Road disaster monitoring method, system and equipment based on distributed optical fiber sensing and medium

Abstract

The invention relates to a road disaster monitoring method, system and device based on distributed optical fiber sensing and a medium. The method comprises the following steps: acquiring an original scattered light time sequence signal of a road; performing denoising processing on the original scattered light time sequence signal to obtain a high-quality sensing signal; performing feature extraction on the high-quality sensing signal to obtain a disaster feature vector; based on a preset disaster signal database, analyzing events of disasters on the road to obtain a disaster event set; based on the disaster feature vector and a preset environmental risk factor, performing time-space correlation analysis processing on the disaster event set to obtain a disaster comprehensive feature vector; and based on the disaster comprehensive feature vector and a preset disaster monitoring grade division threshold set, comprehensively analyzing disaster risks existing in the road, and obtaining a disaster analysis and evaluation report. By adopting the method, the perspectiveness and the accuracy of safety monitoring can be improved.

Description

Technical Field

[0001] This invention belongs to the field of highway safety monitoring, and in particular relates to a method, system, equipment and medium for highway disaster monitoring based on distributed optical fiber sensing. Background Technology

[0002] With the development of highway safety monitoring technology, highway disaster monitoring technology based on distributed optical fiber sensing has emerged. This technology can use a single sensing optical fiber laid along the highway to achieve continuous and real-time sensing of physical parameters such as vibration, strain, and temperature over a distance of tens of kilometers, thus giving rise to online monitoring methods with optical fiber as the nerve.

[0003] In existing online monitoring methods that use fiber optics as the "nerve," the typical processing procedure is as follows: Scattered light signals along the sensing fiber are acquired using demodulation equipment and converted into continuous data reflecting changes in physical quantities. The system typically sets fixed amplitude or frequency thresholds for various disaster signals based on experience. When real-time data exceeds a specific threshold, an alarm is triggered. This process mainly relies on direct judgment of the signal's surface characteristics and matching with simple rules.

[0004] However, existing online monitoring methods based on fiber optics have significant problems: the highway environment is complex, and vibrations generated by normal activities such as vehicle traffic and mechanical construction overlap with the characteristics of disaster precursor signals, making it easy to trigger false alarms by relying solely on fixed thresholds; more importantly, for disasters with hidden and slowly evolving characteristics such as underground cavities and seepage, their early signs are extremely weak signal disturbances, completely submerged in environmental noise, which traditional methods cannot effectively extract and identify, resulting in the system only being able to alarm when a disaster is approaching or after it has occurred, failing to achieve prior risk warning and trend prediction, leading to insufficient foresight and accuracy in safety monitoring. Summary of the Invention

[0005] Therefore, it is necessary to provide a highway disaster monitoring method, system, equipment, and medium based on distributed optical fiber sensing that can improve the foresight and accuracy of safety monitoring, addressing the aforementioned technical problems.

[0006] Acquire the original scattered light time-series signal of the highway, which includes the acquisition location label and acquisition time label;

[0007] The original scattered light time-series signal is denoised to obtain a high-quality sensing signal.

[0008] Feature extraction is performed on high-quality sensing signals to obtain disaster feature vectors; and based on a pre-set disaster signal database, events indicating the presence of disasters on highways are analyzed to obtain a disaster event set; the disaster signal database includes various disaster types and their corresponding disaster feature vectors; the disaster event set is used to indicate the disaster events present on highways; each disaster event includes a disaster type, a data collection location label, and a data collection time label;

[0009] Based on disaster feature vectors and preset environmental risk factors, spatiotemporal correlation analysis is performed on the disaster event set to obtain a comprehensive disaster feature vector. The comprehensive disaster feature vector includes disaster type components, disaster correlation components, and disaster development trend feature components.

[0010] Based on the comprehensive disaster feature vector and the preset disaster monitoring level classification threshold set, the disaster risks existing in the highway are comprehensively analyzed to obtain a disaster analysis and assessment report; the disaster monitoring level classification threshold set is used to indicate the monitoring level of the highway determined according to the comprehensive disaster feature vector.

[0011] Furthermore, the original scattered light time-series signal is denoised to obtain a high-quality sensing signal, including:

[0012] Modal decomposition is performed on the original scattered light time series signal to obtain the intrinsic mode components and residual terms of the original scattered light time series signal;

[0013] The correlation coefficients between the intrinsic mode components and the original scattered light time-series signal are calculated to obtain the correlation coefficients of the intrinsic mode components;

[0014] Based on the correlation coefficient and a preset correlation coefficient threshold, the intrinsic mode components with a correlation coefficient greater than the correlation coefficient threshold are selected as the dominant signal components of the original scattered light time series signal.

[0015] Based on the wavelet denoising algorithm, the dominant component of the signal is denoised to obtain the denoised dominant component of the signal.

[0016] Based on the dominant component, intrinsic mode component, and residual term of the denoised signal, the original scattered light time-series signal is reconstructed to obtain a high-quality sensing signal.

[0017] Furthermore, feature extraction is performed on the high-quality sensing signals to obtain disaster feature vectors; and based on a pre-set disaster signal database, events indicating disasters on highways are analyzed to obtain a set of disaster events, including:

[0018] Based on the preset feature extraction length parameters and acquisition location labels, the original scattered light time series signals are sorted to obtain the length distance sequence number of each original scattered light time series signal;

[0019] The original scattered light time-series signals of the same length distance number are combined to obtain a set of signal segments of each length distance number;

[0020] For each length distance index, the set of signal segments is input into a preset feature extraction network to extract the variation features of the original scattered light time-series signal in the set of signal segments, and obtain the disaster feature vector for each length distance index; the feature extraction network is constructed based on the attention mechanism and a one-dimensional convolutional neural network;

[0021] For each length distance index, calculate the distance between the disaster feature vector of the length distance index and the disaster feature vector of each disaster type in the disaster signal database to obtain the vector distance set of each length distance index;

[0022] Based on the vector distance set of each length distance index and the preset risk distance threshold parameter, the disaster events existing on the highway and the disaster types of the disaster events are determined, and an initial disaster event set is obtained;

[0023] Based on the signal segment set with each length distance number and the initial disaster event set, the collection location label and collection time label of each disaster event in the initial disaster event set are determined, and the disaster event set is obtained based on the disaster type, collection location label and collection time label of each disaster event.

[0024] Furthermore, for each length distance index, the distance between the disaster feature vector of the length distance index and the disaster feature vectors of each disaster type in the disaster signal database is calculated, including:

[0025] For each length distance index, calculate the distance between the disaster feature vector of the length distance index and the disaster feature vectors of each disaster type in the disaster signal database. The form of the distance between the disaster feature vector of the length distance index and the disaster feature vectors of each disaster type in the disaster signal database is:

[0026]

[0027] in, It is any distance index. It is an index for any type of disaster. It is the first A disaster feature vector with a distance index of length, Disaster type The disaster feature vector, It is a disaster feature vector The distance between them.

[0028] Furthermore, based on the disaster feature vector and preset environmental risk factors, spatiotemporal correlation analysis is performed on the disaster event set to obtain a comprehensive disaster feature vector, including:

[0029] Based on the collection location label, collection time label and disaster type, the spatiotemporal correlation between disaster events in the disaster event set is quantified to obtain the correlation degree between each disaster event, and based on the correlation degree between each disaster event, the spatiotemporal correlation degree matrix of disaster events is obtained;

[0030] Based on the spatiotemporal correlation matrix of disaster events and the preset correlation threshold, the pairs of disaster events in the disaster event set are combined to obtain disaster type clusters;

[0031] Based on disaster type clusters, the disaster feature vectors corresponding to each disaster event within the disaster type cluster are fused to obtain disaster association components, and disaster type components are obtained based on the disaster type of each disaster event in the disaster type cluster;

[0032] Based on a set of disaster events, a pre-set historical disaster correlation vector database, disaster type clusters, disaster correlation components, and environmental risk factors, the development trend of each disaster event in the disaster type cluster is predicted, and the characteristic components of disaster development trend are obtained. The historical disaster correlation vector database includes historical disaster correlation vectors of historical processes and the collection location labels of historical disaster correlation vectors.

