether there is adjustable occlusion specifically includes the following steps:

[0082] Acquire inspection images taken by drones during power distribution line inspections;

[0083] The inspection image is subjected to occlusion identification to determine whether there is occlusion.

[0084] When there is obstruction, determine the obstruction area;

[0085] Calculate the occlusion ratio based on the occlusion area;

[0086] The occlusion ratio is compared with the preset occlusion invalid ratio and image invalid ratio;

[0087] When the occlusion ratio is less than the invalid occlusion ratio, it is determined that there is invalid occlusion;

[0088] When the occlusion ratio is greater than the invalid image ratio, the inspection image is determined to be invalid.

[0089] When the occlusion ratio is between the occlusion invalid ratio and the image invalid ratio, it is determined that there is adjustable occlusion.

[0090] Furthermore, the visual feature data optimization method for UAV power line inspection also includes the following steps:

[0091] Step S102: When adjustable occlusion exists, mark the occlusion area and plan the affected area, reconstruct the occlusion area in the inspection image, and generate a reconstructed image.

[0092] In this embodiment of the invention, when adjustable occlusion is determined, the occlusion area in the inspection image is marked, and the occlusion center point of the occlusion area is determined. Then, the occlusion area is multiplied by a preset influence factor to calculate the influence area. After that, an influence area of ​​the influence area is planned outside the occlusion area with the occlusion center point as the center. Based on the influence area, inverse distance influence analysis is performed on the occlusion area to calculate multiple reconstruction values. Then, according to the multiple reconstruction values, each pixel of the occlusion area in the inspection image is reconstructed and filled to generate a reconstructed image. Specifically, the calculation formula for the multiple reconstruction values ​​is as follows:

[0093] ;

[0094] in, For the first in the occluded area The reconstructed value of each pixel; For the affected area The influence values ​​of each pixel are totaled. 1 pixel; For the first in the occluded area The pixel and the first pixel in the affected area The distance between pixels.

[0095] It is understandable that if the inspected image is a grayscale image, the calculated multiple reconstruction values ​​are the grayscale values ​​of multiple pixels in the occluded area, and reconstruction filling is performed; if the inspected image is a color image, it is necessary to calculate the red, green and blue channel values ​​of each pixel in the occluded area, and then perform comprehensive reconstruction filling.

[0096] It is understood that, in this embodiment of the invention, the influence factor can be 10.

[0097] Specifically, in another preferred embodiment provided by the present invention, when adjustable occlusion exists, marking the occlusion area and planning the affected area, and reconstructing the occlusion area in the inspection image to generate a reconstructed image specifically includes the following steps:

[0098] When adjustable occlusion exists, mark the occluded area;

[0099] Determine the center point of the occlusion area;

[0100] The affected area is calculated by multiplying the occlusion area by a preset impact factor.

[0101] Based on the center point of the obstruction and the area of ​​influence, plan the affected region;

[0102] Based on the affected area, an inverse distance influence analysis is performed on the occluded area to calculate multiple reconstruction values;

[0103] Based on the multiple reconstruction values, the occluded areas in the inspection image are reconstructed and filled to generate a reconstructed image.

[0104] Furthermore, the visual feature data optimization method for UAV power line inspection also includes the following steps:

[0105] Step S103: Perform line feature recognition on the reconstructed image to identify multiple line segment objects, compare the lengths of the multiple line segment objects, select the longest line segment target, and calculate the target width.

[0106] In this embodiment of the invention, multiple line segment objects are identified by performing line feature recognition on the reconstructed image, and the line segment lengths of the multiple line segment objects are calculated. The multiple line segment lengths are then compared and arranged, and the largest target length is selected from the multiple line segment lengths. The line segment object corresponding to the target length is then determined as the line segment target, and the target width of the line segment target is then calculated.

[0107] Specifically, in another preferred embodiment provided by the present invention, the step of performing line feature recognition on the reconstructed image to determine multiple line segment objects, comparing the lengths of the multiple line segment objects, selecting the longest line segment target, and calculating the target width specifically includes the following steps:

[0108] Line feature recognition is performed on the reconstructed image to identify multiple line segment objects;

[0109] Calculate the lengths of the line segments from multiple line segment objects;

[0110] The lengths of multiple line segments are compared and sorted, and the longest line segment target is selected from the multiple line segment objects;

[0111] Calculate the target width of the straight line segment target.

