Φ-OTDR vibration event classification method and system based on feature fusion

Abstract

The present invention belongs to the field of optical fiber signal recognition technology and provides a Φ-OTDR vibration event classification method and system based on feature fusion. Its technical solution is to divide the preprocessed spatiotemporal data matrix of multiple types of disturbance events into single-channel time domain data of multiple types of disturbance events; train a constructed vibration event classification model based on the single-channel time domain data of multiple types of disturbance events to obtain a trained vibration event classification model; wherein the construction process of the vibration event classification model includes: extracting a first feature map and a second feature map of each type of disturbance event based on the single-channel time domain data; fusing the first feature map and the second feature map of each event to obtain a fused feature of each type of disturbance event, and classifying the event according to the fused feature; and classifying the data to be detected based on the trained vibration event classification model to obtain a classification result. The method significantly improves the intelligent level of optical cable safety monitoring.

Description

Technical Field

[0001] The present invention belongs to the technical field of optical fiber signal recognition, and in particular relates to a Φ-OTDR system vibration event classification method and system based on feature fusion. Background Art

[0002] The statements in this section merely provide background information related to the present invention and do not necessarily constitute prior art.

[0003] Fiber optic cables are becoming increasingly popular as information communication methods. They offer numerous advantages, including high communication capacity, low transmission loss, and strong resistance to electromagnetic interference, making them the mainstream method of information transmission in modern society. However, in practice, unexpected damage and harm to lines often occur due to factors such as construction machinery and human intervention. Therefore, monitoring the status of optical cables is crucial.

[0004] Distributed Optical Fiber Sensors (DOFS) achieve long-distance distributed monitoring along optical fibers by measuring and demodulating the changes in various parameters during light propagation. DOFS can simultaneously acquire sensor information from tens of thousands of points along a single fiber. Phase-sensitive optical time-domain reflectometry (Φ-OTDR) technology is one method for implementing DOFS. Using Φ-OTDR technology, it can detect ground vibrations near underground optical fibers, identify threat activity around the cables, and provide real-time warnings. These technologies offer advantages such as high real-time performance, immunity to electromagnetic interference, and long-distance distributed detection.

[0005] Currently, methods for optical cable status monitoring using Φ-OTDR technology can be categorized into machine learning algorithms and deep learning algorithms. Machine learning algorithms such as support vector machines (SVMs), random forests (RFs), decision trees, and nearest neighbor classification networks have achieved some classification results. However, traditional machine learning algorithms require manual feature extraction, which is labor-intensive and time-consuming. Since the emergence of deep learning, experts both domestically and internationally have combined deep learning with distributed fiber optic sensors to perform pattern recognition on a variety of events, achieving excellent results. Deep learning methods such as CNNs can automatically extract data features, solving the pattern recognition problem end-to-end.

[0006] However, most neural networks use two-dimensional (2D) convolution kernels to construct convolutional layers. Therefore, researchers often convert spatiotemporal data matrices into two-dimensional images. However, this results in high computational costs. To improve algorithm efficiency, researchers have adopted a real-time distributed deep learning network model based on a one-dimensional convolutional neural network (1DCNN). This model can capture more features of one-dimensional time series and has better real-time processing capabilities and higher computational efficiency than traditional two-dimensional convolutional neural networks. However, some existing methods still cannot effectively capture long-series time signals and have a certain false positive rate (FPR). Summary of the Invention

[0007] In order to solve at least one technical problem existing in the above background technology, the present invention provides a Φ-OTDR system vibration event classification method and system based on feature fusion, which improves the event recognition accuracy and reduces the false alarm rate by improving the feature extraction capability of time series signals.

[0008] In order to achieve the above object, the present invention adopts the following technical solutions:

[0009] A first aspect of the present invention provides a method for classifying vibration events in a Φ-OTDR system based on feature fusion, comprising the following steps:

[0010] Acquire spatiotemporal data of multiple types of disturbance events in a phase-sensitive optical time-domain reflectometry (Φ-OTDR) distributed fiber optic sensing system;

[0011] Divide the preprocessed spatiotemporal data matrix of multiple types of disturbance events into single-channel time domain data of multiple types of disturbance events;

[0012] The vibration event classification model is trained based on single-channel time domain data of multiple types of disturbance events to obtain a trained vibration event classification model; wherein the vibration event classification model is constructed in the following steps:

[0013] The first characteristic map and the second characteristic map of each type of disturbance event are obtained based on the single-channel time domain data extraction;

[0014] The first feature map and the second feature map of each event are fused to obtain the fused features of each type of disturbance event, and classification is performed based on the fused features;

[0015] The detection data is classified based on the trained vibration event classification model to obtain the classification results.