[0033] Based on the disaster correlation component, disaster type component, and disaster development trend characteristic component, a comprehensive disaster feature vector is obtained.

[0034] Furthermore, based on the disaster event set, a pre-defined historical disaster correlation vector database, disaster type clusters, disaster correlation components, and environmental risk factors, the development trend of each disaster event within the disaster type cluster is predicted, yielding disaster development trend characteristic components, including:

[0035] Based on disaster type clusters and disaster event sets, historical disaster association vectors are extracted from the historical disaster association vector database to obtain a sequence of historical disaster association vectors;

[0036] Time series decomposition of historical disaster correlation vector sequences yields long-term trend terms;

[0037] By predicting the long-term trend term, the predicted value of the change trend of the disaster correlation vector is obtained;

[0038] By weighting and fusing the predicted trend values, disaster correlation vectors, and environmental risk factors, the characteristic components of disaster development trends are obtained.

[0039] Furthermore, based on the comprehensive disaster feature vector and the preset disaster monitoring level classification threshold set, a comprehensive analysis of the disaster risks existing in the highway is conducted to obtain a disaster analysis and assessment report, including:

[0040] Based on disaster correlation vectors and disaster development trend feature components, the probability of disasters occurring on highways is quantified to obtain a comprehensive risk index;

[0041] Based on the threshold set for disaster monitoring level classification and the comprehensive risk index, the monitoring level is obtained;

[0042] Based on the comprehensive disaster feature vector and monitoring level, a disaster analysis and assessment report is generated.

[0043] Secondly, this application also provides a highway disaster monitoring system based on distributed optical fiber sensing, comprising:

[0044] The data acquisition module is used to acquire the original scattered light time-series signal of the highway, which includes a collection location label and a collection time label.

[0045] The signal denoising module is used to denoise the original scattered light time-series signal to obtain a high-quality sensing signal;

[0046] The feature extraction module is used to extract features from high-quality sensing signals to obtain disaster feature vectors; and based on a pre-set disaster signal database, it analyzes events indicating the presence of disasters on highways to obtain a disaster event set; the disaster signal database includes various disaster types and their corresponding disaster feature vectors; the disaster event set indicates the disaster events present on highways; the disaster events include disaster type, data collection location label, and data collection time label;

[0047] The feature analysis module is used to perform spatiotemporal correlation analysis on a set of disaster events based on disaster feature vectors and preset environmental risk factors to obtain a comprehensive disaster feature vector. The comprehensive disaster feature vector includes disaster type components, disaster correlation components, and disaster development trend feature components.

[0048] The report generation module is used to comprehensively analyze the disaster risks existing in the highway based on the comprehensive disaster feature vector and the preset disaster monitoring level classification threshold set, and obtain a disaster analysis and assessment report; the disaster monitoring level classification threshold set is used to indicate the monitoring level of the highway determined according to the comprehensive disaster feature vector.

[0049] Thirdly, this application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement any of the highway disaster monitoring methods based on distributed optical fiber sensing of the first aspect of this application.

[0050] Fourthly, this application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements any of the highway disaster monitoring methods based on distributed optical fiber sensing of the first aspect of this application.

[0051] The aforementioned highway disaster monitoring method, system, equipment, and medium based on distributed optical fiber sensing acquires the original scattered light time-series signal of the highway, which includes a collection location label and a collection time label. The original scattered light time-series signal is denoised to obtain a high-quality sensing signal. Features are extracted from the high-quality sensing signal to obtain a disaster feature vector. Based on a pre-set disaster signal database, events indicating the presence of disasters on the highway are analyzed to obtain a disaster event set. The disaster signal database includes various disaster types and their corresponding disaster feature vectors. The disaster event set indicates the presence of disaster events on the highway. Each disaster event includes a disaster type, a collection location label, and a collection time label. Based on the disaster feature vector and pre-set environmental risk factors, the disaster event set undergoes spatiotemporal correlation analysis to obtain a comprehensive disaster feature vector. The comprehensive disaster feature vector includes a disaster type component, a disaster correlation component, and a disaster development trend component. Based on the comprehensive disaster feature vector and a pre-set disaster monitoring level classification threshold set, the disaster risk on the highway is comprehensively analyzed to obtain a disaster analysis and assessment report. The disaster monitoring level classification threshold set indicates the highway monitoring level determined based on the comprehensive disaster feature vector. It achieves fully automated and intelligent analysis of hidden highway hazards, from detecting weak signals to clearly classifying risks. This effectively overcomes the shortcomings of traditional threshold alarm methods, such as high false alarm rates and inability to identify early, weak signals, thus improving the monitoring system's foresight and accuracy. Attached Figure Description

[0052] To more clearly illustrate the technical solutions in the embodiments or related technologies of this application, the accompanying drawings used in the description of the embodiments or related technologies will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0053] Figure 1 A flowchart illustrating a highway disaster monitoring method based on distributed optical fiber sensing, provided as an embodiment of this application;

[0054] Figure 2 A flowchart illustrating a highway disaster monitoring method based on distributed optical fiber sensing, provided as an embodiment of this application;

[0055] Figure 3 A schematic diagram of a highway disaster monitoring system based on distributed optical fiber sensing, provided as an embodiment of this application;

[0056] Figure 4 This is a schematic diagram of the structure of a computer device for a highway disaster monitoring method based on distributed optical fiber sensing, provided as an embodiment of this application. Detailed Implementation

[0057] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0058] In one embodiment, such as Figure 1 As shown, a highway disaster monitoring method based on distributed optical fiber sensing is provided. This embodiment illustrates the method applied to a terminal. It is understood that this method can also be applied to a server, and further to a system including both a terminal and a server, and is implemented through interaction between the terminal and the server. In this embodiment, the method includes the following steps S101-S105, wherein:

[0059] S101, acquire the original scattered light time sequence signal of the highway, wherein the original scattered light time sequence signal includes the acquisition location label and the acquisition time label.

[0060] Specifically, the original scattered light time-series signal is a time-varying signal continuously sensed along the length of the optical fiber, and its mathematical form is: ,in To collect time tags, The location tag is used to collect the original scattered light timing signal along the fiber optic distance. The data was collected at various locations, and the longer the distance along the optical fiber, the more likely the data collection location will be tagged. The larger the value, the better. The original scattered light timing signal is acquired by the terminal using this signal.

[0061] S102 performs noise reduction processing on the original scattered light timing signal to obtain a high-quality sensing signal.

[0062] Specifically, the terminal performs denoising and signal enhancement processing on the raw scattered light time-series signal with strong environmental noise acquired in S101. This aims to suppress interference from traffic vibrations, wind noise, and other sources, highlighting weak signal components related to changes in the underground structure, thereby obtaining a high-quality sensing signal. The mathematical form of the high-quality sensing signal is as follows: ,in To collect time tags, This is a tag for the data collection location.

[0063] S103, extract features from high-quality sensing signals to obtain disaster feature vectors; and based on a pre-set disaster signal database, analyze events where disasters exist on the highway to obtain a disaster event set; the disaster signal database includes each disaster type and its corresponding disaster feature vector; the disaster event set is used to indicate disaster events existing on the highway; the disaster event includes disaster type, collection location label, and collection time label.

