A climbing robot and detection method for cable defect detection

Abstract

The present application relates to the technical field of cable defect detection, and provides a climbing robot and detection method for cable defect detection. The climbing robot is used to simultaneously perform surface defect detection and internal defect detection on cables, and includes a main frame with a frame structure, a clamping travel unit installed within the frame structure of the main frame, and at least one main drive wheel and at least two clamping wheels in the clamping travel unit are used to clamp the cable passing through both ends of the frame structure of the main frame. The visual inspection module and magnetic flux leakage detection module installed at both ends of the frame perform surface defect detection and internal defect detection on the cable. During the inspection process, the main drive wheel is driven by the drive unit installed on the frame structure of the main frame, causing the main frame to autonomously move along the length of the cable. In this way, while achieving real-time and synchronous detection of internal and external defects in the cable, comprehensive inspection of the entire length of the cable is achieved through autonomous movement.

Description

Technical Field

[0001] The present application relates to the technical field of cable defect detection, and in particular to a climbing robot and a detection method for cable defect detection. Background Art

[0002] Cables, as important load-bearing or supporting structures, are widely used in various projects, including bridges, towers, power transmission, oil pipelines, and cableways. Their safety directly impacts the operational reliability of these facilities and public safety. Therefore, regular cable inspection and maintenance are particularly important. Traditional cable inspections rely primarily on manual inspections or equipment based on single-technology testing. These methods have numerous limitations in practical applications.

[0003] During manual inspections, inspectors use the naked eye or simple tools to perform visual inspections. However, cables are often deployed at high altitudes or in hard-to-reach areas. Manual inspections are not only time-consuming and labor-intensive, but also highly dangerous, especially in high-risk environments like towers and bridges. Furthermore, the inspector's experience and skill level directly impact the accuracy of inspections, making it difficult for manual inspections to effectively detect hidden defects within cables.

[0004] Current automated cable inspection equipment, which uses nondestructive testing technologies such as magnetic flux leakage, ultrasonic testing, and eddy current testing, often has limited functionality and limited mobility, making it inefficient in addressing complex and changing cable environments. These fixed inspection methods are often only installed in fixed locations to perform localized inspections on cables, failing to cover the entire cable. Manually moving these devices for cable inspection significantly reduces inspection efficiency. Summary of the Invention

[0005] The purpose of this application is to provide a climbing robot and a detection method for cable defect detection, so as to solve or alleviate the problems existing in the above-mentioned prior art.

[0006] In order to achieve the above objectives, this application provides the following technical solutions:

[0007] The present application provides a climbing robot for cable defect detection, which is used to simultaneously perform surface defect detection and internal defect detection on the cable. The climbing robot includes: a main frame, which is a frame structure;

[0008] a clamping travel unit, installed in the frame structure of the main frame, comprising: at least one main drive wheel and at least two clamping wheels, wherein the main drive wheel and the clamping wheels are capable of clamping cables passing through both ends of the frame structure of the main frame;

[0009] The cable detection unit, including a visual detection module and a magnetic flux leakage detection module, is installed at both ends of the main body frame, and is respectively used for surface defect detection and internal defect detection of the cable;

[0010] The driving unit is installed on the frame structure of the main body frame and is used to drive the main driving wheel to rotate, so as to drive the main body frame to move along the length direction on the cable.

[0011] Preferably, the main body frame includes an upper frame and a lower frame. At both ends of the upper frame, upper connecting frames with a 冂-shaped structure are respectively arranged. At both ends of the lower frame, lower connecting frames with a 凵-shaped structure adapted to the upper connecting frames are respectively arranged. The upper connecting frames and the lower connecting frames can be telescopically adjusted along a first direction; wherein, the first direction is the extending direction of the side plates of the upper connecting frame or the lower connecting frame; the main driving wheel is rotatably installed on the bottom surface of the upper frame, and the clamping wheel is installed on the top surface of the lower frame, and the relative distance between the clamping wheel and the main driving wheel can be adaptively adjusted according to the radial dimension of the cable.

[0012] Preferably, the clamping wheel is rotatably installed on the lower frame through a clamping wheel frame. The clamping wheel frame includes a rotating connecting rod and a tension spring. One end of the rotating connecting rod is rotatably installed on the lower frame, and the other end is rotatably installed with the clamping wheel; one end of the tension spring is connected to the middle part of the rotating connecting rod, and the other end is connected to the lower frame, and is used to drive the rotating connecting rod to have a relative movement trend towards the cable around its connection axis with the lower frame, so as to clamp the cable with the clamping wheel.

[0013] Preferably, there are two groups of the clamping wheel frames, and the two groups of clamping wheel frames are symmetrically arranged at both ends of the lower frame to drive the two clamping wheels to have a trend of moving towards each other.

[0014] Preferably, there are two main driving wheels, and the two main driving wheels are respectively arranged opposite to the two clamping wheels; and / or, the two main driving wheels are respectively connected to the output ends of the two driving units through synchronous belt drives.

[0015] Preferably, both the main driving wheel and the clamping wheel are of a spindle-shaped structure to clamp the cable.

[0016] Preferably, the visual detection module includes an image detection bracket and at least three visual detection cameras. The image detection bracket is installed at the head of the main body frame and is circumferentially provided with at least three claws. Three visual detection cameras are respectively arranged on the at least three claws in a one-to-one correspondence, so that the imaging ranges of the three visual detection cameras cover the surface of the cable circumferentially.

[0017] Preferably, the magnetic flux leakage detection module includes: a magnetic detection bracket, an excitation device and a detection device; the magnetic detection bracket is installed at the tail of the main frame; the excitation device is an annular structure, installed on the magnetic detection bracket close to the main frame, and is used to magnetize the internal steel wire of the cable; wherein the cable passes through the center of the annular structure of the excitation device; the detection device is an annular structure, installed on the magnetic detection bracket parallel to the excitation device and away from the main frame, and is used to detect magnetic defects in the magnetized steel wire inside the cable.

[0018] Preferably, the detection device includes: a detection mounting plate and a plurality of Hall sensors, the detection mounting plate is installed on the magnetic detection bracket parallel to the excitation device, and the plurality of Hall sensors are evenly distributed on the detection mounting plate to detect magnetic defects in the magnetized steel wire inside the cable.

