An unmanned power insulation state detection method and system based on machine vision

Abstract

The application is suitable for the technical field of cable detection, and particularly relates to a method and system for detecting the insulation state of power cables based on machine vision, which comprises the following steps: image acquisition, extraction of bending curvature information based on the initial image; generation of a customized structured light image matched with the bending curvature information; during the scanning of the cable, the customized structured light image corresponding to the current position information is called up and projected onto the surface of the insulation layer; collection of the distorted structured light image, comparison and analysis of the distorted structured light image with reference information, and output of the insulation state evaluation result. The application greatly reduces the real-time calculation burden of the unmanned mobile platform during continuous inspection, improves the detection speed and response real-time performance, effectively eliminates the misjudgment caused by accidental textures on the surface of the cable or projection noise, significantly improves the accuracy and reliability of the detection result, and can realize the automatic detection of the physical damage on the surface of the cable insulation layer.

Description

Technical Field

[0001] This invention belongs to the field of cable inspection technology, and in particular relates to an unmanned power insulation status inspection method and system based on machine vision. Background Technology

[0002] In existing technologies, the insulation condition detection of power equipment mainly relies on two methods: manual inspection and non-contact imaging detection. Manual inspection involves visual observation or handheld testing equipment to check insulators, cable terminals, and other parts point by point, which has problems such as high labor intensity, low detection efficiency, and high risk of working at height.

[0003] Non-contact imaging detection is a passive imaging method, and its detection effect is limited by the color, texture and lighting conditions of the cable surface, making it difficult to accurately identify minute surface physical defects in complex backgrounds. Summary of the Invention

[0004] The purpose of this invention is to provide an unmanned power insulation condition detection method based on machine vision, which aims to solve the problem that non-contact imaging detection is a passive imaging method, and the detection effect is limited by the color, texture and lighting conditions of the cable surface, making it difficult to accurately identify tiny surface physical defects in complex backgrounds.

[0005] This invention is implemented as follows: a machine vision-based unmanned power insulation status detection method, the method comprising:

[0006] The cable is continuously imaged using an image acquisition device to obtain an initial image of the insulation layer surface, and the bending curvature information at that location is extracted based on the initial image.

[0007] Based on the curvature information, a customized structured light image is generated that matches the curvature. The customized structured light image contains a grid pattern in which the warp lines are parallel to the axis of the cable and the weft lines overlap with the cross-section of the insulation layer surface at the corresponding position.

[0008] During the cable scanning process, a corresponding customized structured light image is retrieved based on the current location information and projected onto the surface of the insulation layer;

[0009] The distorted structured light image projected onto the surface of the insulating layer is acquired by an image acquisition device. The distorted structured light image is compared and analyzed with reference information to output the insulation status assessment result.

[0010] Preferably, the step of generating a customized structured light image matching the curvature information includes:

[0011] Based on the curvature information extracted from the initial image, the spatial orientation curve of the cable centerline at that location is fitted, and the coverage width of the grid pattern is matched with the cable diameter according to the preset cable diameter parameters.

[0012] Obtain a standard rectangular grid pattern, which contains mutually orthogonal meridians and parallels. Based on the cable's curvature information, control each meridian to bend synchronously with the cable's direction in the projection plane to obtain a customized grid pattern.

[0013] Based on the projection parameters of the projection device, the customized grid pattern is converted into the input image space of the projector to generate the final structured light projection image.

[0014] Preferably, the step of retrieving a corresponding customized structured light image based on the current location information and projecting it onto the surface of the insulation layer during the cable scanning process includes:

[0015] During continuous detection, the displacement of the cable is acquired by a displacement sensor, the current position of the cable in the detection field of view is determined based on the displacement, and a corresponding position index is generated.

[0016] The corresponding customized structured light image is retrieved based on the location index. The customized structured light images are pre-stored locally, and each customized structured light image corresponds one-to-one with each location of the cable.

[0017] Customized structured light images are projected onto the surface of the cable insulation layer, so that the grid pattern projected onto the cable surface matches the cable's orientation and curvature at that location.