[0064] Specifically, the pre-built disaster signal database is a pre-constructed knowledge base that stores disaster feature vectors for various disaster types (such as "karst cavity development", "roadbed seepage activity", "slope instability vibration", etc.). The mathematical form of the pre-built disaster signal database is a [database containing...]. A set of entries: , It is a pre-set disaster signal database; It is the total number of disaster types; Characterizing the first Disaster types; It is the first Disaster feature vectors for different disaster types. The terminal intelligently analyzes the high-quality sensing signals obtained in S102, extracts features from the high-quality sensing signals to obtain the currently monitored disaster feature vectors, and compares these current disaster feature vectors with the disaster feature vectors of each disaster type in the preset disaster signal database to identify the types of disasters that may occur along the highway. Based on the collection time and location labels of the high-quality sensing signals with extracted features, the collection location and collection time labels of each disaster event are determined, resulting in a set of disaster events. The disaster feature vectors of each disaster type in the preset disaster signal database are obtained by extracting features from the high-quality sensing signals collected before the occurrence of each disaster type. The extraction method is the same as the method described above for extracting features from the current high-quality sensing signals. The preset disaster signal database can be compiled based on actual work needs.

[0065] S104, based on the disaster feature vector and the preset environmental risk factors, performs spatiotemporal correlation analysis on the disaster event set to obtain the comprehensive disaster feature vector; the comprehensive disaster feature vector includes disaster type component, disaster correlation component and disaster development trend feature component.

[0066] Specifically, preset environmental risk factors are used to indicate the basic probability of a disaster occurring on each section of the highway. These factors can be set based on the actual roadbed or environmental conditions of each section of the highway. The preset environmental risk factors are in the form of... , These are pre-set environmental risk factors. It is a tag for any collection location. It is a collection location label The environmental risk factor, with a higher value, indicates a higher risk of disaster on the highway corresponding to the data collection location label. The terminal performs in-depth analysis on the discrete disaster event set obtained from S103, aiming to uncover the intrinsic spatial and temporal connections between events, aggregating isolated alarm points and assessing their overall status and development trend to generate a comprehensive disaster feature vector that fully describes the disaster characteristics. The comprehensive disaster feature vector includes a disaster type component, a disaster correlation component, and a disaster development trend feature component. The disaster type component characterizes the main disaster categories present on the highway and is a vector obtained by concatenating various disaster types. The disaster correlation component is a high-dimensional real-number vector obtained by fusing disaster feature vectors from multiple related events; it encodes the overall attributes of each disaster type at the signal feature level. The disaster development trend feature component is a scalar that quantifies the development and change tendency of each disaster type over a future period, such as the severity and direction of the trend (deterioration or mitigation).

[0067] S105, based on the comprehensive disaster feature vector and the preset disaster monitoring level classification threshold set, comprehensively analyzes the disaster risks existing in the highway and obtains a disaster analysis and assessment report; the disaster monitoring level classification threshold set is used to indicate the monitoring level of the highway determined according to the comprehensive disaster feature vector.

[0068] Specifically, the preset disaster monitoring level classification threshold set is used to indicate the determination of the highway monitoring level based on the comprehensive disaster feature vector. This includes the level classification thresholds for each disaster type and the corresponding monitoring level. The monitoring level is used to indicate the need for on-site monitoring of the highway in conjunction with the comprehensive disaster feature vector to verify the existence of real disasters on the highway. Based on the comprehensive disaster feature vector obtained in S104, the terminal maps the disaster correlation component and the disaster development trend feature component to obtain the risk index for that disaster type, according to different disaster types within the disaster type component. Combined with the threshold parameters corresponding to the disaster type in the preset disaster monitoring level classification threshold set, the monitoring level for that disaster type is determined. The monitoring level and comprehensive disaster feature vector for each disaster type are then converted into a structured representation, resulting in a disaster analysis and assessment report.

[0069] This embodiment provides a highway disaster monitoring method based on distributed fiber optic sensing. By sequentially executing a complete chain of signal acquisition, noise suppression, intelligent feature recognition, event spatiotemporal correlation clustering, and comprehensive risk assessment, a disaster analysis and assessment report is obtained. This achieves fully automated and intelligent analysis of hidden highway disasters, from weak signal perception to clear risk classification. It effectively overcomes the shortcomings of traditional threshold alarm methods, such as high false alarm rates and inability to identify early weak signals, thus improving the foresight and accuracy of the monitoring system.

[0070] In one embodiment, the original scattered light timing signal is denoised to obtain a high-quality sensing signal, including:

[0071] S201, perform mode decomposition on the original scattered light time series signal to obtain the intrinsic mode components and residual terms of the original scattered light time series signal.

[0072] Specifically, for each data collection location tag, the original scattered light time-series signal ((Right now Label at any fixed collection location The terminal uses an adaptive mode decomposition algorithm to process the original scattered light time-series signal, obtaining the intrinsic mode components and residual terms. The adaptive mode decomposition algorithm can adaptively decompose complex, non-stationary original signals into a series of intrinsic mode components arranged from high to low frequency. and a residual term representing the average trend or constant of the signal. ,satisfy . This represents the total number of intrinsic mode components (IMFs). IMFs contain the oscillatory characteristics of a signal at different time scales, where high-frequency components are usually related to noise, while low-frequency components may contain hazardous signals. Adaptive mode decomposition algorithms can be empirical mode decomposition or its improved versions, characterized by not requiring preset basis functions and being entirely driven by the data itself.

[0073] S202, calculate the correlation coefficient between the intrinsic mode components and the original scattered light time-series signal to obtain the correlation coefficient of the intrinsic mode components.

[0074] Specifically, for each tag at a collection location, the terminal calculates each intrinsic mode function component of the raw scattered light time-series signal. With the original signal Pearson correlation coefficient The correlation coefficients of each intrinsic mode component are obtained. The calculation formula is as follows: ,in Describing covariance, Represents standard deviation. Correlation coefficient. This is used to measure the similarity and importance of the intrinsic mode component to the original scattered light timing signal waveform. The larger the absolute value, the more important the energy or information occupied by the intrinsic mode component in the original scattered light timing signal; conversely, a component close to zero is likely to represent random noise.

[0075] S203, based on the correlation coefficient and the preset correlation coefficient threshold, select the intrinsic mode component with a correlation coefficient greater than the correlation coefficient threshold as the dominant signal component of the original scattered light time series signal.

[0076] Specifically, a correlation coefficient threshold is preset to measure the magnitude of the correlation coefficient of the intrinsic mode components, which are the main characteristics of the original scattered light time-series signal. The default setting is 0.1, but it can be set according to actual work requirements. For the original scattered light time-series signal corresponding to each acquisition location tag, the terminal selects all intrinsic mode components that satisfy the correlation coefficient greater than the correlation coefficient threshold as the dominant signal components of the original scattered light time-series signal.

[0077] S204, based on the wavelet denoising algorithm, performs denoising processing on the dominant component of the signal to obtain the denoised dominant component of the signal.

[0078] Specifically, for the original scattered light time-series signal corresponding to each data collection location tag, the terminal further performs wavelet thresholding denoising on each dominant signal component. The terminal performs discrete wavelet transform on the dominant signal component by selecting a wavelet basis function (such as 'db4') and the number of decomposition levels to obtain the wavelet coefficients for each level. Then, a soft thresholding function is used to process these wavelet coefficients: for the wavelet coefficients... If its absolute value is less than the threshold If it is greater than or equal to, then set it to zero; if it is greater than or equal to Then let it become Threshold It is based on Confirmed, among which This is an estimate of the noise standard deviation. Let be the signal length. Finally, perform an inverse wavelet transform on the thresholded coefficients to obtain the dominant denoised component of the original scattered light time-series signal. .

[0079] S205 reconstructs the original scattered light time-series signal based on the dominant component, intrinsic mode component, and residual term of the denoised signal to obtain a high-quality sensing signal.

[0080] Specifically, for the original scattered light time-series signal corresponding to each data collection location tag, the terminal will use the following formula to determine the dominant component of the denoised signal after wavelet denoising, the other intrinsic mode components after removing the dominant component, and the residual term: The signal is then reconstructed to obtain a high-quality sensing signal. These are collection location tags. The collection location is tagged as The set of dominant components of the denoised signal. It is a set Any dominant component of the denoised signal, The collection location is tagged as The set of intrinsic mode components remaining after removing the dominant component of the denoised signal. The collection location is tagged as The residual terms. The reconstructed signal. That is, the collection location is tagged as High-quality sensing signals.