[0112] Furthermore, the visual feature data optimization method for UAV power line inspection also includes the following steps:

[0113] Step S104: Compare the target width with a preset width threshold, identify the horizontal / vertical deviation angle of the straight line segment target, and perform correction rotation adjustment on the reconstructed image.

[0114] In this embodiment of the invention, the target width is compared with a preset width threshold. If the target width is less than the width threshold, the straight line segment target is assumed to be a wire and should be horizontal. In this case, the horizontal deviation angle of the straight line segment target is identified, and the reconstructed image is horizontally rotated according to the horizontal deviation angle to make the straight line segment target return to a horizontal state. If the target width is not less than the width threshold, the straight line segment target is assumed to be a pole and should be vertical. In this case, the vertical deviation angle of the straight line segment target is identified, and the reconstructed image is vertically rotated according to the vertical deviation angle to make the straight line segment target return to a vertical state. After that, the invalid borders of the image after horizontal or vertical rotation are cropped to obtain the optimized image.

[0115] It is understandable that invalid borders are blank filled areas at the four corners of an image after it has been rotated, containing no original image information.

[0116] It is understandable that when drones inspect power distribution lines, they maintain a certain distance from the power distribution lines while taking pictures. Therefore, in the images captured, the width of the conductors and towers does not change much, and the width of the conductors and towers themselves are quite different. Therefore, the conductors or towers can be distinguished by comparing the target width with a preset width threshold.

[0117] Specifically, in another preferred embodiment provided by the present invention, the step of comparing the target width with a preset width threshold and identifying the horizontal / vertical deviation angle of the straight line segment target, and performing correction rotation adjustment on the reconstructed image specifically includes the following steps:

[0118] The target width is compared with a preset width threshold;

[0119] When the target width is less than a width threshold, the horizontal deviation angle of the straight line segment target is identified;

[0120] The reconstructed image is horizontally rotated to correct the deviation according to the stated horizontal deviation angle.

[0121] When the target width is not less than the width threshold, identify the vertical deviation angle of the target line segment;

[0122] The reconstructed image is vertically rotated to correct the deviation according to the stated vertical deviation angle.

[0123] The invalid borders of the image after horizontal or vertical skew rotation are cropped to obtain the optimized image.

[0124] Furthermore, Figure 2 The diagram illustrates the application architecture of a visual feature data optimization system for UAV power line inspection provided by an embodiment of the present invention.

[0125] Specifically, in another preferred embodiment of the present invention, a visual feature data optimization system for UAV power line inspection includes:

[0126] The image occlusion judgment module 101 is used to acquire inspection images taken by the UAV during power line inspection, perform occlusion recognition and area comparison on the inspection images, and determine whether there is adjustable occlusion.

[0127] In this embodiment of the invention, when a drone inspects power distribution lines, it takes inspection photos and transmits images. The image occlusion judgment module 101 acquires the inspection images transmitted by the drone and identifies occlusion in the inspection images to determine whether occlusion exists. If occlusion exists, it determines the occlusion area in the inspection image and calculates the occlusion ratio by dividing the occlusion area by the total area of ​​the inspection image. Then, it compares the occlusion ratio with preset invalid occlusion ratios and invalid image ratios to determine the nature of the occlusion. Specifically, if the occlusion ratio is less than the invalid occlusion ratio, invalid occlusion is determined to exist. In this case, the occlusion in the inspection image can be regarded as minor noise and ignored, without the need for subsequent reconstruction processing. If the occlusion ratio is greater than the invalid image ratio, the inspection image is determined to be invalid. In this case, the occlusion in the inspection image is too large, which seriously affects the validity of the image. If the occlusion ratio is between the invalid occlusion ratio and the invalid image ratio, adjustable occlusion is determined to exist, and subsequent reconstruction processing can be performed.

[0128] Specifically, in another preferred embodiment provided by the present invention, the image occlusion determination module 101 specifically includes:

[0129] The inspection image acquisition unit is used to acquire inspection images taken by the drone during the inspection of power distribution lines;

[0130] An occlusion recognition unit is used to perform occlusion recognition on the inspection image and determine whether there is an occlusion.