[0018] Preferably, the step of acquiring a distorted structured light image of the insulating layer surface projected onto the image acquisition device, comparing and analyzing the distorted structured light image with reference information, and outputting an insulation state assessment result includes:

[0019] After projection, a surface image carrying a grid pattern is acquired synchronously as a distorted structured light image. The distorted structured light image is preprocessed, including at least one of grayscale processing, filtering and noise reduction, and contrast enhancement.

[0020] A grid line extraction operation is performed on the preprocessed distorted structured light image to extract all grid lines in the image. Based on the extracted grid lines, the actual pixel coordinates of each grid node in the image are calculated to obtain the actual grid topology.

[0021] Obtain the reference mesh topology, compare the actual mesh topology with the reference mesh topology point by point, calculate the pixel displacement of each mesh node, determine whether there is physical damage at the location, and output the insulation status assessment result.

[0022] Preferably, for the damaged location, based on the customized structured light image corresponding to the current location, the grid pattern in the customized structured light image is rotated by 1 degree each time, and multiple structured light images with different rotation angles are generated sequentially and projected onto the damaged location respectively. At each rotation angle, the distorted structured light image acquisition and the calculation of the pixel displacement of the grid node are performed respectively. The proportion of the location where the pixel displacement exceeds the preset threshold is counted at each rotation angle. When the proportion exceeds the preset value, it is confirmed that there is physical damage at the location.

[0023] Another object of the present invention is to provide an unmanned power insulation condition detection system based on machine vision, the system comprising:

[0024] The image acquisition module is used to continuously acquire images of the cable through the image acquisition device to obtain an initial image of the insulation layer surface, and extract the bending curvature information at that location based on the initial image;

[0025] The structured light customization module is used to generate a customized structured light image that matches the curvature information. The customized structured light image includes a grid pattern, in which the warp lines are parallel to the axis of the cable and the weft lines overlap with the cross-section of the insulation layer surface at the corresponding position.

[0026] The structured light projection module is used to retrieve a customized structured light image based on the current position information during the cable scanning process and project it onto the surface of the insulation layer.

[0027] The damage assessment module is used to acquire distorted structured light images of the insulation layer surface after projection through an image acquisition device, compare and analyze the distorted structured light images with reference information, and output insulation condition assessment results.

[0028] Preferably, the structured light customization module includes:

[0029] The pattern size adjustment unit is used to fit the spatial direction curve of the cable centerline at the location based on the curvature information extracted from the initial image, and to match the coverage width of the grid pattern with the cable diameter according to the preset cable diameter parameters.

[0030] The deformation adjustment unit is used to obtain a standard rectangular grid pattern, which contains mutually orthogonal meridians and parallels. Based on the bending curvature information of the cable, the unit controls each meridian to bend synchronously with the cable in the projection plane to obtain a customized grid pattern.

[0031] The projection parameter conversion unit is used to convert the customized grid pattern to the input image space of the projector according to the projection parameters of the projection device, so as to generate the final structured light projection image.

[0032] Preferably, the structured light projection module includes:

[0033] The displacement detection unit is used to acquire the displacement of the cable through a displacement sensor during continuous detection, determine the current position of the cable in the detection field of view based on the displacement, and generate a corresponding position index.

[0034] The image retrieval unit is used to retrieve the corresponding customized structured light image according to the position index. The customized structured light image is pre-stored locally, and each customized structured light image corresponds one-to-one with each position of the cable.

[0035] An image projection unit is used to project a customized structured light image onto the surface of the cable insulation layer, so that the grid pattern projected onto the cable surface matches the cable's orientation and curvature at that location.

[0036] Preferably, the damage assessment module includes:

[0037] The image preprocessing unit is used to synchronously acquire a surface image carrying a grid pattern after projection, as a distorted structured light image, and to preprocess the distorted structured light image. The preprocessing includes at least one of grayscale processing, filtering and noise reduction, and contrast enhancement.

[0038] The mesh extraction unit is used to perform mesh line extraction on the preprocessed distorted structured light image, extract all the mesh lines in the image, and calculate the actual pixel coordinates of each mesh node in the image based on the extracted mesh lines to obtain the actual mesh topology.

[0039] The damage assessment unit is used to acquire the reference mesh topology, compare the actual mesh topology with the reference mesh topology point by point, calculate the pixel displacement of each mesh node, determine whether there is physical damage at the location, and output the insulation status assessment result.