[0081] This embodiment provides a highway disaster monitoring method based on distributed optical fiber sensing. By adaptively separating the frequency bands of signal and noise using mode decomposition, and then focusing on the dominant signal components using correlation coefficients, a wavelet transform is applied for refined denoising in the time-frequency domain, resulting in high-quality sensing signals covering all data collection location labels. This effectively filters out environmental and system noise while preserving, to the greatest extent possible, weak signal characteristics related to underground disasters, laying a reliable signal foundation for subsequent high-precision disaster identification.

[0082] In one embodiment, features are extracted from the high-quality sensing signals to obtain a disaster feature vector; and based on a pre-set disaster signal database, events indicating disasters on highways are analyzed to obtain a disaster event set, including:

[0083] S301, based on the preset feature extraction length parameters and acquisition location labels, sorts the original scattered light time series signals to obtain the length distance sequence number of each original scattered light time series signal.

[0084] Specifically, the preset feature extraction length parameter defines the basic spatial window length for feature analysis, with a default setting of 50 meters (using 50 meters of optical fiber as a spatial analysis window unit). The preset feature extraction length parameter can be set according to actual work requirements. The terminal sorts the high-quality sensing signal acquisition location labels from smallest to largest, and uses the preset feature extraction length parameter as the window length. It slides across the sorted original scattered light time-series signal sequence with a preset sliding step size, dividing the original scattered light time-series signal into multiple continuous intervals. A unique length-distance sequence number is assigned to the original scattered light time-series signal within each interval. ( ). This serial number This represents a continuous signal segment at a location along the highway. For example, the preset sliding step size is set by default to half of the preset feature extraction length parameter, but it can also be set according to actual work requirements.

[0085] S302, combine the original scattered light timing signals of the same length distance number to obtain a set of signal segments of each length distance number.

[0086] Specifically, the terminal splices the original scattered light time-series signals with the same distance index according to their acquisition location labels from smallest to largest, obtaining a set of signal segments for each distance index. Each set of signal segments for any distance index is in the form of a two-dimensional data matrix. ,in It is any distance index. It is the length distance sequence number A collection of signal segments, This represents the total number of time tags collected. It is the length distance sequence number The total number of collection location tags included within the interval, and any element in the signal segment set. Indicates the first For each signal segment, the acquisition time label is... Collection location tags are High-quality sensing signals.

[0087] S303, for each length distance number, the signal segment set is input into a preset feature extraction network to extract the change features of the original scattered light time sequence signal in the signal segment set, and obtain the disaster feature vector of each length distance number; wherein the feature extraction network is constructed based on the attention mechanism and a one-dimensional convolutional neural network.

[0088] Specifically, for each length distance index, the terminal inputs its set of signal segments into a pre-trained deep learning model—a feature extraction network. This network uses a one-dimensional convolutional neural network as its backbone, enabling it to effectively capture the local correlations and patterns of the signal over time. Simultaneously, a channel attention mechanism is embedded in the network, allowing it to adaptively focus on feature channels more important for distinguishing disaster types and suppress less important channels. The network uses any of the aforementioned length distance indices... signal segment set As input, after passing through multiple layers of convolution, pooling, non-linear activation, and attention weighting, the network outputs a fixed-dimensional real-number vector at the fully connected layer. This vector That is, the length distance sequence number. The disaster feature vector.

[0089] S304. For each length distance index, calculate the distance between the disaster feature vector of the length distance index and the disaster feature vector of each disaster type in the disaster signal database to obtain the vector distance set of each length distance index.

[0090] Specifically, for each length distance index, the terminal traverses each disaster type in the disaster signal database, calculates the distance between the disaster feature vector of the current length distance index and the disaster feature vectors of each disaster type, and obtains the vector distance set for each length distance index. Among these, any length distance index... The vector distance set is in the form of any element in the vector distance set Indicates the length distance sequence number Disaster feature vectors and disaster types The distance between the disaster feature vectors reflects the similarity between the current road segment signal characteristics and various standard disaster patterns.

[0091] S305. Based on the vector distance set of each length distance index and the preset risk distance threshold parameter, determine the disaster events existing on the highway and the disaster types of the disaster events, and obtain the initial disaster event set.

[0092] Specifically, a preset risk distance threshold parameter is used to determine whether the similarity is high enough to confirm the occurrence of a disaster, and can be set according to actual work. For each length distance sequence number, the terminal searches for the smallest distance value and its corresponding disaster type in its vector distance set. If the smallest distance value is less than the preset risk distance threshold parameter, it is determined that a disaster event has been detected for the length distance sequence number, and the disaster type corresponding to the smallest distance value is determined as the disaster type of the disaster event. The terminal combines the disaster events and corresponding disaster types of each length distance sequence number in ascending order to obtain an initial disaster event set.

[0093] S306. Based on the signal segment set with each length distance number and the initial disaster event set, determine the collection location label and collection time label of each disaster event in the initial disaster event set, and obtain the disaster event set based on the disaster type, collection location label and collection time label of each disaster event.

[0094] Specifically, the terminal refines the information for each disaster event in the initial disaster event set. Based on the length-distance sequence number corresponding to the disaster event, it extracts the acquisition location label of the high-quality sensing signal in the middle of the signal segment set corresponding to that length-distance sequence number, using it as the acquisition location label for that disaster event. It also extracts the acquisition time label of the high-quality sensing signal in the middle of the sequence number, using it as the acquisition time label for that disaster event. The terminal then binds each disaster event in the initial disaster event set with its corresponding disaster type, acquisition location label, and acquisition time label, forming a complete disaster event set.

[0095] This embodiment provides a highway disaster monitoring method based on distributed fiber optic sensing. It normalizes continuous spatial signals into standard segments, automatically extracts abstract features using a deep neural network, and achieves preliminary quantitative identification of disaster types by comparing them with prototypes in a knowledge base. Finally, it outputs a set of disaster events with disaster types including spatiotemporal attributes. This achieves automated conversion from raw signals to semantic event descriptions, avoiding the subjectivity of traditional manual threshold setting and providing high-quality, machine-understandable input data for subsequent macro-situation analysis.

[0096] In one embodiment, for each length distance index, the distance between the disaster feature vector of the length distance index and the disaster feature vectors of each disaster type in the disaster signal database is calculated, including:

[0097] S401, for each length distance index, calculate the distance between the disaster feature vector of the length distance index and the disaster feature vectors of each disaster type in the disaster signal database. The form of the distance between the disaster feature vector of the length distance index and the disaster feature vectors of each disaster type in the disaster signal database is:

[0098]

[0099] in, It is any distance index. It is an index for any type of disaster. It is the first A disaster feature vector with a distance index of length, Disaster type The disaster feature vector, It is a disaster feature vector The distance between them.

[0100] Specifically, the terminal calculates the distance between the disaster feature vector of each length distance index and the disaster feature vector of each disaster type in the disaster signal database according to the formula, thus obtaining the distance between the disaster feature vector of each length distance index and the disaster feature vector of each disaster type. Wherein, the... The disaster feature vector of each length distance index can be obtained from S303, and the disaster type Disaster feature vector This can be obtained from a disaster signal database. Symbols This represents calculating the L2 norm of two vectors, which is the square root of the sum of the squares of the differences between the vectors along their dimensions: ,in This is the dimension of the disaster feature vector. Distance The smaller the value, the more characteristic it is of the signal in the current road segment. With disaster type standard mode The more similar they are.

[0101] This embodiment provides a highway disaster monitoring method based on distributed optical fiber sensing. By explicitly using Euclidean distance as the quantification standard for feature vector similarity, the disaster type matching process has a clear mathematical definition and repeatable calculation steps. This metric is simple, effective, and computationally efficient, and is compatible with the intrinsic principles of many machine learning classifiers, providing a foundation for building a stable and interpretable preliminary disaster identification module.

