Perimeter security area abnormal event identification method based on optical fiber signals

Abstract

The invention belongs to the field of distributed optical fiber signal vibration event classification, and discloses a perimeter security area abnormal event identification method based on optical fiber signals. An optical fiber vibration event data set is collected through a phase-sensitive optical time domain reflectometer and preprocessed, and a training set is formed; the 1DCNN model is improved; inputting the training sample into the improved 1DCNN deep learning model for training, and obtaining a trained improved 1DCNN model; and abnormal event identification is carried out. The method improves the recognition accuracy of the abnormal events in the perimeter security area, and solves the problems that a traditional feature extraction method for optical fiber signals needs to manually manufacture features, seriously depends on expert knowledge and is weak in generalization ability, and the calculation amount is large in the process that an existing model carries out recognition through the optical fiber signals.

Description

Technical Field

[0001] The invention belongs to the field of distributed optical fiber signal vibration event classification, and in particular relates to an abnormal event recognition method for a perimeter security zone based on optical fiber signals. Background Art

[0002] Nowadays, with the rapid development of the information age, more and more cables, submarine optical cables, oil and gas pipelines, and railway tracks are being laid, and the resulting safety monitoring and other issues have become a major problem that needs to be solved urgently. Vibration information is the carrier of various safety hazards. If the external vibration information can be monitored in real time, and the characteristics of different events can be distinguished from the vibration signal and then identified, then it will inevitably play an important role in safety monitoring. For example, it can identify whether there is vibration of large mechanical devices near the cable to prevent the cable from being damaged due to construction, and identify whether someone has climbed over the protective net around the railway to achieve the purpose of safety alert.

[0003] Distributed Optical Fiber Vibration Sensing (DVS) technology can continuously measure the vibration signals around the optical fiber along the sensing optical fiber. It is a distributed optical fiber sensing technology. Compared with traditional sensors, this type of technology has the characteristics of being distributed, anti-electromagnetic interference, corrosion-resistant, and power-free. At the same time, the distributed optical fiber vibration sensing system also has the characteristics of high sensitivity, simple structure, and long detection distance. It is very sensitive to external vibrations and has a large dynamic range of monitoring, which is very suitable for detecting vibration events in a large range. Its unique application scenarios make it the first choice in border security, oil and gas pipeline leakage detection and positioning, building structure damage monitoring, border monitoring in multiple environments, and other fields. Phase-sensitive optical time domain reflectometer (Φ-OTDR) is a typical representative of distributed optical fiber sensing systems and is widely used in security monitoring fields such as national defense borders and military bases. However, high nuisance alarm rate (NAR) is the main disadvantage of the Φ-OTDR system. Due to the sensitivity of the sensing fiber to the environment and the similarity between some intrusion signals and non-intrusion signals, the system is prone to make poor judgments during the vibration signal recognition process, resulting in a high nuisance alarm rate and a large waste of manpower and material resources.

[0004] Traditional shallow fiber event classification methods require manual extraction of features from the collected vibration signals during feature extraction. These extraction methods are cumbersome and complex, and it requires a certain amount of experience and knowledge to determine whether the extracted features can help the classifier identify. In addition, some extracted features may be invalid in other scenarios, resulting in unstable generalization of the model. Once the environment changes, the false alarm rate of the system will increase significantly. Summary of the invention

[0005] The purpose of the present invention is to provide a method for identifying abnormal events in perimeter security zones based on optical fiber signals. The method can directly extract features from an original optical fiber vibration event data set and identify the type of vibration event without human intervention, thereby improving the accuracy of identifying abnormal events in perimeter security zones and solving the problems that traditional feature extraction methods for optical fiber signals require artificially created features, are heavily dependent on expert knowledge, have weak generalization capabilities, and have a large amount of computational complexity in the process of using optical fiber signals for identification in existing models.