[0040] Preferably, for the damaged location, based on the customized structured light image corresponding to the current location, the grid pattern in the customized structured light image is rotated by 1 degree each time, and multiple structured light images with different rotation angles are generated sequentially and projected onto the damaged location respectively. At each rotation angle, the distorted structured light image acquisition and the calculation of the pixel displacement of the grid node are performed respectively. The proportion of the location where the pixel displacement exceeds the preset threshold is counted at each rotation angle. When the proportion exceeds the preset value, it is confirmed that there is physical damage at the location.

[0041] This invention is not limited by the color, texture, and ambient lighting conditions of the cable surface, and does not require complex online curvature calculations. It significantly reduces the real-time calculation burden of unmanned mobile platforms during continuous inspection, improves detection speed and real-time response, effectively eliminates misjudgments caused by accidental textures or projection noise on the cable surface, and significantly improves the accuracy and reliability of detection results. It can realize automated detection of physical damage to the surface of cable insulation. Attached Figure Description

[0042] Figure 1 A flowchart illustrating an unmanned power insulation status detection method based on machine vision, provided for an embodiment of the present invention;

[0043] Figure 2 This is an architecture diagram of an unmanned power insulation status detection system based on machine vision, provided for an embodiment of the present invention. Detailed Implementation

[0044] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0045] like Figure 1 As shown, in a preferred embodiment of the present invention, the method includes:

[0046] S100: The cable is continuously image acquired by the image acquisition device to obtain an initial image of the insulation layer surface, and the bending curvature information at that location is extracted based on the initial image.

[0047] In this step, before the detection is performed, the image acquisition device moves along the cable under test, or controls the cable under test to move relative to the image acquisition device, and takes pictures of the surface of the cable insulation layer to obtain an initial image sequence covering the entire section of the cable under test. Since the cable is curved, the fixed grid structured light cannot adapt to this curvature. For each frame of the initial image acquired, the edge detection algorithm extracts the two side edge contours of the cable insulation layer in the image, fits the pixel coordinates of the cable centerline, calculates the tangent direction of the centerline point by point along the length of the cable, and calculates the curvature value at that position based on the rate of change of direction between adjacent points. When the cable exhibits a significant arc-shaped bend in a certain section, the curvature radius of each point in that arc segment is output by fitting, such as 0.5 meters, 1.2 meters, etc., and saved as the cable curvature information at that position. The curvature information is indexed by position and forms a corresponding relationship with each position on the cable.

[0048] S200, Based on the curvature information, a customized structured light image is generated to match it. The customized structured light image includes a grid pattern, in which the warp lines are parallel to the axis of the cable, and the weft lines overlap with the cross-section of the insulation layer surface at the corresponding position.

[0049] In this step, a customized structured light image is generated for each detection location. Based on the direction curve of the cable centerline at that location, the warp lines in the grid pattern bend with the cable, while the weft lines are straight. The spacing between the weft lines is determined according to the cable diameter. The resulting grid pattern covers the surface area of ​​the cable insulation layer. When the customized grid pattern is projected onto the cable surface, the warp lines extend along the cable axis, and the weft lines are aligned with the cross-sectional direction of the cable. For example, if the cable bends to the right at a certain point, the warp lines also bend to the right at that point, and the degree of bending is consistent with the cable. The generated customized structured light images are stored according to the cable location number.

[0050] During the cable scanning process, the S300 retrieves the corresponding customized structured light image based on the current location information and projects it onto the surface of the insulation layer.

[0051] In this step, during the scanning and detection phase, a customized structured light image corresponding to the current position is retrieved from the local storage and transmitted to the projection device. The projection device synchronously projects the corresponding customized structured light image the instant the cable moves to the corresponding detection position. For example, when the cable moves from position A to position B, after the position change is detected based on the displacement feedback from the encoder, the customized grid pattern corresponding to position B is retrieved and projected onto the cable surface, so that the projected grid pattern always maintains precise consistency with the actual direction and curvature of the cable at the current detection position.

[0052] The S400 acquires a distorted structured light image of the surface of the insulation layer after it has been projected onto the image acquisition device. It then compares and analyzes the distorted structured light image with reference information and outputs the insulation status assessment result.

