Optical cable early warning method, system and device and storage medium

Abstract

The invention discloses an optical cable early warning method, system and device and a storage medium, and the method comprises the steps: obtaining monitoring data, and predicting a risk event and a risk point according to the monitoring data and a trained cross-modal large model; the monitoring data comprises any one or more of vibration data, image data or audio data; detecting a fault point and a fault event of the optical cable by using an optical time domain reflectometer; and if the risk point is matched with the fault point, evaluating a risk level according to the risk event and the fault event, and performing early warning according to the risk level. The embodiment of the invention can quickly and accurately monitor and early warn the optical cable, and can be widely applied to the technical field of optical cable transmission.

Description

Technical Field

[0001] The present invention relates to the field of optical cable transmission technology, and in particular to an optical cable early warning method, system, device and storage medium. Background Art

[0002] In modern communication networks, the safe operation of optical cables is of vital importance. First of all, the optical cable network is the foundation of national communication facilities and undertakes important information transmission tasks. Since optical cables are usually laid outdoors, they are affected by environmental factors and external activities and are prone to damage. For example, construction excavation, traffic accidents and natural disasters can cause direct or indirect damage to optical cables, thereby affecting the stable operation of communication networks. Secondly, the existing optical cable monitoring and maintenance methods mainly rely on manual inspections, which not only consumes a lot of manpower and material resources, but also cannot achieve real-time monitoring and rapid response of optical cables. With the continuous expansion of the scale of optical cable networks, traditional inspection methods have been unable to meet the needs of modern communication networks for efficient and accurate monitoring. Summary of the invention

[0003] In view of this, an object of the embodiments of the present invention is to provide an optical cable early warning method, system, device and storage medium, which can quickly and accurately monitor and warn the optical cable.

[0004] On the one hand, an embodiment of the present invention provides an optical cable early warning method, comprising:

[0005] Acquire monitoring data, and predict risk events and risk points based on the monitoring data and the trained cross-modal large model; the monitoring data includes any one or more of vibration data, image data or audio data;

[0006] Use optical time domain reflectometer to detect fault points and fault events of optical cables;

[0007] If the risk point matches the fault point, the risk level is evaluated according to the risk event and the fault event, and an early warning is issued according to the risk level.

[0008] Optionally, the cross-modal large model is trained by the following method:

[0009] Acquire monitoring sample data, perform first preprocessing on the monitoring sample data, and add labels; the monitoring sample data includes any one or more of vibration sample data, image sample data, or audio sample data;

[0010] The monitoring sample data are fused according to the labels to obtain a training sample set, and the cross-modal large model is trained according to the training sample set.

[0011] Optionally, the vibration sample data is obtained by the following method:

[0012] Acquire an initial vibration sample signal through an optical fiber vibration sensor;

[0013] The initial vibration sample signal is subjected to a second preprocessing, and the initial vibration sample signal having a signal-to-noise ratio lower than a first threshold is subjected to a recovery process to obtain vibration sample data.

[0014] Optionally, the image sample data is obtained by the following method:

[0015] Acquiring ground image data within the range of the optical cable, and performing a third preprocessing on the ground image data;

[0016] The ground image data after the third preprocessing is classified and located to obtain image sample data.

[0017] Optionally, the method further comprises:

[0018] If risk events and risk points cannot be predicted based on monitoring data and the trained cross-modal big model due to missing data, identify the missing data segments;

[0019] According to the characteristics of the missing data segment and the first preset weight, similar data segments are searched in the database, and the missing data segment is supplemented according to the confidence of the similar data segment.

[0020] Optionally, the method further comprises:

[0021] Predict risk events and risk points based on supplemented monitoring data and trained cross-modal large models;

[0022] Obtain on-site detection results. If the risk event or the risk point is inconsistent with the on-site detection results, replenish the missing data segments.

[0023] Optionally, the assessing the risk level according to the risk event and the failure event includes:

[0024] Determine a risk score according to the risk event, and determine a failure score according to the failure event;

[0025] Determine the fiber impact score and the signal impact score;

[0026] The risk level is evaluated according to the risk score, the fault score, the fiber impact score, the signal impact score and a second preset weight.

[0027] On the other hand, an embodiment of the present invention provides an optical cable early warning system, including:

[0028] The first module is used to obtain monitoring data and predict risk events and risk points based on the monitoring data and the trained cross-modal large model; the monitoring data includes any one or more of vibration data, image data or audio data;

[0029] The second module is used to detect the fault point and fault event of the optical cable using an optical time domain reflectometer;

[0030] The third module is used to evaluate the risk level according to the risk event and the fault event if the risk point matches the fault point, and issue an early warning according to the risk level.

[0031] On the other hand, an embodiment of the present invention provides an optical cable early warning device, including:

[0032] at least one processor;

[0033] at least one memory for storing at least one program;

[0034] When the at least one program is executed by the at least one processor, the at least one processor implements the above method.

[0035] On the other hand, an embodiment of the present invention provides a computer-readable storage medium, in which a program executable by a processor is stored. When the program executable by the processor is executed by the processor, it is used to perform the above method.

