Communication optical cable fault identification method and system based on machine vision

Abstract

The invention relates to the technical field of computer vision, in particular to a communication optical cable fault identification method and system based on machine vision. The method comprises the following steps: acquiring an original visible light optical cable image and an original near-infrared optical cable image; performing dual-spectrum registration and pixel-level fusion on the original visible light optical cable image and the original near-infrared optical cable image to obtain a registered communication optical cable image; performing local contrast improvement on the registered communication optical cable image to obtain an enhanced communication optical cable image; constructing a communication optical cable motion prediction model; performing dynamic tracking and geometric constraint screening on the communication optical cable target section based on the communication optical cable motion prediction model to obtain optical cable detection area positioning data; and performing surface morphological feature extraction on the enhanced communication optical cable image to obtain optical cable surface depression feature data. According to the method, the problem of recognition blind areas formed by single spectrum imaging under complex illumination and shielding conditions is effectively solved.

Description

Technical Field

[0001] The present invention relates to the field of computer vision technology, and in particular to a communication optical cable fault identification method and system based on machine vision. Background Art

[0002] In mountainous backbone communication networks, overhead fiber optic cables are often installed on steep slopes, in canyons, or at high altitudes, carrying the brunt of information transmission. These environments are characterized by complex terrain and unpredictable climates. Traditional manual inspections require climbing towers or traversing uninhabited areas, resulting in low efficiency and high risk. In recent years, drones equipped with visible light / infrared cameras for aerial inspection have become a mainstream alternative, enabling rapid acquisition of surface images of optical cables. However, existing machine vision-based recognition methods still suffer from invisibility limitations in mountainous environments. When drones take aerial photos of suspended optical cables in undulating terrain, localized areas of the cable surface create blind spots due to multi-angle illumination interference and physical obstructions. For example, cracks or sheath damage in locations such as recessed areas on the shady side of the cable and shadowed areas at the connection points of tower hardware can be easily misidentified as light and shadow noise in single-view imaging. Furthermore, strong winds in mountainous areas cause optical cables to oscillate slightly at high frequencies, making it difficult for conventional target tracking algorithms to lock onto the same detection point. This results in drifting fault feature positions in consecutive image frames, making it impossible to enhance recognition confidence through time-series comparison. Summary of the Invention

[0003] Based on this, it is necessary for the present invention to provide a communication optical cable fault identification method and system based on machine vision to solve at least one of the above technical problems.

[0004] To achieve the above object, a communication optical cable fault identification method based on machine vision includes the following steps:

[0005] Step S1: obtaining an original visible light cable image and an original near-infrared cable image; performing dual-spectral registration and pixel-level fusion on the original visible light cable image and the original near-infrared cable image to obtain a registered communication cable image; performing local contrast enhancement on the registered communication cable image to obtain an enhanced communication cable image;

[0006] Step S2: constructing a communication optical cable motion prediction model; dynamically tracking and geometrically constraining the target section of the communication optical cable based on the communication optical cable motion prediction model to obtain positioning data of the optical cable detection area;

[0007] Step S3: extracting surface morphological features of the enhanced communication optical cable image to obtain optical cable surface depression feature data; locating the fault candidate area based on the optical cable surface depression feature data to obtain located optical cable fault candidate area data;

[0008] Step S4: continuously multi-frame tracking verification is performed on the positioning optical cable fault candidate area data to obtain optical cable fault area time sequence stability data; communication optical cable motion characteristic data is acquired; optical cable motion constraint condition parameters are generated based on the communication optical cable motion characteristic data; optical cable fault authenticity evaluation is performed according to the optical cable motion constraint condition parameters and the optical cable fault area time sequence stability data, and verified optical cable fault feature data is obtained;

[0009] Step S5: fault severity evaluation is performed according to the verified optical cable fault feature data, and an optical cable fault severity level is obtained; a structured optical cable fault detection report is generated based on the optical cable fault severity level.

[0010] The present application can complement the advantages of visible light images and near-infrared images through dual-spectrum registration and pixel-level fusion. Visible light images provide rich texture details, while near-infrared images enhance the sensitivity to surface material and temperature differences. This not only effectively solves the recognition blind area problem formed by single-spectrum imaging under complex lighting and shielding conditions, but also significantly improves the visibility and recognition of optical cable surface features, making small cracks, depressions and other fault features more clearly identifiable in the fused image. Through the combination of dynamic tracking and geometric constraint filtering, precise positioning of the optical cable detection area is provided with strong technical support. In strong wind environments in mountainous areas, the high-frequency and small-amplitude swing of the optical cable poses a great challenge to conventional target tracking algorithms. By constructing a communication optical cable motion prediction model, the motion trajectory of the optical cable can be predicted in real time, and the detection area is verified by combining geometric constraint conditions. This not only overcomes the interference of optical cable swing on target tracking, but also ensures the stability and continuity of fault detection, even in the case of rapid swing of the optical cable, the fault area can be accurately locked. Through continuous multi-frame tracking verification and motion constraint condition evaluation of the fault candidate area, the present application improves the accuracy and confidence of fault identification. Through time series analysis and motion characteristic constraints, false positives caused by optical cable swing or lighting changes can be effectively filtered. For example, the pseudo-image caused by temperature difference deformation on the surface of the optical cable is similar to the real fault in a single frame image, but through multi-frame tracking and motion constraint verification, real faults and pseudo-images can be accurately distinguished, thereby significantly improving the accuracy and confidence of fault identification. The structured detection report generated based on the fault feature data provides comprehensive and accurate information support for optical cable maintenance. The report not only includes the severity level of the fault, but also provides the precise geographic coordinates of the fault through the coordinate mapping of the UAV GPS coordinates and the camera internal parameter table. This structured detection report can directly guide the maintenance personnel to carry out precise repair, greatly improving the efficiency and reliability of optical cable inspection, while reducing the cost and risk of manual re-inspection.

[0011] Preferably, step S1 comprises the following steps:

[0012] Step S11: Visible light image acquisition is performed on the target segment of the communication optical cable to obtain an original visible light cable image, wherein a CMOS sensor with a resolution of 4096*3072 pixels is used for image acquisition, an exposure time of 1 / 500 seconds is set, and an ISO value of 100 is set;

[0013] Step S12: Near-infrared image acquisition is performed on the target segment of the communication optical cable to obtain an original near-infrared cable image, wherein a near-infrared CMOS sensor with a wavelength range of 780-1000 nm is used for image acquisition, a gain of 2 times is set, an exposure time of 1 / 250 seconds is set, and a filter transmittance of 85% is set;

[0014] Step S13: SIFT feature point extraction is performed on the original visible light cable image to obtain a visible light image feature point set, and SIFT feature point extraction is performed on the original near-infrared cable image to obtain a near-infrared image feature point set;

[0015] Step S14: Feature matching is performed on the visible light image feature point set and the near-infrared image feature point set to obtain a dual-spectrum feature matching point pair;

[0016] Step S15: Affine transformation matrix construction is performed according to the dual-spectrum feature matching point pair to obtain dual-spectrum registration transformation parameters;

[0017] Step S16: Geometric correction is performed on the original near-infrared cable image based on the dual-spectrum registration transformation parameters to obtain a registered near-infrared cable image, and geometric correction is performed on the original near-infrared cable image based on the dual-spectrum registration transformation parameters to obtain a registered near-infrared cable image;

[0018] Step S17: Pixel-level fusion is performed on the original visible light cable image and the registered near-infrared cable image to obtain a registered communication cable image;

[0019] Step S18: Local contrast enhancement is performed on the registered communication cable image to obtain an enhanced communication cable image.

[0020] Preferably, the step S2 of constructing the communication cable motion prediction model comprises the following steps:

[0021] Step S21: Obtain three-axis gyroscope original data in the flight process of the unmanned aerial vehicle to obtain unmanned aerial vehicle angular velocity measurement data;

[0022] Step S22: Obtain three-axis accelerometer original data in the flight process of the unmanned aerial vehicle to obtain unmanned aerial vehicle linear acceleration measurement data;

[0023] Step S23: zero offset correction is performed on the unmanned aerial vehicle angular velocity measurement data to obtain corrected unmanned aerial vehicle angular velocity data, wherein a static correction method is adopted, and the average value of 500 sampling points in a static state is calculated as a zero offset compensation value;

[0024] Step S24: gravity component separation is performed on the unmanned aerial vehicle linear acceleration measurement data to obtain unmanned aerial vehicle net acceleration data;

[0025] Step S25: attitude angle integral calculation is performed based on the corrected unmanned aerial vehicle angular velocity data to obtain unmanned aerial vehicle attitude angle data, wherein a quaternion integral method is adopted, and the integral step is 0.001 seconds, including pitch angle, roll angle and yaw angle;

[0026] Step S26: coordinate system conversion is performed according to the unmanned aerial vehicle attitude angle data and the unmanned aerial vehicle net acceleration data to obtain unmanned aerial vehicle motion state data;

[0027] Step S27: optical cable relative motion solving is performed on the unmanned aerial vehicle motion state data to obtain optical cable swing relative displacement data;

[0028] Step S28: frequency domain feature extraction is performed based on the optical cable swing relative displacement data to obtain optical cable swing spectrum feature parameters; and a communication optical cable motion prediction model is constructed according to the optical cable swing spectrum feature parameters.

[0029] Preferably, the dynamic tracking and geometric constraint screening of the target section of the communication optical cable based on the communication optical cable motion prediction model in step S2 comprises the following steps:

[0030] Step S291: Harris corner point detection is performed on the enhanced communication optical cable image to obtain an optical cable image feature corner point set; feature point state initialization is performed based on the optical cable image feature corner point set to obtain optical cable feature point state initial data;

[0031] Step S292: motion prediction is performed on the optical cable feature point state initial data according to the communication optical cable motion prediction model to obtain optical cable feature point predicted positions;

[0032] Step S293: continuous frame image acquisition is performed on the target section of the communication optical cable to obtain communication optical cable continuous frame images, and gradient domain enhancement is performed on the communication optical cable continuous frame images to obtain enhanced communication optical cable continuous frame images;

[0033] Step S294: optical cable feature point observation update is performed on the enhanced communication optical cable continuous frame images to obtain optical cable feature point observed positions;

[0034] Step S295: fusion is performed on the optical cable feature point predicted positions and the optical cable feature point observed positions to obtain optical cable stable tracking coordinates;

[0035] Step S296: Normalized correlation template matching is performed on the optical cable stable tracking coordinates to obtain an optical cable detection region candidate position set;

[0036] Step S297: Geometric feature analysis is performed on the optical cable detection region candidate position set to obtain optical cable detection region geometric feature data, and an optical cable geometric constraint condition set is constructed based on the optical cable detection region geometric feature data;

[0037] Step S298: The optical cable detection region candidate position set is screened and verified based on the optical cable geometric constraint condition set to obtain optical cable detection region positioning data.