[0053] In this step, after projection is completed, the cable surface is photographed to obtain a surface image of the projected area. The surface image contains the grid pattern projected onto the cable surface. If the cable insulation layer surface is undamaged, the grid pattern is consistent with the customized structured light. If the cable insulation layer surface has physical damage such as scratches, pits, cracks, or bulges, the grid pattern will be locally distorted at the damaged location. The acquired distorted structured light image is preprocessed to extract the center line position of each grid line in the image and calculate the actual pixel coordinates of each grid node to form the actual grid topology structure of the current frame. This is compared with the reference grid topology structure, and an insulation status assessment report is output based on the comparison results.

[0054] In a preferred embodiment of the present invention, the step of generating a customized structured light image matching the curvature information includes:

[0055] S201, based on the curvature information extracted from the initial image, fit the spatial orientation curve of the cable centerline at that location, and according to the preset cable diameter parameters, make the coverage width of the grid pattern match the cable diameter.

[0056] In this step, after obtaining the cable's bending curvature information, the tangent direction and bending radius at the current detection position are extracted, and the spatial orientation curve of the cable centerline in the projection plane is fitted. The preset cable diameter parameter is read, which is a pre-entered preset value. The grid pattern is matched with this diameter parameter. If the grid width is too narrow, it cannot completely cover the surface area of ​​the cable, resulting in some missing detection areas. For example, for a 10 kV power cable with a diameter of 30 mm, the coverage width of the grid pattern is set to be slightly larger than the pixel width corresponding to 30 mm to ensure that the grid just covers the cable surface. When the customized structured light image is projected onto the actual cable surface, the entire grid falls within the surface area of ​​the cable insulation layer.

[0057] S202, Obtain a standard rectangular grid pattern. The standard rectangular grid pattern contains mutually orthogonal meridians and parallels. Based on the cable's curvature information, control each meridian to bend synchronously with the cable's direction in the projection plane to obtain a customized grid pattern.

[0058] In this step, after determining the cable's routing curve and grid coverage width, a standard rectangular grid pattern is used as the base template. This standard rectangular grid pattern consists of multiple orthogonal meridians and parallels arranged at fixed intervals. The meridians extend along the length of the cable, and the intersections of the meridians and parallels form grid nodes. Based on the fitted cable centerline routing curve, each meridian in the standard rectangular grid is controlled to bend synchronously with the routing curve point by point in the projection plane, so that the relative spacing between the meridians remains unchanged after bending, and the parallels maintain a straight shape. For example, when the cable bends to the right at a certain point with a radius of 1.2 meters, all meridians bend to the right at that point with a bending radius of 1.2 meters. The meridians and parallels are always perpendicular. After processing, the bending shape of the meridians in the customized grid pattern is consistent with the cable routing, and the parallels are orthogonal to the meridians.

[0059] S203, based on the projection parameters of the projection device, the customized grid pattern is converted to the input image space of the projector to generate the final structured light projection image.

[0060] In this step, the customized mesh pattern is converted to the input image space of the projection device to generate a structured light projection image for direct projection by the projector. Since there is a fixed relative position between the projection device and the image acquisition device during actual installation, and the projection lens itself has optical distortion characteristics, it needs to be corrected. Based on the calibration parameters between the projection device and the image acquisition device, the rotation matrix and translation vector between the two are determined, and the mapping relationship between the projection coordinate system and the image coordinate system is established. Inverse distortion compensation is applied to each mesh node in the customized mesh pattern to generate a binarized structured light projection image. For example, when the projection device has a barrel distortion of 2 pixels at the edge field of view, inverse compensation is applied to the coordinates of the mesh nodes in that area in the projection image to make the projected image consistent with the customized mesh pattern.

[0061] In a preferred embodiment of the present invention, the step of retrieving a corresponding customized structured light image based on the current position information and projecting it onto the surface of the insulation layer during the cable scanning process includes:

[0062] S301: During continuous detection, the displacement of the cable is obtained through a displacement sensor, the current position of the cable in the detection field of view is determined based on the displacement, and a corresponding position index is generated.