[0006] To achieve the above object, the technical solution adopted by the present invention is: A method for identifying abnormal events in a perimeter security zone based on optical fiber signals, the method comprising: The optical fiber vibration event data set is collected by phase-sensitive optical time domain reflectometry and preprocessed to form a training set, wherein the training samples in the training set are the preprocessed optical fiber vibration event data; Inputting the training samples into an improved 1DCNN deep learning model, the improved 1DCNN deep learning model includes a four-layer convolution module, a long short-term memory network, a global mean pooling layer, a fully connected layer, and an SVM classifier optimized by the gray wolf optimization algorithm, wherein each layer of the convolution module includes a convolution layer, a batch normalization layer, a Relu activation function layer, and a maximum pooling layer, and an attention mechanism module is connected between the first layer of the convolution module and the second layer of the convolution module; Calculate a first loss function based on the output of the fully connected layer in the improved 1DCNN deep learning model and the label of the training sample, and update the weight parameters of the improved 1DCNN deep learning model before the SVM classifier optimized by the gray wolf optimization algorithm according to the first loss function until the training is completed; The weight parameters before the SVM classifier optimized by the gray wolf optimization algorithm in the improved 1DCNN deep learning model are fixed, the second loss function is calculated according to the output of the SVM classifier optimized by the gray wolf optimization algorithm and the labels of the training samples, and the parameters of the SVM classifier optimized by the gray wolf optimization algorithm are adjusted according to the second loss function until the training is completed; Take the trained improved 1DCNN deep learning model and output the abnormal event recognition results for the optical fiber vibration event data to be identified.

[0007] Several optional methods are also provided below, but they are not intended to be additional limitations on the above-mentioned overall solution, but are merely further supplements or preferences. Under the premise that there are no technical or logical contradictions, each optional method can be combined with the above-mentioned overall solution separately, and multiple optional methods can also be combined.

[0008] Preferably, the optical fiber vibration event data includes multiple abnormal event categories, and each training sample under each abnormal event category consists of m time domain points and n adjacent space points, that is, each training sample is a matrix with m rows and n columns.

[0009] Preferably, the step of inputting the training samples into the improved 1DCNN deep learning model comprises: The data points in the matrix of m rows and n columns are spliced ​​row by row to obtain a spliced ​​signal; The spliced ​​signal is input into the improved 1DCNN deep learning model.

[0010] As a preferred method, the process of optimizing the SVM classifier using the Gray Wolf Optimization Algorithm is as follows: Set the wolf pack size and the maximum number of iterations, and set the penalty parameter in the SVM classifier and kernel function parameters Set as the two-dimensional coordinates of the individual positions of the wolf pack, and randomly initialize the penalty parameters and kernel function parameters ; The training samples of the training set are input into the SVM classifier, the fitness value is calculated according to the output of the SVM classifier, and the wolf levels are classified according to the fitness values. The wolf levels include Wolf, Wolf, Wolf and Wolf, the formula for calculating the fitness value is as follows: In the formula, is the fitness value, is the number of correct classifications, is the number of misclassifications; The mathematical model of gray wolf hunting is as follows: In the formula, is the distance between the gray wolf and its prey, is the number of iterations, is the current position of the prey, Indicates the current location of the wolf being searched. To search for the updated location of the wolf, is the first coefficient, is the second coefficient; it is calculated using the following formula: In the formula, is the convergence factor, as the number of iterations increases Update, linearly decrease from 2 to 0; and is a random variable with a range of [0,1]; Wolf and Wolf, Wolf, The distances between the wolves are shown below: The mathematical model of the wolf's step length and direction is as follows: In the formula, , , They are Wolf, Wolf, Wolf and The distance of the wolf, , , They are Wolf, Wolf, The first coefficient corresponding to the wolf, , , They are Wolf, Wolf, The wolf's current location, , , For Wolf, Wolf, The second coefficient corresponding to the wolf, , , They are Wolf Target Wolf, Wolf, The updated position of the wolf; The mathematical formula for the wolf's final updated position is: Calculate the fitness value of the updated individual and record the optimal fitness value of the current iteration as ,if , for The fitness value of the wolf is Updated to , and record the optimal position as The wolf's position; if , for The fitness value of the wolf is Assign to Wolf, and record the optimal position as The position of the wolf; if , for The fitness value of the wolf is Assign to Wolf, and record the optimal position as The location of the wolf; Determine whether the number of iterations reaches the maximum number of iterations. If so, the loop terminates and the optimal penalty parameter is obtained. and the optimal kernel function parameters , and use the optimal penalty parameter and the optimal kernel function parameters Get the SVM classifier optimized by the gray wolf optimization algorithm; if it is not reached, return to continue iterating.

[0011] The present invention provides a method for identifying abnormal events in a perimeter security zone based on optical fiber signals, which has the following beneficial effects compared with the prior art: (1) The present invention inserts BN (batch normalization) between the convolution layer and the activation function layer of the convolution module of the 1DCNN deep learning model. Adding the batch normalization layer can speed up the training of the network and enhance the generalization ability.