[0038] Preferably, the surface morphology feature extraction on the enhanced communication optical cable image in step S3 comprises the following steps:

[0039] Step S31: A texture extraction filter is constructed according to the enhanced communication optical cable image to obtain an optical cable surface texture extraction filter;

[0040] Step S32: Convolution operation is performed on the enhanced communication optical cable image by using the optical cable surface texture extraction filter to obtain optical cable multi-scale texture response data;

[0041] Step S33: Amplitude normalization is performed on the optical cable multi-scale texture response data to obtain standard optical cable texture feature data;

[0042] Step S34: Multi-directional texture energy evaluation is performed based on the standard optical cable texture feature data to obtain optical cable surface texture energy distribution data;

[0043] Step S35: Self-adaptive threshold segmentation is performed on the optical cable surface texture energy distribution data to obtain optical cable texture salient region data;

[0044] Step S36: A morphological structure element is constructed according to the optical cable texture salient region data to obtain an optical cable special morphological operator;

[0045] Step S37: Top-Hat transformation is performed on the enhanced communication optical cable image by using the optical cable special morphological operator to obtain optical cable surface recess feature data.

[0046] Preferably, the fault candidate region positioning based on the optical cable surface recess feature data in step S38 comprises the following steps:

[0047] Step S381: Morphological reconstruction is performed on the optical cable surface recess feature data to obtain optical cable surface abnormal morphology data; edge detection is performed based on the optical cable surface abnormal morphology data to obtain optical cable abnormal edge contour data;

[0048] Step S382: logical AND operation is performed on the optical cable abnormal edge profile data and the optical cable surface indentation feature data to obtain an optical cable surface abnormal feature map;

[0049] Step S383: connected domain labeling is performed on the optical cable surface abnormal feature map to obtain an optical cable abnormal connected domain label;

[0050] Step S384: geometric feature parameter extraction is performed based on the optical cable abnormal connected domain label to obtain optical cable abnormal region geometric feature data;

[0051] Step S385: multi-condition screening filtering is performed on the optical cable abnormal region geometric feature data according to a preset screening condition to obtain optical cable suspected fault region data;

[0052] Step S386: spatial position matching is performed on the optical cable suspected fault region data according to the optical cable detection region positioning data to obtain optical cable fault candidate region coordinates;

[0053] Step S387: boundary extraction is performed based on the optical cable fault candidate region coordinates to obtain optical cable fault candidate region boundary data; and the center of mass coordinates of the optical cable fault candidate region boundary data is calculated to obtain positioning optical cable fault candidate region data.

[0054] Preferably, step S4 comprises the following steps:

[0055] Step S41: time sequence window division is performed on the positioning optical cable fault candidate region data to obtain an optical cable fault region time sequence window;

[0056] Step S42: region center trajectory extraction is performed based on the optical cable fault region time sequence window to obtain optical cable fault region center trajectory data;

[0057] Step S43: trajectory smoothing filtering is performed on the optical cable fault region center trajectory data to obtain smoothed optical cable fault region trajectory data;

[0058] Step S44: trajectory displacement variance calculation is performed based on the smoothed optical cable fault region trajectory data to obtain an optical cable fault region displacement stability parameter;

[0059] Step S45: morphological feature time sequence change monitoring is performed on the positioning optical cable fault candidate region data to obtain optical cable fault region morphological change data;

[0060] Step S46: morphological stability evaluation is performed according to the optical cable fault region morphological change data to obtain an optical cable fault region morphological stability parameter;

[0061] Step S47: comprehensive stability evaluation is performed based on the optical cable fault region displacement stability parameter and the optical cable fault region morphological stability parameter to obtain optical cable fault region time sequence stability data;

[0062] Step S48: Obtain communication cable motion characteristic data; generate cable motion constraint condition parameters based on the communication cable motion characteristic data; perform cable fault authenticity evaluation according to the cable motion constraint condition parameters and the cable fault region time sequence stability data, and obtain verified cable fault feature data.

[0063] Preferably, step S48 comprises the following steps:

[0064] Step S481: Obtain enhanced communication cable continuous frame images; perform optical flow field feature extraction on the enhanced communication cable continuous frame images to obtain cable image optical flow field data;

[0065] Step S482: Extract local optical flow vectors from the cable image optical flow field data based on the positioning cable fault candidate region data to obtain cable fault region optical flow vectors;

[0066] Step S483: Perform vector consistency evaluation on the cable fault region optical flow vectors to obtain cable fault region optical flow consistency parameters;

[0067] Step S484: Perform motion characteristic feature collection on the communication cable to obtain communication cable motion characteristic data, and construct motion constraint conditions based on the communication cable motion characteristic data to obtain cable motion constraint condition parameters;

[0068] Step S485: Perform constraint verification on the cable fault region optical flow consistency parameters based on the cable motion constraint condition parameters to obtain constraint verified cable fault optical flow data;

[0069] Step S486: Perform spatio-temporal correlation on the cable fault region time sequence stability data and the constraint verified cable fault optical flow data to obtain cable fault spatio-temporal correlation feature data;

[0070] Step S487: Perform fault authenticity scoring based on the cable fault spatio-temporal correlation feature data to obtain cable fault authenticity scoring;

[0071] Step S488: Perform threshold value judgment screening on the cable fault authenticity scoring to obtain screened cable fault region data; perform fault feature parameter statistics based on the screened cable fault region data to obtain verified cable fault feature data.

[0072] Preferably, step S5 comprises the following steps:

[0073] Step S51: Perform fault type feature extraction on the verified cable fault feature data to obtain a cable fault type feature vector;

[0074] Step S52: Perform fault classification rule set construction based on the cable fault type feature vector to obtain a cable fault classification discrimination rule;

[0075] Step S53: According to the optical cable fault classification judgment rule, the fault type of the verification optical cable fault feature data is identified, and optical cable fault type identification data is obtained.

[0076] Step S54: According to the optical cable fault type identification data, the defect geometric size is quantified, and the optical cable fault geometric size parameter is obtained.

[0077] Step S55: Based on the optical cable fault geometric size parameter, the defect risk degree is evaluated, and the optical cable fault severity level is obtained.

[0078] Step S56: Obtain the UAV GPS coordinates and the camera internal parameter table carried;

[0079] Step S57: According to the UAV GPS coordinates and the camera internal parameter table carried, the coordinate mapping of the verification optical cable fault feature data is carried out, and the communication optical cable fault geographic coordinates are obtained; According to the optical cable fault severity level and the communication optical cable fault geographic coordinates, a structured optical cable fault detection report is generated.

[0080] Preferably, the present application also provides a machine vision-based communication optical cable fault identification system for executing the machine vision-based communication optical cable fault identification method as described above, which comprises:

[0081] An image preprocessing module is configured to obtain original visible light cable images and original near-infrared cable images; perform dual-spectrum registration and pixel-level fusion on the original visible light cable images and the original near-infrared cable images to obtain registered communication cable images; and perform local contrast enhancement on the registered communication cable images to obtain enhanced communication cable images.

[0082] A dynamic tracking module is configured to construct a communication cable motion prediction model; and perform dynamic tracking and geometric constraint filtering on a communication cable target segment based on the communication cable motion prediction model to obtain cable detection area positioning data.

[0083] A feature extraction module is configured to extract surface morphology features from the enhanced communication cable images to obtain cable surface depression feature data; and perform fault candidate area positioning based on the cable surface depression feature data to obtain positioned cable fault candidate area data.

[0084] A fault verification module is configured to perform continuous multi-frame tracking verification on the positioned cable fault candidate area data to obtain cable fault area time sequence stability data; obtain communication cable motion characteristic data; generate cable motion constraint condition parameters based on the communication cable motion characteristic data; and perform cable fault authenticity evaluation based on the cable motion constraint condition parameters and the cable fault area time sequence stability data to obtain verification optical cable fault feature data.

[0085] The evaluation and reporting module is used to evaluate the severity of the fault based on the verified optical cable fault characteristic data to obtain the optical cable fault severity level; and generate a structured optical cable fault detection report based on the optical cable fault severity level.

[0086] In the present invention, the image preprocessing module effectively solves the limitations of single spectrum imaging in complex environments through dual-spectral registration and pixel-level fusion, and significantly improves the quality and feature recognition of optical cable images. The dynamic tracking module, combined with the communication optical cable motion prediction model and geometric constraint screening, can accurately locate the optical cable detection area, overcome the interference of optical cable swinging caused by strong winds in mountainous areas on target tracking, and ensure the stability and continuity of fault detection. The feature extraction module can accurately extract the concave features on the surface of the optical cable and accurately locate the candidate fault area, providing a reliable basis for further fault verification. The fault verification module effectively filters out misjudgments caused by optical cable swinging or lighting changes through multi-frame tracking verification and motion constraint condition evaluation, significantly improving the accuracy and confidence of fault identification. The evaluation and reporting module can generate a structured detection report containing the fault severity level and geographic coordinates based on the verified fault feature data, providing comprehensive and accurate information support for optical cable maintenance. In summary, the present invention not only improves the efficiency and reliability of optical cable fault detection, but also reduces operation and maintenance costs and the risk of manual re-inspection. It is particularly suitable for communication optical cable inspection tasks in complex environments such as mountainous areas, and has broad application prospects and important practical significance. BRIEF DESCRIPTION OF THE DRAWINGS

[0087] Other features, objects and advantages of the present invention will become more apparent from reading the detailed description made with reference to the following drawings:

[0088] Fig. 1 A schematic flow chart of the steps of a method for identifying communication optical cable faults based on machine vision according to an embodiment is shown.

[0089] Fig. 2 A detailed flowchart of step S3 of an embodiment is shown.

[0090] Fig. 3 A detailed flowchart of step S38 of an embodiment is shown. DETAILED DESCRIPTION

[0091] The following is a clear and complete description of the technical method of the present invention in conjunction with the accompanying drawings. It is obvious that the embodiments described are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without making any creative efforts are within the scope of protection of the present invention.

[0092] Furthermore, the accompanying drawings are included to provide a further understanding of the present application, and are incorporated in and constitute a part of this specification. The drawings illustrate embodiments of the present application and, together with the description, serve to explain the principles of the present application. In the drawings:

[0093] It is to be understood that, although terms such as "first", "second", and so on can be used herein to describe various elements, these elements should not be limited by these terms. These terms are only used to distinguish one element from another. For example, without departing from the scope of the exemplary embodiments, a first element can be referred to as a second element, and similarly a second element can be referred to as a first element. The term "and / or" as used herein includes any and all combinations of one or more of the associated associated items.