[0063] In this step, during the continuous scanning detection phase, the physical position of the cable currently at the center of the detection field of view is monitored in real time. This physical position is quantized into discrete position index values. For example, by detecting the agitation angle of the walking mechanism through the encoder, each output pulse corresponds to a displacement of 0.1 mm. When the cumulative pulse amount corresponds to 1.25 meters, combined with the optical field of view offset of 0.15 meters, the cable segment at the current detection position of 1.10 meters is obtained and quantified as the position index "P1100".

[0064] S302, retrieve the corresponding customized structured light image according to the location index. The customized structured light images are pre-stored locally, and each customized structured light image corresponds one-to-one with each location of the cable.

[0065] In this step, the current position is detected in real time. When an index change is detected, a customized structured light image sequence of each position of the cable under test is stored locally. A mapping relationship between the cable position and the structured light image is established. Images within a certain range before and after the current position are preloaded. When moving to the corresponding position, the corresponding customized structured light image data is retrieved using the position index as the key value.

[0066] S303 projects a customized structured light image onto the surface of the cable insulation layer, so that the grid pattern projected onto the cable surface matches the cable's orientation and curvature at that location.

[0067] In this step, after retrieving the customized structured light image, the image data is transmitted to the projection device. At the end of the current projection cycle, the device automatically switches to a new image, so that the projected pattern keeps updated in sync with the actual position. For example, when detecting a cable section with a radius of 1.5 meters that bends to the right, the corresponding grid image is retrieved, where the meridian bends to the right with a radius of 1.5 meters. When the device moves to a straight section, it immediately switches to a grid image with a straight meridian, thus achieving dynamic tracking between the projected pattern and the surface morphology of the cable.

[0068] In a preferred embodiment of the present invention, the step of acquiring a distorted structured light image of the surface of the insulating layer after projection using an image acquisition device, comparing and analyzing the distorted structured light image with reference information, and outputting an insulation state assessment result includes:

[0069] S401, after projection, synchronously acquire a surface image carrying a grid pattern as a distorted structured light image, and preprocess the distorted structured light image, including at least one of grayscale processing, filtering and noise reduction, and contrast enhancement.

[0070] In this step, when the customized structured light image is projected onto the cable surface, the image acquisition device acquires an image of the cable surface carrying a grid pattern, resulting in a distorted structured light image. The acquired distorted structured light image is then preprocessed. The preprocessing process includes grayscale conversion, filtering and denoising, and contrast enhancement. Grayscale conversion converts the color image into a grayscale image to reduce the computational complexity of subsequent image processing. Gaussian filtering is used to eliminate high-frequency noise introduced during image acquisition. Contrast enhancement improves the grayscale difference between the grid lines and the cable background through histogram equalization.

[0071] S402, perform a grid line extraction operation on the preprocessed distorted structured light image to extract all grid lines in the image. Based on the extracted grid lines, calculate the actual pixel coordinates of each grid node in the image to obtain the actual grid topology.

[0072] In this step, a grid line extraction operation is performed on the distorted structured light image to separate the grid lines from the background. The grid line extraction uses an edge detection algorithm, which extracts all edge pixels in the image using the Canny edge detection operator. The Hough transform is used to identify continuous lines belonging to the longitude and latitude of the grid from the edge pixels, resulting in a set of grid lines. Since the cable has a cylindrical surface, the grid lines projected onto it appear as a series of continuous curves with specific directions and spacing in the image. Taking a distorted structured light image of a section of cable as an example, there are a total of 12 latitude lines distributed horizontally and 20 longitude lines distributed vertically in this image. The extraction algorithm outputs the centerline coordinates of each longitude and latitude line respectively. By calculating the intersection points of the longitude and latitude lines, the actual pixel coordinates of all grid nodes are obtained. The coordinate data of all nodes are organized into a two-dimensional array structure according to the longitude and latitude directions, which is the actual grid topology structure.

[0073] S403: Obtain the reference mesh topology, compare the actual mesh topology with the reference mesh topology point by point, calculate the pixel displacement of each mesh node, determine whether there is physical damage at the location, and output the insulation status assessment result.