[0012] (2) The present invention inserts a CBAM attention mechanism module between the first convolution module and the second convolution module of the 1DCNN deep learning model, assigning different weights to each channel data, thereby solving the problem that the 1DCNN deep learning model regards each channel data as equally important and enhancing the local feature extraction capability of the 1DCNN deep learning model.

[0013] (3) Since the fiber optic vibration event data is time series data, and the 1DCNN deep learning model cannot capture the time series characteristics of the data, the present invention uses a long short-term memory network (LSTM) that can capture the time correlation between the data after the 1DCNN deep learning model extracts the spatial characteristics between the data, thereby fully extracting the spatiotemporal characteristics of the fiber optic vibration data, solving the problem that the 1DCNN deep learning model cannot capture the time correlation between the fiber optic data, thereby greatly improving the accuracy of vibration event recognition and solving the problem of high false alarm rate in the current fiber optic vibration measurement perimeter security.

[0014] (4) After the 1DCNN deep learning model extracts features, the traditional method is to directly input them into the fully connected layer, flatten these feature maps, and then use 2 to 3 layers of fully connected layers for dimensionality reduction. However, the amount of fiber vibration data is too large, and the amount of data after the features extracted by the model are flattened is too large, resulting in problems such as too many model parameters, large amount of calculation, and overfitting. The present invention uses global mean pooling to replace the first few layers of the fully connected layer, and only retains the last layer, which greatly reduces the parameters in the network and avoids the problem of too many parameters caused by the fully connected layer.

[0015] (5) Since the Softmax classifier has a slow calculation speed, is greatly affected by noise interference, and only performs a simple normalization operation on the fully connected output data in accordance with the probability distribution, it is unable to further improve the classification performance of the convolutional neural network. These problems ultimately lead to unsatisfactory classification results and are not suitable for optical fiber vibration data sets. Therefore, the present invention uses the SVM classifier optimized by the Gray Wolf Optimization Algorithm to replace the conventional Softmax classifier in 1DCNN to identify vibration events based on the extracted features, further improving the recognition accuracy of vibration events. The Gray Wolf Optimization Algorithm is used to improve the SVM classifier and find the optimal parameter combination of the SVM classifier, which improves the classification performance compared to the conventional SVM classifier.

[0016] (6) The method of the present invention is an end-to-end recognition method. The entire process does not require human intervention. It can directly extract features from the original optical fiber signal and identify the type of vibration event. It solves the problems of traditional feature extraction methods that require artificial feature creation, are heavily dependent on expert knowledge, and have weak generalization capabilities. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 It is a flow chart of a method for identifying abnormal events in a perimeter security zone based on optical fiber signals of the present invention; Figure 2 This is a training flow chart of the improved 1DCNN deep learning model of the present invention; Figure 3 It is a visualization effect diagram of the features extracted from the first experimental object in the experiment of the present invention; Figure 4 It is a visualization effect diagram of the features extracted from the second experimental object in the experiment of the present invention; Figure 5 This is a visualization effect diagram of the features extracted from the third experimental object in the experiment of the present invention; Figure 6 This is a schematic diagram of the confusion matrix of the 1DCNN-Softmax() model for optical fiber signal classification in the experiment of the present invention; Figure 7This is a schematic diagram of the confusion matrix of the optical fiber signal classification effect of the 1DCNN-CBAM-Softmax() model in the experiment of the present invention; Figure 8 This is a schematic diagram of the confusion matrix of the 1DCNN-CBAM-LSTM-Softmax() model for optical fiber signal classification in the experiment of the present invention; Fig. 9 This is a schematic diagram of the confusion matrix of the optical fiber signal classification effect of the 1DCNN-CBAM-LSTM-GWO-SVM model in the experiment of the present invention. DETAILED DESCRIPTION

[0018] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0019] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art of the present invention. The terms used in the specification of the present invention herein are only for the purpose of describing specific embodiments and are not intended to limit the present invention.

[0020] like Figure 1 As shown, this embodiment provides a method for identifying abnormal events in a perimeter security zone based on optical fiber signals, comprising the following steps: First, we set up a deep learning model computing environment, using a computer as a server, building a fiber vibration event database on the server, and creating a data table to save the data. Then, we prepared a Windows operating system workbench locally, downloaded the data on the server to the local workbench, and built a deep learning model and classifier using Python based on the dimension and length of the data.