[0094] To achieve the above object, there is provided Figs. 1 to 3 The present application provides a communication optical cable fault identification method based on machine vision, comprising the following steps:

[0095] Step S1: Obtain an original visible light cable image and an original near-infrared light cable image; perform dual-spectrum registration and pixel-level fusion on the original visible light cable image and the original near-infrared light cable image to obtain a registered communication optical cable image; and perform local contrast enhancement on the registered communication optical cable image to obtain an enhanced communication optical cable image;

[0096] Step S2: Construct a communication optical cable motion prediction model; perform dynamic tracking and geometric constraint filtering on a communication optical cable target section based on the communication optical cable motion prediction model to obtain optical cable detection region positioning data;

[0097] Step S3: Extract surface morphological features from the enhanced communication optical cable image to obtain cable surface depression feature data; and perform fault candidate region positioning based on the cable surface depression feature data to obtain positioned cable fault candidate region data;

[0098] Step S4: Perform continuous multi-frame tracking verification on the positioned cable fault candidate region data to obtain cable fault region time sequence stability data; obtain communication optical cable motion characteristic data; generate cable motion constraint condition parameters based on the communication optical cable motion characteristic data; and perform cable fault authenticity evaluation according to the cable motion constraint condition parameters and the cable fault region time sequence stability data to obtain verified cable fault feature data;

[0099] Step S5: performing fault severity assessment according to the verified optical cable fault feature data to obtain an optical cable fault severity level; and generating a structured optical cable fault detection report based on the optical cable fault severity level.

[0100] Preferably, step S1 comprises the following steps:

[0101] Step S11: performing visible light image acquisition on the target segment of the communication optical cable to obtain an original visible light optical cable image, wherein a CMOS sensor with a resolution of 4096*3072 pixels is used for image acquisition, an exposure time of 1 / 500 seconds is set, and an ISO value of 100 is set.

[0102] Step S12: performing near-infrared image acquisition on the target segment of the communication optical cable to obtain an original near-infrared optical cable image, wherein a near-infrared CMOS sensor with a wavelength range of 780-1000 nm is used for image acquisition, a gain of 2 times is set, an exposure time of 1 / 250 seconds is set, and a filter transmittance of 85% is set.

[0103] Step S13: performing SIFT feature point extraction on the original visible light optical cable image to obtain a visible light image feature point set, and performing SIFT feature point extraction on the original near-infrared optical cable image to obtain a near-infrared image feature point set.

[0104] Step S14: performing feature matching on the visible light image feature point set and the near-infrared image feature point set to obtain a dual-spectrum feature matching point pair.

[0105] Step S15: constructing an affine transformation matrix according to the dual-spectrum feature matching point pair to obtain a dual-spectrum registration transformation parameter.

[0106] Step S16: performing geometric correction on the original near-infrared optical cable image based on the dual-spectrum registration transformation parameter to obtain a registered near-infrared optical cable image, and performing geometric correction on the original near-infrared optical cable image based on the dual-spectrum registration transformation parameter to obtain a registered near-infrared optical cable image.

[0107] Step S17: performing pixel-level fusion on the original visible light optical cable image and the registered near-infrared optical cable image to obtain a registered communication optical cable image.

[0108] Step S18: performing local contrast enhancement on the registered communication optical cable image to obtain an enhanced communication optical cable image.

[0109] In this embodiment, in the communication cable inspection scene, a CMOS sensor camera with a resolution of 4096x3072 pixels is used for visible light image acquisition. The camera is mounted on a drone for shooting the target section of the communication cable. During shooting, the exposure time of the camera is set to 1 / 500 seconds and the ISO value is set to 100. For example, in the communication cable inspection task in mountainous areas, the drone flies along the preset flight path, and when approaching the target section of the cable, the camera automatically triggers the shooting function to obtain clear raw visible light cable images. Similarly, using the equipment carried by the drone, a near-infrared CMOS sensor with a wavelength range of 780-1000 nm is used for near-infrared image acquisition. During acquisition, the gain of the sensor is set to 2 times, the exposure time is set to 1 / 250 seconds, and a filter with a transmittance of 85% is selected. Taking the communication cable inspection in mountainous areas as an example, the near-infrared sensor is aimed at the target section of the cable during the flight of the drone. This setting can enhance the sensitivity to the surface material and temperature difference of the cable, especially in areas that are difficult to distinguish under visible light conditions, such as the shady side or shadow area, the near-infrared image can provide clearer details. The SIFT feature points of the collected raw visible light cable image and raw near-infrared cable image are extracted respectively by using image processing software (such as Open CV). Taking the visible light image as an example, the image is imported into the Open CV software, and the SIFT detector function is called. The software will automatically detect a series of feature points in the image and form a visible light image feature point set. Similarly, the same operation is performed on the near-infrared image to obtain a near-infrared image feature point set. For example, in a complex cable image, the SIFT algorithm can identify key feature points such as texture and edge on the surface of the cable. These feature points have scale invariance and rotation invariance, and can remain stable under different viewing angles and lighting conditions. With the help of the feature matching function in the Open CV software, the visible light image feature point set and the near-infrared image feature point set are matched. In specific operation, the software compares the descriptors of the feature points for similarity, finds similar feature point pairs in the two images, and forms a dual-spectrum feature matching point pair. For example, in the cable image, one feature point in the visible light image corresponds to one feature point in the near-infrared image. These two points are corresponding in spatial position, but they show different features due to different imaging spectra. Using the functions in the Open CV software, an affine transformation matrix is constructed based on the dual-spectrum feature matching point pair. Specifically, the software calculates the affine transformation parameters that can transform the near-infrared image to align with the visible light image based on the positional relationship of the matching point pair. For example, after the matching point pair is determined, the software solves a best affine transformation matrix through least squares method or other optimization algorithms. The matrix can describe the geometric transformation relationship between the two images, including translation, rotation and scaling operations.Based on the dual-spectrum registration transformation parameters, the original near-infrared cable image is geometrically corrected using image processing software (such as MATLAB or Open CV). In the software, the coordinates of each pixel point in the near-infrared image are transformed according to the affine transformation matrix, so that it is aligned with the visible light image in spatial position, thereby obtaining the registered near-infrared cable image. For example, for a pixel point in the near-infrared image, its corresponding position in the visible light image coordinate system is calculated through the affine transformation matrix, and the pixel value of the point is mapped to the new position. The pixel-level fusion is performed on the registered near-infrared cable image and the original visible light cable image by using an image fusion algorithm (such as a wavelet transform-based fusion algorithm). In the fusion process, the two images are first decomposed by wavelet to obtain detailed information in different scales and directions; the detailed information is fused according to certain fusion rules, such as taking the absolute value or weighted average; and finally, wavelet reconstruction is performed to obtain the fused registered communication cable image. For example, in the fusion process, the visible light image provides clearer information for the texture details on the surface of the cable, while the near-infrared image is more advantageous for the manifestation of material and temperature differences. Through pixel-level fusion, the advantages of the two images can be complementary. The registered communication cable image is subjected to local contrast enhancement using Adobe Photo shop software. In Photo shop, select the "Adjustment" option in the "Image" menu and click the "Curve" tool. By adjusting the shape of the curve, the contrast of the image can be enhanced. For example, slightly press the dark part of the curve and slightly raise the bright part to form a curve similar to "S", which can enhance the overall contrast of the image while preserving details. Using the "Shadow / Highlight" tool, move the "Shadow" slider to the right and the "Highlight" slider to the left to adjust the dark and bright parts of the image, respectively, so that the details such as recesses and cracks on the surface of the cable are more clearly visible. Through these operations, the enhanced communication cable image is finally obtained.

[0110] Preferably, the step S2 of constructing the communication cable motion prediction model comprises the following steps:

[0111] Step S21: Obtain the three-axis gyroscope original data in the flight process of the unmanned aerial vehicle to obtain the unmanned aerial vehicle angular velocity measurement data;

[0112] Step S22: Obtain the three-axis accelerometer original data in the flight process of the unmanned aerial vehicle to obtain the unmanned aerial vehicle linear acceleration measurement data;

[0113] Step S23: Perform zero offset correction on the unmanned aerial vehicle angular velocity measurement data to obtain corrected unmanned aerial vehicle angular velocity data, wherein a static correction method is adopted to calculate the average value of 500 sampling points in a static state as the zero offset compensation value;

[0114] Step S24: gravity component separation is performed on the unmanned aerial vehicle linear acceleration measurement data to obtain unmanned aerial vehicle net acceleration data;

[0115] Step S25: attitude angle integral calculation is performed based on the corrected unmanned aerial vehicle angular velocity data to obtain unmanned aerial vehicle attitude angle data, wherein a quaternion integral method is adopted, the integral step is 0.001 seconds, and the integral result contains pitch angle, roll angle and yaw angle;

[0116] Step S26: coordinate system conversion is performed according to the unmanned aerial vehicle attitude angle data and the unmanned aerial vehicle net acceleration data to obtain unmanned aerial vehicle motion state data;

[0117] Step S27: optical cable relative motion calculation is performed on the unmanned aerial vehicle motion state data to obtain optical cable swing relative displacement data;

[0118] Step S28: frequency domain feature extraction is performed based on the optical cable swing relative displacement data to obtain optical cable swing frequency spectrum feature parameters; and a communication optical cable motion prediction model is constructed according to the optical cable swing frequency spectrum feature parameters.