[0074] In this step, after obtaining the actual mesh topology, the corresponding reference mesh topology is retrieved. The reference mesh topology is directly calculated based on the customized structured light image and is used to reflect the standard mesh shape of the cable in a damage-free state. During comparative analysis, the actual mesh topology and the reference mesh topology are compared node by node. For each mesh node, the Euclidean distance between its pixel coordinates in the actual image and the reference coordinates is calculated to obtain the pixel displacement of the node. For example, in the comparison results of a certain detection position, the displacement of most mesh nodes is concentrated in the range of 0.1 to 0.3 pixels, which is within the normal range of image noise and quantization error. However, six consecutive nodes in a certain local area all show a displacement of more than 1.5 pixels. This area highly coincides with the location of a scratch on the cable with a width of about 0.8 mm and a length of about 5 mm. Based on this, it is determined that there is physical damage at this location. The degree of damage is graded and evaluated according to the size and distribution range of the displacement of each node. The displacement range of 1 to 2 pixels is marked as slight damage, the range of 2 to 5 pixels is marked as moderate damage, and the displacement of more than 5 pixels is marked as severe damage. Finally, the insulation status evaluation result is output.

[0075] In one embodiment of the present invention, after determining that physical damage exists at a certain location, damage confirmation is performed:

[0076] Based on the customized structured light image corresponding to the current detection position, the grid pattern is rotated successively in a step of 1° within a preset angle range, such as -10° to +10°, generating a total of 21 structured light images with different rotation angles and projecting them onto the damage position. Distorted structured light image acquisition and grid node pixel displacement calculation are performed at each rotation angle, and it is recorded whether the pixel displacement exceeds the preset threshold at each angle.

[0077] Let the total number of rotation angles be The number of angles that were determined to be damaged was... Then the confirmation rate ,when The system confirms physical damage at a given location; otherwise, it includes the location in the re-inspection queue. For example, in a certain confirmation process, the scratch was nearly aligned with the meridian direction of the grid only when the grid rotation angle was -1°, 0°, and 1°. At the other 18 angles, the grid distortion characteristics were weakened to below the threshold due to the relative angle change, resulting in a confirmation rate of only 14.3%, which is lower than the preset threshold. This location was then judged as suspected damage and included in the re-inspection queue for manual review by maintenance personnel. Through the multi-angle rotation confirmation mechanism, single misjudgments caused by projection noise can be eliminated, significantly improving the accuracy and reliability of damage detection results.

[0078] like Figure 2 As shown in the figure, an unmanned power insulation status detection system based on machine vision is provided in an embodiment of the present invention. The system includes:

[0079] The image acquisition module 100 is used to continuously acquire images of the cable through the image acquisition device to obtain an initial image of the surface of the insulation layer, and extract the bending curvature information at that location based on the initial image.

[0080] The structured light customization module 200 is used to generate a customized structured light image that matches the curvature information. The customized structured light image includes a grid pattern in which the warp lines are parallel to the axis of the cable and the weft lines overlap with the cross-section of the insulation layer surface at the corresponding position.

[0081] In this system, the structured light customization module 200 includes:

[0082] The pattern size adjustment unit is used to fit the spatial direction curve of the cable centerline at the location based on the curvature information extracted from the initial image, and to match the coverage width of the grid pattern with the cable diameter according to the preset cable diameter parameters.

[0083] The deformation adjustment unit is used to obtain a standard rectangular grid pattern, which contains mutually orthogonal meridians and parallels. Based on the bending curvature information of the cable, the unit controls each meridian to bend synchronously with the cable in the projection plane to obtain a customized grid pattern.

[0084] The projection parameter conversion unit is used to convert the customized grid pattern to the input image space of the projector according to the projection parameters of the projection device, so as to generate the final structured light projection image.

[0085] The structured light projection module 300 is used to retrieve a corresponding customized structured light image based on the current position information during the cable scanning process and project it onto the surface of the insulation layer.

[0086] In this system, the structured light projection module 300 includes:

[0087] The displacement detection unit is used to acquire the displacement of the cable through a displacement sensor during continuous detection, determine the current position of the cable in the detection field of view based on the displacement, and generate a corresponding position index.

[0088] The image retrieval unit is used to retrieve the corresponding customized structured light image according to the position index. The customized structured light image is pre-stored locally, and each customized structured light image corresponds one-to-one with each position of the cable.