[0036] Implementation of the embodiments of the present invention includes the following beneficial effects: the embodiment first acquires monitoring data, predicts risk events and risk points based on the monitoring data and a trained cross-modal large model, the monitoring data includes any one or more of vibration data, image data or audio data, predicts risk events and risk points based on the multimodal monitoring data and the cross-modal large model, and improves the accuracy of the prediction results. Then, an optical time domain reflectometer is used to detect the fault point and fault event of the optical cable, and the fault of the optical cable is detected by another method. If the risk point matches the fault point, the risk level is evaluated based on the risk event and the fault event, and an early warning is issued based on the risk level. A comprehensive analysis is performed through the prediction results of the monitoring data and the detection results of the optical time domain reflectometer to improve the accuracy of optical cable monitoring and early warning, while improving efficiency, reducing human intervention and maintenance costs, and ensuring the stable operation of the communication network. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 It is a schematic diagram of the steps of an optical cable early warning method provided by an embodiment of the present invention;

[0038] Figure 2 It is a schematic diagram of a process flow of obtaining vibration sample data provided by an embodiment of the present invention;

[0039] Figure 3 This is an installation position distribution diagram of an optical cable camera provided by an embodiment of the present invention;

[0040] Figure 4 is a schematic diagram of a process flow of obtaining image sample data provided by an embodiment of the present invention;

[0041] Figure 5 It is a schematic diagram of a step flow of training a cross-modal large model provided by an embodiment of the present invention;

[0042] Figure 6 is a schematic diagram of a process flow of supplementing missing data segments provided by an embodiment of the present invention;

[0043] Figure 7 is another schematic flow chart of steps for supplementing missing data segments provided by an embodiment of the present invention;

[0044] Figure 8 It is a schematic diagram of a multimodal data fusion process provided by an embodiment of the present invention;

[0045] Fig. 9 This is a schematic diagram of a process flow of refilling missing data segments provided by an embodiment of the present invention;

[0046] Fig.10 This is a schematic diagram of a process flow for assessing risk levels provided by an embodiment of the present invention;

[0047] Fig.11 It is a structural block diagram of an optical cable early warning system provided by an embodiment of the present invention;

[0048] Fig.12 It is a structural block diagram of an optical cable early warning device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0049] The present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments. The step numbers in the following embodiments are only provided for the convenience of explanation and description, and the order between the steps is not limited in any way. The execution order of each step in the embodiment can be adaptively adjusted according to the understanding of those skilled in the art.

[0050] It should be noted that, although the functional module division is performed in the device schematic diagram and the logical order is shown in the flow chart, in some cases, the steps shown or described can be performed in a different order from the module division in the device or the order in the flow chart. The terms "first", "second", etc. in the specification and claims and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described here can be implemented in an order other than those illustrated or described here. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0051] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art to which this application belongs. The terms used herein are only for the purpose of describing the embodiments of this application and are not intended to limit this application.

[0052] Some technical terms in this embodiment are explained below.

[0053] The linear Sagnac structure is an optical structure. Compared with other distributed sensing structures, the vibration sensing system based on the Sagnac structure has the characteristics of simple structure, long sensing distance and stable polarization.

[0054] Cross-modal models are a type of artificial intelligence model that can process and understand multiple types (modal) of data (such as text, images, audio, video, etc.). Its goal is to make data of different modalities interrelated and then jointly understand and analyze them. For example, a cross-modal model can process images and text at the same time, understand objects in a picture and their corresponding text descriptions, or match vibration signals with scenes in a video. Cross-modal models are widely used in multimodal fusion, information retrieval, image generation and other fields.

[0055] Cross-modal retrieval is a technology that can retrieve data in one modality (such as images or videos) through data in another modality (such as text). Its purpose is to improve the effect of data retrieval by analyzing and understanding the semantic relationship between different modalities without being limited to a single modality. For example, you can use natural language descriptions to search for related images in an image library, or use images to search for related text content. This technology relies on the cross-modal model to learn and understand the common feature space of multiple modal data.

[0056] An Optical Time-Domain Reflectometer (OTDR) is a measurement device used to detect faults, losses, and other performance issues in optical fiber links. OTDR analyzes the characteristics of optical fiber by sending laser pulses into the optical fiber and measuring the intensity and time of the returned reflected light signal. Through these reflected signals, OTDR can identify loss points, breakpoints, connection points, and changes in reflectivity in the optical fiber, thereby locating fault points and damage in the optical fiber link. OTDR technology is widely used in the installation, maintenance, and troubleshooting of optical fiber networks.

[0057] YOLO (You Only Look Once) is a real-time object detection algorithm that can detect objects in an image with a single forward pass of the neural network. YOLO can recognize multiple objects at the same time and predict the category of each object and its location (bounding box) in the image to achieve fast and efficient object detection. Its main feature is fast detection speed and is suitable for real-time applications.

[0058] Mask R-CNN (Mask Region-Based Convolutional Neural Networks) is an object detection and image segmentation algorithm, which is an improvement on the classic Faster R-CNN. Mask R-CNN can not only detect objects in an image, but also generate pixel-level segmentation masks for each object, thus achieving more sophisticated image analysis. It excels in accuracy and is suitable for scenarios that require more detailed object information, such as autonomous driving and medical image analysis.

[0059] See also Figure 1 The embodiment of the present invention provides an optical cable early warning method, comprising:

[0060] S100, acquiring monitoring data, and predicting risk events and risk points based on the monitoring data and the trained cross-modal large model; the monitoring data includes any one or more of vibration data, image data or audio data.

[0061] The type of monitoring data is determined according to the actual application, and prediction can be made using data from one or two modes or data from more than two modes.

[0062] In a specific embodiment, in a certain location range, if only vibration data captured by distributed optical fiber can be obtained, the vibration data is input into the trained cross-modal large model. If a risk event, such as construction excavation, is predicted, then if the same prediction result is reached more than three times, it can be preliminarily confirmed that a risk event of construction excavation has occurred at the location. For example: when the number exceeds the set 3 times, a manual search is notified, and after the search, the personnel feedback information: construction is within the enclosure 20 meters away from the line, which does not affect the line patrol, and colleagues are aware of it.

[0063] In a specific embodiment, in a certain position range, multimodal environmental data such as vibration data and image data of the current position can be obtained in real time. These multimodal data are input into the trained cross-modal large model for real-time prediction. The number of predictions is calculated based on the predicted results. If the number of predictions of the same result reaches the set threshold upper limit, such as more than 2 times, then it can be confirmed that a risk event has occurred at the location.

[0064] S200, using an optical time domain reflectometer to detect fault points and fault events of the optical cable.