[0119] In this example, during a communication cable inspection task, a DJI drone is used for flight operations. The DJI drone is equipped with a built-in three-axis gyroscope sensor that can measure the angular velocity of the drone in real-time. By connecting the drone to its companion software (such as DJI Pilot or DJI Fly), the raw three-axis gyroscope data during flight can be obtained. These data are recorded at a sampling frequency of 100 Hz, including angular velocity values around the X, Y, and Z axes. For example, during a mountainous communication cable inspection task, the gyroscope sensor records angular velocity data during the flight of the drone, which is transmitted to the ground control station in real-time through the software. Similarly, a DJI drone is used for flight operations, and its built-in three-axis accelerometer can measure the linear acceleration of the drone. Through DJI Pilot or DJI Fly software, the raw three-axis accelerometer data during the flight of the drone can be obtained. These data are also recorded at a sampling frequency of 100 Hz, including acceleration values along the X, Y, and Z axes. For example, during flight, the accelerometer records acceleration data of the drone at different flight attitudes, which will be used for subsequent gravity component separation and motion state calculation. Through the software interface, the accelerometer data is viewed in real-time. In MATLAB, load the angular velocity data file and select the static correction method. The specific operation is to calculate the average value of 500 sampling points of the drone in a stationary state as the zero offset compensation value. For example, before the drone takes off, place it on a stable ground and record the angular velocity data of 500 sampling points, calculate the average value. Subtract this average value from the angular velocity data of the entire flight process to obtain the corrected angular velocity data of the drone. Use MATLAB software to perform gravity component separation on the linear acceleration measurement data of the drone. In MATLAB, load the accelerometer data file and use the functions in the built-in signal processing toolbox to decompose the acceleration data into gravity components and net acceleration components. The specific operation is to extract the gravity component through a low-pass filter and subtract the gravity component from the original acceleration data to obtain the net acceleration data of the drone. For example, select a low-pass filter with a cutoff frequency of 0.1 Hz to filter the acceleration data. Use the quaternion integration method in MATLAB software to perform attitude angle integration calculation on the corrected angular velocity data of the drone. In MATLAB, load the corrected angular velocity data file and call the quaternion integration function with an integration step size of 0.001 seconds. This function will calculate the pitch angle, roll angle, and yaw angle of the drone according to the angular velocity data. For example, in a flight task, the attitude angle data of the drone at different time points is calculated through the quaternion integration method and saved as a new data file. In MATLAB, load the attitude angle data and net acceleration data files and use the coordinate transformation matrix to convert the acceleration data from the body coordinate system of the drone to the geographic coordinate system.The specific operation is to construct a rotation matrix according to the pitch angle, roll angle and yaw angle, multiply the net acceleration data by the rotation matrix, and obtain the motion state data of the unmanned aerial vehicle in the geographical coordinate system. For example, through coordinate system conversion, the acceleration components of the unmanned aerial vehicle in the north, east and ground directions can be obtained, so that the motion state of the unmanned aerial vehicle can be more intuitively described. The motion state data of the unmanned aerial vehicle is calculated by using the MATLAB software. In MATLAB, the motion state data file of the unmanned aerial vehicle is loaded, and the swing displacement data of the optical cable relative to the unmanned aerial vehicle is calculated according to the initial position of the optical cable and the motion trajectory of the unmanned aerial vehicle. The specific operation is to convert the motion state data of the unmanned aerial vehicle into the relative motion data of the optical cable through geometric relationship and kinematic equation. For example, in a flight task, the swing displacement data of the optical cable at different time points is obtained by calculation and saved as a new data file. The MATLAB software is used to extract the frequency domain features of the optical cable swing relative displacement data. In MATLAB, the optical cable swing displacement data file is loaded, and the fast Fourier transform (FFT) function is used to convert it from time domain to frequency domain. By analyzing the frequency spectrum, the frequency spectrum characteristic parameters of the optical cable swing are extracted, such as the main frequency and amplitude. For example, in a flight task, the frequency spectrum of the optical cable swing is obtained by FFT analysis, and it is found that the main frequency is 2Hz and the amplitude is 0.05 meters. According to these frequency spectrum characteristic parameters, a communication optical cable motion prediction model is constructed.

[0120] Preferably, the step S2 of dynamically tracking and geometrically filtering the target segment of the communication optical cable based on the communication optical cable motion prediction model comprises the following steps:

[0121] Step S291: Harris corner point detection is performed on the enhanced communication optical cable image to obtain a set of optical cable image feature corner points; feature point state initialization is performed based on the set of optical cable image feature corner points to obtain initial data of the optical cable feature point state;

[0122] Step S292: motion prediction is performed on the initial data of the optical cable feature point state according to the communication optical cable motion prediction model to obtain a predicted position of the optical cable feature point;

[0123] Step S293: continuous frame images of the target segment of the communication optical cable are acquired to obtain communication optical cable continuous frame images, and the communication optical cable continuous frame images are enhanced in the gradient domain to obtain enhanced communication optical cable continuous frame images;

[0124] Step S294: optical cable feature point observation update is performed on the enhanced communication optical cable continuous frame images to obtain an observed position of the optical cable feature point;

[0125] Step S295: the predicted position of the optical cable feature point and the observed position of the optical cable feature point are fused to obtain a stable tracking coordinate of the optical cable;

[0126] Step S296: Normalized correlation template matching is performed on the optical cable stable tracking coordinates to obtain a set of optical cable detection region candidate position sets;

[0127] Step S297: Geometric feature analysis is performed on the set of optical cable detection region candidate position sets to obtain optical cable detection region geometric feature data, and a set of optical cable geometric constraint conditions is constructed based on the optical cable detection region geometric feature data.

[0128] Step S298: The set of optical cable detection region candidate position sets is screened and verified based on the set of optical cable geometric constraint conditions to obtain optical cable detection region positioning data.

[0129] In this embodiment, the enhanced communication cable image is imported into Open CV, and the Harris corner detection function is called. The sensitivity parameter (e.g., 0.04) and window size (e.g., 3x3 pixels) for corner detection are set. The tool automatically detects the corner points of the cable in the image and generates a set of feature corner points for the cable image. Based on these corner point sets, the feature point state initialization is performed, i.e., each corner point is assigned initial position, velocity, and acceleration state parameters, and the initial data of the cable feature point state is obtained. For example, in a cable image, multiple corner points are detected, which are located at the edges or texture changes of the cable. In MATLAB, the initial data of the cable feature point state is loaded, and the predicted position of each feature point at the next time is calculated based on the pre-constructed communication cable motion prediction model (based on frequency domain feature parameters). For example, assuming that the model predicts that the cable will oscillate at a certain frequency and amplitude in a strong wind environment, based on these parameters, the approximate position of each feature point in the next frame image can be predicted. During the flight of the unmanned aerial vehicle, a high-definition camera of the DJI unmanned aerial vehicle is used to continuously capture frame images of the target segment of the communication cable. The camera takes pictures of the cable at a frequency of 30 frames per second, obtaining continuous frame images of the communication cable. Open CV software is used to enhance the gradient domain of these continuous frame images. The specific operation is to enhance the edge and texture information of the cable in the image by calculating the gradient amplitude of the image. For example, the horizontal and vertical gradients of the image are calculated by the Sobel operator, and the gradient amplitude map is fused with the original image by weighting, obtaining the enhanced continuous frame image of the communication cable. Open CV software is used to update the observation of the cable feature points in the enhanced continuous frame image of the communication cable. The specific operation is to search for the actual feature point position in the enhanced continuous frame image by taking the predicted feature point position as the initial search point. For example, the template matching method is used to search for the best matching point within a certain range centered on the predicted position, thereby updating the observed position of the cable feature points. MATLAB software is used to fuse the predicted position and observed position of the cable feature points. In MATLAB, the predicted position and observed position data are loaded, and a Kalman filter is used for data fusion. The Kalman filter combines the predicted value and the observed value to calculate the stable tracking coordinates of the cable feature points. For example, the initial state and noise covariance matrix of the Kalman filter are set, and the state of the filter is updated iteratively to obtain the optimal estimated position of each feature point. Open CV software is used to perform normalized correlation template matching on the stable tracking coordinates of the cable. The specific operation is to take the image region around the stable tracking coordinates as a template and perform normalized correlation matching in the enhanced continuous frame image. For example, the template size is set to 15x15 pixels, and the matching threshold is set to 0.8. By calculating the normalized correlation coefficient between the template and the image region, the most similar region to the template is found, thereby obtaining a set of candidate positions for the cable detection region.In MATLAB, load the candidate position set, calculate the geometric features of each candidate region, such as area, perimeter, and aspect ratio. For example, extract the contour of the candidate region by the contour detection algorithm, and calculate its geometric feature data. Based on these geometric feature data, construct a set of geometric constraints for the optical cable, such as area range (e.g. 100-500 pixels2) and aspect ratio (e.g. 1.5-3). Compare the geometric features of each candidate region with the constraints, and the regions that meet the conditions are retained, and the regions that do not meet the conditions are eliminated. For example, if a candidate region has an area of 300 pixels2 and an aspect ratio of 2.5, it meets the preset constraints and is confirmed as an optical cable detection region; otherwise, it is determined as an interference region and is eliminated. Through this screening and verification, the accurate positioning data of the optical cable detection region is obtained.

[0130] Preferably, the surface morphology feature extraction of the enhanced communication optical cable image in step S3 includes the following steps:

[0131] Step S31: Construct a texture extraction filter according to the enhanced communication optical cable image to obtain an optical cable surface texture extraction filter;

[0132] Step S32: Perform convolution operation on the enhanced communication optical cable image using the optical cable surface texture extraction filter to obtain optical cable multi-scale texture response data;

[0133] Step S33: Perform amplitude normalization on the optical cable multi-scale texture response data to obtain standard optical cable texture feature data;

[0134] Step S34: Perform multi-directional texture energy evaluation based on the standard optical cable texture feature data to obtain optical cable surface texture energy distribution data;

[0135] Step S35: Perform adaptive threshold segmentation on the optical cable surface texture energy distribution data to obtain optical cable texture salient region data;

[0136] Step S36: Construct a morphological structure element according to the optical cable texture salient region data to obtain an optical cable special morphological operator;

[0137] Step S37: Perform Top-Hat transformation on the enhanced communication optical cable image using the optical cable special morphological operator to obtain optical cable surface recess feature data.

[0138] In this embodiment, the image processing toolbox in MATLAB is used, and a Gabor filter is selected as the texture extraction tool. The Gabor filter can effectively extract the texture features in the image, and its parameters include direction, frequency and bandwidth. For example, the direction of the Gabor filter is set to 0°, 45°, 90° and 135°, the frequency is set to 0.1 and 0.2 (unit: reciprocal of pixels), and the bandwidth is set to 1.5. Through these parameter settings, a set of Gabor texture extraction filters with multiple directions and frequencies is constructed. The convolution function in MATLAB software is used to apply the Gabor texture extraction filter to the enhanced communication cable image. The specific operation is to perform convolution operation between each Gabor filter and the enhanced image to obtain cable multi-scale texture response data under different directions and frequencies. For example, for a Gabor filter with a direction of 0° and a frequency of 0.1, a response image will be obtained after convolution operation, in which the highlighted part represents the texture features under the direction and frequency. Through the convolution of multiple filters, texture response data can be obtained. The pixel value of each response image is normalized to the range of 0 to 1. For example, for a response image, the maximum value and the minimum value are calculated, and the formula: The pixel values are normalized. The normalized data is called standard cable texture feature data. The normalized standard cable texture feature data is evaluated using multi-directional texture energy estimation with MATLAB software. The specific operation is to calculate the texture energy in each direction, that is, to perform a sum of squares operation on the normalized response image of each direction. For example, for the response image with a direction of 0°, the sum of squares of all pixel values is calculated to obtain the texture energy of this direction. By evaluating the texture energy of multiple directions, cable surface texture energy distribution data can be obtained. The cable surface texture energy distribution data is segmented using an adaptive threshold with MATLAB software. The specific operation is to automatically calculate the best threshold using the Otsu method to segment the texture energy distribution data into significant and non-significant regions. For example, the Otsu method will find a threshold according to the histogram of the texture energy data, so that the inter-class variance of the two regions after segmentation is maximized. Through adaptive threshold segmentation, cable texture significant region data can be obtained, which contains the texture details of the cable surface, such as recesses and crack features. The MATLAB software is used to construct morphological structure elements according to the cable texture significant region data. The specific operation is to select appropriate morphological structure elements, such as circular or rectangular structure elements. For example, a circular structure element with a radius of 3 pixels is selected, and the shape and size of the structure element can be adjusted according to the size of the cable surface features. By constructing a dedicated morphological operator, the texture features of the cable surface can be better processed. The Top-Hat transformation of the enhanced communication cable image is performed using the morphological operation function in MATLAB software and the dedicated morphological operator for the cable. First, the opening operation (erosion followed by dilation) is performed on the image, and then the original image is subtracted from the image after the opening operation to obtain the image after the Top-Hat transformation. For example, the Top-Hat transformation is performed using a circular structure element with a radius of 3 pixels, which can highlight the recess features of the cable surface.