[0089] An image projection unit is used to project a customized structured light image onto the surface of the cable insulation layer, so that the grid pattern projected onto the cable surface matches the cable's orientation and curvature at that location.

[0090] The damage assessment module 400 is used to acquire distorted structured light images of the surface of the insulation layer after projection through an image acquisition device, compare and analyze the distorted structured light images with reference information, and output insulation status assessment results.

[0091] In this system, the damage assessment module 400 includes:

[0092] The image preprocessing unit is used to synchronously acquire a surface image carrying a grid pattern after projection, as a distorted structured light image, and to preprocess the distorted structured light image. The preprocessing includes at least one of grayscale processing, filtering and noise reduction, and contrast enhancement.

[0093] The mesh extraction unit is used to perform mesh line extraction on the preprocessed distorted structured light image, extract all the mesh lines in the image, and calculate the actual pixel coordinates of each mesh node in the image based on the extracted mesh lines to obtain the actual mesh topology.

[0094] The damage assessment unit is used to acquire the reference mesh topology, compare the actual mesh topology with the reference mesh topology point by point, calculate the pixel displacement of each mesh node, determine whether there is physical damage at the location, and output the insulation status assessment result.

[0095] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A machine vision-based unmanned power insulation status detection method, characterized in that, The method includes: The cable is continuously imaged using an image acquisition device to obtain an initial image of the insulation layer surface, and the bending curvature information at that location is extracted based on the initial image. Based on the curvature information, a customized structured light image is generated that matches it. The customized structured light image contains a grid pattern in which the warp lines are parallel to the axis of the cable and the weft lines overlap with the cross-section of the insulation layer surface at the corresponding position. During the cable scanning process, a corresponding customized structured light image is retrieved based on the current location information and projected onto the surface of the insulation layer; The distorted structured light image projected onto the surface of the insulating layer is acquired by an image acquisition device. The distorted structured light image is compared and analyzed with reference information to output the insulation status assessment result.

2. The machine vision-based unmanned power insulation status detection method according to claim 1, characterized in that, The step of generating a customized structured light image that matches the curvature information includes: Based on the curvature information extracted from the initial image, the spatial orientation curve of the cable centerline at that location is fitted, and the coverage width of the grid pattern is matched with the cable diameter according to the preset cable diameter parameters. Obtain a standard rectangular grid pattern, which contains mutually orthogonal meridians and parallels. Based on the cable's curvature information, control each meridian to bend synchronously with the cable's direction in the projection plane to obtain a customized grid pattern. Based on the projection parameters of the projection device, the customized grid pattern is converted into the input image space of the projector to generate the final structured light projection image.

3. The unmanned power insulation status detection method based on machine vision according to claim 1, characterized in that, The step of retrieving a corresponding customized structured light image based on the current location information and projecting it onto the surface of the insulation layer during the cable scanning process includes: During continuous detection, the displacement of the cable is acquired by a displacement sensor, the current position of the cable in the detection field of view is determined based on the displacement, and a corresponding position index is generated. The corresponding customized structured light image is retrieved based on the location index. The customized structured light images are pre-stored locally, and each customized structured light image corresponds one-to-one with each location of the cable. Customized structured light images are projected onto the surface of the cable insulation layer, so that the grid pattern projected onto the cable surface matches the cable's orientation and curvature at that location.

4. The unmanned power insulation status detection method based on machine vision according to claim 1, characterized in that, The steps of acquiring a distorted structured light image of the insulating layer surface projected onto the image acquisition device, comparing and analyzing the distorted structured light image with reference information, and outputting an insulation state assessment result include: After projection, a surface image carrying a grid pattern is acquired synchronously as a distorted structured light image. The distorted structured light image is preprocessed, including at least one of grayscale processing, filtering and noise reduction, and contrast enhancement. A grid line extraction operation is performed on the preprocessed distorted structured light image to extract all grid lines in the image. Based on the extracted grid lines, the actual pixel coordinates of each grid node in the image are calculated to obtain the actual grid topology. Obtain the reference mesh topology, compare the actual mesh topology with the reference mesh topology point by point, calculate the pixel displacement of each mesh node, determine whether there is physical damage at the location, and output the insulation status assessment result.