[0131] The occlusion area determination unit is used to determine the occlusion area when occlusion exists.

[0132] An occlusion ratio calculation unit is used to calculate the occlusion ratio based on the occlusion area;

[0133] A ratio comparison unit is used to compare the occlusion ratio with a preset occlusion invalid ratio and an image invalid ratio;

[0134] An invalid occlusion determination unit is used to determine that there is invalid occlusion when the occlusion ratio is less than the invalid occlusion ratio;

[0135] An image invalidity determination unit is used to determine that the inspection image is invalid when the occlusion ratio is greater than the image invalidity ratio;

[0136] An adjustable occlusion determination unit is used to determine that adjustable occlusion exists when the occlusion ratio is between the invalid occlusion ratio and the invalid image ratio.

[0137] Furthermore, the visual feature data optimization system for UAV power line inspection also includes:

[0138] The occlusion area reconstruction module 102 is used to mark the occlusion area and plan the affected area when there is adjustable occlusion, and to reconstruct the occlusion area in the inspection image to generate a reconstructed image.

[0139] In this embodiment of the invention, when adjustable occlusion is determined, the occlusion area reconstruction module 102 marks the occlusion area in the inspection image and determines the occlusion center point of the occlusion area. Then, according to a preset influence factor, the occlusion area is multiplied to calculate the influence area. After that, with the occlusion center point as the center, an influence area of ​​the influence area is planned outside the occlusion area. Based on the influence area, inverse distance influence analysis is performed on the occlusion area to calculate multiple reconstruction values. Then, according to the multiple reconstruction values, each pixel of the occlusion area in the inspection image is reconstructed and filled to generate a reconstructed image. Specifically, the calculation formula for the multiple reconstruction values ​​is as follows:

[0140] ;

[0141] in, For the first in the occluded area The reconstructed value of each pixel; For the affected area The influence values ​​of each pixel are totaled. 1 pixel; For the first in the occluded area The pixel and the first pixel in the affected area The distance between pixels.

[0142] Specifically, in another preferred embodiment provided by the present invention, the occlusion area reconstruction module 102 specifically includes:

[0143] An occlusion area marking unit is used to mark the occlusion area when adjustable occlusion exists;

[0144] A center point determination unit is used to determine the occlusion center point of the occlusion area;

[0145] An impact area calculation unit is used to multiply the shading area by a preset impact factor to calculate the impact area.

[0146] An impact area planning unit is used to plan the impact area based on the shading center point and the impact area;

[0147] The reconstruction value calculation unit is used to perform inverse distance influence analysis on the occluded area based on the influence area and calculate multiple reconstruction values;

[0148] The reconstruction filling unit is used to perform relevant reconstruction filling on the occluded areas in the inspection image according to multiple reconstruction values ​​to generate a reconstructed image.

[0149] Furthermore, the visual feature data optimization system for UAV power line inspection also includes:

[0150] The line segment recognition and comparison module 103 is used to perform line feature recognition on the reconstructed image, determine multiple line segment objects, compare the lengths of the multiple line segment objects, select the longest line segment target, and calculate the target width.

[0151] In this embodiment of the invention, the line segment recognition and comparison module 103 identifies multiple line segment objects by performing line feature recognition on the reconstructed image, calculates the line segment lengths of the multiple line segment objects, compares and arranges the multiple line segment lengths, selects the largest target length from the multiple line segment lengths, determines the line segment object corresponding to the target length as the line segment target, and then calculates the target width of the line segment target.

[0152] Specifically, in another preferred embodiment provided by the present invention, the straight line segment recognition and comparison module 103 specifically includes:

[0153] A line feature recognition unit is used to perform line feature recognition on the reconstructed image to determine multiple line segment objects;

[0154] A line segment length calculation unit is used to calculate the line segment length of multiple line segment objects;

[0155] A line segment length comparison unit is used to compare and arrange the lengths of multiple line segments, and select the longest line segment target from the multiple line segment objects;

[0156] The target width calculation unit is used to calculate the target width of the straight line segment target.

[0157] Furthermore, the visual feature data optimization system for UAV power line inspection also includes:

[0158] The skew correction and rotation adjustment module 104 is used to compare the target width with a preset width threshold, identify the horizontal / vertical deviation angle of the straight line segment target, and perform skew correction and rotation adjustment on the reconstructed image.