[0102] In one embodiment, based on the disaster feature vector and preset environmental risk factors, a spatiotemporal correlation analysis is performed on the disaster event set to obtain a comprehensive disaster feature vector, including:

[0103] S501 quantifies the spatiotemporal correlation between disaster events in the disaster event set based on the collection location label, collection time label and disaster type, obtains the correlation degree between each disaster event, and obtains the spatiotemporal correlation degree matrix of disaster events based on the correlation degree between each disaster event.

[0104] Specifically, the terminal iterates through the disaster event set. For any two disaster events in the set, based on their collection location label, collection time label, and disaster type, the following formula is used: The correlation degree between two disaster types was calculated. The correlation degrees between each disaster type were then combined to obtain a spatiotemporal correlation matrix of disaster types. It is an index to any disaster event. It is an index to another disaster event. It is a disaster event Collection location tags, It is a disaster event Collection location tags, It is a disaster event The collection time tag, It is a disaster event The collection time tag, These are preset space influence control parameters. These are preset time-dependent control parameters. It is a disaster event Types of disasters It is a disaster event Types of disasters It is an indicator function, when and Output 1 if they are the same, otherwise output 1. The preset spatial and temporal impact control parameters can be set according to actual work conditions. This embodiment does not further limit the setting of these preset spatial and temporal impact control parameters. The spatiotemporal correlation matrix of the disaster event is a two-dimensional matrix, in which any element... It is a disaster event and disaster events The degree of correlation.

[0105] S502, based on the spatiotemporal correlation matrix of disaster events and the preset correlation threshold, combines each disaster event pair in the disaster event set to obtain a disaster type cluster.

[0106] Specifically, a preset correlation threshold is used to determine whether the correlation is high enough to confirm a link between two disaster events. The terminal analyzes the correlation matrix and selects all events that meet the threshold. The event These events are considered strongly correlated, and clustering algorithms from graph theory are used to divide them into at least one disaster type cluster. Each disaster type cluster may contain multiple disaster events of the same type or a few closely related types. All events in a cluster are highly correlated in terms of time, space, and type, while the correlation between events in different clusters is weak. Each disaster type cluster is considered to represent a spatially concentrated disaster area along the highway that may be continuously developing.

[0107] S503, based on disaster type clusters, fuses the disaster feature vectors corresponding to each disaster event within the disaster type cluster to obtain disaster association components, and obtains disaster type components based on the disaster type of each disaster event in the disaster type cluster.

[0108] Specifically, for each disaster type cluster, the terminal extracts the disaster feature vectors (calculated based on S303) of each disaster event within the cluster and performs a weighted average calculation. The disaster feature vectors of each disaster event are then integrated to obtain a disaster association component. Furthermore, the terminal performs a weighted average calculation on the collection location labels of each disaster event within the cluster to obtain the collection location label for that disaster association component. The terminal also statistically analyzes the disaster types of each disaster event within the cluster and selects the most frequently occurring disaster type to form the disaster type component.

[0109] S504, based on a set of disaster events, a pre-set historical disaster correlation vector database, disaster type clusters, disaster correlation components, and environmental risk factors, predicts the development trend of each disaster event in the disaster type cluster and obtains the disaster development trend characteristic components; the historical disaster correlation vector database includes historical disaster correlation vectors of historical processes and the collection location labels of historical disaster correlation vectors.

[0110] Specifically, the pre-set historical disaster correlation vector database includes historical disaster correlation vectors calculated for all identified disaster type clusters in each past monitoring period, as well as the collection location labels for these historical disaster correlation vectors. The database also includes the collection time labels for these historical disaster correlation vectors. Based on the disaster event set, the terminal determines the collection location labels for disaster events included in the disaster type clusters, extracts historical disaster correlation vectors from the pre-set historical disaster correlation vector database, and performs time series analysis and trend prediction in conjunction with environmental risk factors to obtain the characteristic components of disaster development trends.

[0111] S505, based on disaster correlation components, disaster type components, and disaster development trend characteristic components, obtains a comprehensive disaster feature vector.

[0112] Specifically, the terminal splices together the disaster-related components, disaster-type components, and disaster-development-trend characteristic components to obtain a comprehensive disaster feature vector.

[0113] This embodiment provides a highway disaster monitoring method based on distributed fiber optic sensing. By quantitatively analyzing the spatiotemporal correlation between discrete disaster events, it intelligently aggregates them into disaster type clusters with greater physical meaning. Then, it fuses the internal characteristics of each cluster, clarifies its main type, and obtains disaster correlation components and disaster type components. It then initiates an assessment process for their development trends, ultimately generating a comprehensive disaster feature vector that fully describes the disaster. This represents a cognitive upgrade from point-based event alarms to a comprehensive understanding of the overall state potential, providing crucial data for accurate risk assessment and differentiated management.

[0114] In one embodiment, based on a set of disaster events, a pre-set historical disaster correlation vector database, disaster type clusters, disaster correlation components, and environmental risk factors, the development trend of each disaster event in the disaster type cluster is predicted to obtain disaster development trend feature components, including:

[0115] S601, based on disaster type clusters and disaster event sets, extracts historical disaster association vectors from the historical disaster association vector database to obtain a historical disaster association vector sequence.

[0116] Specifically, the terminal extracts the location tags of each disaster event within a disaster type cluster from the disaster event set, forming a type cluster location tag set. Based on this set, the terminal extracts historical disaster association vectors from the historical disaster association vector database whose absolute differences between the location tags and any location tag in the type cluster location tag set meet a preset distance requirement, forming a historical disaster association vector sequence. The form of this historical disaster association vector sequence is... any one of them Characterizing any acquisition time tag as The historical disaster correlation vector. The preset distance requirement can be set according to the actual requirements for prediction accuracy; this embodiment does not further limit the setting of this preset distance requirement.

[0117] S602, perform time series decomposition on the historical disaster correlation vector sequence to obtain the long-term trend term.

[0118] Specifically, the terminal obtains the historical disaster correlation vector sequence Time series decomposition is performed to separate the long-term trend, periodic fluctuations, and random noise components. This process transforms the sequence decomposition of a multidimensional vector into a process of decomposing and then synthesizing each scalar dimension separately. For the vector's... Each dimension yields its scalar value sequence. For this scalar series, a locally weighted regression method is used to extract the trend: at each time point... A weighted linear regression model is used to fit the data within its neighborhood window. The weights are determined by the time distance between the point and the center point using a kernel function (such as a cubic function), so that nearby points have high weights and distant points have low weights. This regression model predicts the smoothed value at that time point, and the smoothed values ​​of all time points are concatenated to form the long-term trend term of the sequence in that dimension. By performing this operation in parallel across all dimensions of the vector, the original sequence of vectors is eventually decomposed into a sequence of vectors representing the long-term trend. And the sequence of residual term vectors.

[0119] S603 predicts the long-term trend term to obtain the predicted value of the change trend of the disaster correlation vector.

[0120] Specifically, the long-term trend term vector sequence obtained from S602 As input, a temporal convolutional network model is used for temporal modeling and prediction. This model consists of multiple layers of causal dilated one-dimensional convolutions, ensuring that the prediction time is accurate. The output depends only on time. Previous historical inputs conform to the causal relationship of time-series prediction. The network, as described above... Taking a continuous trend term vector as input, the network learns its evolution pattern across time steps through multiple convolutional layers and nonlinear activation functions, and outputs the result for the next time step (i.e., time). Predicted value of the trend term vector The predicted value This is a direct estimate of the smoothing trend of the disaster status in the next monitoring cycle. The terminal further calculates this predicted trend. Compared with the most recent actual trend The change vector between Furthermore, by calculating the norm of this change vector (such as the L2 norm) or by weighting and aggregating the rates of change of its various dimensions, it is compressed into a scalar value. This scalar value This is the predicted value of the changing trend of the disaster correlation vector.

[0121] S604 weights and fuses the predicted value of the change trend, the disaster correlation vector and the environmental risk factor to obtain the characteristic components of the disaster development trend.

