Optical cable defect detection method and terminal

Abstract

The invention discloses an optical cable defect detection method and a terminal, and the method comprises the steps: obtaining multi-sensor data of an optical cable, and extracting data features from the multi-sensor data; acquiring a preset model library, and associating and filtering the data features by using the preset model library to generate a fusion feature packet; the fusion feature packet is analyzed based on the improved D-S evidence theory, the defect type and the comprehensive confidence coefficient of the optical cable are obtained, an optical cable inspection report is generated according to the multi-sensor data corresponding to the fusion feature packet, the defect type and the comprehensive confidence coefficient, accurate detection of optical cable defects can be achieved, and the inspection accuracy and efficiency are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of defect detection, in particular to a method for detecting defects of an optical cable and a terminal. BACKGROUND

[0002] The traditional manual high-altitude optical cable inspection mode has long been plagued by many bottlenecks, including: the detection efficiency is limited by terrain and distance, making it difficult to achieve high-frequency, full-coverage rapid inspection; the safety risk of high-altitude and complex environment operation is prominent, and the personnel safety hazard is significant; the detection method relying on manual experience and naked eye observation is easy to miss subtle defects and may have subjective misjudgment; the scattered inspection data records are difficult to synchronize and trace back, which restricts the digitalization process of operation and maintenance. SUMMARY

[0003] The technical problem to be solved by the present application is to provide a method and terminal for detecting defects of an optical cable, which can accurately detect defects of an optical cable and improve the accuracy and efficiency of inspection.

[0004] To solve the above technical problems, the technical scheme adopted by the present application is: A method for detecting defects of an optical cable, comprising the steps of: obtaining multi-sensor data of the optical cable, and extracting data features from the multi-sensor data; obtaining a preset model library, using the preset model library to associate and filter the data features, and generating a fusion feature package; analyzing the fusion feature package based on an improved D-S evidence theory to obtain a defect type and a comprehensive confidence of the optical cable, and generating an optical cable inspection report according to the multi-sensor data corresponding to the fusion feature package, the defect type and the comprehensive confidence.

[0005] To solve the above technical problems, another technical scheme adopted by the present application is: A terminal for detecting defects of an optical cable, comprising a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program and realizes each step of the above-mentioned method for detecting defects of an optical cable.

[0006] The application has the beneficial effects that the application provides an optical cable defect detection method and terminal, realizes comprehensive perception of the optical cable state by acquiring multi-sensor data and extracting data features, and ensures comprehensiveness of defect detection; the data features are associated and filtered based on a preset model library to generate a fusion feature package, effectively improving correlation and consistency between features, reducing data redundancy, and thus reducing false alarms caused by a single sensor or environmental interference; the fusion feature package is analyzed based on the improved D-S evidence theory, which can comprehensively evaluate multi-source information, output a defect type and comprehensive confidence of the optical cable, and improve accuracy and reliability of defect diagnosis; and an optical cable inspection report is generated according to the multi-sensor data, defect type and comprehensive confidence corresponding to the fusion feature package, thereby providing an intuitive and accurate reliable basis for operation and maintenance decision. BRIEF DESCRIPTION OF DRAWINGS

[0007] Figure 1 A flowchart of an optical cable defect detection method according to an embodiment of the application; Figure 2 A schematic diagram of an optical cable defect detection terminal according to an embodiment of the application; KEY 1. An optical cable defect detection terminal; 2. A memory; 3. A processor. DETAILED DESCRIPTION

[0008] To explain the technical content, purposes and effects of the application in detail, the following describes the embodiments in conjunction with the drawings.

[0009] Before the embodiments of the application are described in detail, some related concepts are first explained: (1) Preset model library: refers to a pre-established set of "abnormality-feature" physical correlation rules containing various typical defects of optical cables, each model defines a cross-modal feature combination that a certain specific defect should present on multi-source sensors, and is used for cross-verification and logical filtering of preliminary features from different sensors.