[0159] In this embodiment of the invention, the skew correction and rotation adjustment module 104 compares the target width with a preset width threshold. If the target width is less than the width threshold, the straight line segment target is assumed to be a wire and should be horizontal. At this time, the horizontal deviation angle of the straight line segment target is identified, and the reconstructed image is horizontally skewed and rotated according to the horizontal deviation angle so that the straight line segment target returns to a horizontal state. If the target width is not less than the width threshold, the straight line segment target is assumed to be a pole and should be vertical. At this time, the vertical deviation angle of the straight line segment target is identified, and the reconstructed image is vertically skewed and rotated according to the vertical deviation angle so that the straight line segment target returns to a vertical state. After that, the invalid borders of the image after horizontal or vertical skew correction and rotation are cropped to obtain an optimized image.

[0160] It should be noted that those skilled in the art can make various modifications and improvements without departing from the inventive concept, and these all fall within the scope of protection of this invention. Therefore, the scope of protection of this patent should be determined by the appended claims.

Claims

1. A method for optimizing visual feature data for UAV power line inspection, characterized in that, The method specifically includes the following steps: The system acquires inspection images taken by a drone during power line inspection, performs occlusion identification and area comparison on the inspection images, and determines whether there is adjustable occlusion. When adjustable occlusion exists, the occlusion area is marked and the affected area is planned. The occlusion area in the inspection image is reconstructed to generate a reconstructed image. Line feature recognition is performed on the reconstructed image to identify multiple line segment objects, and the lengths of the multiple line segment objects are compared to select the longest line segment target, and the target width is calculated. The target width is compared with a preset width threshold, and the horizontal / vertical deviation angle of the target line segment is identified. The reconstructed image is then rotated and adjusted to correct the deviation.

2. The visual feature data optimization method for UAV power line inspection according to claim 1, characterized in that, The process of acquiring inspection images taken by a drone during power line inspection, and performing occlusion identification and area comparison on the inspection images to determine whether there is adjustable occlusion, specifically includes the following steps: Acquire inspection images taken by drones during power distribution line inspections; The inspection image is subjected to occlusion identification to determine whether there is occlusion. When there is obstruction, determine the obstruction area; Calculate the occlusion ratio based on the occlusion area; The occlusion ratio is compared with the preset occlusion invalid ratio and image invalid ratio; When the occlusion ratio is less than the invalid occlusion ratio, it is determined that there is invalid occlusion; When the occlusion ratio is greater than the invalid image ratio, the inspection image is determined to be invalid. When the occlusion ratio is between the occlusion invalid ratio and the image invalid ratio, it is determined that there is adjustable occlusion.

3. The visual feature data optimization method for UAV power line inspection according to claim 2, characterized in that, When adjustable occlusion exists, marking the occlusion area and planning the affected area, and reconstructing the occlusion area in the inspection image to generate a reconstructed image specifically includes the following steps: When adjustable occlusion exists, mark the occluded area; Determine the center point of the occlusion area; The affected area is calculated by multiplying the occlusion area by a preset impact factor. Based on the center point of the obstruction and the area of ​​influence, plan the affected region; Based on the affected area, an inverse distance influence analysis is performed on the occluded area to calculate multiple reconstruction values; Based on the multiple reconstruction values, the occluded areas in the inspection image are reconstructed and filled to generate a reconstructed image.

4. The visual feature data optimization method for UAV power line inspection according to claim 3, characterized in that, The formulas for calculating the multiple reconstructed values ​​are as follows: ； in, For the first in the occluded area The reconstructed value of each pixel; For the affected area The influence values ​​of each pixel are totaled. 1 pixel; For the first in the occluded area The pixel and the first pixel in the affected area The distance between pixels.

5. The visual feature data optimization method for UAV power line inspection according to claim 1, characterized in that, The process of performing line feature recognition on the reconstructed image to identify multiple line segment objects, comparing the lengths of the multiple line segment objects, selecting the longest line segment target, and calculating the target width specifically includes the following steps: Line feature recognition is performed on the reconstructed image to identify multiple line segment objects; Calculate the lengths of the line segments from multiple line segment objects; The lengths of multiple line segments are compared and sorted, and the longest line segment target is selected from the multiple line segment objects; Calculate the target width of the straight line segment target.