[0019] The present application also provides a method for detecting cable defects, wherein any of the aforementioned climbing robots for detecting cable defects is used to simultaneously detect surface defects and internal defects on the cable. The method includes:

[0020] Step S101: performing image enhancement processing on the acquired surface image of the cable, and extracting surface defect features of the cable based on the Canny edge detection algorithm and a pre-built composite deep learning model;

[0021] as well as,

[0022] De-noising the collected cable magnetic flux leakage signal and obtaining the cable magnetic flux leakage defect characteristics based on spectrum analysis method;

[0023] Step S102: inputting the surface defect features and the magnetic flux leakage defect features of the cable into a generator of a generative adversarial network at the same time, fusing the surface defect features and the magnetic flux leakage defect features to obtain comprehensive defect data of the cable;

[0024] Step S103 : acquiring depth information of the cable defects based on the comprehensive defect data, and generating a three-dimensional point cloud of the cable defects, so as to simultaneously acquire surface defects and internal defects of the cable.

[0025] Preferably, step S101 includes: performing image enhancement processing on the surface image after Gaussian filtering and denoising based on a contrast-limited adaptive histogram equalization method to obtain a surface enhanced image of the cable;

[0026] performing gradient calculation on the surface enhancement image of the cable, removing edge redundant pixels from the obtained gradient image based on a non-maximum suppression method, and performing edge detection and connection on the image from which the redundant edge pixels have been removed based on a double threshold method to obtain an edge image of the cable surface defects;

[0027] The edge image is input into a composite deep learning model based on a convolutional neural network and a Transformer architecture to extract surface defect features of the cable.

[0028] Preferably, in step S101, the collected magnetic leakage signal of the cable is denoised by wavelet transform and low-pass filtering in sequence; based on the spectrum analysis method, the denoised signal obtained by the denoising process is subjected to fast Fourier transform to obtain the distribution characteristics of the denoised signal at different frequencies, so as to obtain the magnetic leakage defect characteristics of the cable.

[0029] Preferably, in step S102, the generator of the generative adversarial network performs feature splicing and weighted fusion on the surface defect features and the magnetic flux leakage defect features in sequence to obtain the comprehensive defect data.

[0030] Preferably, it also includes: based on the discriminator of the generative adversarial network, respectively calculating the loss function, mean square error and cross entropy loss between the generated comprehensive defect data and the pre-acquired cable defect sample data to evaluate the generated comprehensive defect data, and iteratively optimizing the generator of the generative adversarial network according to the evaluation results.

[0031] Beneficial effects:

[0032] The climbing robot for cable defect detection provided in an embodiment of the present application is used to simultaneously perform surface defect detection and internal defect detection on the cable. The climbing robot includes a main frame with a frame structure, a clamping walking unit installed in the frame structure of the main frame, and at least one main drive wheel and at least two clamping wheels in the clamping walking unit are used to clamp the cable passing through the two ends of the frame structure of the main frame. The visual inspection module and the magnetic flux leakage inspection module installed at both ends of the frame are used to perform surface defect detection and internal defect detection on the cable. During the inspection process, the main drive wheel is driven by the drive unit installed on the frame structure of the main frame to rotate, causing the main frame to self-propelled along the length of the cable. In this way, while achieving real-time and synchronous detection of internal and external defects of the cable, comprehensive inspection of the entire length of the cable is achieved through autonomous movement.

[0033] At the same time, the cable surface image acquired by the climbing robot is enhanced, and the cable's surface defect features are extracted based on the Canny edge detection algorithm and a pre-built composite deep learning model. The collected cable's magnetic flux leakage signal is denoised, and the cable's magnetic flux leakage defect features are obtained based on spectral analysis. The cable's surface defect features and magnetic flux leakage defect features are then simultaneously input into the generator of a generative adversarial network, where they are fused to obtain comprehensive cable defect data. Finally, based on the comprehensive defect data, the cable's defect depth information is obtained, and a 3D point cloud of the cable's defects is generated to simultaneously identify both surface and internal defects. This allows for simultaneous analysis and detection of both internal and external cable defects based on the cable surface image and magnetic flux leakage information acquired by the climbing robot. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] The drawings and descriptions that constitute part of this application are intended to provide a further understanding of this application. The exemplary embodiments and descriptions of this application are intended to explain this application and do not constitute an improper limitation on this application. Among them:

[0035] Figure 1 This is a schematic structural diagram of a climbing robot for cable defect detection provided according to some embodiments of the present application;

[0036] Figure 2 A schematic structural diagram of a main frame provided according to some embodiments of the present application;

[0037] Figure 3 This is a schematic diagram of installing a clamping walking unit on a main frame according to some embodiments of the present application;

[0038] Figure 4 This is a schematic diagram of assembling a clamping wheel frame on a lower frame according to some embodiments of the present application;

[0039] Figure 5 This is a schematic diagram of assembling a cable detection unit on a main frame according to some embodiments of the present application;

[0040] Figure 6 A schematic structural diagram of a main drive wheel provided according to some embodiments of the present application;

[0041] Figure 7 A schematic flow chart of a detection method for cable defect detection provided according to some embodiments of the present application.

[0042] Description of reference numerals:

[0043] 100, main frame; 200, clamping and traveling unit; 300, cable detection unit; 400, driving unit;

[0044] 101, upper frame; 102, lower frame; 111, install top plate; 121, upper connecting frame; 112, install bottom plate; 122, lower connecting frame;

[0045] 201, main driving wheel; 202, clamping wheel; 203, clamping wheel frame; 213, rotating connecting rod; 223, tension spring;

[0046] 301. Visual detection module; 302. Magnetic flux leakage detection module; 312. Magnetic detection bracket; 322. Excitation device; 332. Detection device. DETAILED DESCRIPTION

[0047] The present application will be described in detail below with reference to the accompanying drawings and in conjunction with embodiments. Each example is provided by way of explanation of the present application and does not limit the present application. In fact, it will be clear to those skilled in the art that modifications and variations can be made in the present application without departing from the scope or spirit of the present application. For example, a feature shown or described as part of one embodiment can be used in another embodiment to produce yet another embodiment. All other embodiments obtained by those of ordinary skill in the art based on the embodiments in the present invention should fall within the scope of protection of the embodiments of the present invention.