5. The machine vision-based unmanned power insulation status detection method according to claim 4, characterized in that, For the damaged location, using the customized structured light image corresponding to the current location as a reference, the grid pattern in the customized structured light image is rotated by 1 degree each time, generating multiple structured light images with different rotation angles and projecting them onto the damaged location. At each rotation angle, the distorted structured light image acquisition and grid node pixel displacement calculation are performed. The proportion of the location where the pixel displacement exceeds a preset threshold is counted at each rotation angle. When the proportion exceeds the preset value, it is confirmed that there is physical damage at the location.

6. A machine vision-based unmanned power insulation condition detection system, characterized in that, The system includes: The image acquisition module is used to continuously acquire images of the cable through the image acquisition device to obtain an initial image of the insulation layer surface, and extract the bending curvature information at that location based on the initial image; The structured light customization module is used to generate a customized structured light image that matches the curvature information. The customized structured light image includes a grid pattern, in which the warp lines are parallel to the axis of the cable and the weft lines overlap with the cross-section of the insulation layer surface at the corresponding position. The structured light projection module is used to retrieve a customized structured light image based on the current position information during the cable scanning process and project it onto the surface of the insulation layer. The damage assessment module is used to acquire distorted structured light images of the insulation layer surface after projection through an image acquisition device, compare and analyze the distorted structured light images with reference information, and output insulation condition assessment results.

7. The machine vision-based unmanned power insulation status detection system according to claim 6, characterized in that, The structured light customization module includes: The pattern size adjustment unit is used to fit the spatial direction curve of the cable centerline at the location based on the curvature information extracted from the initial image, and to match the coverage width of the grid pattern with the cable diameter according to the preset cable diameter parameters. The deformation adjustment unit is used to obtain a standard rectangular grid pattern, which contains mutually orthogonal meridians and parallels. Based on the bending curvature information of the cable, the unit controls each meridian to bend synchronously with the cable in the projection plane to obtain a customized grid pattern. The projection parameter conversion unit is used to convert the customized grid pattern to the input image space of the projector according to the projection parameters of the projection device, so as to generate the final structured light projection image.

8. The machine vision-based unmanned power insulation status detection system according to claim 6, characterized in that, The structured light projection module includes: The displacement detection unit is used to acquire the displacement of the cable through a displacement sensor during continuous detection, determine the current position of the cable in the detection field of view based on the displacement, and generate a corresponding position index. The image retrieval unit is used to retrieve the corresponding customized structured light image according to the position index. The customized structured light image is pre-stored locally, and each customized structured light image corresponds one-to-one with each position of the cable. An image projection unit is used to project a customized structured light image onto the surface of the cable insulation layer, so that the grid pattern projected onto the cable surface matches the cable's orientation and curvature at that location.

9. The machine vision-based unmanned power insulation status detection system according to claim 6, characterized in that, The damage assessment module includes: The image preprocessing unit is used to synchronously acquire a surface image carrying a grid pattern after projection, as a distorted structured light image, and to preprocess the distorted structured light image. The preprocessing includes at least one of grayscale processing, filtering and noise reduction, and contrast enhancement. The mesh extraction unit is used to perform mesh line extraction on the preprocessed distorted structured light image, extract all the mesh lines in the image, and calculate the actual pixel coordinates of each mesh node in the image based on the extracted mesh lines to obtain the actual mesh topology. The damage assessment unit is used to acquire the reference mesh topology, compare the actual mesh topology with the reference mesh topology point by point, calculate the pixel displacement of each mesh node, determine whether there is physical damage at the location, and output the insulation status assessment result.

10. The machine vision-based unmanned power insulation status detection system according to claim 9, characterized in that, For the damaged location, using the customized structured light image corresponding to the current location as a reference, the grid pattern in the customized structured light image is rotated by 1 degree each time, generating multiple structured light images with different rotation angles and projecting them onto the damaged location. At each rotation angle, the distorted structured light image acquisition and grid node pixel displacement calculation are performed. The proportion of the location where the pixel displacement exceeds a preset threshold is counted at each rotation angle. When the proportion exceeds the preset value, it is confirmed that there is physical damage at the location.

Images