[0065] Optical Time Domain Reflectometry (OTDR) technology is an effective method for detecting fiber line faults and performance. By analyzing the transmission and reflection characteristics of light in optical fibers, OTDR can accurately locate fault points in optical cables, identify fiber attenuation, breakage, and connection problems, and provide key support for optical cable maintenance and repair.

[0066] OTDR technology is based on the reflection and scattering of light in optical fibers. The system sends a short pulse light signal into the optical fiber. When the light signal propagates in the optical fiber, it will be reflected and scattered at the discontinuous points of the optical fiber (such as break points, joints, and damaged points). The system measures the time and intensity of the reflected light signal to analyze the loss and fault conditions in the optical fiber, thereby accurately locating the fault point. In addition, OTDR can eliminate random noise, improve detection accuracy, and ensure data reliability through multiple measurements and data averaging. The working principle of OTDR is similar to that of radar. Through the time difference and intensity changes of the reflected signal, the system can draw a loss distribution map of the optical fiber and identify abnormal conditions in the optical fiber.

[0067] In a specific embodiment, OTDR technology is widely used in the maintenance and fault detection of optical fiber communication networks. For example, during the laying and maintenance of optical cables, the system can use OTDR equipment to test the optical cables, detect the connection quality and loss of the optical fibers, and ensure the transmission performance of the optical cables. When an optical cable fails, the system quickly detects the fault point through the OTDR equipment, determines the nature and location of the fault, and provides an important reference for the repair work. For example, the OTDR equipment can detect faults such as breakage, bending, and joint damage of the optical fiber. By accurately locating the fault point, the operation and maintenance personnel can quickly carry out repair work and reduce the impact of optical cable faults on the communication network. OTDR technology can also be used for regular inspection and preventive maintenance of optical fibers. By regularly measuring the attenuation characteristics of optical fibers, the system can detect potential fault hazards in advance, take preventive measures, and ensure the long-term reliable operation of optical cables.

[0068] S300. If the risk point matches the fault point, the risk level is evaluated according to the risk event and the fault event, and an early warning is issued according to the risk level.

[0069] If the risk point and the fault point can match each other, that is, the risk point and the fault point can correspond, and the two different methods can determine that the optical cable has a transmission problem at the same location, it means that there is a high probability that the optical cable has a problem at this location. The risk level of the optical cable system is comprehensively evaluated based on risk events and other related events such as fault events, and early warnings are issued according to different risk levels.

[0070] By analyzing the attenuation changes and fault history of optical fibers, the system can predict the location and time of possible optical fiber failures, take maintenance measures in advance, and reduce the occurrence of failures. The system can also combine environmental data and construction activity information to comprehensively evaluate the risk level of optical cables. If the test results show that the fault point is not enough to affect the normal operation of the optical cable or optical fiber, it will issue an early warning signal and arrange maintenance. For example, in areas with frequent construction, the system can increase the OTDR detection frequency, monitor the status of the optical cable in real time, and detect and handle faults in a timely manner. Through OTDR technology and early warning analysis, the system can achieve efficient maintenance and management of optical cables and ensure the stable operation of the optical fiber communication network.

[0071] Optionally, see Figure 2 , the vibration sample data is obtained by the following method:

[0072] S011. Acquire an initial vibration sample signal through an optical fiber vibration sensor.

[0073] In a specific embodiment, the vibration signal x(t) passes through the optical fiber vibration sensor system of the linear Sagnac structure, and the output signal s(t,pos)=A(t)×x(t)×δ(pos-pos0(t)), where A(t)x(t) represents the time-dependent vibration amplitude, δ is the Dirac function, and pos0(t) is the position function of the vibration signal, thereby locating the exact position and time of the vibration signal, providing accurate data for further signal processing. The linear Sagnac structure is an optical structure. Compared with other distributed sensing structures, the vibration sensing system based on the Sagnac structure has a series of advantages such as simple structure, long sensing distance, stable polarization, and low cost. In addition, its light source and detector are located on the same side, which is convenient for cabling.

[0074] S012: Perform a second preprocessing on the initial vibration sample signal, and perform a recovery process on the initial vibration sample signal whose signal-to-noise ratio is lower than the first threshold to obtain vibration sample data.

[0075] It should be noted that the second preprocessing and recovery processing method and the first threshold value are determined according to actual applications, and are not specifically limited in this embodiment. Specifically, the collected vibration signal data is transmitted to the TM-UNe model, so as to reduce the noise of the vibration signal data to improve the signal-to-noise ratio, and mark the vibration signal data with a signal-to-noise ratio lower than the set threshold; and the marked vibration signal data with a signal-to-noise ratio lower than the set threshold is restored by support vector machine (SVM) technology.

[0076] In a specific embodiment, the vibration signal is processed using the TM-UNe model, s′(t, pos)=TM-UNe(s(t, pos)), and the signal whose signal-to-noise ratio SNR(s′(t, pos)) after processing is still lower than the threshold θ is marked, then the segment data is set as s l ' ow (t,pos), reduces background noise through advanced machine learning models and improves the signal-to-noise ratio of the data for subsequent processing. Among them, the TM-UNe model is a deep learning-based model used in the field of image and signal processing, especially in noise reduction and signal-to-noise ratio improvement. It can be understood that the signal-to-noise ratio (SVR) is a metric used to compare the level of the desired signal with the level of the background noise, usually expressed in decibels (dB). A high signal-to-noise ratio means that the signal is stronger than the noise, which is an important indicator of data quality. The support vector machine (SVM) model is used to recover the data with low signal-to-noise ratio, s″(t,pos)=SVM(s l ' ow(t,pos)), where SVM is a recovery function learned from training data, which aims to maximize the availability of recovered data. For data with low signal-to-noise ratio after processing, SVM technology is used to try to restore data quality, improve data availability and reliability, and ensure that all valuable monitoring data are used to the maximum extent.