[0139] Preferably, the step S38 of locating the fault candidate region based on the cable surface recess feature data comprises the following steps:

[0140] Step S381 : morphological reconstruction is performed on the cable surface recess feature data to obtain cable surface abnormal morphology data; edge detection is performed based on the cable surface abnormal morphology data to obtain cable abnormal edge contour data;

[0141] Step S382: logical AND operation is performed on the cable abnormal edge contour data and the cable surface recess feature data to obtain a cable surface abnormal feature map;

[0142] Step S383: connected domain labeling is performed on the cable surface abnormal feature map to obtain cable abnormal connected domain labels;

[0143] Step S384: Geometric feature parameter extraction is performed based on the optical cable abnormal connection domain label to obtain optical cable abnormal region geometric feature data;

[0144] Step S385: Multi-condition screening filtering is performed on the optical cable abnormal region geometric feature data according to a preset screening condition to obtain optical cable suspected fault region data;

[0145] Step S386: Spatial position matching is performed on the optical cable suspected fault region data according to the optical cable detection region positioning data to obtain optical cable fault candidate region coordinates;

[0146] Step S387: Boundary extraction is performed based on the optical cable fault candidate region coordinates to obtain optical cable fault candidate region boundary data; and the optical cable fault candidate region boundary data is subjected to a centroid coordinate calculation to obtain positioning optical cable fault candidate region data.

[0147] In this embodiment, MATLAB software is used to reconstruct the morphological features of the cable surface recess. The specific operation is to use the morphological operation tool of MATLAB to select appropriate structural elements (for example, a square structural element of 3x3) to perform a closing operation (first expansion and then corrosion) to fill the small holes and broken parts in the recess features, and obtain the abnormal morphological data of the cable surface. Based on the reconstructed abnormal morphological data, the edge detection function of MATLAB (such as Canny edge detector) is used for edge detection. The low threshold of Canny detector is set to 0.05 and the high threshold is set to 0.15, so as to obtain the abnormal edge contour data of the cable. MATLAB software is used to perform logical AND operation on the cable abnormal edge contour data and the cable surface recess feature data. The specific operation is to perform pixel-by-pixel logical AND operation on the edge contour data (binary image) and the recess feature data (gray image). Through the above operation, the area that meets the edge condition and the recess feature at the same time can be extracted, and the abnormal feature map of the cable surface is obtained. The highlighted area in the abnormal feature map is the potential fault area. MATLAB software is used to label the connected domain of the cable surface abnormal feature map. The specific operation is to call the bwlabel function of MATLAB to label the connected regions in the abnormal feature map, and assign a unique label to each connected domain. For example, set the connectivity to 8 (representing 8 directions of connectivity), so as to obtain the cable abnormal connected domain label. Each label represents an independent abnormal area. MATLAB software is used to extract the geometric feature parameters based on the cable abnormal connected domain label. The specific operation is to call the regionprops function of MATLAB to extract the geometric feature parameters of each connected domain, such as area, perimeter, aspect ratio, circularity, etc. For example, calculate the area (unit: pixel2) and perimeter (unit: pixel) of each connected domain, and calculate the aspect ratio and circularity. MATLAB software is used to perform multi-condition screening and filtering on the cable abnormal area geometric feature data according to the preset screening conditions. The specific operation is to set the screening conditions, for example, the area is greater than 100 pixels2 and less than 1000 pixels2, the aspect ratio is between 1.5 and 3, and the circularity is less than 0.8. Through these conditions, the area that meets the cable fault feature is screened out, and the interference of other non-fault areas is excluded, and the cable suspected fault area data is obtained. MATLAB software is used to match the spatial position of the cable suspected fault area data based on the cable detection area positioning data. The specific operation is to compare the geometric center coordinates of the suspected fault area with the positioning data of the cable detection area to determine whether the suspected fault area is located within the cable detection area. For example, set a tolerance range (such as 5 pixels), if the distance between the geometric center of the suspected fault area and the center of the cable detection area is within the tolerance range, it is considered that the area is a cable fault candidate area, and the cable fault candidate area coordinates are obtained. MATLAB software is used to extract the boundary based on the cable fault candidate area coordinates.The specific operation is to call the edge detection function of MATLAB (such as Canny edge detector) to perform boundary extraction on the fault candidate area to obtain the boundary data of the cable fault candidate area. The regionprops function of MATLAB is used to calculate the centroid coordinates of each fault candidate area. For example, the formula for calculating the centroid coordinates is: where (x i ,y i ) is the pixel coordinates on the boundary, and N is the total number of pixels on the boundary. Through boundary extraction and centroid coordinate calculation, the positioning cable fault candidate area data is obtained.

[0148] Preferably, step S4 comprises the following steps:

[0149] Step S41: performing time series window division on the positioning cable fault candidate area data to obtain a cable fault area time series window;

[0150] Step S42: performing region centroid trajectory extraction based on the cable fault area time series window to obtain cable fault area centroid trajectory data;

[0151] Step S43: performing trajectory smoothing filtering on the cable fault area centroid trajectory data to obtain smoothed cable fault area trajectory data;

[0152] Step S44: performing trajectory displacement variance calculation based on the smoothed cable fault area trajectory data to obtain a cable fault area displacement stability parameter;

[0153] Step S45: performing morphological feature time series change monitoring on the positioning cable fault candidate area data to obtain cable fault area morphological change data;

[0154] Step S46: performing morphological stability evaluation based on the cable fault area morphological change data to obtain a cable fault area morphological stability parameter;

[0155] Step S47: performing comprehensive stability evaluation based on the cable fault area displacement stability parameter and the cable fault area morphological stability parameter to obtain cable fault area time series stability data;

[0156] Step S48: obtaining communication cable motion characteristic data; generating cable motion constraint condition parameters based on the communication cable motion characteristic data; performing cable fault authenticity evaluation based on the cable motion constraint condition parameters and the cable fault area time series stability data to obtain verified cable fault feature data.

[0157] In this embodiment, MATLAB software is used to divide the time series window of the candidate area data for locating the optical cable fault. The specific operation is to divide the time series data of the candidate area into multiple fixed length time windows. For example, set the length of each time window to 10 frames of images, and the step size to 2 frames of images. In this way, for a time series data containing 100 frames of images, it can be divided into 45 time windows. Through the above operation, the entire time series data is decomposed into multiple local time series. MATLAB software is used to extract the centroid trajectory of the optical cable fault area in each time series window. The specific operation is to calculate the centroid coordinates of the fault area in each window, and connect these centroid coordinates in time order to form the centroid trajectory. For example, for the fault area in each time window, use the regionprops function of MATLAB to calculate the centroid coordinates, arrange these coordinate points in time order to obtain the centroid trajectory data of the optical cable fault area. The centroid trajectory can reflect the motion trend of the fault area in the time series. MATLAB software is used to perform smoothing filter processing on the centroid trajectory data of the optical cable fault area. The specific operation is to use a moving average filter to smooth the centroid trajectory. For example, set the window length of the moving average filter to 5 data points, that is, the value of each data point is replaced by the average value of the previous and next 2 data points. Through the above operation, the noise and jitter in the centroid trajectory can be reduced, and the smooth optical cable fault area trajectory data is obtained. MATLAB software is used to calculate the trajectory displacement variance of the smoothed optical cable fault area trajectory data. The specific operation is to calculate the displacement between adjacent centroid points on the trajectory, and calculate the variance of these displacements. For example, for each pair of adjacent centroid points on the smoothed trajectory, calculate the Euclidean distance between them to obtain a series of displacement values, calculate the variance of these displacement values, and obtain the displacement stability parameter of the optical cable fault area. The smaller the displacement variance, the more stable the motion of the fault area. MATLAB software is used to monitor the morphological feature time series change of the candidate area data for locating the optical cable fault. The specific operation is to calculate the geometric feature parameters (such as area, perimeter, aspect ratio, etc.) of the fault area in each time window, and observe the changes of these parameters over time. For example, for the fault area in each time window, use the regionprops function of MATLAB to calculate its area and aspect ratio, and record the changes of these parameters in different time windows. Through the above operation, the morphological change data of the optical cable fault area is obtained, reflecting the morphological evolution of the fault area in the time series. MATLAB software is used to evaluate the morphological stability of the optical cable fault area morphological change data. The specific operation is to calculate the change rate of the morphological feature parameters, and judge the morphological stability of the fault area according to the change rate. For example, calculate the area change rate and aspect ratio change rate, and set the change rate threshold to 10% and 5% respectively. If the area change rate in a certain time window is less than 10%, and the aspect ratio change rate is less than 5%, it is considered that the morphological stability of the fault area in this time window is stable.Through the above operation, the morphology stability parameter of the optical cable fault area is obtained. The displacement stability parameter and the morphology stability parameter of the optical cable fault area are comprehensively evaluated using MATLAB software. The specific operation is to weight and sum the displacement stability parameter and the morphology stability parameter to obtain a comprehensive stability score. For example, the weight of the displacement stability parameter is set to 0.6, and the weight of the morphology stability parameter is set to 0.4. According to the comprehensive stability score, the fault area is divided into three levels of stable, relatively stable and unstable. Through the above operation, the time sequence stability data of the optical cable fault area is obtained. For the detailed implementation process of step S48, please refer to the sub-steps of step S48.

[0158] Preferably, step S48 comprises the following steps:

[0159] Step S481: Obtain enhanced communication optical cable continuous frame images; perform optical flow field feature extraction on the enhanced communication optical cable continuous frame images to obtain optical cable image optical flow field data;

[0160] Step S482: Extract local optical flow vectors from the optical cable image optical flow field data based on the positioning optical cable fault candidate area data to obtain optical cable fault area optical flow vectors;

[0161] Step S483: Perform vector consistency evaluation on the optical cable fault area optical flow vectors to obtain optical cable fault area optical flow consistency parameters;

[0162] Step S484: Collect motion characteristic features of the communication optical cable to obtain communication optical cable motion characteristic data, and construct motion constraint conditions based on the communication optical cable motion characteristic data to obtain optical cable motion constraint condition parameters;

[0163] Step S485: Based on the optical cable motion constraint condition parameters, the optical cable fault area optical flow consistency parameters are verified to obtain constraint verification optical cable fault optical flow data;

[0164] Step S486: Temporally and spatially correlate the optical cable fault area time sequence stability data and the constraint verification optical cable fault optical flow data to obtain optical cable fault spatiotemporal correlation feature data;

[0165] Step S487: Based on the optical cable fault spatiotemporal correlation feature data, a fault authenticity score is obtained.