[0010] (2) Improved D-S evidence theory: refers to a decision layer fusion method for information fusion, the improvement point of which is to introduce a conflict evidence redistribution basic probability strategy for synthesizing multiple uncertain or conflicting evidences to obtain a high confidence comprehensive judgment.

[0011] (3) Defect recognition model: refers to an algorithm model based on computer vision for automatically recognizing defects (such as cracks, wear, foreign matter, etc.) on the surface of optical cables in visible light images.

[0012] (4) Arbitration evidence: in the fusion process of the improved D-S evidence theory, when there is a high conflict between core evidences from different sensors, an auxiliary evidence with high reliability is additionally introduced to eliminate the conflict.

[0013] (5) Depster combination rule: is the core mathematical rule in D-S evidence theory, which is used to combine the basic probability assignment (BPA) of multiple independent evidence sources, calculate a joint basic probability assignment, and finally obtain the comprehensive support, opposition and uncertainty of the proposition (such as "there is a certain type of defect") by multiplying and normalizing the evidence probability supporting the same proposition.

[0014] (6) Dynamic Bayesian network: refers to a kind of graph model for modeling the probability relationship between random variables changing over time, which can perform causal reasoning and future prediction.

[0015] In the prior art, optical cable inspection mainly relies on manual climbing or riding a carrier to approach for visual inspection and simple instrument assisted detection. Typical application scenarios include daily inspection, fault finding and state evaluation of power transmission lines and communication optical cables crossing complex geographical environments such as mountains, valleys and forests. The working environment often involves high altitude, strong electromagnetic field, severe weather and other high-risk conditions. However, the existing technology has the following outstanding bottlenecks for a long time: first, the inspection efficiency is limited by the accessibility and distance of the terrain, making it difficult to achieve high frequency, wide range and rapid coverage; second, manual work in high altitude and complex environment faces significant safety risks; third, relying on the experience and naked eye observation of the inspection personnel, the detection ability for subtle defects such as cracks and wear is insufficient, and it is easy to produce missed detection and subjective misjudgment; fourth, the data generated during the inspection process are mainly scattered and unstructured manual records, which are difficult to realize systematic synchronization, integration and traceability, and seriously restrict the evolution of operation and maintenance management towards digitalization and intelligentization. Therefore, there is an urgent need for an optical cable defect detection method and terminal that can overcome the above defects.

[0016] To at least solve the above problems, please refer to Figure 1 The embodiment of the present application provides an optical cable defect detection method, comprising the steps of: obtaining multi-sensor data of an optical cable, and extracting data features from the multi-sensor data; obtaining a preset model library, and using the preset model library to associate and filter the data features to generate a fusion feature package; analyzing the fusion feature package based on an improved D-S evidence theory to obtain a defect type and a comprehensive confidence of the optical cable, and generating an optical cable inspection report according to the multi-sensor data corresponding to the fusion feature package, the defect type and the comprehensive confidence.

[0017] As can be seen from the above description, the beneficial effects of the present invention are as follows: by acquiring multi-sensor data of optical cables and extracting data features, a comprehensive perception of the optical cable status is achieved, providing a rich and reliable data foundation for defect analysis; by using a preset model library to correlate and filter data features, a fusion feature package is generated, effectively integrating multi-source heterogeneous data, improving the accuracy and efficiency of defect detection, and reducing false alarms caused by false alarms from a single sensor or environmental interference; based on the improved DS evidence theory, the fusion feature package is analyzed to accurately identify the risk of missed detection and misjudgment, and an optical cable inspection report is generated according to multi-sensor data, defect type, and comprehensive confidence level, realizing the digital and standardized output of inspection results, and improving the level of operation and maintenance management and traceability.

[0018] Furthermore, the multi-sensor data includes visible light images; Extracting data features from the multi-sensor data includes: The visible light image is preprocessed, and the preprocessed visible light image is input into the defect recognition model to output visual features.