[0048] Currently, during the cable inspection process, manual inspections can only be performed periodically, with poor real-time performance. They are unable to detect sudden structural problems in a timely manner and pose a huge safety hazard. Automated inspection equipment using non-destructive testing technology is difficult to move and has low inspection efficiency. Furthermore, it can only detect specific types of defects. For example, magnetic flux leakage detection can only identify broken wires or corrosion points in the cable, but cannot detect surface wear or mechanical damage.

[0049] However, cable damage can manifest in a variety of ways, ranging from internal wire breakage and corrosion caused by material fatigue to surface wear, cracks, and corrosion caused by prolonged exposure to the environment. This requires inspection equipment capable of simultaneously identifying both internal and external defects, a requirement currently limited by single-source inspection technologies. Magnetic flux leakage technology can detect internal cable defects but is unable to detect surface defects. Visual inspection technology can identify external defects but cannot penetrate the cable's interior.

[0050] Based on this, this application proposes a climbing robot for cable defect detection, which can realize real-time and synchronous detection of internal and external defects of cables, and realize comprehensive detection of cables through autonomous movement. Figures 1 to 6As shown in the figure, the climbing robot includes: a main body frame 100, and a clamping walking unit 200, a cable detection unit 300, and a driving unit 400 installed on the main body frame 100. Among them, the clamping walking unit 200 is used to clamp the cable to be detected and, on the premise that power is provided by the driving unit 400, drive the main body frame 100 to move along the length direction on the cable; the cable detection unit 300 is used to perform surface defect detection and internal defect detection on the cable to be detected simultaneously, and during the detection process, move on the cable together with the main body frame 100, so as to achieve real-time and synchronous detection of internal and external defects of the cable while comprehensively detecting the entire length of the cable.

[0051] In this application, the main body frame 100 is of a frame structure, and both ends and the side of the frame structure are hollowed out to reduce the overall mass of the frame structure. The clamping walking unit 200, the cable detection unit 300, and the driving unit 400 are respectively installed on the top plate and the bottom plate of the frame structure. Here, the main body mechanism is assembled by buckling the upper and lower parts, so that the cable with or without cable installed can be detected, improving the overall applicability of the climbing robot.

[0052] In a specific example, the main body frame 100 includes: an upper frame 101 and a lower frame 102. At both ends of the upper frame 101, upper connecting frames 121 with a 冂-shaped structure are respectively provided, and at both ends of the lower frame 102, lower connecting frames 122 with a 凵-shaped structure adapted to the upper connecting frames 121 are respectively provided. On the side plates of the upper connecting frame 121 and the lower connecting frame 122, a plurality of adapted adjustment through holes are respectively provided along the length direction. When detecting cables with different diameters, through the cooperation of the adjustment through holes at different positions on the upper connecting frame 121 and the lower connecting frame 122, the upper connecting frame 121 and the lower connecting frame 122 are telescopically adjusted along the first direction (the extending direction of the side plate of the upper connecting frame 121 or the lower connecting frame 122), so as to adjust the relative distance between the upper frame 101 and the lower frame 102 to adapt to cables of different sizes.

[0053] Here, the upper connecting frame 121 is detachably installed on the installation top plate 111 of the upper frame 101 through a bolt fastening assembly. That is to say, the top plate of the 冂-shaped structure of the upper connecting frame 121 and the installation top plate 111 are tightly connected through a bolt and nut assembly; the lower connecting frame 122 adopts the same installation connection method as the upper connecting frame 121 and is detachably installed on the installation bottom plate 112 of the lower frame 102.

[0054] The main drive wheel 201 is rotatably mounted on the bottom surface of the mounting plate 111 of the upper frame 101. Mounting pads are symmetrically arranged along the bottom surface of the mounting plate 111 along a second direction (the width of the climbing robot, perpendicular to the first direction and the cable axis). The main drive wheel 201 is rotatably mounted on bearing blocks at each end, which are then removably attached to the mounting pads via bolt fastening assemblies. A synchronous belt driven pulley is mounted on the end of the main drive wheel 201's rotating shaft extending from the bearing block. A drive unit 400 (stepper motor) is removably mounted on the top surface of the mounting plate 111. A synchronous belt drive pulley is mounted on the output end of the drive unit 400. A synchronous belt transmission is used between the synchronous belt drive pulley and the synchronous belt driven pulley, transmitting the stepper motor's output power to the main drive wheel 201. Consequently, when the cable is clamped, the friction between the main drive wheel 201 and the cable drives the climbing robot along the cable.

[0055] In this embodiment, a main drive wheel 201 is symmetrically mounted on each end of the mounting plate 111 of the upper frame 101 (in the axial direction of the cable). Specifically, two main drive wheels 201 are mounted on each end of the mounting plate 111. This effectively enhances the contact force between the climbing robot and the cable. Simultaneous driving of the front and rear main drive wheels 201 facilitates smoother movement of the climbing robot on the cable. The two main drive wheels 201 can be driven synchronously by a single drive unit 400, or independently by two independent drive units 400. When two independent drive units 400 are used for independent driving, they are positioned at each end of the mounting plate 111, positioning the center of gravity of the climbing robot closer to the center. This, in turn, provides for improved load-bearing conditions between the climbing robot and the cable during operation.

[0056] In the embodiment of the present application, a clamping wheel 202 is installed on the top surface of the mounting base 112 of the lower frame 102. Specifically, the clamping wheel 202 is rotatably mounted on the lower frame 102 through a clamping wheel frame 203; one end of the clamping wheel frame 203 is rotatably mounted on the lower frame 102, and the other end of the clamping wheel 202 is rotatably mounted. Then, the rotation of the clamping wheel frame 203 around the axis of its rotational connection with the lower frame 102 can drive the clamping wheel 202 to rotate, so that the relative distance of the clamping wheel 202 with respect to the main driving wheel 201 can be adaptively adjusted according to the radial size of the cable.

[0057] In this embodiment, the clamping wheel frame 203 includes two parallel rotating links 213 and a tension spring 223. One end of the two parallel rotating links 213 is symmetrically mounted to the upper surface of the mounting base 112 of the lower frame 102 along the second direction, and the other end of the two parallel rotating links 213 is rotatably mounted (between the two rotating links 213) on the clamping wheel 202. Consequently, the clamping wheel 202 moves as the rotating links 213 rotate about the axis of rotation of the rotating links 213 and the lower frame 102.