[0016] Furthermore, the first characteristic graph of each type of disturbance event is extracted based on the single-channel time domain data, including:

[0017] Extract multi-scale features based on single-channel time domain data and obtain multi-scale fusion features after fusion;

[0018] The multi-scale fusion features are input into the BatchNorm layer, where the mean and variance of each feature dimension are calculated to obtain standardized features.

[0019] Globally pool the standardized features to obtain a C-dimensional vector;

[0020] Then, through a size of k The one-dimensional convolution layer processes the C-dimensional vector and applies the Sigmoid activation function to the processed result to obtain the channel weight;

[0021] Multiply the channel weights by the original feature map channel by channel to obtain the enhanced feature map.

[0022] Furthermore, the multi-scale features of the single-channel time domain data of the multi-scale TCB module are used, including:

[0023] The multi-scale TCB module includes a parallel first causal convolution branch, a second causal convolution branch, and a third causal convolution branch;

[0024] The first causal convolution branch takes as input single-channel time domain data and outputs high-frequency transient event features as The input of the second causal convolution branch is single-channel time domain data, and the output medium-periodic event feature is recorded as The input of the third causal convolution branch is single-channel time domain data, and the output long-period event feature is recorded as ;

[0025] The features output by the first causal convolution branch, the second causal convolution branch, and the third causal convolution branch 、 and After passing through the corresponding global pooling GAP layer, the corresponding scale features are output.

[0026] Furthermore, a second characteristic map of each type of disturbance event is obtained based on the single-channel time domain data extraction, including:

[0027] Process the single-channel time domain data in segments;

[0028] The segmented single-channel time domain data is input into the corresponding forward LSTM layer and reverse LSTM layer respectively. Each forward LSTM layer processes the input sequence in forward order, and each reverse LSTM layer processes the input sequence in reverse order. Finally, the vector formed by concatenating the two output vectors is used as the final feature.

[0029] Furthermore, when classifying based on the fused features, the LogSoftmax layer is used to normalize the classification features to generate the classification probability distribution of the vibration events.

[0030] Furthermore, the preprocessing of spatiotemporal data of multiple types of disturbance events includes normalization processing and first-order difference processing.

[0031] A second aspect of the present invention provides a Φ-OTDR system vibration event classification system based on feature fusion, comprising:

[0032] A data acquisition module for acquiring spatiotemporal data of multiple types of disturbance events in a phase-sensitive optical time domain reflectometer (Φ-OTDR) distributed optical fiber sensing system;

[0033] A data preprocessing module, which is used to divide the preprocessed spatiotemporal data matrix of multiple types of disturbance events into single-channel time domain data of multiple types of disturbance events;

[0034] A vibration event classification model training module is used to train a constructed vibration event classification model based on single-channel time domain data of multiple types of disturbance events to obtain a trained vibration event classification model; wherein the vibration event classification model construction process includes: extracting a first feature map and a second feature map of each type of disturbance event based on the single-channel time domain data; fusing the first feature map and the second feature map of each event to obtain a fused feature of each type of disturbance event, and classifying the disturbance event according to the fused feature;

[0035] The classification module is used to classify the data to be detected based on the trained vibration event classification model to obtain a classification result.

[0036] Furthermore, the Φ-OTDR distributed optical fiber sensing system includes an ultra-narrow linewidth laser, a signal generator, an acousto-optic modulator, an erbium-doped fiber amplifier, a circulator, a sensing fiber, a photodetector, a data acquisition card, and a signal processing module;

[0037] The output ends of the ultra-narrow linewidth laser and the signal generator are connected to the input end of the acousto-optic modulator, the output end of the acousto-optic modulator is connected to the input end of the erbium-doped fiber amplifier, the output end of the erbium-doped fiber amplifier is connected to one end of the circulator, the other end of the circulator is respectively connected to the input ends of the sensing fiber and the photodetector, the output end of the photodetector is connected to the input end of the data acquisition card, and the output end of the data acquisition card is connected to the signal processing module.

[0038] A third aspect of the present invention provides a computer-readable storage medium.