[0019] As described above, preprocessing visible light images effectively improves image quality and standardization, providing a clear and unified input data foundation for subsequent feature extraction. The preprocessed image is then input into the defect recognition model, which outputs visual features, enabling efficient and accurate identification of surface defects in optical cables and improving the objectivity and consistency of feature extraction.

[0020] Furthermore, the multi-sensor data includes infrared thermal images; Extracting data features from the multi-sensor data includes: The infrared thermal image is subjected to temperature calibration, dynamic compensation and radiation correction. Based on temperature threshold segmentation and region growing algorithm, the processed infrared thermal image is analyzed to identify the overheated area on the optical cable surface. Temperature difference analysis and trend prediction are performed on the overheated area on the optical cable surface to obtain temperature characteristics.

[0021] As described above, by performing temperature calibration processing on infrared thermal images, the original grayscale image data is converted into temperature data, providing an accurate and quantitative temperature information basis for subsequent analysis. Dynamic compensation and radiation correction processing are then performed to effectively reduce thermal noise and measurement deviations introduced by environmental factors and equipment factors, resulting in a high-quality corrected temperature image reflecting the thermal state of the optical cable surface. Based on temperature threshold segmentation and region growing algorithms, this image is analyzed to accurately identify and segment potential overheated areas on the optical cable surface, enabling the location of abnormal temperature rise areas. Furthermore, temperature difference analysis and trend prediction are performed on the identified overheated areas, transforming the discrete temperature distribution into temperature features reflecting the degree of thermal anomalies, significantly improving the efficiency of identifying hidden thermal defects in optical cables.

[0022] Furthermore, the multi-sensor data includes acoustic wave data; Extracting data features from the multi-sensor data includes: Abnormal sound wave data is obtained from sound wave data based on voiceprint recognition technology, the abnormal sound wave data is analyzed to locate the location of the abnormal sound source, and acoustic features are generated based on the abnormal sound wave data and the location of the abnormal sound source.

[0023] As described above, by acquiring abnormal acoustic wave data from acoustic wave data based on voiceprint recognition technology, specific acoustic anomalies related to optical cable defects can be identified from complex background noise. The acoustic field characteristics of the abnormal acoustic wave data can be analyzed to accurately locate the specific location of the abnormal sound source on the optical cable. The abstract audio signal is transformed into fault location information, and acoustic features are generated based on the abnormal acoustic wave data and the location of the abnormal sound source, providing reliable data for subsequent fusion diagnosis.

[0024] Further, a preset model library is obtained, and the data features are correlated and filtered using the preset model library to generate a fused feature package, including: The data features are matched with each defect physical model in the preset model library, and each defect physical model includes a combination of data features corresponding to the defect type; Based on the matching results, the data features are associated and filtered to generate a fused feature package.

[0025] As described above, by matching the extracted visual, temperature, acoustic and other multi-source data features with each defect physical model in the preset model library, the preliminary screening and identification of potential defect types can be achieved, effectively focusing the analysis scope, improving the pertinence and efficiency of feature processing, and generating a fusion feature package based on the matching results to provide high-quality input data for subsequent fusion diagnosis, thereby improving the accuracy and reliability of the defect identification process.

[0026] Furthermore, based on the improved D-S evidence theory, the fused feature packet is analyzed to obtain the defect type and its comprehensive confidence level of the optical cable, including: Each data feature in the fused feature package is treated as a piece of evidence, and a basic probability is assigned to each piece of evidence. When there is a conflict between different pieces of evidence, arbitration evidence is introduced, and the basic probability is redistributed according to the arbitration evidence. The defect type of the defect physical model corresponding to the fusion feature is used as the fault hypothesis. The basic probabilities of all evidence are synthesized using the Depster synthesis rule to obtain the comprehensive confidence level of the fault hypothesis. If the comprehensive confidence level reaches the preset confidence threshold, the fault hypothesis is determined to be correct, and the defect type and its comprehensive confidence level of the optical cable are determined.