[0077] See also Figure 3 , install cameras within a certain range of the optical cable to obtain audio and video data. The initial parameters are: set the coverage of a camera to 5,000 square meters, with a length and width of 100 and 50 meters respectively. The area where the optical cable passes, including urban buildings, suburbs and traffic areas, is set to set a camera every 1 kilometer. Underground signal: receive vibration signals from the area where the optical cable passes through through optical fiber; aboveground signal: capture image information of the area where the optical cable passes through through the camera.

[0078] The system collects environmental images along the optical cable through front-end monitoring devices and uses advanced image recognition technology to analyze these images. In order to ensure the accuracy of recognition, deep learning models such as YOLO (You Only Look Once) and Mask R-CNN (Mask Region-Based Convolutional Neural Networks) can be used. These models can accurately classify and locate objects in the picture. For example, in a park environment, the model can identify features such as trees, grass, and amusement facilities; in a road environment, the model can identify features such as vehicles, road signs, and traffic lights; at a construction site, the model can identify features such as excavators, building materials, and construction workers.

[0079] Optionally, see Figure 4 , the image sample data is obtained by the following method:

[0080] S021. Acquire ground image data within the range of the optical cable, and perform third preprocessing on the ground image data.

[0081] The front-end monitoring device obtains environmental images along the optical cable, which are taken regularly or in real time along the optical cable line. The third preprocessing includes but is not limited to size adjustment, denoising, etc. to improve image quality and subsequent recognition accuracy.

[0082] S022. Classify and locate the ground image data after the third preprocessing to obtain image sample data.

[0083] It should be noted that the specific method of classification and positioning is determined according to actual application and is not specifically limited in this embodiment.

[0084] In a specific embodiment, the YOLO model is used to perform fast object detection on the preprocessed image to generate the bounding box and classification label of the object. The input is the preprocessed image, and the output is the object bounding box and classification label, such as trees, vehicles, excavators, etc. The YOLO model generates the bounding box and classification label of the object by quickly locating and classifying the objects in the image. Subsequently, the Mask R-CNN model is used to perform fine segmentation on the detected objects to generate a segmentation mask for each object. The input is the object bounding box and classification label as well as the preprocessed image, and the output is the object segmentation mask. The specific conversion process includes using the Mask R-CNN model to perform fine segmentation on the object and generate a segmentation mask for the object.

[0085] By matching the identified objects and their locations with a pre-set environmental feature library, the system can accurately determine the type of environment in the image. The input is the object classification label and segmentation mask, and the output is the environment type, including specific environment classifications such as parks, roads, and construction sites. The environment type data is then combined with the GPS positioning data to mark the environment type of the optical cable location and store it in a database. The input is the environment type and the GPS positioning data of the image, and the output is an environment type database that records the environment type and related information for each section of the optical cable location.

[0086] Through the above steps, the system can collect and analyze environmental images along the optical cable in real time, use deep learning models for image recognition and classification, accurately determine the type of environment around the optical cable, and establish an environmental type database. This provides reliable basic data for optical cable risk assessment and maintenance, and significantly improves the intelligence level and response speed of the system.

[0087] The data collected in the above steps are integrated together to construct a multimodal environmental dataset containing vibration data, image data, and audio data, and stored in a structured database to provide support for future risk assessment and decision-making.

[0088] Optionally, see Figure 5 , the cross-modal large model is trained by the following method:

[0089] S031. Acquire monitoring sample data, perform first preprocessing on the monitoring sample data, and add labels; the monitoring sample data includes any one or more of vibration sample data, image sample data, or audio sample data.

[0090] The vibration signal data and image data in each group of cross-modal historical data are respectively subjected to a first preprocessing, including denoising, feature extraction and data alignment; a risk analysis is performed on the risk event types corresponding to the collected cross-modal historical data, and a risk event type label is added to the cross-modal historical data according to each risk event type.

[0091] S032. The monitoring sample data is fused according to the labels to obtain a training sample set, and the cross-modal large model is trained according to the training sample set.

[0092] The preprocessed data of different modalities are subjected to feature extraction. The feature vectors of the extracted data of different modalities are fused according to the labels. The feature vectors of different modalities at the same time are integrated together through multimodal fusion technology to form an integrated feature vector. Data of different modalities can provide rich feature information, including vibration, image and other data types, making the feature vector more representative and helpful to discover potential fault characteristics.

[0093] In a specific embodiment, the collected vibration signal data and image data are obtained, and the risk events corresponding to the collected cross-modal historical data are classified. According to each risk event type, the risk event type label is added to the cross-modal historical data, which helps to find the relationship between multiple feature data and risk event types. For example, during construction excavation, the original optical cable surface layer and the interior have been partially damaged, so when collecting multi-mode data, it is input into the multi-modal model for extraction and analysis, and it is found that the vibration signal generated by the construction is inconsistent with the previous detection vibration signal of the area. The original optical time domain reflection technology detected a weak fault signal, but it was not enough to completely confirm that the point was a fault point. Therefore, further detection of the construction vibration signal can detect that a part of the optical cable is damaged. Therefore, extracting multi-modal data and linking it with multiple fault types facilitates better judgment of the degree of damage to the optical cable. The pre-processed different modal data are feature extracted and integrated. When training the model, the training model process takes into account the impact of multiple modal data on the prediction results of risk events more comprehensively, which can improve the accuracy and precision of risk event prediction and better support predictive maintenance and early warning.

[0094] Optionally, see Figure 6 , the optical cable early warning method also includes:

[0095] S410: If risk events and risk points cannot be predicted based on monitoring data and the trained cross-modal big model due to missing data, the missing data segments are determined.

[0096] In the real-time prediction process, there may be two types of input data: the first is image data or vibration data that the cross-modal model has not learned; the second is mixed vibration data. For example, after the vibration signal is denoised and repaired, it is still difficult to identify the vibration signal, and there is a lack of corresponding image data as support. When the confidence of the multimodal prediction result of the input data is lower than the set threshold, the data is marked as a missing key environmental data fragment, and the marked data is extracted for subsequent processing.