[0122] Specifically, the terminal determines the environmental risk factor based on the disaster correlation vector of the current collection location label, and uses the following formula to combine the trend prediction value obtained from S603, the current disaster correlation component obtained from S503, and the preset environmental risk factor: The components of disaster development trend characteristics are obtained by fusion. Among them, It is a characteristic component of the disaster development trend. , and These are preset weights, which sum to 1, and can be set according to actual work requirements. It is a predicted value of the changing trend. It is a disaster correlation vector. It is an environmental risk factor. It is a normalization function.

[0123] This embodiment provides a highway disaster monitoring method based on distributed optical fiber sensing. By retrieving historical state sequences of disaster areas, extracting their inherent trend components, and extrapolating their future trajectory using a predictive model, the method finally combines current intensity and geological background for comprehensive weighting to generate a forward-looking disaster development trend characteristic component. This process combines statistical patterns based on historical data with the current specific context, achieving quantitative prediction of highway disaster evolution trends and giving the monitoring system "foresight" capabilities. This is a key technological step in enabling early warning before disasters occur.

[0124] In one embodiment, based on a comprehensive disaster feature vector and a preset disaster monitoring level classification threshold set, a comprehensive analysis of the disaster risks existing on the highway is conducted to obtain a disaster analysis and assessment report, including:

[0125] S701 quantifies the probability of a disaster on a highway based on disaster correlation vectors and disaster development trend characteristic components, thus obtaining a comprehensive risk index.

[0126] Specifically, the terminal uses the following formula based on the disaster correlation vector and disaster development trend feature components in the comprehensive disaster feature vector: The comprehensive risk index is calculated. It is a comprehensive risk index. It is a disaster correlation vector. It is a characteristic component of the disaster development trend. and It is a preset weight vector with a sum of 1, which can be set according to the actual work.

[0127] S702, based on the threshold set for disaster monitoring level classification and the comprehensive risk index, yields the monitoring level.

[0128] Specifically, the preset disaster monitoring level classification threshold set is a structured knowledge base used to indicate the monitoring level mapped from the comprehensive risk index, and its form is as follows: .in It is a threshold set for classifying disaster monitoring levels. It is a comprehensive risk index. It is a preset low-risk upper limit threshold. The preset upper limit threshold for medium risk, the preset upper limit threshold for low risk, and the preset upper limit threshold for medium risk are all less than 1, and satisfy the following conditions: The default setting is , Alternatively, risk thresholds can be set based on actual work practices. When the monitoring level is at the routine attention level, it indicates that the highway risk is slightly abnormal or in its initial stage, requiring only attention during subsequent routine inspections. When the monitoring level is at the planned monitoring level, it indicates that the highway risk may be in the development stage, with clear abnormal signals, requiring inclusion in recent planned inspections or detailed investigations. When the monitoring level is at the emergency monitoring level, it indicates that the highway risk is sending strong signals or has a severe development trend, requiring immediate initiation of special on-site verification and emergency response.

[0129] S703 generates a disaster analysis and assessment report based on the comprehensive disaster feature vector and monitoring level.

[0130] Specifically, the terminal converts the comprehensive disaster feature vector and monitoring level into structured language according to preset language organization rules, resulting in a disaster analysis and assessment report. The preset language organization rules define how to transform the structured data in the vector into natural language descriptions or clear rules, and these rules can be set according to the format requirements of disaster analysis and assessment reports in actual work.

[0131] This embodiment provides a highway disaster monitoring method based on distributed fiber optic sensing. By combining feature vectors that comprehensively describe disaster clusters with preset, targeted risk assessment rules, a risk index is quantified. This index is then mapped to discrete monitoring levels based on defined thresholds, ultimately generating a structured disaster analysis and assessment report. This process transforms complex, multi-dimensional monitoring data into clear and actionable risk management information, achieving a leap from data perception to decision support and improving the practicality and intelligence of highway safety monitoring.

[0132] To further illustrate the solution of the application embodiment in this embodiment, a specific example is provided below.

[0133] illustrate:

[0134] This application provides a highway disaster monitoring method based on distributed optical fiber sensing, referring to... Figure 2 This includes the following steps:

[0135] S01, acquire raw sensor data.

[0136] Specifically, this includes: using sensing optical fibers in the roadside or roadbed to listen to the sound waves and vibration information inside the underground soil in real time, monitoring multi-source parameters such as strain and temperature, and accurately sensing underground cavities, water seepage and other defects in all weather, real time and across the entire highway area, and obtaining video monitoring systems in key disaster sections to assist in verification.

[0137] S02, Underground disaster detection, early warning and treatment.

[0138] Specifically, it includes: (1) using a signal denoising and enhancement agent to process the input data, wherein the signal denoising and enhancement agent uses a deep learning-based denoising autoencoder to clean and classify the original data, perform signal denoising, interference removal and standardization, and effectively extract weak feature signal parameters related to underground disasters; (2) using a disaster identification and classification agent to identify disasters, wherein the disaster identification and classification agent is based on a convolutional neural network (CNN) and a temporal pattern recognition model; (3) using a disaster spatiotemporal positioning and evolution analysis agent to analyze the development trend of disasters, and by combining the signal arrival time difference and optical cable spatial location information, to achieve meter-level accurate positioning of disaster events. Using models such as Long Short-Term Memory Network (LSTM), analyze and identify the historical development characteristics and evolutionary trends of diseases in time series, and predict the future development trend of diseases; (4) Use risk assessment and decision-making intelligent agents to conduct disaster risk assessment. The risk assessment and decision-making intelligent agents integrate real-time monitoring data, historical data, and external knowledge such as geological and geographical information and expert knowledge bases to construct risk assessment models, conduct disease risk assessment, generate quantifiable monitoring levels, and guide relevant maintenance or treatment work for diseases; (5) Use early warning and disposal intelligent agents to generate risk warnings. When an emergency disease is detected, the intelligent agent automatically generates disposal plans based on past experience for management units to choose from, conducts emergency linkage with highway on-site early warning equipment, and shares information with traffic management, medical, fire and other departments. Emergency linkage drones are used to check the scene, linkage video monitoring systems are used to check the accident situation, linkage surrounding traffic lights and information boards are used to broadcast accident information and remind traffic participants.

[0139] The aforementioned highway disaster monitoring method based on distributed optical fiber sensing acquires the original scattered light time-series signal from the highway, which includes a collection location label and a collection time label. The original scattered light time-series signal is then denoised to obtain a high-quality sensing signal. Feature extraction is performed on the high-quality sensing signal to obtain a disaster feature vector. Based on a pre-set disaster signal database, events indicating potential disasters on the highway are analyzed to obtain a disaster event set. The disaster signal database includes various disaster types and their corresponding disaster feature vectors. The disaster event set indicates the presence of disaster events on the highway. Each disaster event includes a disaster type, a collection location label, and a collection time label. Based on the disaster feature vector and pre-set environmental risk factors, the disaster event set undergoes spatiotemporal correlation analysis to obtain a comprehensive disaster feature vector. This comprehensive disaster feature vector includes a disaster type component, a disaster correlation component, and a disaster development trend component. Based on the comprehensive disaster feature vector and a pre-set disaster monitoring level classification threshold set, the disaster risk on the highway is comprehensively analyzed to obtain a disaster analysis and assessment report. The disaster monitoring level classification threshold set indicates the highway monitoring level determined based on the comprehensive disaster feature vector. This achieves fully automated and intelligent analysis of hidden highway disasters, from weak signal perception to clear risk classification. It effectively overcomes the shortcomings of traditional threshold alarm methods, such as high false alarm rate and inability to identify early weak signals, and improves the foresight and accuracy of the monitoring system.

[0140] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0141] Based on the same inventive concept, this application also provides a highway disaster monitoring system based on distributed optical fiber sensing for implementing the highway disaster monitoring method based on distributed optical fiber sensing described above. The solution provided by this system is similar to the solution described in the above method. Therefore, the specific limitations of one or more embodiments of the highway disaster monitoring system based on distributed optical fiber sensing provided below can be found in the limitations of the highway disaster monitoring method based on distributed optical fiber sensing described above, and will not be repeated here.