[0058] Here, adaptive adjustment of the relative distance of the clamping wheel 202 from the main drive wheel 201 is achieved via a tension spring 223, one end of which is connected to the middle portion of the rotating link 213 and the other end is connected to the mounting plate 111 of the lower frame 102. The contraction of the tension spring 223 drives the rotating link 213 about its connection axis with the lower frame 102 toward the cable, causing the clamping wheel 202 to clamp the cable.

[0059] In one specific example, two parallel rotating links 213 are connected via a wheel frame shaft, i.e., both ends of the wheel frame shaft are detachably mounted to the middle of the two parallel rotating links 213. Furthermore, one end of a tension spring 223 is connected to the wheel frame shaft, and the other end is connected to the middle of the mounting base 112. Furthermore, there are two tension springs 223, each arranged parallel to and near the two parallel rotating links 213, thereby providing a more uniform force condition on the clamping wheel frame 203.

[0060] Here, the rotating link 213 is connected to the mounting base 112 via a mounting base. The rotating link 213 is rotatably mounted on the mounting base, which is removably connected to the mounting base 112 via a bolt fastening assembly. Simultaneously, the tension spring 223 is also connected to the mounting base 112 via another mounting base. The mounting base, which is compatible with the tension spring 223, is removably mounted to a mounting pad located on the mounting base 112 via a bolt fastening assembly. When two tension springs 223 are used to act on the clamping wheel frame 203, a wheel seat shaft is disposed between the two mounting bases (compatible with the tension springs 223) arranged symmetrically along the second direction, and the two tension springs 223 are respectively connected to the wheel seat shafts. In this embodiment, retaining rings are used at the connection between the tension spring 223 and the wheel frame shaft and the wheel seat shaft to limit the displacement of both ends of the tension spring 223 in the second direction, effectively preventing the tension spring 223 from deviating and preventing the tension spring 223 from imparting an unbalanced load on the clamping wheel frame 203.

[0061] In the present application, the two clamping wheels 202 are mounted on the mounting base 112 of the lower frame 102 via two sets of clamping wheel frames 203. The two sets of clamping wheel frames 203 are symmetrically arranged at both ends of the lower frame 102 to drive the two clamping wheels 202 to move toward each other, thereby clamping the cable through the coordinated cooperation between the two clamping wheels 202 and the main drive wheel 201. When two independent drive units 400 are used to drive the two main drive wheels 201, the two main drive wheels 201 are arranged directly opposite the two clamping wheels 202, respectively, to provide better cable stress conditions.

[0062] After the main drive wheel 201 and the clamping wheel 202 clamp the cable, the drive unit 400 drives the main drive wheel 201 to rotate, enabling the climbing robot to move along the cable. The main drive wheel 201 and the clamping wheel 202 do not have a conical structure, which allows for automatic centering during movement when the cable is clamped, ensuring that the climbing robot always moves along the cable axis, effectively preventing deviation.

[0063] In the embodiment of the present application, the visual inspection module 301 is used to detect surface defects of the cable. The visual inspection module 301 includes an image inspection bracket and at least three visual inspection cameras. The image inspection bracket is mounted on the head of the main bracket and has at least three claws evenly distributed along the circumference. Three visual inspection cameras are arranged on each of the at least three claws in a one-to-one correspondence, so that the imaging range of the three visual inspection cameras covers the cable surface along the circumference. Thus, by imaging the cable surface along the circumference of the cable through at least three visual inspection cameras, all circumferential surface defects of the cable can be detected without omission.

[0064] In the embodiment of the present application, the magnetic flux leakage detection module 302 is used to detect internal defects in the cable. Specifically, the magnetic flux leakage detection module 302 includes: a magnetic detection bracket 312, an excitation device 322, and a detection device 332; the magnetic detection bracket 312 is mounted at the rear of the main frame 100, and the excitation device 322 is an annular structure, mounted on the magnetic detection bracket 312 near the main frame 100, and is used to magnetize the internal steel wires of the cable. Specifically, a permanent magnet is arranged within the excitation device 322. When the cable passes through the center of the annular structure of the excitation device 322, the permanent magnet forms a uniform magnetic field inside the cable, thereby magnetizing the internal steel wires of the cable.

[0065] The ring-shaped detection device 332 is mounted parallel to the excitation device 322 on the magnetic detection bracket 312 and away from the main frame 100, and is used to detect magnetic defects in the magnetized steel wire inside the cable. The detection device 332 includes: a detection mounting plate and multiple Hall sensors. The detection mounting plate and the excitation device 322 are mounted on the magnetic detection bracket 312; multiple Hall sensors are evenly distributed along the circumference on the detection mounting plate to detect magnetic defects in the magnetized steel wire inside the cable. When the detection device passes over damaged or defective parts inside the cable, due to the change in magnetic permeability of the damaged or defective parts, part of the magnetic flux leaks into the air to form a leakage magnetic field. The Hall sensor captures this leakage magnetic change signal and converts it into an electrical signal to detect magnetic defects inside the cable, such as broken wires and rust.

[0066] Data acquired by the visual inspection module 301 and the magnetic flux leakage detection module 302 is exchanged through a data communication unit mounted on the main frame 100 and transmitted to a data processing unit. The data processing unit then determines cable defects and damage based on the received surface and internal data. The data communication unit is mounted on the mounting top plate 111 of the upper frame 101 and located between the two drive units 400. A main control unit is also located between the two drive units 400. This main control unit contains a control program for the climbing robot, enabling its self-movement and detection. During its movement, the climbing robot's real-time motion is monitored by multiple position, attitude, and speed sensors mounted on the main frame 100. These sensors transmit this monitoring data to the main control module, which then autonomously adjusts the climbing robot's position, attitude, and speed in real time according to the control program.

[0067] Here, data storage, data processing and other unit modules can also be deployed on the main frame 100 of the climbing robot. The data storage unit can be used to store the data collected by the climbing robot, and the data processing unit can be used to perform noise filtering, signal enhancement, image recognition and other operations on the collected data to realize the identification and judgment of cracks, corrosion, wear and tear on the cable surface and defects such as broken wires and rust inside the cable.