[0097] For example, when a vibration signal detects the passing of a large truck, if the vibration signal fails to match the image information, that is, the image data corresponding to the vibration signal is not obtained, it can be marked as missing information. If the amount of such missing information is large, and an abnormal vibration signal is detected at the same time (such as matching the characteristics of a risk event), the vibration signal should also be marked as key information. This shows that the missing image information is also key missing environmental information, and its existence needs to be further verified. Another type of key missing information is that the detected vibration signal cannot be identified, that is, the signal is too mixed and effective features cannot be extracted from it.

[0098] S420: Search for similar data segments in a database according to the characteristics of the missing data segment and the first preset weight, and supplement the missing data segment according to the confidence of the similar data segment.

[0099] It should be noted that the output of the cross-modal model usually includes a confidence score, which indicates the prediction accuracy of the cross-modal model.

[0100] In order to capture the features of the missing data fragment through the cross-modal model, it is necessary to extract relevant data from the multimodal environmental data set. These data come from the same location at different times, the same time period of adjacent locations, or historical data with similar features. Then, the existing data fragments s1 or s2 or sk (if it is unknown data) of the missing data fragment are inferred based on the relevant data to capture similar features. If the feasibility of the reasoning reaches the set range, more useful data fragments can be searched from the relevant data to infer the missing key environmental information fragment. If the existing data cannot directly infer the existing data fragment of the fragment, data fusion reasoning is required. The data fusion process requires selecting relevant data from the same time period of adjacent locations or historical data with similar features. By performing fusion reasoning on these relevant data, an attempt is made to infer the existing data fragments. If the fused data can infer the missing data fragment S, the missing fragment information can also be restored based on the fused data. If a credible result cannot be obtained after multiple reasonings (that is, the number of reasonings reaches the set range), the data fragment at the location is expanded, and the data is re-extracted to continue reasoning until the credibility of the reasoning result reaches the preset threshold.

[0101] In a specific embodiment, see Figure 7First, the cross-modal model marks the data segment that is missing key environmental information. The length of the data segment is S. Then, by identifying the information of this data segment, extract S1, S2, .., Si, ..., SK, and analyze the features of these segments S1, S2, .., Si, ..., SK. Then, based on these features, use the cross-modal retrieval model to retrieve features similar to these segments. The features include but are not limited to time points, geographic locations, vibration signals, and image signals. Similar features have different weights. For example, the similarity calculation formula is as follows:

[0102] Similarity = a1*time proximity + b1*geographical proximity + c1*vibration signal fitting degree (similarity) + d1*image information similarity. This formula can also automatically adjust the weights of a1, b1, c1, and d1 later (accuracy will be set based on the accuracy of the cross-modal model inference, with a score of 1 to 10).

[0103] When retrieval can be used to supplement key data, the least and most relevant data is preferred, whether it is historical data or recent data. In the process of finding similar feature data, cross-modal retrieval will be used to handle unknown modal data and low-confidence cross-modal retrieval. Retrieval techniques include: searching for similar feature data through a cross-modal retrieval model, combined with the following retrieval techniques:

[0104] Multi-level similarity matching: not only matching similar data in a single mode, but also comprehensively considering the similarity of multi-modal data. For example, historical data that matches both vibration signals and images may be more valuable for reference. A multi-level similarity matching strategy can be used to perform hierarchical retrieval of spatial features (such as similar geographical locations) and temporal features (such as similar time periods) to ensure high relevance of retrieval results.

[0105] Spatial proximity retrieval: Based on geospatial data (such as GPS coordinates), retrieve multimodal data of monitoring points adjacent to the current detection point. These data usually have similar environmental and risk characteristics and can be used to supplement the insufficient data of the current point.

[0106] Time proximity analysis: Considering the temporal nature of events, historical data close to the current time point is used to supplement. For example, if the currently detected vibration signal is weak, the strong vibration signal at the same location in the adjacent time period can be used as a reference to assist in determining the nature of the risk event.

[0107] There are two inference processes for determining missing data:

[0108] Reasoning process 1: Identify the missing key environmental data in the data segment, and then use the above-mentioned data retrieval technology to find the most reasonable data in a certain dimension through the similarity calculation formula to supplement it. The data found is not fused but used directly. The segment of the supplemented data is used as input and passed into the cross-modal model (X-VILA model). If the confidence carried by the model prediction result is lower than the set threshold, it means that the data searched out without input data is not accurate enough and needs to be fused, then reasoning process 2 is performed.

[0109] Reasoning process 2: Through the fusion process, the data of the missing segment is newly supplemented and predicted again through the cross-modal model. If the confidence carried by the result is lower than the set threshold, the missing data fusion process needs to be repeated. The number of repetitions is set to 100 or 150 times. If the confidence is higher than the set threshold within 100 times, it stops.

[0110] For example: suppose an optical cable line is being monitored, and a construction excavation occurs during the process, resulting in possible damage to the optical cable. During the monitoring process, an abnormal vibration signal is detected within a certain period of time, but the image data is missing and the risk cannot be accurately determined. The cross-modal model first identifies the missing data segment. Assuming the data length is S, the features of this data segment may include vibration signals and timestamps. Next, the system extracts relevant fragments S1, S2, .., Si, ..., SK from the historical data. These fragments may contain vibration and image signal features of the same location under similar circumstances (such as during construction) before, and are retrieved and screened using the above similarity calculation formula.