[0027] As described above, by treating each data feature in the fused feature package as an independent piece of evidence and assigning it a basic probability, a quantitative foundation for fusion decision-making is laid. When there are conflicts between evidence from different sources, arbitration evidence is introduced and the basic probabilities are reallocated accordingly, effectively resolving evidence contradictions caused by sensor errors or environmental interference and improving decision reliability. The defect physical model type corresponding to the fused feature package is used as a fault hypothesis to be verified. The basic probabilities of all evidence are synthesized using Depster's synthesis rule, and the comprehensive confidence of the fault hypothesis is calculated. This achieves deep fusion and uncertainty quantification of multi-source information at the probability level. When the comprehensive confidence reaches a preset confidence threshold, an objective judgment of the defect type is achieved, effectively reducing the risk of misjudgment caused by the limitations of single-source information.

[0028] Furthermore, it also includes: Construct a dynamic Bayesian network, which includes environmental data, load, historical status data of optical cables, and multi-sensor data; Based on the fused feature package and load trend data, the dynamic Bayesian network is used to predict the risk level of defect evolution in future time periods.

[0029] As described above, by constructing a dynamic Bayesian network that integrates environmental data, load, historical optical cable status, and multi-sensor data, multi-source heterogeneous information is integrated into a unified probabilistic graphical model. This establishes the causal dependency between state variables and influencing factors, providing a comprehensive and reasonable mathematical model foundation for defect evolution analysis. Based on the current fused feature package and trend-compliant data, this network is used for time-series reasoning and probability calculation to predict the risk level of defect evolution in future time periods, achieving a forward-looking assessment of the optical cable's health status. This enables proactive early warning for optical cable operation and maintenance, improving the accuracy and planning of risk management.

[0030] Furthermore, acquire multi-sensor data, including: Each type of sensor data is assigned a unique timestamp, and the multi-sensor data is aligned according to the time dimension based on the timestamp.

[0031] As described above, by assigning a unified timestamp to each type of sensor data, all multi-sensor data can be aligned and synchronized in the time dimension based on the timestamp, thereby achieving consistent integration of multi-modal data in the time series and improving the accuracy and effectiveness of data fusion analysis.

[0032] Furthermore, acquiring multi-sensor data also includes: Each type of sensor data is transformed into a unified coordinate system, and the multi-sensor data is aligned and correlated according to spatial location based on the coordinate system.

[0033] As described above, by transforming each type of sensor data into a unified coordinate system, and precisely aligning and associating multi-sensor data according to their corresponding actual spatial locations, visible light images, infrared thermal images, sound source localization, and other data can be accurately mapped to the same physical segment of the optical cable. This achieves the fusion and registration of multi-source heterogeneous data in the spatial dimension, thereby providing an accurate and reliable spatial data foundation for subsequent defect feature extraction, correlation analysis, and comprehensive diagnosis based on the same spatial location. This significantly improves the accuracy of defect localization and the effectiveness of multi-evidence collaborative judgment.

[0034] Please refer to Figure 2 Another embodiment of the present invention provides an optical cable defect detection terminal, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the various steps of the optical cable defect detection method described above.

[0035] The optical cable defect detection method and terminal described above are applicable to optical cable inspection scenarios, enabling accurate detection of optical cable defects and improving inspection accuracy and efficiency. The specific implementation methods are described below: Please refer to Figure 1 One embodiment of the present invention is as follows: A method for detecting defects in optical cables, comprising the following steps: S1. Acquire multi-sensor data of the optical cable and extract data features from the multi-sensor data, specifically including S1.1-S1.3.

[0036] In this embodiment, the multi-sensor data includes visible light images, infrared thermal images, and acoustic wave data. The sensor data is fused to achieve comprehensive detection and precise correlation positioning of the surface condition, temperature distribution, and abnormal sounds of high-voltage cables under a unified time and space reference.