[0068] When using the climbing robot of any of the above embodiments to simultaneously detect surface defects and internal defects on the cable, first, before the detection operation begins, the climbing robot is assembled, including installing the clamping walking unit 200, cable detection unit 300, drive unit 400, data communication unit, data storage unit, power supply, sensor, etc. on the main frame 100, and each part is inspected to ensure that it can work normally; then, the upper and lower parts of the main frame 100 are disassembled and symmetrically wrapped around the two sides of the cable, and the upper and lower parts of the main frame 100 are pushed closer to the center of gravity of the cable until the main drive wheel 201 and the clamping wheel 202 clamp the cable, and the upper and lower parts are fastened together by selecting a suitable adjustment through hole to complete the clamping of the climbing robot on the cable to be inspected.

[0069] After the climbing robot is mounted on the cable to be inspected and the inspection unit and sensors are initialized, the climbing robot's movement path and inspection parameters are set in the main control unit based on the actual cable conditions. The climbing robot is then activated, allowing it to autonomously walk on the cable while capturing images of the cable surface using the visual inspection module 301 and detecting changes in the cable's internal magnetic field using the magnetic flux leakage detection module 302. These images and magnetic field change data are then transmitted in real time to the data storage and processing unit via the data communication unit. The data processing unit then performs denoising and signal enhancement on the cable's magnetic field change data, as well as edge detection, morphological processing, and image recognition on the captured images, thereby identifying and determining the type, location, and severity of surface and internal defects in the cable. Finally, after completing the full-length inspection of the cable, the climbing robot automatically returns to its starting position. Alternatively, after completing the inspection of a specific area, it can proceed to inspect other areas of the cable. Furthermore, the data communication module can transmit the inspection data to other terminals, enabling parallel storage and processing of the data for analysis, decision-making, and maintenance of the cable.

[0070] The climbing robot for cable defect detection can simultaneously detect surface defects and internal defects of the cable. After obtaining the surface image data and magnetic flux leakage signal data of the cable, the cable defect analysis program deployed in the data processing unit can be used to perform synchronous analysis of the surface defects and internal defects of the cable. Figure 7 As shown, the detection method for cable defect detection deployed in the data processing unit includes:

[0071] Step 101: Perform image enhancement processing on the acquired surface image of the cable, and extract the surface defect characteristics of the cable based on the Canny edge detection algorithm and a pre-built composite deep learning model; and perform denoising processing on the collected leakage magnetic signal of the cable, and obtain the leakage magnetic defect characteristics of the cable based on the spectrum analysis method.

[0072] In this application, when processing surface images acquired by a climbing robot to extract surface defect features of the cable, the Canny edge detection algorithm is used to extract edge information from the surface image of the cable acquired by the climbing robot. Image gradients are calculated to find image edges and perform refinement processing. Specifically, first, based on a contrast-limited adaptive histogram equalization method, image enhancement processing is performed on the surface image after Gaussian filtering and denoising to obtain a surface-enhanced image of the cable.

[0073] Here, a surface image is denoised and smoothed using Gaussian filtering, making details clearer and improving image usability and visual quality. After obtaining the smoothed image through Gaussian filtering, the image is further enhanced using histogram equalization to optimize the visual effect, increase image contrast, and generate a surface-enhanced image. Specifically, the number of pixels at each grayscale level in the smoothed image is counted to obtain the image's grayscale distribution histogram. Based on the grayscale distribution histogram, the cumulative probability distribution of the image is calculated to obtain the cumulative frequency of the corresponding grayscale value. Then, the grayscale values ​​of the smoothed image are mapped to a new grayscale value range using the CDF method, ensuring that the pixel values ​​are evenly distributed throughout the entire grayscale range. Finally, the grayscale values ​​in the smoothed image are replaced with the mapped grayscale values ​​to generate a surface-enhanced image with higher contrast.

[0074] Then, the gradient of the cable's surface enhancement image is calculated, and the edge redundant pixels are removed from the obtained gradient image based on the non-maximum suppression method. Specifically, the Sobel operator is used to convolve the surface enhancement image to obtain the horizontal gradient of the image. and vertical gradients , and according to the formula:

[0075]

[0076] Determine the gradient strength of a pixel in a gradient image and gradient direction .

[0077] Then, the edges of the gradient image are processed by performing non-maximum suppression on the gradient image. The gradient direction is used to determine the direction for each pixel when performing neighborhood comparison. That is, when performing non-maximum suppression, the neighboring pixels of the current pixel are selected based on the gradient inverse and the gradient strength in that direction is compared. The gradient strength of each pixel is used to determine whether to retain the pixel. Only when the gradient strength of the pixel is greater than the gradient strength of the neighboring pixels along the gradient direction is the pixel retained as an edge point of the image.

[0078] Among them, after determining the gradient direction of each pixel in the gradient image, the neighborhood of each pixel is divided into two parts along the gradient direction (i.e., the two neighboring pixels before and after) according to the gradient direction of the pixel; then the gradient strength of the current pixel is compared with the gradient strength of the neighboring pixels. If the gradient strength of the current pixel is greater than the gradient strength of its two neighboring pixels, the current pixel is retained; if the gradient strength of the current pixel is less than the gradient strength of its two neighboring pixels, the current pixel is suppressed, that is, the pixel value of the current pixel is set to 0.

[0079] The gradient image is refined using non-maximum suppression to remove redundant edge pixels and retain the most significant edge pixels, facilitating edge connection and defect location. In this application, edge detection and connection are performed on the image with redundant edge pixels removed using a dual-threshold method to obtain an edge image of the cable surface defect. Furthermore, edge detection is performed on the image with redundant edge pixels removed using a dual-threshold method to determine the strength of the image edge and separate strong edges, weak edges, and non-edge areas.

[0080] Specifically, the gradient intensities of pixels in the image from which redundant edge pixels have been removed are compared using two preset intensity thresholds (a first intensity threshold and a second intensity threshold, with the first intensity threshold being greater than the second intensity threshold). Pixels with gradient intensities higher than the first intensity threshold are considered strong edge pixels and marked as edge points of the image; pixels with gradient intensities lower than the second intensity threshold are considered non-edge pixels and are suppressed or removed.