[0039] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of the above-mentioned method for classifying vibration events of a Φ-OTDR system based on feature fusion.

[0040] A fourth aspect of the present invention provides a computer device.

[0041] A computer device includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, the steps of the above-mentioned method for classifying vibration events of a Φ-OTDR system based on feature fusion are implemented.

[0042] Compared with the prior art, the present invention has the following beneficial effects:

[0043] The present invention introduces the MTCB-ECA module, extracts multi-scale features based on single-channel time domain data, dynamically adjusts the contribution of each channel to the final classification result by calculating the importance of feature channels, and uses multi-level feature fusion to pass the fused features through the BiLSTM layer and then combine them with contextual features to obtain the final fusion. This fully leverages their advantages in temporal feature extraction, channel attention mechanism, and context modeling, effectively improving the model's ability to extract Φ-OTDR signal features.

[0044] Advantages of additional aspects of the present invention will be given in part in the following description and in part will be obvious from the following description, or will be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] The accompanying drawings, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

[0046] Figure 1 This is a flow chart of a vibration event classification method for a Φ-OTDR system based on feature fusion provided by an embodiment of the present invention;

[0047] Figure 2 The invention provides a Φ-OTDR distributed optical fiber sensing system;

[0048] Figure 3 This is a time domain diagram of watering events provided by an embodiment of the present invention;

[0049] Figure 4 This is a time domain diagram of mining events provided by an embodiment of the present invention;

[0050] Figure 5 is a time domain diagram of a knock event provided by an embodiment of the present invention;

[0051] Figure 6 This is a time domain diagram of walking events provided by an embodiment of the present invention;

[0052] Figure 7 This is a time domain diagram of a shaking event provided by an embodiment of the present invention;

[0053] Figure 8This is the feature fusion process provided by the embodiment of the present invention;

[0054] Figure 9 MTCB-ECA module diagram provided by an embodiment of the present invention;

[0055] Figure 10 This is the BiLSTM layer provided by the embodiment of the present invention;

[0056] Figure 11 This is the gated spatiotemporal feature fusion mechanism provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0057] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0058] It should be noted that the following detailed descriptions are illustrative and intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the art to which the present invention belongs.

[0059] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit the exemplary embodiments according to the present invention. As used herein, unless the context clearly indicates otherwise, the singular form is intended to include the plural form. In addition, it should be understood that when the terms "comprise" and / or "include" are used in this specification, they indicate the presence of features, steps, operations, devices, components and / or combinations thereof.

[0060] In response to the problem that the existing model is insufficient in capturing the features of long-sequence time series signals, the present invention provides a Φ-OTDR system vibration event classification method recognition algorithm based on feature fusion, which improves the event recognition accuracy and reduces the false alarm rate by improving the feature extraction capability of time series signals, thereby improving the protection of optical cables. First, through multi-level feature fusion, feature extraction and fusion are performed in terms of temporal feature extraction, channel attention mechanism and context modeling, which effectively improves the model's ability to extract Φ-OTDR signal features. In addition, the MTCB-ECA module is introduced to dynamically adjust the contribution of each channel to the final classification result by calculating the importance of the feature channel, so that the model can focus more on key signal features and effectively suppress the influence of redundant information, thereby improving the recognition accuracy of the model. It effectively solves the problems of low recognition accuracy and high false alarm rate in the existing technology, and significantly improves the intelligent level of optical cable safety monitoring.

[0061] Example 1

[0062] like Figure 1 As shown, this embodiment provides a vibration event classification method for a Φ-OTDR system based on feature fusion, comprising the following steps:

[0063] Step 1: Acquire raw data of multiple types of disturbance events in the phase-sensitive optical time domain reflectometry Φ-OTDR distributed optical fiber sensing system;

[0064] like Figure 2 As shown, the Φ-OTDR distributed optical fiber sensing system includes an ultra-narrow linewidth laser, a signal generator, an acousto-optic modulator, an erbium-doped fiber amplifier, a circulator, a sensing fiber, a photodetector, a data acquisition card and a signal processing module;

[0065] The output ends of the ultra-narrow linewidth laser and the signal generator are connected to the input end of the acousto-optic modulator, the output end of the acousto-optic modulator is connected to the input end of the erbium-doped fiber amplifier, the output end of the erbium-doped fiber amplifier is connected to one end of the circulator, the other end of the circulator is respectively connected to the sensing fiber and the input end of the photodetector, the output end of the photodetector is connected to the input end of the data acquisition card, and the output end of the data acquisition card is connected to the signal processing module;