[0166] Step S488: Threshold value judgment screening is performed on the optical cable fault authenticity score to obtain screened optical cable fault area data; and based on the screened optical cable fault area data, fault feature parameter statistics are performed to obtain verification optical cable fault feature data.

[0167] In this embodiment, the Open CV software is used to extract the optical flow field features of the enhanced communication cable continuous frame images. The specific operation is to load the continuous frame image sequence and call the optical flow algorithm in Open CV (such as Lucas-Kanade optical flow algorithm). The parameters of the optical flow algorithm are set, for example, the window size is 15×15 pixels, the maximum number of iterations is 10, and the error threshold is 0.03. Through these parameters, the algorithm can calculate the motion vector of each pixel point in each frame image, so as to obtain the cable image optical flow field data. These data reflect the motion direction and speed of the cable between consecutive frames. The local optical flow vector of the cable fault area is extracted from the optical flow field data using MATLAB software. The specific operation is to determine the location of the fault area in each frame image according to the positioning of the cable fault candidate area data, and extract the optical flow vector corresponding to these areas from the optical flow field data. For example, if the location of the fault area in a certain frame image is a rectangular area, all the optical flow vectors in the rectangular area are extracted to obtain the optical flow vector of the cable fault area. These local optical flow vectors can reflect the motion characteristics of the fault area. The vector consistency of the optical flow vector of the cable fault area is evaluated using MATLAB software. The specific operation is to calculate the direction consistency of all optical flow vectors in the fault area. For example, the cosine value of the vector angle is used as the consistency index, the cosine value of the angle between each optical flow vector and other vectors is calculated, and the average value is taken. If the average cosine value is greater than 0.8, it means that the direction of the optical flow vector in the fault area is relatively consistent, and the optical flow consistency parameter of the cable fault area is obtained. This parameter reflects the uniformity of the motion of the fault area. The motion characteristic feature of the communication cable is collected using MATLAB software. The specific operation is to analyze the motion trajectory, speed and acceleration of the cable in the continuous frame image. For example, the motion trajectory of the cable centroid and the speed and acceleration change on the trajectory are calculated. According to these motion characteristic data, the motion constraint condition parameters of the cable are constructed. For example, the maximum swing angle of the cable is set to 10 degrees, and the maximum speed is 5 pixels / frame, which are used as the motion constraint conditions. The constraint verification of the optical flow consistency parameter of the cable fault area is performed using MATLAB software. The specific operation is to compare the optical flow consistency parameter with the motion constraint condition parameter of the cable. For example, if the optical flow consistency parameter indicates that the motion direction of the fault area is consistent, but its speed exceeds the maximum speed in the motion constraint condition of the cable, then the optical flow data of this fault area will be marked as not meeting the constraint condition. Through the above operation, the constraint verification cable fault optical flow data is obtained, which excludes the false judgment that does not meet the motion characteristics of the cable. The spatiotemporal correlation of the timing stability data of the cable fault area and the constraint verification cable fault optical flow data is performed using MATLAB software. The specific operation is to combine the timing stability data (such as displacement stability parameter) with the optical flow data (such as speed and direction) to analyze the correlation of the fault area in time and space.For example, if a fault region shows stable displacement in time series, and its optical flow direction and speed also meet the constraint conditions, the spatiotemporal correlation feature data of the region will indicate that it is a high-confidence fault region. The MATLAB software is used to score the fault authenticity based on the spatiotemporal correlation feature data of the optical cable fault. The specific operation is to calculate the authenticity score of the fault region according to the spatiotemporal correlation feature data. For example, the scoring standard is set as: high time series stability (weight 0.5), high optical flow direction consistency (weight 0.3), and speed meeting the constraint (weight 0.2). According to these standards, the comprehensive score of each fault region is calculated to obtain the optical cable fault authenticity score. The higher the score, the higher the authenticity of the fault region. The MATLAB software is used to perform threshold judgment and screening on the optical cable fault authenticity score. The specific operation is to set a threshold, for example, 0.7, and retain the fault regions with a score higher than the threshold and eliminate the regions with a score lower than the threshold. Through the above operation, the screened optical cable fault region data is obtained. The fault feature parameters of the screened fault region are counted, for example, the area, shape, position, and other parameters of the fault region are counted to obtain the verification optical cable fault feature data.

[0168] Preferably, step S5 comprises the following steps:

[0169] Step S51: extracting fault type features from the verification optical cable fault feature data to obtain an optical cable fault type feature vector;

[0170] Step S52: constructing a fault classification rule set based on the optical cable fault type feature vector to obtain an optical cable fault classification discrimination rule;

[0171] Step S53: identifying the fault type of the verification optical cable fault feature data according to the optical cable fault classification discrimination rule to obtain optical cable fault type identification data;

[0172] Step S54: quantifying the defect geometric size according to the optical cable fault type identification data to obtain optical cable fault geometric size parameters;

[0173] Step S55: evaluating the defect risk degree based on the optical cable fault geometric size parameters to obtain the optical cable fault severity level;

[0174] Step S56: obtaining the UAV GPS coordinates and the camera internal parameter table;

[0175] Step S57: mapping the verification optical cable fault feature data according to the UAV GPS coordinates and the camera internal parameter table to obtain the communication optical cable fault geographic coordinates; generating a structured optical cable fault detection report according to the optical cable fault severity level and the communication optical cable fault geographic coordinates.

[0176] In this embodiment, MATLAB software is used to extract fault type features from the validation optical cable fault feature data. The specific operation is to call the image processing toolbox of MATLAB to extract the texture features, shape features and gray level features of the fault area. For example, use the gray level co-occurrence matrix (GLCM) to extract texture features, calculate contrast, correlation, energy and homogeneity parameters; use the regionprops function to extract shape features such as area, perimeter, aspect ratio and circularity; and at the same time, count the mean and standard deviation of the gray level of the fault area. Combine these features into a feature vector to obtain the optical cable fault type feature vector. Use MATLAB software to construct the fault classification rule set based on the optical cable fault type feature vector. The specific operation is to use the labeled fault sample data to train the classification model through machine learning algorithms (such as support vector machine SVM or decision tree). For example, use the fitcsvm function of MATLAB to train the SVM classifier, use the fault type feature vector as the input, and use the fault type label (such as "crack", "indentation", "break") as the output. Through the trained classification model, the optical cable fault classification discrimination rule can be obtained. These rules define the decision boundary of different fault types in the feature space. Use MATLAB software to identify the fault type of the validation optical cable fault feature data according to the optical cable fault classification discrimination rule. The specific operation is to input the validation optical cable fault feature data into the trained classification model, and the model will output the fault type identification data according to the feature vector. For example, for a fault area, its feature vector is input into the SVM classifier, and the classifier will output the fault type as "crack". Through the above operation, the type of each fault area can be accurately identified, and the optical cable fault type identification data can be obtained. Use MATLAB software to quantify the defect geometric dimensions according to the optical cable fault type identification data. The specific operation is to call the regionprops function to extract the geometric dimension parameters of different types of fault areas. For example, for the crack type fault area, extract its length and width; for the indentation type fault area, extract its depth and area. These geometric dimension parameters can quantitatively describe the severity of the fault. The detailed implementation process of step S55 can be referred to the sub-steps of step S55. In the process of unmanned aerial vehicle inspection, use the matching software (such as DJI Pilot or DJI Fly) of DJI unmanned aerial vehicle to obtain the GPS coordinates of the unmanned aerial vehicle. The specific operation is to record the real-time GPS position data of the unmanned aerial vehicle in the flight log of the software, including longitude, latitude and height. At the same time, use the camera calibration tool (such as the camera calibration module of Open CV) to calibrate the internal parameters of the camera carried by the unmanned aerial vehicle. The specific operation is to shoot a series of images of the calibration board with known geometric shapes, and calculate the internal parameters of the camera through the cv2.calibrateCamera function of Open CV, including focal length, principal point coordinates and distortion coefficient. Save these internal parameter data as camera internal parameter table.The detailed implementation procedure of step S57 can refer to the sub-steps of step S57.

[0177] Especially important is that step S55 further comprises the following steps:

[0178] Step S551: Based on the optical cable fault geometric size parameter, the defect danger degree is evaluated to obtain optical cable fault danger level preliminary evaluation data;

[0179] Step S552: According to the optical cable fault type identification data and the optical cable fault danger level preliminary evaluation data, the fault position importance weight is distributed to obtain the optical cable fault position importance weight;

[0180] Step S553: Based on the preset fuzzy logic membership function, the optical cable fault geometric size parameter is fuzzed to obtain the optical cable fault fuzzy membership data;

[0181] Step S554: According to the optical cable fault fuzzy membership data and the optical cable fault position importance weight, the fuzzy reasoning operation is performed to obtain the optical cable fault evaluation fuzzy result;

[0182] Step S555: The optical cable fault evaluation fuzzy result is de-fuzzed to obtain the optical cable fault severity level.

[0183] In this embodiment, the MATLAB software is used to evaluate the defect risk level of the geometric size parameters of the optical cable fault. The specific operation is to score the geometric size parameters of each fault area according to the preset evaluation standard. For example, for the fault area of crack type, set the crack length exceeding 10 centimeters as high risk (score 10 points), 5-10 centimeters as medium risk (score 5 points), and less than 5 centimeters as low risk (score 1 point). For the fault area of indentation type, set the indentation depth exceeding 2 millimeters as high risk (score 10 points), 1-2 millimeters as medium risk (score 5 points), and less than 1 millimeter as low risk (score 1 point). According to these standards, the initial evaluation data of the risk level of each fault area is calculated. The MATLAB software is used to allocate the importance weight of the fault position according to the optical cable fault type identification data and the initial evaluation data of the optical cable fault risk level. The specific operation is to allocate an importance weight of the position for each fault area according to the fault type and the initial evaluation data. For example, for the high-risk crack fault, the weight is allocated as 0.8; for the medium-risk indentation fault, the weight is allocated as 0.5; and for the low-risk other fault, the weight is allocated as 0.2. The MATLAB software is used to fuzz the geometric size parameters of the optical cable fault. The specific operation is to convert the geometric size parameters into fuzzy membership data based on the preset fuzzy logic membership function. For example, for the crack length, set the membership function as follows: the membership function of short crack (0-5 centimeters) is trapezoidal function, the membership function of medium crack (5-10 centimeters) is triangular function, and the membership function of long crack (more than 10 centimeters) is trapezoidal function. Through these membership functions, the specific geometric size parameters are converted into fuzzy membership data, for example, a crack with a length of 7 centimeters has a membership of 0.6 for medium crack and 0.4 for long crack. The MATLAB software is used for fuzzy inference operation. The specific operation is to perform fuzzy inference according to the fuzzy membership data of the optical cable fault and the importance weight of the fault position. For example, use the fuzzy logic toolbox of MATLAB to define fuzzy rules, such as "if the crack length is medium and the position importance is high, then the risk level is high". Through these rules, combined with the fuzzy membership data and the weight, the inference operation is performed to obtain the fuzzy result of the optical cable fault evaluation. The MATLAB software is used to de-fuzz the fuzzy result of the optical cable fault evaluation. The specific operation is to use de-fuzzing methods (such as center of gravity method or maximum membership method) to convert the fuzzy result into specific severity level. For example, use the center of gravity method to calculate the center of gravity position of the fuzzy result, and map it to the specific severity level, such as 1-3 for low risk, 4-6 for medium risk, and 7-10 for high risk. Through de-fuzzing, the severity level of the optical cable fault is obtained.