[0081] Pixels whose gradient intensity is between the first intensity threshold and the second intensity threshold are regarded as weak edge pixels. The gradient intensity of weak edge pixels is low and is not enough to form a clear edge independently. In this application, an edge tracking method is adopted to scan along the gradient direction of the edge pixel, and check whether the neighborhood pixels of the edge pixel contain strong edge pixels. If the weak edge pixel is directly adjacent to a strong edge pixel, the weak edge pixel is marked as the edge point of the image, that is, marked as a valid edge and added to the edge of the image; if the weak edge pixel is not adjacent to the strong edge pixel, the weak edge pixel is considered to be noise or an incomplete edge, and the weak edge pixel is marked as an isolated edge pixel, and the weak edge pixel is removed to ensure the edge continuity and accuracy of the image, thereby obtaining an edge image of the cable surface defect.

[0082] After obtaining an edge image of the cable surface defect, the edge image is fed into a hybrid deep learning model based on a convolutional neural network and Transformer architecture for feature extraction. This model extracts surface defect features from the edge image, which contains clear edge information. Specifically, the convolutional neural network extracts local image features, such as edge texture patterns and geometric structures, through multiple layers of convolution kernels. Pooling layers are used to reduce the data dimensionality and enhance the stability of the extracted local features. The Transformer module utilizes a self-attention mechanism to capture the correlations and global characteristics between different regions in the image, ensuring that the features reflect not only local details but also global context. Next, deep learning is used to integrate the local and global features extracted by the convolutional neural network and Transformer architecture to generate a high-dimensional feature vector that comprehensively describes the shape, size, and distribution of the cable surface defects. Finally, the extracted high-dimensional features, known as surface defect features, are output for defect detection and classification.

[0083] When processing the magnetic flux leakage signals acquired by the climbing robot to identify the cable's magnetic flux leakage defect characteristics, the signals are de-noised using wavelet transforms and low-pass filtering. Specifically, a discrete wavelet transform is first performed on the signals to decompose them into low-frequency and high-frequency components. The low-frequency components retain the signal's primary characteristics and trends, while the high-frequency components contain more noise information. During this process, an appropriate wavelet basis and number of decomposition levels are selected to ensure that the decomposition results accurately represent the signal characteristics.

[0084] Then, by setting a threshold (calculated based on statistical methods such as VisuShrink or determined empirically), soft or hard thresholding is applied to suppress noise in the high-frequency components. Soft thresholding can smoothly adjust components exceeding the threshold, while hard thresholding directly sets the noise component below the threshold to zero, ensuring that the necessary detail characteristics are retained in the denoised high-frequency components.

[0085] Next, the low-frequency components are further processed using a low-pass filter. By designing filter parameters (such as the cutoff frequency and order), any high-frequency residual noise is removed, while also enhancing signal smoothness and preserving the key trends of the cable magnetic flux leakage signal. Finally, an inverse wavelet transform is performed on the processed high- and low-frequency components to reconstruct the denoised signal. This ensures overall signal integrity and significantly improves the signal-to-noise ratio, providing high-quality input data for subsequent spectrum analysis and feature extraction.

[0086] After denoising the collected cable's magnetic flux leakage signal, the denoised signal is subjected to a fast Fourier transform based on a spectrum analysis method to obtain the distribution characteristics of the denoised signal at different frequencies, thereby obtaining the cable's magnetic flux leakage defect characteristics. Specifically, the denoised signal is first divided into several segments for spectrum analysis using a selected signal window to avoid frequency leakage. Then, a fast Fourier transform is applied to each segment of the signal, converting it from the time domain to the frequency domain. In the frequency domain, the frequency components of the signal are represented in the form of amplitude and phase. By calculating the Fourier transform of each signal segment, the corresponding spectrum is obtained, which displays the amplitude (intensity) distribution of the signal at each frequency. Here, in order to remove the influence of noise, the spectrum is smoothed, such as using a sliding window to smooth the amplitude. By observing the amplitude changes in the spectrum, the main frequency components in the signal are identified.

[0087] Surface defects in cables can cause noticeable changes in the spectrum, especially significant increases or changes in frequency components within the spectrum. For example, defects such as cracks and corrosion typically exhibit specific peaks or change patterns in the frequency domain, allowing the frequency distribution characteristics of the signal to be extracted through the spectrum graph. Extracting key signal features from the spectrum, such as frequency peaks, angles, and form factors, helps identify possible defects in the signal; for example, changes in frequency peaks indicate the presence of a specific type of defect on or within the cable. Finally, based on the distribution characteristics of the spectrum and the frequency changes of the signal, combined with known defect patterns, magnetic flux leakage defects in the cable can be identified and located by analyzing information such as the signal's amplitude, phase, and frequency band.

[0088] Step S102: The surface defect features and magnetic flux leakage defect features of the cable are simultaneously input into the generator of the generative adversarial network, and the surface defect features and magnetic flux leakage defect features are fused to obtain comprehensive defect data of the cable.

[0089] In this application, in the generative adversarial network, the fusion of surface defect features and magnetic leakage defect features is achieved through feature splicing and weighted fusion. Specifically, after the surface defect features and magnetic leakage defect features of the cable are input into the generative adversarial network, the generator of the generative adversarial network performs feature splicing on the input surface defect features and magnetic leakage defect features, and directly merges the surface defect features and magnetic leakage defect features, that is, the surface defect features and magnetic leakage defect features are spliced ​​in the feature vector dimension according to the columns or rows of the feature vector to form a new feature vector. For example, the vector dimension of the surface defect feature is , the vector dimension of the leakage magnetic defect feature is When the surface defect features and the magnetic flux leakage defect features are spliced ​​according to the rows of the feature vector, the dimension of the new feature matrix is When the surface defect features and the magnetic flux leakage defect features are spliced ​​according to the columns of the feature vector, the dimension of the new feature matrix is .

[0090] After concatenating the surface defect features and magnetic flux leakage defect features in the feature vector dimension, different weights are assigned to each feature based on its importance, adjusting its contribution to the fusion result (comprehensive defect data). This achieves a weighted fusion of the surface defect and magnetic flux leakage defect features. During the weighted fusion process, each component (surface feature and magnetic flux leakage feature) of the new feature vector (the concatenated feature vector) is assigned a weight based on its importance, adjusting its contribution to the final comprehensive defect data during the fusion process. Weighted fusion reflects the importance of features in the weights, optimizes feature representation, and improves defect detection accuracy.