[0066] The specific data acquisition principle is as follows: the continuous light signal emitted by the ultra-narrow linewidth laser is input into the acousto-optic modulator, and the pulse signal generated by the signal generator is input into the acousto-optic modulator; the acousto-optic modulator modulates the continuous light signal into a pulse light signal; the erbium-doped fiber amplifier amplifies the pulse light signal, and the amplified pulse light signal is connected to the sensing fiber through the circulator; the back Rayleigh scattered light signal returned from the sensing fiber is transmitted to the photodetector after passing through the circulator, and the photodetector converts the received light signal into an electrical signal and transmits the electrical signal to the data acquisition card; the data acquisition card samples the electrical signal and transmits the obtained sampled signal to the signal processing module; the signal processing module identifies and marks the sampled signal to obtain the original data corresponding to multiple types of disturbance events.

[0067] The final collected data includes six events: background noise, digging, knocking, shaking, and walking. Non-background events include digging, knocking, shaking, and walking. Each event consists of 10,000 points in the time domain and 10 adjacent spatial nodes. The number of samples in the collected raw data is shown in Table 1:

[0068] Table 1. Number of samples collected for various events

[0069]

[0070] Step 2: Preprocess the acquired raw data of multiple types of disturbance events and divide the preprocessed data into training sets and test sets according to a certain ratio to form independent single-channel time domain data sets;

[0071] In this embodiment, the raw data is preprocessed, including normalization and first-order difference processing;

[0072] Min-Max normalization is used to obtain normalized data. The calculation formula is:

[0073] ,

[0074] in, x Represents the original data, Indicates the minimum value of the data. Indicates the maximum value of the data. represents the normalized data;

[0075] Perform first-order difference processing on the normalized data with an original time series signal length of 10,000. The calculation formula is:

[0076] ,

[0077] in, Represents the original data, represents the data after first-order difference, Represents the index of the data point.

[0078] Then the sample sequence is labeled, shuffled and divided into data sets in a ratio of 7:2:1 to obtain single-channel time domain data. .

[0079] In this embodiment, the pre-processed spatiotemporal data matrix is divided into independent single-channel time domain data. Figure 3 is the time domain diagram of watering events, Figure 4 To mine event time domain graphs, Figure 5 is the time domain diagram of the knock event, Figure 6 is the time domain diagram of walking events, Figure 7 This is the time domain diagram of the shaking event.

[0080] Step 3: Training the constructed vibration event classification model based on the single-channel time domain data to obtain a trained vibration event classification model;

[0081] like Figure 8 As shown in Figure 2, the construction process of the vibration event classification model specifically includes the following steps:

[0082] Step 301: Input the single-channel time domain data into the MTCB-ECA module to extract the first feature map of each event;

[0083] like Figure 9 As shown, the MTCB-ECA module includes a multi-scale TCB module, a BatchNorm layer, an ECA layer, a weight normalization layer, and a Dropout layer;

[0084] like Figure 9 As shown, specifically including:

[0085] Step 3011: Combine single-channel time domain data and the multi-scale TCB module to extract multi-scale features;

[0086] The multi-scale TCB module includes a first causal convolution branch, a second causal convolution branch, and a third causal convolution branch in parallel; the input of the first causal convolution branch is a single-channel time domain data, and the output feature is recorded as , the input of the second causal convolution branch is single-channel time domain data, and the output feature is recorded as ,

[0087] The input of the third causal convolution branch is single-channel time domain data, and the output feature is recorded as ;

[0088] In this embodiment, the expansion rates of the first causal convolution branch, the second causal convolution branch, and the third causal convolution branch are 1, 2, and 4, respectively;

[0089] Specifically, the first causal convolution branch is used to capture high-frequency transient events, such as mining.

[0090] Causal Dilated Convolution: , padded with 2 zeros on the left , the output is: ,in, is the convolution kernel weight, The input sequence is in time steps tk value.

[0091] The second causal convolution branch is used to capture medium-periodic events such as walking.

[0092] Causal Dilated Convolution: , left-padded with 4 zeros ( )*2=4), the output is: ;

[0093] The third causal convolution branch is used to capture long-period events.