[0142] In one exemplary embodiment, such as Figure 3 As shown, a highway disaster monitoring system 200 based on distributed optical fiber sensing is provided, comprising:

[0143] Data acquisition module 201 is used to acquire the original scattered light time-series signal of the highway, wherein the original scattered light time-series signal includes a collection location label and a collection time label;

[0144] The signal denoising module 202 is used to denoise the original scattered light timing signal to obtain a high-quality sensing signal.

[0145] The feature extraction module 203 is used to extract features from high-quality sensing signals to obtain disaster feature vectors; and based on a preset disaster signal database, it analyzes events where disasters exist on the highway to obtain a disaster event set; the disaster signal database includes various disaster types and corresponding disaster feature vectors; the disaster event set is used to indicate disaster events existing on the highway; the disaster events include disaster type, collection location label, and collection time label;

[0146] The feature analysis module 204 is used to perform spatiotemporal correlation analysis on the disaster event set based on the disaster feature vector and preset environmental risk factors to obtain the comprehensive disaster feature vector; the comprehensive disaster feature vector includes disaster type component, disaster correlation component and disaster development trend feature component;

[0147] The report generation module 205 is used to comprehensively analyze the disaster risks existing in the highway based on the comprehensive disaster feature vector and the preset disaster monitoring level classification threshold set, and obtain a disaster analysis and assessment report; the disaster monitoring level classification threshold set is used to indicate the monitoring level of the highway determined according to the comprehensive disaster feature vector.

[0148] Furthermore, the signal denoising module is also used for:

[0149] Modal decomposition is performed on the original scattered light time series signal to obtain the intrinsic mode components and residual terms of the original scattered light time series signal;

[0150] The correlation coefficients between the intrinsic mode components and the original scattered light time-series signal are calculated to obtain the correlation coefficients of the intrinsic mode components;

[0151] Based on the correlation coefficient and a preset correlation coefficient threshold, the intrinsic mode components with a correlation coefficient greater than the correlation coefficient threshold are selected as the dominant signal components of the original scattered light time series signal.

[0152] Based on the wavelet denoising algorithm, the dominant component of the signal is denoised to obtain the denoised dominant component of the signal.

[0153] Based on the dominant component, intrinsic mode component, and residual term of the denoised signal, the original scattered light time-series signal is reconstructed to obtain a high-quality sensing signal.

[0154] Furthermore, the feature extraction module includes:

[0155] The sorting unit is used to sort the original scattered light time series signals based on the preset feature extraction length parameters and acquisition location labels, and obtain the length distance sequence number of each original scattered light time series signal;

[0156] The sorting and combining unit is used to combine the original scattered light time-series signals with the same length distance index to obtain a set of signal segments with each length distance index;

[0157] The signal set feature extraction unit is used to input the signal segment set into a preset feature extraction network for each length distance number, extract the variation features of the original scattered light time series signal in the signal segment set, and obtain the disaster feature vector for each length distance number; the feature extraction network is constructed based on the attention mechanism and a one-dimensional convolutional neural network;

[0158] The distance calculation unit is used to calculate the distance between the disaster feature vector of each length distance number and the disaster feature vector of each disaster type in the disaster signal database, so as to obtain the vector distance set of each length distance number;

[0159] The initial disaster event set determination unit is used to determine the disaster events existing on the highway and the disaster types of the disaster events based on the vector distance set of each length distance index and the preset risk distance threshold parameter, so as to obtain the initial disaster event set;

[0160] The disaster event set determination unit is used to determine the collection location label and collection time label of each disaster event in the initial disaster event set based on the signal segment set with each length distance number and the initial disaster event set, and to obtain the disaster event set based on the disaster type, collection location label and collection time label of each disaster event.

[0161] Furthermore, the distance calculation unit is also used for:

[0162] For each length distance index, calculate the distance between the disaster feature vector of the length distance index and the disaster feature vectors of each disaster type in the disaster signal database. The form of the distance between the disaster feature vector of the length distance index and the disaster feature vectors of each disaster type in the disaster signal database is:

[0163]

[0164] in, It is any distance index. It is an index for any type of disaster. It is the first A disaster feature vector with a distance index of length, Disaster type The disaster feature vector, It is a disaster feature vector The distance between them.

[0165] Furthermore, the feature analysis module includes:

[0166] The correlation calculation unit is used to quantify the spatiotemporal correlation between disaster events in the disaster event set based on the collection location label, collection time label and disaster type, to obtain the correlation between each disaster event, and to obtain the spatiotemporal correlation matrix of disaster events based on the correlation between each disaster event;

[0167] The disaster type aggregation unit is used to combine pairs of disaster events in the disaster event set based on the spatiotemporal correlation matrix of disaster events and a preset correlation threshold to obtain disaster type clusters;

[0168] The disaster feature aggregation unit is used to fuse the disaster feature vectors corresponding to each disaster event within a disaster type cluster to obtain the disaster association component, and to obtain the disaster type component based on the disaster type of each disaster event in the disaster type cluster.

[0169] The disaster trend prediction unit is used to predict the development trend of each disaster event in the disaster type cluster based on the disaster event set, the preset historical disaster correlation vector database, the disaster type cluster, the disaster correlation components and environmental risk factors, and to obtain the disaster development trend feature components; the historical disaster correlation vector database includes historical disaster correlation vectors of historical processes and the collection location labels of historical disaster correlation vectors.

[0170] The comprehensive feature vector generation unit is used to obtain a comprehensive disaster feature vector based on disaster correlation components, disaster type components, and disaster development trend feature components.

[0171] Furthermore, the disaster trend prediction unit is also used for:

[0172] Based on disaster type clusters and disaster event sets, historical disaster association vectors are extracted from the historical disaster association vector database to obtain a sequence of historical disaster association vectors;

[0173] Time series decomposition of historical disaster correlation vector sequences yields long-term trend terms;

[0174] By predicting the long-term trend term, the predicted value of the change trend of the disaster correlation vector is obtained;

[0175] By weighting and fusing the predicted trend values, disaster correlation vectors, and environmental risk factors, the characteristic components of disaster development trends are obtained.

[0176] Furthermore, the report generation module is also used for:

[0177] Based on disaster correlation vectors and disaster development trend feature components, the probability of disasters occurring on highways is quantified to obtain a comprehensive risk index;

[0178] Based on the threshold set for disaster monitoring level classification and the comprehensive risk index, the monitoring level is obtained;

[0179] Based on the comprehensive disaster feature vector and monitoring level, a disaster analysis and assessment report is generated.

[0180] In one embodiment, such as Figure 4 A computer device is provided, comprising:

[0181] At least one processor 301, and a memory 302 communicatively connected to at least one of the processors 301: the memory stores application code executable by at least one of the processors, the application code being executed by at least one of the processors to enable at least one of the processors to perform the previously described highway disaster monitoring method based on distributed fiber optic sensing.

[0182] Computer equipment may also include: sensor 303.

[0183] The processor 301, memory 302 and sensor 303 can be connected via a bus or other means, with the bus being an example in the figure.

[0184] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the steps in the above method embodiments.

[0185] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to in the description of the method embodiments. The device embodiments described above are merely illustrative. The components described as separate parts may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this disclosure according to actual needs. Those skilled in the art can understand and implement this without creative effort.

[0186] The above-described embodiments are merely illustrative of several implementation methods of the embodiments of this application, and their descriptions are relatively specific and detailed. However, they should not be construed as limiting the scope of the patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the embodiments of this application, and these modifications and improvements all fall within the protection scope of the embodiments of this application.