[0037] S1.1 Preprocess the visible light image, input the preprocessed visible light image into the defect recognition model, and output visual features.

[0038] In this embodiment, a visible light image of the optical cable is acquired using a visible light camera. The preprocessing of the visible light image includes: enhancing the image using CLAHE (contrast-limited adaptive histogram equalization) to improve the visibility of details under uneven lighting conditions; extracting the edge structure of the cable using the Canny edge detection algorithm and combining it with Hough transform to achieve accurate localization of the cable region; and reducing image noise using a wavelet denoising algorithm to obtain a high-quality, high-contrast preprocessed image, thereby improving the accuracy of subsequent recognition.

[0039] The pre-processed visible light image is input into the defect recognition model, which is based on the YOLO framework. By embedding a CBAM (Convolutional Attention Module), the model focuses on key areas of the cable surface, enhancing its ability to extract features of minor defects (such as cracks, wear, broken strands, and foreign objects hanging on the cable surface). Transfer learning and data augmentation techniques are used to train a detection model adapted to complex field environments using limited samples. The model outputs visual features such as defect type, location coordinates, and confidence level, and combines them with data from the laser ranging unit to achieve three-dimensional spatial localization of the defect.

[0040] S1.2. Perform temperature calibration, dynamic compensation and radiation correction processing on the infrared thermal image. Analyze the processed infrared thermal image based on temperature threshold segmentation and region growing algorithm to identify the overheated area on the optical cable surface. Perform temperature difference analysis and trend prediction on the overheated area on the optical cable surface to obtain temperature characteristics.

[0041] In this embodiment, the infrared thermal image is subjected to temperature calibration, dynamic compensation, and radiation correction. Specifically, the raw thermal radiation data is obtained by configuring a vanadium oxide uncooled detector, real-time temperature calibration is performed using a blackbody radiation source, and dynamic compensation is performed in the 8~14μm response band in combination with ambient temperature and humidity sensor data to achieve a temperature measurement error of ≤±1℃. At the same time, a radiation correction algorithm is used to eliminate the influence of factors such as atmospheric transmittance and background radiation.

[0042] Based on temperature threshold segmentation and region growing algorithms, processed infrared thermal images can be analyzed to automatically identify and segment overheated areas on the surface of optical cables and connecting components, such as abnormal temperature rise areas caused by poor joint contact or partial discharge. For the identified overheated areas on the optical cable surface, combined with real-time acquired cable load current data, temperature difference analysis and trend prediction are performed to obtain temperature characteristics and provide early warnings of potential faults.

[0043] S1.3. Obtain abnormal sound wave data from sound wave data based on voiceprint recognition technology, analyze the abnormal sound wave data to locate the abnormal sound source, and generate acoustic features based on the abnormal sound wave data and the location of the abnormal sound source.

[0044] In this embodiment, abnormal sound wave data is obtained from sound wave data based on voiceprint recognition technology. Specifically, a high-sensitivity acoustic imaging device (ultrasonic microphone array) is used to collect abnormal sound wave data generated by corona discharge, arcing, or mechanical loosening on the surface of cables and insulators using voiceprint recognition technology. Beamforming algorithms and sound field imaging technology are used to process the identified abnormal sound waves, enabling precise location of the sound source in three-dimensional space. Acoustic features are generated based on the abnormal sound wave data and the location of the abnormal sound source. These acoustic features are then fused with visual and infrared data to achieve multi-dimensional cross-verification and comprehensive diagnosis of faults such as discharge.