[0094] Causal Dilated Convolution: , left-padded with 8 zeros ( )*4=8), the output is: ;

[0095] The features output by the first causal convolution branch, the second causal convolution branch, and the third causal convolution branch 、 and After passing through the corresponding global pooling GAP layer, the output , , ;

[0096] Specifically, the calculation of the global pooling GAP layer is:

[0097] ,

[0098] Then , , Input to the fully connected layer to dynamically generate fusion weights :

[0099] ,

[0100] Among them, the input weight ,satisfy , is bias;

[0101] Weighted summation outputs multi-scale fusion features:

[0102] ,

[0103] Step 3012: Multi-scale fusion features Input BatchNorm layer, in which the mean and variance of each feature dimension are calculated to obtain standardized features ;

[0104] Step 3013: Perform global pooling (GAP) on the standardized features of step 3012 to obtain a C-dimensional vector (C is the number of channels):

[0105] ,

[0106] Step 3014, then through a size of k The one-dimensional convolution layer processes this C-dimensional vector, k The channel weight is obtained by applying the Sigmoid activation function to the kernel size of the cross-channel interaction;

[0107] Specifically, a one-dimensional convolutional layer (kernel size k) is used to perform cross-channel interaction on the input first feature map, where k is determined by an adaptive function:

[0108] ,

[0109] Among them, C is the number of channels, b and γ are hyperparameters, which are designed to ensure that models with different numbers of channels can maintain sufficient coverage of cross-channel interaction.

[0110] Specifically, applying the Sigmoid activation function, the channel weight is obtained as:

[0111] ,

[0112] in, is the weight parameter of the one-dimensional convolutional layer, which determines the importance of each input feature in cross-channel interaction. The part of the C-dimensional vector obtained by the previous operation that is related to the current channel is used to calculate the channel weight;

[0113] Finally, feature weighting is performed and these weights are Multiply the original feature map channel by channel to obtain the enhanced feature map , specifically, for the original feature map The c feature maps of channels , after weighting, the enhanced feature map is obtained , the formula is:

[0114] ,

[0115] Where c=1,2,...C, C is the number of channels in the feature map. Combining the weighted results of all channels will give us the complete enhanced feature map.

[0116] The ECA module is introduced to perform weighted adjustments on the fused features. By calculating the importance of feature channels, the ECA module dynamically adjusts each channel's contribution to the final classification result, further improving the model's recognition accuracy. This mechanism allows the model to focus more on key signal features, suppress the influence of redundant information, and ultimately output the classification results of the vibration event.

[0117] Step 302: Input the single-channel time domain data into the BiLSTM layer to extract the second feature map of each event;

[0118] like Figure 10 As shown, the following steps are included:

[0119] Step 3021: Convert the single channel time domain data Segment processing;

[0120] Assume segment length T=100: , where B is the batch dimension, T is the number of time steps, and 1 is the single-channel input feature dimension;

[0121] Step 3022: Input the segmented single-channel time domain data into the corresponding forward LSTM layer and reverse LSTM layer respectively. Each forward LSTM layer processes the input sequence in a forward order, and each reverse LSTM layer processes the input sequence in a reverse order. Finally, the vector formed by concatenating the two output vectors is used as the final feature expression to obtain the output result of the BiLSTM layer. ;

[0122] In this embodiment, the forward LSTM layer and the reverse LSTM layer adopt the existing LSTM layer structure.

[0123] Step 303: Fusing the first feature map and the second feature map of each event to obtain a third feature map of each event;

[0124] like Figure 11 Specifically, the local time domain features extracted by MTCB-ECA are As , global spatiotemporal features extracted by BiLSTM As Input into the gated spatiotemporal feature fusion mechanism, firstly linearly project the local time domain features and the global spatiotemporal features to the same dimension, and then pass the Sigmoid gate weight Dynamically control the ratio of features extracted by the two modules:

[0125] ,

[0126] in, is the encoding feature of the space, is the coding feature of the time domain, is the feature weighted fusion according to the gated ratio;

[0127] Compared with traditional feature splicing, the dynamic gating mechanism can adaptively suppress noise-dominated modes (such as instantaneous interference in vibration signals); it automatically learns modal correlations through end-to-end training without the need for manual weight setting.

[0128] Step 304: normalize the third feature map to generate a classification probability distribution of the vibration event;

[0129] Specifically, the third feature map is normalized using the LogSoftmax layer to generate the classification probability distribution of the vibration event, which is expressed as: ,

[0130] in, It is i The raw scores of the classes, represents the sum of all categories, It is jThe raw scores of the classes.