[0091] In the weighted fusion process, the initial weights are randomly generated first, and the weight values ​​of the surface features and the magnetic flux leakage features are preliminarily set. Subsequently, the weights of the surface features and the magnetic flux leakage features are automatically adjusted according to the discriminator results through back propagation through the generative adversarial network, so that the contribution of the two types of features to the fusion results matches the actual detection requirements. Specifically, the concatenated feature vectors are first Surface features in and magnetic flux leakage characteristics Assign weights separately and ; Then, according to the formula:

[0092]

[0093] Surface feature parts and magnetic flux leakage characteristics Perform weighted calculation to obtain the fused comprehensive feature vector In this process, the expression ability of fusion features can be improved through nonlinear changes, and the fused feature vector That is, comprehensive defect data, which includes key information of the cable's surface defects and magnetic flux leakage defects, and provides input for subsequent defect type identification or quantitative analysis.

[0094] In this application, cable defect sample data are collected through cable fault detection (reporting, maintenance, etc.) in the historical period, including various defects of the cable during long-term use marked by surface detection, including defect types such as cracks, corrosion, and fractures, defect sizes, defect locations, etc., as well as internal defects of the cable obtained by traditional detection methods (visual inspection, ultrasonic testing, magnetic flux leakage testing, X-ray, etc.).

[0095] Then, based on the discriminator of the generative adversarial network, the loss function, mean squared error, and cross-entropy loss are calculated between the generated synthetic defect data and the previously acquired cable defect sample data. The generated synthetic defect data is evaluated, and the generator of the generative adversarial network is iteratively optimized based on the evaluation results. Specifically, the discriminator calculates the loss function, mean squared error, and cross-entropy loss, determines the difference (loss error) between the generated synthetic defect data and the defect sample data, and feeds the judgment result back to the generator. The generator then performs adaptive iterative optimization based on this feedback information. Furthermore, through the iterative optimization of the generator, the quality and accuracy of the generated synthetic defect data are further improved.

[0096] During this process, the discriminator learns the characteristics of the defect sample data and determines whether the generated synthetic defect data conforms to the common defect pattern for cables. Specifically, the discriminator extracts defect features (defect morphology, size, distribution, edge characteristics, and signal spectral characteristics) from the synthetic defect data and compares them with the defect sample data. If the defect features extracted from the generated synthetic defect data match those in the defect sample data, the discriminator automatically determines that the generated synthetic defect data conforms to the common defect pattern for cables. Otherwise, the discriminator automatically determines that the generated synthetic defect data does not conform to the common defect pattern for cables and provides feedback to optimize the generator.

[0097] The loss function, mean square error, and cross entropy loss calculated by the discriminator on the produced comprehensive defect data characterize the difference between the generated comprehensive defect data and the defect sample data; the generator is iteratively optimized based on the feedback results of the discriminator to reduce the difference between the generated comprehensive defect data and the defect sample data and improve the quality of the generated comprehensive defect data.

[0098] After receiving feedback (loss error) from the discriminator, the generator uses the backpropagation algorithm to calculate the gradients of its internal parameters. After calculating the error for each neural network layer based on the loss error, the calculated errors are propagated layer by layer using the chain rule. The generator adjusts parameter weights and biases based on this gradient information, thereby optimizing the generation of synthetic defect data.

[0099] During each training session, the generator adjusts the generated synthetic defect data based on the discriminator's feedback, bringing it closer to the defect sample data in terms of morphology, size, and distribution. As training continues, the generator continuously adjusts its generated measurements, making the generated synthetic defect data more similar to common cable defect patterns. Simultaneously, the discriminator gradually improves its judgment accuracy based on the generator's optimization. Ultimately, the synthetic defect data generated by the generator will pass the discriminator's high standards and reach a level closest to the defect sample data.

[0100] Step S103: Acquire the depth information of the cable based on the comprehensive defect data, and generate a three-dimensional point cloud of the cable defects to simultaneously acquire the surface defects and internal defects of the cable.

[0101] In this application, the edges of surface and internal defects in cables typically appear as high-contrast areas in images, such as cracks and scratches. Depth information for these edges can be derived from the corresponding magnetic flux leakage signal. Changes in the cable's magnetic flux leakage signal effectively reflect the depth and location of the defect. Furthermore, using wavelet transforms and spectral analysis techniques, features such as signal amplitude changes and frequency response are extracted from the cable's magnetic flux leakage signal. These features reflect the distinct characteristics of defects inside and outside the cable, providing information about the defect's depth.

[0102] In the two-dimensional image coordinate system In the process of combining the depth information in the magnetic flux leakage signal , convert these defect points into coordinates in three-dimensional space The location of each defect point is a point cloud, and all defect points together form a three-dimensional point cloud. The density of the point cloud represents the severity or density of the defect, while the distribution range of the point cloud can reflect the expansion of the defect.

[0103] The RANSAC algorithm is then used to fit the extracted 3D point cloud data into 3D geometric models (such as spheres, cylinders, and cracks) to describe the shape of the cable defects. This 3D defect model identifies critical areas (such as deep cracks and localized fatigue damage that could lead to fatigue fractures), assesses the impact of defects on the cable structure, and supports subsequent maintenance decisions.

[0104] In the description of the present invention, it should be understood that the terms "center", "longitudinal", "lateral", "length", "width", "thickness", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside", "clockwise", "counterclockwise", "axial", "radial", "circumferential" and the like to indicate orientations or positional relationships based on the orientations or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be understood as limiting the present invention.

[0105] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of the technical features being referred to. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one such feature. In the description of the present invention, "plurality" means at least two, such as two, three, etc., unless otherwise specifically defined.

[0106] In the present invention, unless otherwise specified or limited, the terms "installed," "connected," "connect," "fixed," etc. should be understood in a broad sense. For example, they can refer to fixed connection, detachable connection, or integration; mechanical connection, electrical connection, or communication; direct connection or indirect connection through an intermediate medium; internal communication between two elements or interaction between two elements, unless otherwise specified. Those skilled in the art will understand the specific meanings of the above terms in the present invention based on specific circumstances.

[0107] In the present invention, unless otherwise expressly specified or limited, when a first feature is "above" or "below" a second feature, it may mean that the first and second features are in direct contact, or that the first and second features are in indirect contact through an intermediary. Furthermore, when a first feature is "above," "above," or "above" a second feature, it may mean that the first feature is directly above or diagonally above the second feature, or simply means that the first feature is at a higher level than the second feature. When a first feature is "below," "below," or "below" a second feature, it may mean that the first feature is directly below or diagonally below the second feature, or simply means that the first feature is at a lower level than the second feature.