Claims

1. A method for monitoring highway disasters based on distributed optical fiber sensing, characterized in that, The method includes: Acquire the original scattered light time-series signal of the highway, wherein the original scattered light time-series signal includes a collection location label and a collection time label; The original scattered light time-series signal is denoised to obtain a high-quality sensing signal; Feature extraction is performed on the high-quality sensing signals to obtain disaster feature vectors; and based on a preset disaster signal database, events indicating the presence of disasters on the highway are analyzed to obtain a disaster event set; the disaster signal database includes each disaster type and the corresponding disaster feature vector; the disaster event set is used to indicate the disaster events present on the highway; the disaster event includes the disaster type, the collection location label, and the collection time label; Based on the disaster feature vector and preset environmental risk factors, spatiotemporal correlation analysis is performed on the disaster event set to obtain a comprehensive disaster feature vector; the comprehensive disaster feature vector includes a disaster type component, a disaster correlation component, and a disaster development trend feature component. Based on the comprehensive disaster feature vector and the preset disaster monitoring level classification threshold set, the disaster risks existing in the highway are comprehensively analyzed to obtain a disaster analysis and assessment report; the disaster monitoring level classification threshold set is used to indicate the monitoring level of the highway based on the comprehensive disaster feature vector.

2. The method according to claim 1, characterized in that, The denoising process of the original scattered light time-series signal to obtain a high-quality sensing signal includes: The original scattered light time series signal is subjected to mode decomposition to obtain the intrinsic mode components and residual terms of the original scattered light time series signal; The correlation coefficient between the intrinsic mode component and the original scattered light time-series signal is calculated to obtain the correlation coefficient of the intrinsic mode component; Based on the correlation coefficient and a preset correlation coefficient threshold, the intrinsic mode components with a correlation coefficient greater than the correlation coefficient threshold are selected as the dominant signal components of the original scattered light time series signal; Based on the wavelet denoising algorithm, the dominant component of the signal is denoised to obtain the denoised dominant component of the signal. Based on the dominant component of the denoised signal, the intrinsic mode component, and the residual term, the original scattered light time-series signal is reconstructed to obtain the high-quality sensing signal.

3. The method according to claim 1, characterized in that, The high-quality sensing signal is subjected to feature extraction to obtain a disaster feature vector; Based on a pre-set disaster signal database, the system analyzes events indicating potential disasters along the highway to obtain a set of disaster events, including: Based on preset feature extraction length parameters and acquisition location labels, the original scattered light time series signals are sorted to obtain the length distance sequence number of each original scattered light time series signal; The original scattered light time-series signals of the same length distance number are combined to obtain a set of signal segments of each length distance number; For each of the aforementioned length distance indices, the set of signal segments is input into a preset feature extraction network to extract the variation features of the original scattered light temporal signal in the set of signal segments, thereby obtaining the disaster feature vector for each of the aforementioned length distance indices; wherein the feature extraction network is constructed based on an attention mechanism and a one-dimensional convolutional neural network; For each of the length distance indices, the distance between the disaster feature vector of the length distance indices and the disaster feature vector of each disaster type in the disaster signal database is calculated to obtain the vector distance set of each of the length distance indices; Based on the vector distance set of each of the length distance indices and the preset risk distance threshold parameter, the disaster events existing on the highway and the disaster types of the disaster events are determined to obtain an initial disaster event set; Based on the set of signal segments with the specified length distance indices and the initial set of disaster events, the collection location label and collection time label of each disaster event in the initial set of disaster events are determined, and the set of disaster events is obtained based on the disaster type, the collection location label, and the collection time label of each disaster event.

4. The method according to claim 3, characterized in that, The step of calculating the distance between the disaster feature vector of each of the length distance indices and the disaster feature vectors of each of the disaster types in the disaster signal database includes: For each of the aforementioned length distance indices, the distance between the disaster feature vector of the aforementioned length distance indices and the disaster feature vectors of each of the aforementioned disaster types in the disaster signal database is calculated. The form of the distance between the disaster feature vector of the aforementioned length distance indices and the disaster feature vectors of each of the aforementioned disaster types in the disaster signal database is: in, It is any distance index. It is an index for any type of disaster. It is the first A disaster feature vector with a distance index of length, Disaster type The disaster feature vector, It is a disaster feature vector The distance between them.

5. The method according to claim 3, characterized in that, The process of performing spatiotemporal correlation analysis on the disaster event set based on the disaster feature vector and preset environmental risk factors to obtain a comprehensive disaster feature vector includes: Based on the collection location label, the collection time label, and the disaster type, the spatiotemporal correlation degree between each disaster event in the disaster event set is quantified to obtain the correlation degree between each disaster event, and based on the correlation degree between each disaster event, a disaster event spatiotemporal correlation degree matrix is ​​obtained; Based on the spatiotemporal correlation matrix of the disaster events and the preset correlation threshold, the pairs of disaster events in the disaster event set are combined to obtain a disaster type cluster; Based on the disaster type cluster, the disaster feature vectors corresponding to each disaster event within the disaster type cluster are fused to obtain the disaster association component, and the disaster type component is obtained based on the disaster type of each disaster event in the disaster type cluster; Based on the disaster event set, the preset historical disaster correlation vector database, the disaster type cluster, the disaster correlation components, and the environmental risk factors, the development trend of each disaster event in the disaster type cluster is predicted to obtain the disaster development trend feature components; the historical disaster correlation vector database includes historical disaster correlation vectors of historical processes and the collection location labels of the historical disaster correlation vectors; Based on the disaster correlation component, the disaster type component, and the disaster development trend feature component, the comprehensive disaster feature vector is obtained.

6. The method according to claim 5, characterized in that, The method of predicting the development trend of each disaster event in the disaster type cluster based on the disaster event set, a preset historical disaster correlation vector database, the disaster type cluster, the disaster correlation components, and the environmental risk factors, and obtaining the disaster development trend feature components, includes: Based on the disaster type cluster and the disaster event set, the historical disaster association vectors are extracted from the historical disaster association vector database to obtain a historical disaster association vector sequence; The historical disaster correlation vector sequence is decomposed into a time series to obtain the long-term trend term; The long-term trend term is predicted to obtain the predicted value of the change trend of the disaster correlation vector; The predicted value of the change trend, the disaster correlation vector, and the environmental risk factor are weighted and fused to obtain the characteristic component of the disaster development trend.

7. The method according to claim 1, characterized in that, The disaster analysis and assessment report is obtained by comprehensively analyzing the disaster risks existing in the highway based on the comprehensive disaster feature vector and the preset disaster monitoring level classification threshold set, including: Based on the disaster correlation vector and the disaster development trend feature components, the probability of the highway being affected by a disaster is quantified to obtain a comprehensive risk index; Based on the disaster monitoring level classification threshold set and the comprehensive risk index, the monitoring level is obtained; Based on the comprehensive disaster feature vector and the monitoring level, the disaster analysis and assessment report is generated.

8. A highway disaster monitoring system based on distributed optical fiber sensing, characterized in that, The system includes: The data acquisition module is used to acquire the original scattered light time-series signal of the highway, wherein the original scattered light time-series signal includes a collection location label and a collection time label; The signal denoising module is used to denoise the original scattered light timing signal to obtain a high-quality sensing signal. The feature extraction module is used to extract features from the high-quality sensing signal to obtain a disaster feature vector; and based on a preset disaster signal database, analyze the events of disasters existing on the highway to obtain a disaster event set; the disaster signal database includes each disaster type and the corresponding disaster feature vector; the disaster event set is used to indicate the disaster events existing on the highway; the disaster event includes the disaster type, the collection location label, and the collection time label; The feature analysis module is used to perform spatiotemporal correlation analysis on the disaster event set based on the disaster feature vector and preset environmental risk factors to obtain a comprehensive disaster feature vector; the comprehensive disaster feature vector includes a disaster type component, a disaster correlation component, and a disaster development trend feature component; The report generation module is used to comprehensively analyze the disaster risks existing in the highway based on the disaster comprehensive feature vector and the preset disaster monitoring level classification threshold set, and obtain a disaster analysis and assessment report; the disaster monitoring level classification threshold set is used to indicate the monitoring level of the highway based on the disaster comprehensive feature vector.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 7.

Images