[0045] Furthermore, in this embodiment, before feature extraction from the multi-sensor data, the acquired multi-sensor data is synchronized in time and space. The specific processing procedure is as follows: 1. Assign a unified timestamp to each type of sensor data, and align the multi-sensor data according to the time dimension based on the timestamp. Specifically, a hardware clock synchronization mechanism based on the IEEE 1588 PTP protocol is adopted to assign a unified microsecond-level high-precision timestamp to all image frames, thermal imaging frames, point cloud frames, and audio stream data acquired by all sensors; for non-periodic or triggered data (such as snapshots taken when an anomaly is detected), the interrupt signal and the corresponding timestamp are used to strictly align it with the main data stream in the time dimension to ensure that all data have a consistent time reference.

[0046] 2. Transform the data from each type of sensor to a unified coordinate system, and align and correlate the multi-sensor data according to their spatial positions based on this coordinate system. Specifically, through a joint calibration method, using a combined target containing checkerboard and heat source feature points, accurately obtain the relative position and attitude transformation matrix between the visible light camera, infrared thermal imager, lidar, and acoustic array, and complete the extrinsic parameter calibration of each sensor; transform the raw data from all sensors to the robot body coordinate system centered on lidar, and then further map the data to the global geographic coordinate system through a robot localization system that integrates GPS, IMU, and lidar SLAM, achieving precise alignment and unification of all data in spatial position.

[0047] Based on the above spatiotemporal synchronization, feature association and fusion analysis based on spatial location can be realized. For example, when the visual system identifies a "damaged insulator" area, the temperature sequence can be extracted from the corresponding position in the infrared thermogram through coordinate transformation, and the acoustic spectrum at that moment and in the spatial direction can be retrieved simultaneously to check whether there are characteristic frequency components of partial discharge. This enables multi-dimensional evidence association and comprehensive judgment of the same physical phenomenon. When the visual system detects a protrusion on the cable surface, the point cloud data can be used to immediately determine whether the protrusion is due to an attached foreign object (exogenous) or the cable itself expanding (endogenous), and its precise size can be measured.

[0048] S2. Obtain a preset model library, and use the preset model library to associate and filter the data features to generate a fused feature package, specifically including: The data features are matched with each defect physical model in the preset model library, and each defect physical model includes a combination of data features corresponding to the defect type; Based on the matching results, the data features are associated and filtered to generate a fused feature package.

[0049] In this embodiment, the pre-built model library is a physical correlation model library of "anomalies-features," such as the corona discharge model (correlated with a combination of features including specific acoustic spectra, weak local temperature rise, and no visible surface anomalies) and the bolt loosening model (correlated with a combination of features including changes in structural gaps, increased contact resistance leading to overheating, and mechanical noise). Matching multi-dimensional data features such as visual, temperature, and acoustic data with each defect physical model in the pre-built model library essentially utilizes the prior knowledge of multi-sensor feature combinations defined by the model to filter and classify primary features. Based on the matching results, the data features are correlated and filtered to generate a fused feature package. By performing physical mechanism-based cross-validation on multi-source alarms, the probability of misjudging environmental interference (such as bird droppings) as real defects is effectively reduced.

[0050] S3. Based on the improved D-S evidence theory, analyze the fused feature packet to obtain the defect type and its comprehensive confidence level of the optical cable, specifically including: Each data feature in the fused feature package is treated as a piece of evidence, and a basic probability is assigned to each piece of evidence. When there is a conflict between different pieces of evidence, arbitration evidence is introduced, and the basic probability is redistributed according to the arbitration evidence. The defect type of the defect physical model corresponding to the fusion feature is used as the fault hypothesis. The basic probabilities of all evidence are synthesized using the Depster synthesis rule to obtain the comprehensive confidence level of the fault hypothesis. If the comprehensive confidence level reaches the preset confidence threshold, the fault hypothesis is determined to be correct, and the defect type and its comprehensive confidence level of the optical cable are determined.