[0021] Step 1: Collect a fiber vibration event data set through a Φ-OTDR (Phase-Sensitive Optical Time Domain Reflectometer) distributed fiber sensing system, and use a preprocessing method corresponding to the fiber vibration event data set collected by the Φ-OTDR distributed fiber sensing system.

[0022] Specifically, the optical fiber vibration event data is collected, including six typical events: background noise, digging, knocking, watering, shaking fences, and walking. In the data set, each sample of each event consists of 10,000 time domain points and 12 adjacent spatial points. Therefore, the data format of each vibration event is a matrix of 10,000 rows and 12 columns. Due to the change in signal amplitude, the signal is normalized to frame these data within a certain range of values. Its main purpose is to eliminate the dimensional influence between the various evaluation indicators and the negative impact caused by singular sample data. The formula is as follows:

[0023] in, Indicates that each dimension data is linearly normalized. Represents the raw data of each dimension data, Indicates the maximum value in each dimension data. It represents the minimum value in each dimension of data. After linear normalization, the data of each dimension is mapped to the unified metric space of [0,1].

[0024] Step 2. Input the training samples into the improved 1DCNN deep learning model. The improved 1DCNN deep learning model includes a four-layer convolution module, a long short-term memory network, a global mean pooling layer, a fully connected layer, and an SVM classifier optimized by the gray wolf optimization algorithm. Each layer of the convolution module includes a convolution layer, a batch normalization layer, a Relu activation function layer, and a maximum pooling layer, and an attention mechanism module is connected between the first layer of the convolution module and the second layer of the convolution module.

[0025] like Figure 2 As shown in the figure, the normalized signals of the 12 nodes are first connected and input into the 1DCNN network to extract the local features of all sensing points and perform dimensionality reduction. The improved 1DCNN consists of 4 convolution modules, each of which includes a convolution layer, a batch normalization layer, a maximum pooling layer, and a relu activation function layer. The convolution layer is used to extract the local features in the vibration signal and the feature structure hidden in the data. The formula is as follows:

[0026] in, For the Layer convolutional layer; For the The output of the convolutional layer, is the activation function; For the The convolutional layer corresponds to the The input of the convolution kernel; For the The convolutional layer The weight of the convolution kernel; For the The convolutional layer The bias of the convolution kernel; is the convolution operation; is the total number of convolution kernels.

[0027] The batch normalization layer is an optimization method to solve the internal covariate shift caused by input transformation in convolutional neural networks. By calculating the variance and standard deviation on a small batch training set, the output distribution of each layer is standardized, which speeds up the network training and enhances the generalization ability. The calculation process of batch normalization is as follows:

[0028] in, represents the size of the mini-batch, and They are respectively The input and output of observations, represents the mean of the small batch samples, is the standard deviation of the mini-batch samples. is a constant close to zero to ensure numerical stability, represents the data after regularization, Express The data obtained by scaling and translating, and is a parameter, Used to adjust the variance of the numerical distribution. When model training starts, set , , and then gradually adjust to the appropriate value through the learning process.

[0029] The activation layer processes the operation results of the convolution layer to obtain the corresponding output results. It is usually nonlinear, which can fit the nonlinear relationship between the number of samples and labels in the neural network. The Relu activation function maps negative numbers to 0 and keeps positive numbers unchanged, which can effectively reduce the problem of gradient disappearance, reduce the difficulty of network training, and speed up convergence. The formula is as follows:

[0030] In the formula, Represents the output of the Relu activation function, and max represents the maximum value of the two numbers. It is the input of Relu activation function.

[0031] The pooling layer, also known as the downsampling layer, can play a role in information screening. It performs pooling processing on the feature map through a sliding window, which reduces the dimension of the feature map and the number of parameters while retaining significant feature information. The maximum pooling layer extracts the maximum value of the data in the local area. The formula is as follows:

[0032] In the formula, For the Tier The feature matrix Pooling results; For the Tier The feature matrix Secondary pooling, the area processed by the pooling kernel; is the number of pooling operations; is the width of the pooling kernel.