[0131] When training the vibration event classification model, the training set data is input into the network model for training. Parameters are optimized by adjusting parameters such as the training batch and learning rate. The weights with the best effect during the training process are saved for intelligent monitoring of optical cables.

[0132] The trained network model is tested using the test set. Precision, Recall, and F1 Score are used to evaluate the performance of the classification model.

[0133] The calculation formula for precision is:

[0134] ,

[0135] The calculation formula of sensitivity (Recall) is:

[0136] ,

[0137] The calculation formula of F1 value (F1Score) is:

[0138] ,

[0139] The category whose predicted label is positive sample and the true label is also positive sample is abbreviated as TP, the category whose predicted label is positive sample and the true label is negative sample is abbreviated as FP, the category whose predicted label is negative sample and the true label is positive sample is abbreviated as FN, and the category whose predicted label is negative sample and the true label is negative sample is abbreviated as TN.

[0140] Step 4: Use the trained vibration event classification model to classify the data to be detected to obtain the classification results of the vibration events.

[0141] In order to verify the classification effect of the method of the present invention, an ablation experiment was conducted to compare the classification effect of the existing algorithm. The ablation experiment comparison results are shown in Table 2:

[0142] Table 2. Comparison results of ablation experiments

[0143]

[0144] Table 2 shows that the proposed method outperforms existing classifiers such as TCN, TCN-ECA, TCN-BiLSTM, TCN-Improve, and TCN-Improve-ECA. This demonstrates that the proposed method improves event recognition accuracy and reduces false alarm rates by enhancing the ability to extract features from time series signals.

[0145] Example 2

[0146] This embodiment provides a Φ-OTDR vibration event classification system based on feature fusion, including:

[0147] A data acquisition module for acquiring spatiotemporal data of multiple types of disturbance events in a phase-sensitive optical time domain reflectometer (Φ-OTDR) distributed optical fiber sensing system;

[0148] A data preprocessing module, which is used to divide the preprocessed spatiotemporal data matrix of multiple types of disturbance events into single-channel time domain data of multiple types of disturbance events;

[0149] A vibration event classification model training module is used to train a constructed vibration event classification model based on single-channel time domain data of multiple types of disturbance events to obtain a trained vibration event classification model; wherein the vibration event classification model construction process includes: extracting a first feature map and a second feature map of each type of disturbance event based on the single-channel time domain data; fusing the first feature map and the second feature map of each event to obtain a fused feature of each type of disturbance event, and classifying the disturbance event according to the fused feature;

[0150] The classification module is used to classify the data to be detected based on the trained vibration event classification model to obtain a classification result.

[0151] It should be noted that the specific implementation method of the Φ-OTDR vibration event classification system based on feature fusion in an embodiment of the present invention is similar to the specific implementation method of the Φ-OTDR vibration event classification method based on feature fusion in an embodiment of the present invention. Please refer to the description of the method part for details. In order to reduce redundancy, it will not be repeated here.

[0152] Example 3

[0153] This embodiment provides a computer-readable storage medium having a computer program stored thereon. When the program is executed by a processor, the steps in the above-mentioned method for classifying Φ-OTDR vibration events based on feature fusion are implemented.

[0154] Example 4

[0155] This embodiment provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, the steps in the feature fusion-based Φ-OTDR vibration event classification method described above are implemented.

[0156] The foregoing description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Those skilled in the art will readily appreciate that various modifications and variations of the present invention are possible. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention are intended to be within the scope of protection of the present invention.