[0184] It is particularly important that step S57 further comprises the following steps:

[0185] Step S571: Obtain GPS positioning original data in the flight process of the unmanned aerial vehicle, to obtain the GPS coordinates of the unmanned aerial vehicle;

[0186] Step S572: Obtain the internal parameter calibration data of the camera carried by the unmanned aerial vehicle, to obtain the internal parameter table of the carried camera;

[0187] Step S573: Construct a geographic coordinate mapping model based on the GPS coordinates of the unmanned aerial vehicle and the internal parameter table of the carried camera, to obtain a cable fault geographic coordinate conversion model;

[0188] Step S574: According to the cable fault geographic coordinate conversion model, convert the pixel coordinates in the verification cable fault feature data into geographic coordinates, to obtain the communication cable fault geographic coordinates;

[0189] Step S575: Based on the communication cable fault geographic coordinates, perform spatial clustering to obtain cable fault spatial distribution clustering data;

[0190] Step S576: Generate a cable fault maintenance priority according to the cable fault severity level and the cable fault type identification data;

[0191] Step S577: Based on the cable fault maintenance priority, the communication cable fault geographic coordinates, and the cable fault spatial distribution clustering data, organize the report structure, to obtain a structured cable fault detection report.

[0192] In this embodiment, during the UAV inspection process, the original GPS positioning data during the UAV flight is obtained using the matching software of DJI UAV (such as DJI Pilot or DJI Fly). The specific operation is to record the real-time GPS position data of the UAV in the flight log of the software, including longitude, latitude and height. For example, the UAV records the GPS coordinates once every second during the inspection process, generating a log file containing the timestamp and GPS coordinates. The camera mounted on the UAV is calibrated using Open CV software. The specific operation is to shoot a series of images of a calibration board of known geometric shape, and calculate the camera's intrinsic parameters, including focal length, principal point coordinates and distortion coefficients, through the cv2.calibrateCamera function of Open CV. For example, a checkerboard calibration board is used, and multiple images at different angles are shot, and these images are imported into Open CV for calibration. After calibration, the camera's intrinsic parameter table is obtained. MATLAB software is used to construct a geographic coordinate mapping model based on the UAV GPS coordinates and the camera intrinsic parameter table. The specific operation is to combine the UAV's GPS coordinates with the camera's intrinsic parameters, and convert the pixel coordinates in the image into geographic coordinates through the projection transformation formula. For example, the fitgeotrans function of MATLAB is used to input the UAV's GPS coordinates and the camera's intrinsic parameters, and a mapping model from pixel coordinates to geographic coordinates is constructed. This model can convert the location of the fault features in the image into actual geographic coordinates.

[0193] The pixel coordinates in the verification optical cable fault feature data are converted into geographic coordinates using the MATLAB software according to the geographic coordinate mapping model. The specific operation is to input the pixel coordinates of the fault feature into the mapping model to obtain the corresponding geographic coordinates. For example, for a detected fault area, its pixel coordinates are (100, 200), and after conversion by the mapping model, its geographic coordinates are (longitude: 116.3974, latitude: 39.9092). The MATLAB software is used for spatial clustering of the communication optical cable fault geographic coordinates. The specific operation is to use the K-means clustering algorithm to cluster the fault points according to the spatial distribution. For example, set the number of cluster centers to 5, and divide all the fault points into 5 clusters. Through clustering analysis, the spatial distribution clustering data of the optical cable fault can be obtained, and the concentrated area of the fault points can be understood. The MATLAB software is used to generate the optical cable fault maintenance priority according to the optical cable fault severity level and optical cable fault type identification data. The specific operation is to assign priorities according to the severity and type of the fault. For example, the priority of high-risk crack fault is 1 (highest), the priority of medium-risk indentation fault is 2, and the priority of low-risk other faults is 3. Through the above operation, it can be quickly determined which faults need to be handled first. The MATLAB software is used to organize the report structure based on the optical cable fault maintenance priority, the communication optical cable fault geographic coordinates and the optical cable fault spatial distribution clustering data. The specific operation is to integrate all the data into a structured report, including fault location, type, severity, priority and spatial distribution, etc. For example, a table is generated to list the detailed information of each fault point, including geographic coordinates, fault type, severity level, maintenance priority and cluster area. The finally generated structured optical cable fault detection report can help the operation and maintenance personnel quickly understand the fault situation and develop a maintenance plan.

[0194] Preferably, the present application also provides a machine vision-based communication optical cable fault identification system for performing the machine vision-based communication optical cable fault identification method as described above, which comprises:

[0195] An image preprocessing module is configured to acquire original visible light cable images and original near-infrared cable images, perform dual-spectrum registration and pixel-level fusion on the original visible light cable images and the original near-infrared cable images to obtain registered communication cable images, and perform local contrast enhancement on the registered communication cable images to obtain enhanced communication cable images.

[0196] A dynamic tracking module is configured to construct a communication cable motion prediction model and perform dynamic tracking and geometric constraint filtering on a target segment of the communication cable based on the communication cable motion prediction model to obtain optical cable detection area positioning data.

[0197] The feature extraction module is configured to perform surface morphology feature extraction on the enhanced communication cable image to obtain cable surface depression feature data, and perform fault candidate region positioning based on the cable surface depression feature data to obtain positioning cable fault candidate region data.

[0198] The fault verification module is configured to perform continuous multi-frame tracking verification on the positioning cable fault candidate region data to obtain cable fault region time sequence stability data, acquire communication cable motion characteristic data, generate cable motion constraint condition parameters based on the communication cable motion characteristic data, and perform cable fault authenticity evaluation according to the cable motion constraint condition parameters and the cable fault region time sequence stability data to obtain verification cable fault feature data.

[0199] The evaluation and reporting module is configured to perform fault severity evaluation according to the verification cable fault feature data to obtain a cable fault severity level, and generate a structured cable fault detection report based on the cable fault severity level.

[0200] Therefore, from any point of view, the embodiments should be regarded as exemplary and non-limiting, the scope of the application being defined by the appended claims and not by the above description, and it is intended to encompass all variations falling within the meaning and the scope of the equivalent elements of the claims.

[0201] The above description is merely illustrative of the application and is not intended to limit the scope of the application. Various modifications of the embodiments described herein will be apparent to those with skill in the art, and it is intended to use all equivalents falling within the spirit and scope of the application. Therefore, the present application is not to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for identifying communication optical cable faults based on machine vision, characterized in that: The following steps are involved: Step S1: obtaining an original visible light cable image and an original near-infrared cable image; Perform dual-spectral registration and pixel-level fusion on the original visible light cable image and the original near-infrared cable image to obtain a registered communication cable image; perform local contrast enhancement on the registered communication cable image to obtain an enhanced communication cable image; Step S2: constructing a communication optical cable motion prediction model; dynamically tracking and geometrically constraining the target section of the communication optical cable based on the communication optical cable motion prediction model to obtain positioning data of the optical cable detection area; Step S3: extracting surface morphological features of the enhanced communication optical cable image to obtain surface depression feature data of the optical cable; Positioning the candidate fault area based on the surface depression feature data of the optical cable to obtain the candidate fault area data of the positioned optical cable; Step S4: Conduct continuous multi-frame tracking verification on the data of the candidate area of ​​the located optical cable fault to obtain the temporal stability data of the optical cable fault area; Acquire communication optical cable motion characteristic data; generate optical cable motion constraint parameters based on the communication optical cable motion characteristic data; evaluate the authenticity of the optical cable fault based on the optical cable motion constraint parameters and the temporal stability data of the optical cable fault area to obtain verified optical cable fault characteristic data; Step S5: performing a fault severity assessment based on the verified optical cable fault characteristic data to obtain an optical cable fault severity level; Generates structured optical cable fault detection reports based on the severity level of the optical cable fault.

2. The method for identifying communication optical cable faults based on machine vision according to claim 1, characterized in that: Step S1 includes the following steps: Step S11: performing visible light image acquisition on the target section of the communication optical cable to obtain an original visible light cable image, wherein a CMOS sensor with a resolution of 4096×3072 pixels is used for image acquisition, and the exposure time is set to 1 / 500 second and the ISO value is 100; Step S12: performing near-infrared image acquisition on the target section of the communication optical cable to obtain an original near-infrared optical cable image, wherein a near-infrared CMOS sensor with a wavelength range of 780-1000 nm is used for image acquisition, the gain is set to 2 times, the exposure time is set to 1 / 250 second, and the filter transmittance is set to 85%; Step S13: performing SIFT feature point extraction on the original visible light cable image to obtain a visible light image feature point set, and performing SIFT feature point extraction on the original near-infrared cable image to obtain a near-infrared image feature point set; Step S14: performing feature matching on the visible light image feature point set and the near-infrared image feature point set to obtain a dual-spectrum feature matching point pair; Step S15: constructing an affine transformation matrix based on the bispectral feature matching point pairs to obtain bispectral registration transformation parameters; Step S16: geometrically correcting the original near-infrared cable image based on the dual-spectrum registration transformation parameters to obtain a registered near-infrared cable image, and geometrically correcting the original near-infrared cable image based on the dual-spectrum registration transformation parameters to obtain a registered near-infrared cable image; Step S17: performing pixel-level fusion on the original visible light cable image and the registered near-infrared cable image to obtain a registered communication cable image; Step S18: performing local contrast enhancement on the registered communication optical cable image to obtain an enhanced communication optical cable image.