[0033] The CBAM attention mechanism is inserted between the convolution modules. 1DCNN considers each channel feature to be equally important during the convolution pooling process, but in reality, the information carried by each channel feature is of different importance. It is unreasonable to simply assume that each channel feature is the same. The convolution attention module CBAM improves the information flow within the network by understanding the information to be emphasized or suppressed, and finds more important feature factors, focusing all attention on these important areas to improve classification accuracy. By differentiating the weights assigned to different channel information, important features are emphasized and unnecessary feature information is suppressed.

[0034] Fiber vibration data is time series data, and 1DCNN cannot learn the time-related characteristics in the signal. The spatial local features extracted by 1DCNN are input into the long short-term memory network LSTM, and LSTM extracts time-related features from it. The calculation formula is as follows:

[0035] in, for The importance of the results of the input gate at all times; Calculate the sigmoid function; is the weight matrix of the input gate; is the offset matrix of the input gate; for The degree to which the results of the moment-forget gate are retained; is the weight matrix of the forget gate; is the offset matrix of the forget gate; Select information for the memory cell's history; is the weight matrix of the memory cell; is the offset matrix of the memory cells; for Memory cells of the moment; for Memory cells of the moment; for The degree of information selection of the output gate results at each moment; is the weight matrix of the output gate; is the offset matrix of the output gate; for Output at the moment; for Output at the moment; is the tanh activation function, express The local characteristics of the space input at each moment.

[0036] The present invention uses global mean pooling to replace the first few layers of the fully connected layer in the original model, retaining only the last fully connected layer, and inputs the spatiotemporal features extracted by the long short-term memory network LSTM into the global mean pooling layer. The global mean pooling layer is proposed to solve the problem of too many parameters in the fully connected network. The formula is as follows: In the formula, In the output feature map The information of each channel after global mean pooling operation; Indicates In the channel eigenvalues; is the dimension of the feature vector.

[0037] Furthermore, this embodiment introduces an SVM classifier and optimizes the SVM classifier using the Grey Wolf Optimization (GWO) algorithm to replace the traditional Softmax classifier in the 1DCNN model, and inputs the features extracted by the deep learning model into the optimized SVM classifier for event recognition.

[0038] Specifically, we first construct a support vector machine (SVM) classifier, map the input sample set to a high-dimensional feature space, and construct an optimal hyperplane in the space that can be regarded as a decision boundary. The optimal hyperplane is used to effectively split the sample set and find support vectors on both sides of the hyperplane. These support vectors are the sample points closest to the hyperplane, which have good generalization ability and robustness. The expression of the input sample set is: in, is the input sample set, For the input variables; For the output variables; is the vector dimension, Indicates the number of samples in the sample set.

[0039] The optimal hyperplane for partitioning can be expressed as: in, is the optimal hyperplane normal vector, is the distance between the hyperplane and the origin, and to find the appropriate and The following questions need to be answered: in, is the penalty factor; is an insensitive factor; and is the relaxation factor.

[0040] In order to solve some sample sets that cannot be separated in linear space, the samples are mapped from the input space to a higher-dimensional feature space so that the samples can be linearly separated in the new space. The kernel function is introduced. And the Lagrange equation is solved to obtain the high-dimensional space classification hyperplane: in, Lagrange multipliers, represents the number of support vectors, For the data points, For the data points.

[0041] The kernel function uses the Gaussian radial basis kernel function, and the formula is as follows: In the formula, is the influencing factor of the training sample, that is, the kernel function parameter.

[0042] Then, the features output by the fully connected layer of the improved 1DCNN deep learning model are input into the constructed SVM classifier. At the same time, the Grey Wolf Optimization Algorithm GWO is used to optimize the SVM classifier to find the optimal parameters of the SVM classifier. The specific steps include:

[0043] Step A: First, normalize the extracted features, set the wolf pack size and the maximum number of iterations, and set the parameters in the SVM classifier. and Set as the two-dimensional coordinates of the individual positions of the wolf pack, randomly initialized and ; Step B: Train the SVM classifier using the training set samples to calculate the fitness value: Wolf, Wolf, Wolf and Wolves) are classified according to their fitness values, and the formula for calculating the fitness value is as follows: in, is the fitness value, is the number of correct classifications, is the number of misclassifications.