[0111] See also Figure 8 , assuming that the data segments retrieved by the system include: vibration signal x1(t) and image s1(t) (related to construction, close in time), vibration signal x2(t) and image s2(t) (related to other construction, slightly farther in time). Select the data segments with the highest similarity (such as x1(t) and s1(t)) and directly pass them into the cross-modal model (X-VILA model) as input, and the model will output the prediction results. Assuming that the confidence of the model output is lower than the set threshold (such as 70%), it indicates that the supplementary data is not accurate enough, and then enter the reasoning process 2. In this process, the model needs to fuse the currently detected vibration signal with the retrieved vibration signal and image. Assuming that after the fusion process, new data is generated: new vibration signal xf(t), new image data sf(t). The fused new data xf(t) and sf(t) are passed into the cross-modal model for re-prediction. If the confidence obtained again is still lower than the set threshold, the system will repeat the fusion process until the confidence is higher than the set value or the maximum number of iterations (such as 100) is reached.

[0112] After completing the information, all positions are shielded in order. That is, if the position shields the image data at that position, and after shielding, the position is filled with environmental information. For the filled position, the risk events and risk points are predicted through the cross-modal model. The prediction results and confidence levels are compared with those before shielding. If they are within the threshold, the camera at that position can be cancelled.

[0113] The key environmental data of each position is filled, so that each position can fully observe the situation of each area, and can also predict risk events and risk points through the cross-modal model. At the same time, the position where the key information has been filled is marked, and the label is temporarily named xi (x1, x2, ...). These positions do not need to be operated next, that is, image information shielding. Image information shielding is an operation performed on the unmarked area, that is, the image information of the area is not involved in the multimodal prediction. And whether this shielded area participates in the part of the information supplement of the marked xi, if so, the area does not participate in the shielding. In this case, the marked xi area and its associated areas are all removed, and the image information of the unmarked area is shielded, and the shielding is calculated for one position alone, not two or more shieldings. After that, the shielded position lacks the image data captured by the camera. The key environmental data of this position is filled, and the position involved in the filled data is marked with xi. If it has been previously, it does not need to be marked again. Then, the risk events and risk points are predicted through the cross-modal model, and the confidence value attached to the predicted result is compared with the confidence value before the image information is masked. If the confidence value after the two is less than the established threshold, the information needs to be supplemented again until the difference in the confidence value of the two is within the established threshold after multiple cycles.

[0114] In a specific embodiment, in an urban construction area, it is assumed that an optical cable line is monitored. A camera was previously installed at a certain location (location A) to obtain real-time image data. However, due to construction activities, the image data of this location is missing. It is hoped that the missing data can be filled through environmental information, and whether the camera can be cancelled can be evaluated. The system completes the key environmental data of location A, and the system decides to shield the image information of location A. The shielding operation is as follows: mark the location, the system marks location A as xi (for example, x1); fill the image data, after completion, location A no longer relies on the original image data, but uses the supplemented environmental information to predict risk events and risk points through a cross-modal model. Shield other locations, check other locations associated with location A (such as locations B and C), if these locations are not marked as xi and are not involved in the information supplement of location A, then shield their image information. The system predicts the risk of location A through a cross-modal model and generates prediction results and confidence. Assume that the prediction results are as follows: The prediction result shows that the probability of a risk event occurring is 80% and the confidence is 70%. Compare the confidence of the new prediction result (70%) with the confidence when the image information is not shielded (assuming it is 75%). If the difference is less than the set threshold (for example, 10%), the camera at position A can be considered for cancellation. If the difference between the confidence before and after shielding is within the set range (for example, 10%), it means that the system can still effectively predict risks even in the absence of image data, and the system can decide to cancel the camera at position A. If the confidence difference exceeds the set threshold (for example, the confidence drops to 60%), the system will supplement the information again and re-predict the risk until the conditions are met or the maximum number of attempts is reached.

[0115] When there is insufficient information at a certain location, distributed reasoning and collaborative analysis can be performed through fused data to complete the missing key parts of the data fragment, thereby improving the accuracy of risk and fault identification.

[0116] Optionally, see Fig. 9 , the optical cable early warning method also includes:

[0117] S510. Predict risk events and risk points based on the supplemented monitoring data and the trained cross-modal large model.

[0118] S520: Obtain on-site detection results. If the risk event or risk point is inconsistent with the on-site detection results, replenish the missing data segments.

[0119] The multimodal environmental data acquired in real time at the current location is used to predict risk events through the trained cross-modal large model. If it is detected that the multimodal environmental data is missing key data during the prediction, the missing key data is supplemented by reasoning through the above method. After that, the results are predicted and uploaded to the server through manual on-site inspection. If the analysis finds that the predicted results are inconsistent with the manual on-site inspection, or the predicted event distance is inconsistent with the manual on-site inspection, the missing data is supplemented again, and the supplemented data is re-predicted through the trained cross-modal large model, and the cross-modal model is updated and improved until the results meet the manual inspection results.

[0120] For example, the prediction result is inconsistent with the manual on-site inspection. The line patrol personnel just passed by (Zhongshan Road East #004) and found no signs of construction nearby. There was no construction within 200 meters of the site. For another example, the predicted event distance is inconsistent with the manual on-site inspection. The corrected construction point is 296#, which is within the input construction site range (the input construction site is 296-297#), and about 100 meters away from the alarm point 297#.

[0121] Optionally, see Fig.10 , assess the risk level based on risk events and failure events, including:

[0122] S310, determining a risk score according to the risk event, and determining a fault score according to the fault event;

[0123] S320, determining an optical fiber impact score and a signal impact score;

[0124] S330, assessing the risk level according to the risk score, the fault score, the fiber impact score, the signal impact score, and the second preset weight.

[0125] It should be noted that the risk level is determined according to the actual application and is not specifically limited in this embodiment. For example, the risk level is divided into three levels: high, medium and low. The fiber impact score includes but is not limited to the fiber aging degree score and the fiber failure number score, and the signal impact score includes but is not limited to the vibration signal uncertainty score.

[0126] The score of each influencing factor of the risk level is determined according to the degree of influence. For example, the scores of construction intensity and places gnawed by rats are high; the scores of many vehicles on the road are medium; the scores of many people walking in the park are low. The higher the assessment score, the greater the risk probability of damage to the optical cable and the higher the risk level. For example, the risk score is determined according to the type of risk event, the fault score is determined according to the type of fault event, the fiber aging score is determined according to the use time of the optical fiber and the results of manual troubleshooting, the vibration signal uncertainty score is determined according to the ratio of the first vibration signal to the recorded vibration signal, and the fiber fault number score is determined according to the number of fiber faults recorded by the system.