3. The method for identifying communication optical cable faults based on machine vision according to claim 1, characterized in that: Constructing the communication optical cable motion prediction model in step S2 includes the following steps: Step S21: Acquire the original data of the three-axis gyroscope during the flight of the UAV to obtain the angular velocity measurement data of the UAV; Step S22: obtaining the original data of the three-axis accelerometer during the flight of the UAV to obtain the linear acceleration measurement data of the UAV; Step S23: performing zero bias correction on the drone angular velocity measurement data to obtain corrected drone angular velocity data, wherein a static correction method is used to calculate the average value of 500 sampling points in a static state as a zero bias compensation value; Step S24: Separating the gravity component of the UAV linear acceleration measurement data to obtain the UAV net acceleration data; Step S25: performing attitude angle integral calculation based on the corrected drone angular velocity data to obtain the drone attitude angle data, wherein the quaternion integration method is used with an integration step of 0.001 seconds, including the pitch angle, roll angle, and yaw angle; Step S26: performing coordinate system conversion based on the drone attitude angle data and the drone net acceleration data to obtain the drone motion state data; Step S27: performing optical cable relative motion calculation on the UAV motion state data to obtain optical cable swing relative displacement data; Step S28: extracting frequency domain features based on the relative displacement data of the optical cable swing to obtain characteristic parameters of the optical cable swing spectrum; and constructing a communication optical cable motion prediction model based on the characteristic parameters of the optical cable swing spectrum.

4. The method for identifying communication optical cable faults based on machine vision according to claim 1, wherein: In step S2, the dynamic tracking and geometric constraint screening of the target segment of the communication optical cable based on the communication optical cable motion prediction model includes the following steps: Step S291: performing Harris corner point detection on the enhanced communication optical cable image to obtain a set of characteristic corner points of the optical cable image; initializing the characteristic point state based on the set of characteristic corner points of the optical cable image to obtain initial data of the characteristic point state of the optical cable; Step S292: performing motion prediction on the initial data of the optical cable feature point state according to the communication optical cable motion prediction model to obtain the predicted position of the optical cable feature point; Step S293: performing continuous frame image acquisition on the target section of the communication optical cable to obtain a continuous frame image of the communication optical cable, and performing gradient domain enhancement on the continuous frame image of the communication optical cable to obtain an enhanced continuous frame image of the communication optical cable; Step S294: performing optical cable feature point observation and updating on the continuous frame images of the enhanced communication optical cable to obtain the observed positions of the optical cable feature points; Step S295: fusing the predicted position of the optical cable feature point with the observed position of the optical cable feature point to obtain the stable tracking coordinates of the optical cable; Step S296: performing normalized correlation template matching on the optical cable stable tracking coordinates to obtain a set of candidate positions of the optical cable detection area; Step S297: performing geometric feature analysis on the set of candidate locations of the optical cable detection area to obtain geometric feature data of the optical cable detection area, and constructing an optical cable geometric constraint condition set based on the geometric feature data of the optical cable detection area; Step S298: Screening and verifying the optical cable detection area candidate position set based on the optical cable geometric constraint condition set to obtain the optical cable detection area positioning data.

5. The method for identifying communication optical cable faults based on machine vision according to claim 1, wherein: The step S3 of extracting surface morphological features from the enhanced communication optical cable image includes the following steps: Step S31: constructing a texture extraction filter based on the enhanced communication optical cable image to obtain an optical cable surface texture extraction filter; Step S32: performing a convolution operation on the enhanced communication optical cable image using an optical cable surface texture extraction filter to obtain optical cable multi-scale texture response data; Step S33: performing amplitude normalization on the optical cable multi-scale texture response data to obtain standard optical cable texture feature data; Step S34: performing multi-directional texture energy evaluation based on the standard optical cable texture feature data to obtain optical cable surface texture energy distribution data; Step S35: performing adaptive threshold segmentation on the optical cable surface texture energy distribution data to obtain optical cable texture significant area data; Step S36: constructing morphological structural elements based on the data of the significant area of ​​the optical cable texture to obtain a morphological operator dedicated to the optical cable; Step S37: using a cable-specific morphological operator to perform Top-Hat transformation on the enhanced communication cable image to obtain cable surface depression feature data.

6. The method for identifying communication optical cable faults based on machine vision according to claim 1, characterized in that: The step S38 of locating the candidate fault area based on the optical cable surface depression feature data includes the following steps: Step S381: performing morphological reconstruction on the optical cable surface depression feature data to obtain optical cable surface abnormal morphological data; performing edge detection based on the optical cable surface abnormal morphological data to obtain optical cable abnormal edge contour data; Step S382: performing a logical AND operation on the optical cable abnormal edge contour data and the optical cable surface depression feature data to obtain an optical cable surface abnormal feature map; Step S383: marking the connected domain of the abnormal feature map of the optical cable surface to obtain the abnormal connected domain label of the optical cable; Step S384: extracting geometric feature parameters based on the optical cable abnormal connected domain label to obtain geometric feature data of the optical cable abnormal area; Step S385: performing multi-condition filtering on the optical cable abnormal area geometric feature data according to preset filtering conditions to obtain the optical cable suspected fault area data; Step S386: performing spatial position matching on the optical cable suspected fault area data based on the optical cable detection area positioning data to obtain the coordinates of the optical cable fault candidate area; Step S387: performing boundary extraction based on the coordinates of the optical cable fault candidate area to obtain boundary data of the optical cable fault candidate area; performing centroid coordinate calculation on the boundary data of the optical cable fault candidate area to obtain positioning optical cable fault candidate area data.

7. The method for identifying communication optical cable faults based on machine vision according to claim 1, characterized in that: Step S4 includes the following steps: Step S41: dividing the data of the candidate area of ​​the located optical cable fault into time series windows to obtain a time series window of the optical cable fault area; Step S42: extracting the centroid trajectory of the optical cable fault area based on the time series window of the optical cable fault area to obtain the centroid trajectory data of the optical cable fault area; Step S43: performing trajectory smoothing filtering on the centroid trajectory data of the optical cable fault area to obtain smoothed trajectory data of the optical cable fault area; Step S44: Calculating the trajectory displacement variance based on the smoothed optical cable fault area trajectory data to obtain the displacement stability parameter of the optical cable fault area; Step S45: monitoring the temporal changes of morphological features of the candidate area for locating the optical cable fault, and obtaining morphological change data of the optical cable fault area; Step S46: performing a morphological stability assessment based on the morphological change data of the optical cable fault area to obtain a morphological stability parameter of the optical cable fault area; Step S47: performing a comprehensive stability evaluation based on the displacement stability parameter and the morphological stability parameter of the optical cable fault area to obtain temporal stability data of the optical cable fault area; Step S48: Acquire the communication optical cable motion characteristic data; generate the optical cable motion constraint condition parameters based on the communication optical cable motion characteristic data; evaluate the authenticity of the optical cable fault based on the optical cable motion constraint condition parameters and the optical cable fault area temporal stability data to obtain verified optical cable fault characteristic data.

8. The method for identifying communication optical cable faults based on machine vision according to claim 7, characterized in that: Step S48 includes the following steps: Step S481: Acquire continuous frame images of the enhanced communication optical cable; perform optical flow field feature extraction on the continuous frame images of the enhanced communication optical cable to obtain optical flow field data of the optical cable image; Step S482: extracting a local optical flow vector from the optical flow field data of the optical cable image based on the data of the candidate area of ​​the located optical cable fault, and obtaining an optical flow vector of the optical cable fault area; Step S483: performing vector consistency evaluation on the optical flow vector of the optical cable fault area to obtain an optical flow consistency parameter of the optical cable fault area; Step S484: collecting motion characteristic features of the communication optical cable to obtain motion characteristic data of the communication optical cable, and constructing motion constraint conditions based on the motion characteristic data of the communication optical cable to obtain optical cable motion constraint condition parameters; Step S485: performing constraint verification on the optical flow consistency parameters of the optical cable fault area based on the optical cable motion constraint condition parameters to obtain constraint verification optical cable fault optical flow data; Step S486: performing spatiotemporal correlation on the optical cable fault area temporal stability data and the constraint verification optical cable fault optical flow data to obtain spatiotemporal correlation feature data of the optical cable fault; Step S487: performing a fault authenticity score based on the temporal and spatial correlation feature data of the optical cable fault to obtain an optical cable fault authenticity score; Step S488: performing threshold judgment screening on the authenticity score of the optical cable fault to obtain screened optical cable fault area data; performing fault characteristic parameter statistics based on the screened optical cable fault area data to obtain verified optical cable fault characteristic data.

9. The method for identifying communication optical cable faults based on machine vision according to claim 1, wherein: Step S5 includes the following steps: Step S51: extracting the fault type feature of the verification optical cable fault feature data to obtain an optical cable fault type feature vector; Step S52: constructing a fault classification rule set based on the optical cable fault type feature vector to obtain an optical cable fault classification rule; Step S53: performing fault type identification on the verified optical cable fault feature data according to the optical cable fault classification and discrimination rules to obtain optical cable fault type identification data; Step S54: quantifying the defect geometric size according to the optical cable fault type identification data to obtain the optical cable fault geometric size parameters; Step S55: performing a defect risk assessment based on the optical cable fault geometric size parameters to obtain the optical cable fault severity level; Step S56: Obtain the GPS coordinates of the drone and the internal reference table of the onboard camera; Step S57: coordinate mapping is performed on the verification optical cable fault feature data according to the GPS coordinates of the drone and the internal reference table of the onboard camera to obtain the geographical coordinates of the communication optical cable fault; and a structured optical cable fault detection report is generated according to the severity level of the optical cable fault and the geographical coordinates of the communication optical cable fault.

10. A communication optical cable fault identification system based on machine vision, characterized in that: For executing the method for identifying a communication optical cable fault based on machine vision according to claim 1, the communication optical cable fault identification system based on machine vision comprises: An image preprocessing module is used to obtain the original visible light cable image and the original near-infrared cable image; perform dual-spectral registration and pixel-level fusion on the original visible light cable image and the original near-infrared cable image to obtain a registered communication cable image; and perform local contrast enhancement on the registered communication cable image to obtain an enhanced communication cable image. The dynamic tracking module is used to build a communication cable motion prediction model; based on the communication cable motion prediction model, the target section of the communication cable is dynamically tracked and geometrically constrained to obtain the positioning data of the cable detection area; A feature extraction module is used to extract surface morphological features of the enhanced communication optical cable image to obtain optical cable surface depression feature data; locate the fault candidate area based on the optical cable surface depression feature data to obtain located optical cable fault candidate area data; The fault verification module is used to continuously track and verify the data of the candidate area of ​​the located optical cable fault to obtain the time-series stability data of the optical cable fault area; obtain the motion characteristic data of the communication optical cable; generate the optical cable motion constraint condition parameters based on the communication optical cable motion characteristic data; and evaluate the authenticity of the optical cable fault based on the optical cable motion constraint condition parameters and the time-series stability data of the optical cable fault area to obtain the verified optical cable fault characteristic data; The evaluation and reporting module is used to evaluate the severity of the fault based on the verified optical cable fault characteristic data to obtain the optical cable fault severity level; and generate a structured optical cable fault detection report based on the optical cable fault severity level.

Images