[0044] The mathematical model of gray wolf hunting is as follows: In the formula, is the distance between the gray wolf and its prey, is the current iteration number, is the current position of the prey, Indicates the current location of the wolf being searched. To search for the updated location of the wolf, is the first coefficient, is the second coefficient. It can be calculated using the following formula:

[0045] In the formula, is the convergence factor, as the number of iterations increases The update is linearly decreased from 2 to 0; and is a random variable with a range of [0,1].

[0046] Wolf and Wolf, Wolf, The distances between the wolves are shown below: The mathematical model of the wolf's step length and direction is as follows: In the formula, , , They are Wolf, Wolf, The location of the wolf; , , They are Wolf, Wolf, Wolf and The distance of the wolf.

[0047] The mathematical formula for the wolf's final updated position is: Step C: Calculate the fitness value of the updated individual and record the optimal fitness value of the current generation as ,if ( The fitness of the wolf), then Updated to , and record the corresponding position. , then Assign to Wolf, the corresponding position is updated to The wolf's position. , then Assign to Wolf, the corresponding position is also updated to The position of the wolf. Obviously, when iterating to the current generation, The position of the wolves is the optimal position for the population.

[0048] Step D: Determine whether the number of iterations reaches the maximum number of iterations. If so, the loop terminates and the optimal parameters are obtained. and ; If not reached, return to step B and continue iterating.

[0049] Step E: Use the best parameters and Build the optimized SVM classifier.

[0050] Step 3: Calculate the first loss function based on the output of the fully connected layer in the improved 1DCNN deep learning model and the label of the training sample, and update the weight parameters of the improved 1DCNN deep learning model before the SVM classifier optimized by the gray wolf optimization algorithm according to the first loss function, and repeatedly calculate the first loss function based on the training sample and update until the training is completed. The first loss function is the cross entropy loss function.

[0051] Step 4: fix the weight parameters before the SVM classifier optimized by the gray wolf optimization algorithm in the improved 1DCNN deep learning model, calculate the second loss function according to the output of the SVM classifier optimized by the gray wolf optimization algorithm and the label of the training sample, adjust the parameters of the SVM classifier optimized by the gray wolf optimization algorithm according to the second loss function, and repeatedly calculate the second loss function based on the training sample until the training is completed. The second loss function is the loss function of the SVM classifier itself.

[0052] Step 5: Take the trained improved 1DCNN deep learning model and output the abnormal event recognition result for the optical fiber vibration event data to be recognized. In the recognition process, data is also collected and normalized based on time domain points and spatial points. The data of all spatial points are connected and input into the improved 1DCNN deep learning model to obtain the abnormal event recognition result.

[0053] In order to verify the classification and recognition performance of the model of the present invention, the experiment was carried out on a computer with 8GB memory, Inter(R)Core(TM)i7-14700HX, and RTX4070GPU. The model was trained on python3.9 and verified using an authoritative dataset.

[0054] The experimental data contains 15,612 samples, including 6 typical events: background noise, digging, knocking, watering, shaking fences, and walking. The number of background noise samples is 3,094, numbered as event 0; the number of digging samples is 2,512, numbered as event 1; the number of knocking samples is 2,530, numbered as event 2; the number of watering samples is 2,298, numbered as event 3; the number of shaking fence samples is 2,728, numbered as event 4; the number of walking samples is 2,450, numbered as event 5. 80% of them are randomly selected as training sets and 20% as test sets. The data is input into the model to evaluate the model performance.

[0055] The classification effect depends largely on the quality of feature extraction. In order to prove that the addition of CBAM attention mechanism module and long short-term memory network LSTM improves the feature extraction ability of 1DCNN model, the feature distinguishability is compared. The final feature vector obtained by the trained model is mapped into a three-dimensional space vector through the Linear Discriminant Analysis (LDA) algorithm. Figure 3-Figure 5As shown, each color corresponds to a vibration event. The stronger the discriminability of the model feature extraction, the stronger the discreteness between events. It can be seen that 1DCNN-CBAM (recorded as the second experimental object, on the basis of the first experimental object, a CBAM attention mechanism module is connected between the first layer convolution module and the second layer convolution module) has a stronger feature extraction capability than 1DCNN (recorded as the first experimental object, which is the part of the present invention including four layers of convolution modules, global mean pooling layer and fully connected layer, and the first layer of convolution module and the second layer of convolution module are not connected with the attention mechanism module). 1DCNN-CBAM-LSTM (recorded as the third experimental object, on the basis of the second experimental object, a long short-term memory network LSTM is added between the last layer of convolution module and the global mean pooling layer) has captured the temporal correlation between features due to the addition of LSTM, and the feature extraction capability is further improved.