[0127] In a specific embodiment, the risk level is calculated as follows: risk level = a2*risk score+b2*fault score+c2*fiber aging score+d2*fiber fault number score+e2*vibration signal uncertainty score. Wherein, a2, b2, c2, d2 and e2 are determined according to actual applications.

[0128] In a specific embodiment, first, the system evaluates the environmental risk of the optical cable location based on the environmental image and vibration data. For example, in a park environment, if a large amount of natural sounds and little construction noise are detected, the system will evaluate the area as a low-risk area; in a road environment, if a large amount of traffic noise is detected, the system will evaluate the area as a medium-risk area; in a construction site environment, if a large amount of construction noise and mechanical equipment are detected, the system will evaluate the area as a high-risk area. Next, the system will evaluate the fault risk of the optical cable based on the OTDR fault point data. For example, if the OTDR data shows that the optical cable has a significant signal reflection change at a certain location, and there are construction equipment and construction noise near the fault point, the system will evaluate the fault point as a high-risk area. By combining environmental risks and fault risks, the system ultimately determines the comprehensive risk level of the optical cable. During the risk assessment process, the system will also combine historical data for trend analysis. For example, by analyzing the data of the past few months, the system can identify the frequency and intensity of construction activities in certain areas, thereby predicting future construction risks. This trend analysis can help the system to warn in advance and improve the safety of the optical cable.

[0129] Finally, the system will generate a detailed risk assessment report for reference by relevant personnel. The report includes the specific analysis process and results of various types of data, as well as the comprehensive risk level of the optical cable and recommended maintenance measures. For example, for low-risk optical cable locations, the system recommends regular monitoring and maintenance; for medium-risk optical cable locations, the system recommends increasing the density of monitoring equipment and collecting data in real time; for high-risk optical cable locations, the system recommends key monitoring, timely warning and maintenance. In this way, the system can ensure the safe operation of the optical cable and provide comprehensive data support and decision-making basis for subsequent warning and maintenance.

[0130] After completing the risk assessment, the system will provide the generated risk assessment report to the relevant personnel. The web and mobile clients are very important in receiving and processing alarm information in real time. For potential damage detected, the system needs to quickly assess its urgency and authenticity, and decide whether to immediately dispatch personnel to inspect and repair. The web terminal displays relevant pictures in real time and configures image acquisition parameters to ensure data accuracy. When potential damage is detected, the system needs to quickly assess its urgency and authenticity, and quickly decide whether to dispatch personnel to the site for further inspection and repair. By instantly pushing alarm information to the person in charge's mobile client, the team can receive timely notifications and respond quickly, effectively avoiding potential losses.

[0131] The mobile client supports real-time reception and viewing of alarm information, provides one-click navigation to the construction point or fault point, and can take photos and record on-site evidence to further enhance real-time handling capabilities. This function not only helps operation and maintenance personnel to quickly and accurately locate problems, but also records on-site details, providing important references for subsequent analysis and processing. In the early warning system, the input data includes image data, vibration data, and OTDR fault point data. Image data is processed by convolutional neural networks to identify construction equipment and personnel activities. Vibration data is processed by convolutional neural networks and long short-term memory networks (LSTM) to identify construction noise and mechanical vibration. OTDR data is processed by convolutional neural networks to identify changes in signal reflection of optical cables. The system evaluates the risk level based on the analysis results of these data and determines the corresponding warning level.

[0132] Implementation of the embodiments of the present invention includes the following beneficial effects: the embodiment first acquires monitoring data, predicts risk events and risk points based on the monitoring data and a trained cross-modal large model, the monitoring data includes any one or more of vibration data, image data or audio data, predicts risk events and risk points based on the multimodal monitoring data and the cross-modal large model, and improves the accuracy of the prediction results. Then, an optical time domain reflectometer is used to detect the fault point and fault event of the optical cable, and the fault of the optical cable is detected by another method. If the risk point matches the fault point, the risk level is evaluated based on the risk event and the fault event, and an early warning is issued based on the risk level. A comprehensive analysis is performed through the prediction results of the monitoring data and the detection results of the optical time domain reflectometer to improve the accuracy of optical cable monitoring and early warning, while improving efficiency, reducing human intervention and maintenance costs, and ensuring the stable operation of the communication network.

[0133] like Fig.11 As shown, an embodiment of the present invention provides an optical cable early warning system, comprising:

[0134] The first module is used to obtain monitoring data and predict risk events and risk points based on the monitoring data and the trained cross-modal large model; the monitoring data includes any one or more of vibration data, image data or audio data;

[0135] The second module is used to detect the fault point and fault event of the optical cable using an optical time domain reflectometer;

[0136] The third module is used to evaluate the risk level according to the risk events and fault events if the risk point matches the fault point, and issue an early warning based on the risk level.

[0137] It can be seen that the contents of the above method embodiments are all applicable to the present system embodiments, the functions specifically implemented by the present system embodiments are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0138] like Fig.12 As shown, an embodiment of the present invention further provides an optical cable early warning device, comprising:

[0139] at least one processor;

[0140] at least one memory for storing at least one program;

[0141] When at least one program is executed by at least one processor, the at least one processor implements the optical cable early warning method steps described in the above method embodiment.