[0108] Throughout the present invention, terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples" mean that a specific feature, structure, material, or characteristic described in conjunction with that embodiment or example is included in at least one embodiment or example of the present invention. In this specification, illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples.

[0109] The foregoing description is merely a preferred embodiment of the present application and is not intended to limit the present application. Various modifications and variations are readily apparent to those skilled in the art. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present application shall be included within the scope of protection of the present application.

Claims

1. A climbing robot for cable defect detection, characterized in that: For simultaneously detecting surface defects and internal defects of a cable, the climbing robot includes: A main body frame, which is a frame structure; A clamping and walking unit installed within the frame structure of the main body frame, including: at least one main driving wheel and at least two clamping wheels. Both the main driving wheel and the clamping wheels are spindle-shaped structures, and the main driving wheel and the clamping wheels can adaptively clamp the cable passing through both ends of the frame structure of the main body frame; A cable detection unit, including a visual detection module and a magnetic flux leakage detection module, installed at both ends of the main body frame, respectively used to obtain the surface image and magnetic flux leakage signal of the cable, so as to simultaneously detect the surface defects and internal defects of the cable based on the surface image and magnetic flux leakage signal of the cable; Wherein, Perform image enhancement processing on the obtained surface image of the cable, and extract the surface defect features of the cable based on the Canny edge detection algorithm and a pre-constructed composite deep learning model; and, perform denoising processing on the collected magnetic flux leakage signal of the cable, and obtain the magnetic flux leakage defect features of the cable based on the spectral analysis method; Simultaneously input the surface defect features and the magnetic flux leakage defect features of the cable into the generator of the generative adversarial network. The generator of the generative adversarial network sequentially performs feature splicing and weighted fusion on the surface defect features and the magnetic flux leakage defect features to obtain the comprehensive defect data of the cable; According to the comprehensive defect data, obtain the depth information of the cable defect and generate the three-dimensional point cloud of the cable defect, so as to simultaneously obtain the surface defects and internal defects of the cable; A driving unit installed on the frame structure of the main body frame, used to drive the main driving wheel to rotate, so as to drive the main body frame to move along the length direction on the cable.

2. The climbing robot for cable defect detection according to claim 1, characterized in that: The main body frame includes: an upper frame and a lower frame, Upper connecting frames with a 冂-shaped structure are respectively arranged at both ends of the upper frame, and lower connecting frames with a 凵-shaped structure adapted to the upper connecting frames are respectively arranged at both ends of the lower frame. The upper connecting frames and the lower connecting frames can be telescopically adjusted along a first direction; wherein, the first direction is the extending direction of the side plate of the upper connecting frame or the lower connecting frame; The main driving wheel is rotatably installed on the bottom surface of the upper frame, and the clamping wheels are installed on the top surface of the lower frame, and the relative distance of the clamping wheels relative to the main driving wheel can be adaptively adjusted according to the radial dimension of the cable.

3. The climbing robot for cable defect detection according to claim 2, characterized in that: The clamping wheels are rotatably installed on the lower frame through a clamping wheel frame. The clamping wheel frame includes: a rotating connecting rod and a tension spring, One end of the rotating connecting rod is rotatably installed on the lower frame, and the other end is rotatably installed with the clamping wheel; One end of the tension spring is connected to the middle part of the rotating connecting rod, and the other end is connected to the lower frame, used to drive the rotating connecting rod to have a relative movement trend towards the cable around its connection axis with the lower frame, so as to make the clamping wheels clamp the cable; There are two groups of the clamping wheel frames, and the two groups of clamping wheel frames are symmetrically arranged at both ends of the lower frame to drive the two clamping wheels to have a trend of moving towards each other.

4. The climbing robot for cable defect detection according to claim 1, characterized in that: There are two main drive wheels, and the two main drive wheels are arranged opposite to the two clamping wheels respectively; And / or, the two main driving wheels are respectively connected to the output ends of the two driving units through synchronous belt transmission.

5. The climbing robot for cable defect detection according to claim 1, characterized in that: The visual inspection module includes: an image detection bracket and at least three visual inspection cameras, wherein the image detection bracket is mounted on the head of the main frame and has at least three claws evenly distributed along the circumference, and three visual inspection cameras are arranged on the at least three claws in a one-to-one correspondence, so that the imaging range of the three visual inspection cameras covers the surface of the cable along the circumference; The magnetic flux leakage detection module includes: a magnetic detection bracket, an excitation device and a detection device; The magnetic detection bracket is installed at the tail of the main frame; The excitation device is an annular structure, mounted on the magnetic detection bracket close to the main frame, and is used to magnetize the internal steel wire of the cable; wherein the cable passes through the center of the annular structure of the excitation device; The detection device is an annular structure, installed on the magnetic detection bracket in parallel with the excitation device and away from the main frame, and is used to detect magnetic defects in the magnetized steel wire inside the cable.

6. The climbing robot for cable defect detection according to claim 1, characterized in that: Based on the contrast-limited adaptive histogram equalization method, the surface image after Gaussian filtering and denoising is enhanced to obtain the surface enhanced image of the cable. performing gradient calculation on the surface enhancement image of the cable, removing edge redundant pixels from the obtained gradient image based on a non-maximum suppression method, and performing edge detection and connection on the image from which the redundant edge pixels have been removed based on a double threshold method to obtain an edge image of the cable surface defects; The edge image is input into a composite deep learning model based on a convolutional neural network and a Transformer architecture to extract surface defect features of the cable.

7. The climbing robot for cable defect detection according to claim 1, characterized in that: The collected cable magnetic flux leakage signal is de-noised by wavelet transform and low-pass filtering. Based on the spectrum analysis method, the denoised signal obtained by the denoising process is subjected to fast Fourier transform to obtain the distribution characteristics of the denoised signal at different frequencies, so as to obtain the magnetic leakage defect characteristics of the cable.

8. The climbing robot for cable defect detection according to claim 1, characterized in that: Also includes: Based on the discriminator of the generative adversarial network, the loss function, mean square error and cross entropy loss between the generated comprehensive defect data and the pre-acquired cable defect sample data are calculated respectively to evaluate the generated comprehensive defect data, and the generator of the generative adversarial network is iteratively optimized according to the evaluation results.

Images