[0056] The confusion matrix is ​​an error matrix that is often used to visually evaluate the performance of a classification model. The main diagonal represents the number of correctly classified samples, while the other cells represent the number of incorrectly classified samples. The darker the color, the larger the number of samples. The confusion matrix obtained after inputting the data into the model is as follows Figure 6-Figure 9As shown in the figure, the accuracy, precision, recall rate and F1 score evaluation criteria of each model are calculated according to the confusion matrix as shown in Tables 1 to 4. From the confusion matrix diagram and the evaluation criteria table, it can be seen that different models have different performances in the identification of optical fiber vibration events. The 1DCNN-Softmax() model (combined with the Softmax classifier for the first experimental object) has low precision on events 1, 2 and 5, low recall rates on events 0 and 3, and an accuracy rate of only 95%. This may be due to its insufficient ability to extract local features. The simple 1DCNN is difficult to undertake the task of classifying optical fiber vibration events. The addition of the CBAM hybrid attention mechanism module enables the model to emphasize important features and suppress unnecessary feature information, enhances the feature extraction ability of 1DCNN, and significantly improves the evaluation indicators of each event. However, the precision of event 5 and the recall rate of event 3 are still not high. This may be because the 1DCNN-CBAM-Softmax() model (combined with the Softmax classifier for the second experimental object) cannot learn the time-related characteristics in the signal. After adding the LSTM neural network, since the 1DCNN-CBAM-LSTM-Softmax() model (combined with the Softmax classifier for the third experimental object) fully extracted the local and temporal features of the signal, the evaluation indicators of each event, except for the recall rate and F1 score of event 5, exceeded 98%, and the final accuracy rate reached 98.7%, indicating that the model has excellent feature extraction capabilities. In the last two groups of experiments, the classification effects of the Softmax() classifier and the GWO-SVM classifier on optical fiber vibration events were compared. It can be seen that the GWO-SVM classifier has improved the evaluation indicators of almost every event compared with the Softmax() classifier, especially the recall rate of event 5, which has increased from 96.7% to 99.0%, and the evaluation indicators of each event have exceeded 98%, and the accuracy rate has reached 99.3%, indicating that the GWO-SVM classifier has achieved better classification effects than the Softmax classifier, and is more suitable for the classification and identification of optical fiber vibration events. Among them, the 1DCNN-CBAM-LSTM-GWO-SVM model is an improved 1DCNN deep learning model of the present invention.

[0057] Table 1 1DCNN-Softmax() model performance indicators Event Type Accuracy Recall F1 score Accuracy 0 0.973 0.931 0.951 0.950 1 0.933 0.950 0.942 0.950 2 0.913 0.996 0.953 0.950 3 0.981 0.898 0.938 0.950 4 0.961 0.982 0.971 0.950 5 0.939 0.939 0.939 0.950 Table 2 1DCNN-CBAM-Softmax() model performance indicators Event Type Accuracy Recall F1 score Accuracy 0 0.985 0.981 0.983 0.966 1 0.966 0.962 0.964 0.966 2 0.955 1 0.977 0.966 3 0.980 0.917 0.947 0.966 4 0.967 0.978 0.973 0.966 5 0.943 0.953 0.948 0.966 Table 3 1DCNN-CBAM-LSTM-Softmax() model performance indicators Event Type Accuracy Recall F1 score Accuracy 0 0.994 0.989 0.991 0.987 1 0.982 0.984 0.983 0.987 2 0.986 1 0.993 0.987 3 0.989 0.985 0.987 0.987 4 0.985 0.993 0.989 0.987 5 0.981 0.967 0.974 0.987 Table 4 1DCNN-CBAM-LSTM-GWO-SVM model performance indicators Event Type Accuracy Recall F1 score Accuracy 0 0.997 0.997 0.997 0.993 1 0.994 0.986 0.990 0.993 2 0.996 0.998 0.997 0.993 3 0.989 0.996 0.993 0.993 4 0.996 0.993 0.994 0.993 5 0.986 0.990 0.988 0.993 The accuracy rate is the overall accuracy rate of the model for the six types of events.