[0142] Among them, the memory, as a non-transient computer-readable storage medium, can be used to store non-transient software programs and non-transient computer executable programs. The memory may include a high-speed random access memory, and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage device. In some embodiments, the memory may optionally include a remote memory remotely arranged relative to the processor, and these remote memories may be connected to the processor via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0143] It can be seen that the contents of the above method embodiments are all applicable to the present device embodiments, the functions specifically implemented by the present device embodiments are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0144] In addition, the embodiment of the present application also discloses a computer program product or a computer program, and the computer program product or the computer program is stored in a computer-readable storage medium. The processor of the computer device can read the computer program from the computer-readable storage medium, and the processor executes the computer program, so that the computer device performs the above method. Similarly, the contents in the above method embodiment are all applicable to the storage medium embodiment, and the functions specifically implemented by the storage medium embodiment are the same as those in the above method embodiment, and the beneficial effects achieved are also the same as those achieved by the above method embodiment.

[0145] An embodiment of the present invention further provides a computer-readable storage medium, which stores a program executable by a processor. The program executable by the processor is used to implement the above method when executed by the processor.

[0146] It is understood that all or some steps and systems in the disclosed method above can be implemented as software, firmware, hardware and appropriate combinations thereof. Some physical components or all physical components can be implemented as software executed by a processor, such as a central processing unit, a digital signal processor or a microprocessor, or implemented as hardware, or implemented as an integrated circuit, such as an application-specific integrated circuit. Such software can be distributed on a computer-readable medium, and the computer-readable medium can include a computer storage medium (or a non-transitory medium) and a communication medium (or a temporary medium). As known to those of ordinary skill in the art, the term computer storage medium includes volatile and non-volatile, removable and non-removable media implemented in any method or technology for storing information (such as computer-readable instructions, data structures, program modules or other data). Computer storage media include, but are not limited to, RAM, ROM, EEPROM, flash memory or other memory technology, CD-ROM, digital versatile disk (DVD) or other optical disk storage, magnetic cassette, magnetic tape, disk storage or other magnetic storage device, or any other medium that can be used to store desired information and can be accessed by a computer. Furthermore, it is well known to those skilled in the art that communication media typically embodies computer readable instructions, data structures, program modules, or other data in a modulated data signal such as a carrier wave or other transport mechanism, and may include any information delivery media.

[0147] It should be understood that in the present application, "at least one (item)" means one or more, and "plurality" means two or more. "And / or" is used to describe the association relationship of associated objects, indicating that three relationships may exist. For example, "A and / or B" can mean: only A exists, only B exists, and A and B exist at the same time, where A and B can be singular or plural. The character " / " generally indicates that the objects associated before and after are in an "or" relationship. "At least one of the following" or similar expressions refers to any combination of these items, including any combination of single or plural items. For example, at least one of a, b or c can mean: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, c can be single or multiple.

[0148] In the several embodiments provided in the present application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or units, which can be electrical, mechanical or other forms.

[0149] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0150] The above is a specific description of the preferred implementation of the present invention, but the invention is not limited to the embodiments. Those skilled in the art may make various equivalent modifications or substitutions without violating the spirit of the present invention. These equivalent modifications or substitutions are all included in the scope defined by the claims of this application.

Claims

1. An optical cable early warning method, characterized in that: include: Acquire monitoring data, and predict risk events and risk points based on the monitoring data and the trained cross-modal large model; the monitoring data includes any one or more of vibration data, image data or audio data; Use optical time domain reflectometer to detect fault points and fault events of optical cables; If the risk point matches the fault point, the risk level is evaluated according to the risk event and the fault event, and an early warning is issued according to the risk level.

2. The method according to claim 1, characterized in that The cross-modal large model is trained by the following method: Acquire monitoring sample data, perform first preprocessing on the monitoring sample data, and add labels; the monitoring sample data includes any one or more of vibration sample data, image sample data, or audio sample data; The monitoring sample data are fused according to the labels to obtain a training sample set, and the cross-modal large model is trained according to the training sample set.

3. The method according to claim 2, characterized in that The vibration sample data is obtained by the following method: Acquire an initial vibration sample signal through an optical fiber vibration sensor; The initial vibration sample signal is subjected to a second preprocessing, and the initial vibration sample signal having a signal-to-noise ratio lower than a first threshold is subjected to a recovery process to obtain vibration sample data.

4. The method according to claim 2, characterized in that: The image sample data is obtained by the following method: Acquiring ground image data within the range of the optical cable, and performing a third preprocessing on the ground image data; The ground image data after the third preprocessing is classified and located to obtain image sample data.

5. The method according to claim 1, characterized in that The method further comprises: If risk events and risk points cannot be predicted based on monitoring data and trained cross-modal big models due to missing data, identify the missing data segments; According to the characteristics of the missing data segment and the first preset weight, similar data segments are searched in the database, and the missing data segment is supplemented according to the confidence of the similar data segment.

6. The method according to claim 5, characterized in that The method further comprises: Predict risk events and risk points based on supplemented monitoring data and trained cross-modal large models; Obtain on-site detection results. If the risk event or the risk point is inconsistent with the on-site detection results, replenish the missing data segments.

7. The method according to claim 1, characterized in that The assessing the risk level according to the risk event and the failure event includes: Determine a risk score according to the risk event, and determine a failure score according to the failure event; Determine the fiber impact score and the signal impact score; The risk level is evaluated according to the risk score, the fault score, the fiber impact score, the signal impact score and a second preset weight.

8. An optical cable early warning system, characterized in that: include: The first module is used to obtain monitoring data and predict risk events and risk points based on the monitoring data and the trained cross-modal large model; the monitoring data includes any one or more of vibration data, image data or audio data; The second module is used to detect the fault point and fault event of the optical cable using an optical time domain reflectometer; The third module is used to evaluate the risk level according to the risk event and the fault event if the risk point matches the fault point, and issue an early warning according to the risk level.

9. An optical cable early warning device, characterized in that: include: at least one processor; at least one memory for storing at least one program; When the at least one program is executed by the at least one processor, the at least one processor implements the method according to any one of claims 1 to 7.

10. A computer-readable storage medium storing a program executable by a processor, characterized in that: The processor-executable program is used to perform the method according to any one of claims 1 to 7 when executed by the processor.

Images