[0051] In this embodiment, each data feature in the fused feature package (such as overheating of a cable node) is treated as an independent piece of evidence, and a Basic Probability Assignment (BPA) is assigned to each piece of evidence. The determination of this BPA comprehensively considers multiple factors, including algorithm confidence (such as the target confidence of YOLO), the real-time health status of the sensor, and current environmental conditions (such as the impact of fog on visual recognition). When there is a significant conflict between evidence from different sensors (e.g., no abnormality is detected visually but a strong infrared alarm is triggered), a conflict evidence redistribution strategy is introduced. This involves calling additional data (such as "surface morphology changes") provided by arbitration units such as laser ranging, and re-evaluating and redistributing the basic probabilities of the relevant evidence accordingly, in order to resolve contradictions and improve the internal consistency of the evidence set.

[0052] The defect type of the defect physical model corresponding to the fused feature package is used as the fault hypothesis to be verified. The basic probabilities after updating all evidence are synthesized by Dempster synthesis rules to calculate the overall confidence level of the hypothesis. If the overall confidence level reaches the preset confidence threshold, the fault hypothesis is determined to be valid, thereby determining the specific defect type of the optical cable and its quantified overall confidence level, and outputting a high-reliability diagnostic conclusion such as "There is a moderate overheating defect caused by poor contact, with an overall confidence level of 94%".

[0053] Furthermore, in this embodiment, the method further includes: constructing a dynamic Bayesian network, which includes environmental data (temperature, humidity, wind speed), load (current), historical status data of optical cables, and the multi-sensor data; based on the fused feature package and load trend data, the dynamic Bayesian network is used to predict the risk level of defect evolution in future time periods. For example, if "slight temperature increase + specific noise + sudden load increase" is observed, the possibility of "early corona discharge" can be inferred, even if the infrared hotspot has not yet reached the alarm threshold at that moment, thus achieving predictive maintenance.

[0054] S4. Generate an optical cable inspection report based on the multi-sensor data corresponding to the fused feature package, the defect type, and the comprehensive confidence level.

[0055] In this embodiment, an optical cable inspection report is generated based on the multi-sensor data corresponding to the fused feature package, the defect type, and its overall confidence level, thus outputting a unified, multi-dimensional digital report. This report not only includes core judgment results such as defect type, geographic coordinates (based on a global coordinate system), severity level, and overall confidence level, but also integrates multi-sensor data snapshots as key evidence, such as spatiotemporally aligned visible light images, infrared thermal images, and acoustic spectrograms, and provides targeted preliminary maintenance suggestions for each type of defect.

[0056] Furthermore, in this embodiment, a closed-loop feedback learning mechanism is established to continuously integrate the confirmation results of on-site manual review or long-term operating data into the model (especially the dynamic Bayesian network (DBN) for risk prediction and the basic probability assignment (BPA) rule for evidence synthesis), thereby enabling the diagnostic model to continuously iterate and self-optimize in practical applications, and continuously improve the overall accuracy and reliability of the system.

[0057] According to another aspect of the invention, Figure 2 This is a schematic diagram illustrating a fiber optic cable defect detection terminal according to an embodiment of the present invention. The electronic device includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the various steps of the fiber optic cable defect detection method described above.

[0058] In summary, this invention provides a method and terminal for optical cable defect detection. By acquiring multi-sensor data, performing spatiotemporal fusion of the multi-sensor data, and extracting data features, it achieves comprehensive perception of the optical cable status, ensuring the comprehensiveness of defect detection. Based on a preset model library, it correlates and filters data features to generate a fused feature package, effectively improving the correlation and consistency between features, reducing data redundancy, and thus reducing false alarms caused by single sensor false alarms or environmental interference. Based on the improved DS evidence theory, it analyzes the fused feature package, comprehensively evaluating multi-source information, outputting the optical cable defect type and its comprehensive confidence level, improving the accuracy and reliability of defect diagnosis. Based on the multi-sensor data, defect type, and comprehensive confidence level corresponding to the fused feature package, it generates an optical cable inspection report, providing an intuitive and accurate reliable basis for operation and maintenance decisions.