Claims

1. A Φ-OTDR vibration event classification method based on feature fusion, characterized in that: The steps include: Acquire spatiotemporal data of multiple types of disturbance events in a phase-sensitive optical time-domain reflectometry (Φ-OTDR) distributed fiber optic sensing system; Divide the preprocessed spatiotemporal data matrix of multiple types of disturbance events into single-channel time domain data of multiple types of disturbance events; The vibration event classification model is trained based on single-channel time domain data of multiple types of disturbance events to obtain a trained vibration event classification model; wherein the vibration event classification model is constructed in the following steps: The first characteristic map and the second characteristic map of each type of disturbance event are obtained based on the single-channel time domain data extraction; The first feature map and the second feature map of each event are fused to obtain the fused features of each type of disturbance event, and classification is performed based on the fused features; Classify the data to be detected based on the trained vibration event classification model to obtain a classification result; The first characteristic graph of each type of disturbance event extracted based on single-channel time domain data includes: Extract multi-scale features based on single-channel time domain data and obtain multi-scale fusion features after fusion; The multi-scale fusion features are input into the BatchNorm layer, where the mean and variance of each feature dimension are calculated to obtain standardized features. Globally pool the standardized features to obtain a C-dimensional vector; Then, through a size of k The one-dimensional convolution layer processes the C-dimensional vector and applies the Sigmoid activation function to the processed result to obtain the channel weight; Multiply the channel weights by the original feature map channel by channel to obtain the enhanced feature map; The multi-scale features of single-channel time domain data using the multi-scale TCB module include: The multi-scale TCB module includes a parallel first causal convolution branch, a second causal convolution branch, and a third causal convolution branch; The first causal convolution branch takes as input single-channel time domain data and outputs high-frequency transient event features as The input of the second causal convolution branch is single-channel time domain data, and the output medium-periodic event feature is recorded as The input of the third causal convolution branch is single-channel time domain data, and the output long-period event feature is recorded as ; The features output by the first causal convolution branch, the second causal convolution branch, and the third causal convolution branch 、 and After passing through the corresponding global pooling GAP layer, the corresponding scale features are output; The single-channel time domain data is input into the BiLSTM layer to extract the second feature map of each event.

2. The Φ-OTDR vibration event classification method based on feature fusion as claimed in claim 1, wherein The second characteristic map of each type of disturbance event is obtained based on the single-channel time domain data extraction, including: Process the single-channel time domain data in segments; The segmented single-channel time domain data is input into the corresponding forward LSTM layer and reverse LSTM layer respectively. Each forward LSTM layer processes the input sequence in forward order, and each reverse LSTM layer processes the input sequence in reverse order. Finally, the vector formed by concatenating the two output vectors is used as the final feature.

3. The Φ-OTDR vibration event classification method based on feature fusion as claimed in claim 1, wherein When classifying based on the fused features, the LogSoftmax layer is used to normalize the classification features to generate the classification probability distribution of the vibration events.

4. The Φ-OTDR vibration event classification method based on feature fusion as claimed in claim 1, wherein The preprocessing of spatiotemporal data of multiple types of disturbance events includes normalization and first-order difference processing.

5. The Φ-OTDR vibration event classification system based on feature fusion is characterized by: The Φ-OTDR vibration event classification method based on feature fusion according to any one of claims 1 to 4 comprises: A data acquisition module for acquiring spatiotemporal data of multiple types of disturbance events in a phase-sensitive optical time domain reflectometer (Φ-OTDR) distributed optical fiber sensing system; A data preprocessing module, which is used to divide the preprocessed spatiotemporal data matrix of multiple types of disturbance events into single-channel time domain data of multiple types of disturbance events; A vibration event classification model training module is used to train a constructed vibration event classification model based on single-channel time domain data of multiple types of disturbance events to obtain a trained vibration event classification model; wherein the vibration event classification model construction process includes: extracting a first feature map and a second feature map of each type of disturbance event based on the single-channel time domain data; fusing the first feature map and the second feature map of each event to obtain a fused feature of each type of disturbance event, and classifying the disturbance event according to the fused feature; The classification module is used to classify the data to be detected based on the trained vibration event classification model to obtain a classification result.

6. The Φ-OTDR vibration event classification system based on feature fusion as claimed in claim 5, characterized in that, The Φ-OTDR distributed optical fiber sensing system includes an ultra-narrow linewidth laser, a signal generator, an acousto-optic modulator, an erbium-doped fiber amplifier, a circulator, a sensing fiber, a photodetector, a data acquisition card, and a signal processing module; The output ends of the ultra-narrow linewidth laser and the signal generator are connected to the input end of the acousto-optic modulator, the output end of the acousto-optic modulator is connected to the input end of the erbium-doped fiber amplifier, the output end of the erbium-doped fiber amplifier is connected to one end of the circulator, the other end of the circulator is respectively connected to the input ends of the sensing fiber and the photodetector, the output end of the photodetector is connected to the input end of the data acquisition card, and the output end of the data acquisition card is connected to the signal processing module.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps of the Φ-OTDR vibration event classification method based on feature fusion as described in any one of claims 1 to 4 are implemented.

8. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, the steps of the Φ-OTDR vibration event classification method based on feature fusion are implemented as described in any one of claims 1 to 4.

Images