6. The visual feature data optimization method for UAV power line inspection according to claim 1, characterized in that, The step of comparing the target width with a preset width threshold and identifying the horizontal / vertical deviation angle of the straight line segment target, and then performing correction and rotation adjustments on the reconstructed image, specifically includes the following steps: The target width is compared with a preset width threshold; When the target width is less than a width threshold, the horizontal deviation angle of the straight line segment target is identified; The reconstructed image is horizontally rotated to correct the deviation according to the stated horizontal deviation angle. When the target width is not less than the width threshold, identify the vertical deviation angle of the target line segment; The reconstructed image is vertically rotated to correct the deviation according to the stated vertical deviation angle. The invalid borders of the image after horizontal or vertical skew rotation are cropped to obtain the optimized image.

7. A visual feature data optimization system for unmanned aerial vehicle (UAV) power line inspection, characterized in that, The system includes an image occlusion detection module, an occlusion region reconstruction module, a line segment recognition and comparison module, and a deviation correction and rotation adjustment module, wherein: The image occlusion judgment module is used to acquire inspection images taken by the drone during power line inspection, perform occlusion recognition and area comparison on the inspection images, and determine whether there is adjustable occlusion. The occlusion region reconstruction module is used to mark the occlusion region and plan the affected region when there is adjustable occlusion, and to reconstruct the occlusion region in the inspection image to generate a reconstructed image. The line segment recognition and comparison module is used to perform line feature recognition on the reconstructed image, identify multiple line segment objects, compare the lengths of the multiple line segment objects, select the longest line segment target, and calculate the target width. The skew correction and rotation adjustment module is used to compare the target width with a preset width threshold, identify the horizontal / vertical deviation angle of the straight line segment target, and perform skew correction and rotation adjustment on the reconstructed image.

8. The visual feature data optimization system for UAV power line inspection according to claim 7, characterized in that, The image occlusion determination module specifically includes: The inspection image acquisition unit is used to acquire inspection images taken by the drone during the inspection of power distribution lines; An occlusion recognition unit is used to perform occlusion recognition on the inspection image and determine whether there is an occlusion. The occlusion area determination unit is used to determine the occlusion area when occlusion exists. An occlusion ratio calculation unit is used to calculate the occlusion ratio based on the occlusion area; A ratio comparison unit is used to compare the occlusion ratio with a preset occlusion invalid ratio and an image invalid ratio; An invalid occlusion determination unit is used to determine that there is invalid occlusion when the occlusion ratio is less than the invalid occlusion ratio; An image invalidity determination unit is used to determine that the inspection image is invalid when the occlusion ratio is greater than the image invalidity ratio; An adjustable occlusion determination unit is used to determine that adjustable occlusion exists when the occlusion ratio is between the invalid occlusion ratio and the invalid image ratio.

9. The visual feature data optimization system for UAV power line inspection according to claim 8, characterized in that, The occlusion area reconstruction module specifically includes: An occlusion area marking unit is used to mark the occlusion area when adjustable occlusion exists; A center point determination unit is used to determine the occlusion center point of the occlusion area; An impact area calculation unit is used to multiply the shading area by a preset impact factor to calculate the impact area. An impact area planning unit is used to plan the impact area based on the shading center point and the impact area; The reconstruction value calculation unit is used to perform inverse distance influence analysis on the occluded area based on the influence area and calculate multiple reconstruction values; The reconstruction filling unit is used to perform relevant reconstruction filling on the occluded areas in the inspection image according to multiple reconstruction values ​​to generate a reconstructed image.

10. The visual feature data optimization system for UAV power line inspection according to claim 7, characterized in that, The line segment recognition and comparison module specifically includes: A line feature recognition unit is used to perform line feature recognition on the reconstructed image to determine multiple line segment objects; A line segment length calculation unit is used to calculate the line segment length of multiple line segment objects; A line segment length comparison unit is used to compare and arrange the lengths of multiple line segments, and select the longest line segment target from the multiple line segment objects; The target width calculation unit is used to calculate the target width of the straight line segment target.

Images