[0059] The above description is merely an embodiment of the present invention and does not limit the patent scope of the present invention. Any equivalent modifications made based on the content of the present invention specification and drawings, or direct or indirect applications in related technical fields, are similarly included within the patent protection scope of the present invention.

Claims

1. A method for detecting defects in optical cables, characterized in that, Including the following steps: Acquire multi-sensor data from the optical cable and extract data features from the multi-sensor data; Obtain a preset model library, and use the preset model library to associate and filter the data features to generate a fused feature package; Based on the improved D-S evidence theory, the fused feature package is analyzed to obtain the defect type and its comprehensive confidence level of the optical cable. Based on the multi-sensor data corresponding to the fused feature package, the defect type, and the comprehensive confidence level, an optical cable inspection report is generated.

2. The method for detecting defects in optical cables according to claim 1, characterized in that, The multi-sensor data includes visible light images; Extracting data features from the multi-sensor data includes: The visible light image is preprocessed, and the preprocessed visible light image is input into the defect recognition model to output visual features.

3. The optical cable defect detection method according to claim 1, characterized in that, The multi-sensor data includes infrared thermal images; Extracting data features from the multi-sensor data includes: The infrared thermal image is subjected to temperature calibration, dynamic compensation and radiation correction. Based on temperature threshold segmentation and region growing algorithm, the processed infrared thermal image is analyzed to identify the overheated area on the optical cable surface. Temperature difference analysis and trend prediction are performed on the overheated area on the optical cable surface to obtain temperature characteristics.

4. The optical cable defect detection method according to claim 1, characterized in that, The multi-sensor data includes acoustic wave data; Extracting data features from the multi-sensor data includes: Abnormal sound wave data is obtained from sound wave data based on voiceprint recognition technology, the abnormal sound wave data is analyzed to locate the location of the abnormal sound source, and acoustic features are generated based on the abnormal sound wave data and the location of the abnormal sound source.

5. The optical cable defect detection method according to claim 1, characterized in that, Obtain a preset model library, use the preset model library to associate and filter the data features, and generate a fused feature package, including: The data features are matched with each defect physical model in the preset model library, and each defect physical model includes a combination of data features corresponding to the defect type; Based on the matching results, the data features are associated and filtered to generate a fused feature package.

6. The method for detecting defects in optical cables according to claim 1, characterized in that, Based on the improved D-S evidence theory, the fused feature packet is analyzed to obtain the defect type and its comprehensive confidence level of the optical cable, including: Each data feature in the fused feature package is treated as a piece of evidence, and a basic probability is assigned to each piece of evidence. When there is a conflict between different pieces of evidence, arbitration evidence is introduced, and the basic probability is redistributed according to the arbitration evidence. The defect type of the defect physical model corresponding to the fusion feature is used as the fault hypothesis. The basic probabilities of all evidence are synthesized using the Depster synthesis rule to obtain the comprehensive confidence level of the fault hypothesis. If the comprehensive confidence level reaches the preset confidence threshold, the fault hypothesis is determined to be correct, and the defect type and its comprehensive confidence level of the optical cable are determined.

7. The method for detecting defects in optical cables according to claim 1, characterized in that, Also includes: Construct a dynamic Bayesian network, which includes environmental data, load, historical status data of optical cables, and multi-sensor data; Based on the fused feature package and load trend data, the dynamic Bayesian network is used to predict the risk level of defect evolution in future time periods.

8. The method for detecting defects in optical cables according to claim 1, characterized in that, Acquire multi-sensor data, including: Each type of sensor data is assigned a unique timestamp, and the multi-sensor data is aligned according to the time dimension based on the timestamp.

9. The optical cable defect detection method according to claim 8, characterized in that, Acquiring multi-sensor data also includes: Each type of sensor data is transformed into a unified coordinate system, and the multi-sensor data is aligned and correlated according to spatial location based on the coordinate system.

10. A fiber optic cable defect detection terminal, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements each step of the optical cable defect detection method according to any one of claims 1 to 9.

Images