[0058] The technical features of the above-described embodiments may be arbitrarily combined. To make the description concise, not all possible combinations of the technical features in the above-described embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0059] The above-mentioned embodiments only express several implementation modes of the present invention, and the description thereof is relatively specific and detailed, but it cannot be understood as limiting the scope of the invention. It should be pointed out that, for a person of ordinary skill in the art, several modifications and improvements can be made without departing from the concept of the present invention, and these all belong to the protection scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the attached claims.

Claims

1. A method for identifying abnormal events in a perimeter security zone based on optical fiber signals, characterized in that: The method for identifying abnormal events in a perimeter security zone based on optical fiber signals comprises: The optical fiber vibration event data set is collected by phase-sensitive optical time domain reflectometry and preprocessed to form a training set, wherein the training samples in the training set are the preprocessed optical fiber vibration event data; Inputting the training samples into an improved 1DCNN deep learning model, the improved 1DCNN deep learning model includes a four-layer convolution module, a long short-term memory network, a global mean pooling layer, a fully connected layer, and an SVM classifier optimized by the gray wolf optimization algorithm, wherein each layer of the convolution module includes a convolution layer, a batch normalization layer, a Relu activation function layer, and a maximum pooling layer, and an attention mechanism module is connected between the first layer of the convolution module and the second layer of the convolution module; Calculate a first loss function based on the output of the fully connected layer in the improved 1DCNN deep learning model and the label of the training sample, and update the weight parameters of the improved 1DCNN deep learning model before the SVM classifier optimized by the gray wolf optimization algorithm according to the first loss function until the training is completed; The weight parameters before the SVM classifier optimized by the gray wolf optimization algorithm in the improved 1DCNN deep learning model are fixed, the second loss function is calculated according to the output of the SVM classifier optimized by the gray wolf optimization algorithm and the labels of the training samples, and the parameters of the SVM classifier optimized by the gray wolf optimization algorithm are adjusted according to the second loss function until the training is completed; Take the trained improved 1DCNN deep learning model and output the abnormal event recognition results for the optical fiber vibration event data to be identified.

2. The method for identifying abnormal events in perimeter security zones based on optical fiber signals according to claim 1, characterized in that: The optical fiber vibration event data includes multiple abnormal event categories, and each training sample under each abnormal event category consists of m time domain points and n adjacent space points, that is, each training sample is a matrix with m rows and n columns.

3. The method for identifying abnormal events in perimeter security zones based on optical fiber signals according to claim 2, characterized in that: The step of inputting the training samples into the improved 1DCNN deep learning model comprises: The data points in the matrix of m rows and n columns are spliced ​​row by row to obtain a spliced ​​signal; The spliced ​​signal is input into the improved 1DCNN deep learning model.

4. The method for identifying abnormal events in perimeter security zones based on optical fiber signals according to claim 1, characterized in that: The process of optimizing the SVM classifier using the Gray Wolf Optimization Algorithm is as follows: Set the wolf pack size and the maximum number of iterations, and set the penalty parameter in the SVM classifier and kernel function parameters Set as the two-dimensional coordinates of the individual positions of the wolf pack, and randomly initialize the penalty parameters and kernel function parameters ; The training samples of the training set are input into the SVM classifier, the fitness value is calculated according to the output of the SVM classifier, and the wolf levels are classified according to the fitness values. The wolf levels include Wolf, Wolf, Wolf and Wolf, the formula for calculating the fitness value is as follows: In the formula, is the fitness value, is the number of correct classifications, is the number of misclassifications; The mathematical model of gray wolf hunting is as follows: In the formula, is the distance between the gray wolf and its prey, is the number of iterations, is the current position of the prey, Indicates the current location of the wolf being searched. To search for the updated location of the wolf, is the first coefficient, is the second coefficient; it is calculated using the following formula: In the formula, is the convergence factor, as the number of iterations increases Update, linearly decrease from 2 to 0; and is a random variable with a range of [0,1]; Wolf and Wolf, Wolf, The distances between the wolves are shown below: The mathematical model of the wolf's step length and direction is as follows: In the formula, , , They are Wolf, Wolf, Wolf and The distance of the wolf, , , They are Wolf, Wolf, The first coefficient corresponding to the wolf, , , They are Wolf, Wolf, The wolf's current location, , , For Wolf, Wolf, The second coefficient corresponding to the wolf, , , They are Wolf Target Wolf, Wolf, The updated position of the wolf; The mathematical formula for the wolf's final updated position is: Calculate the fitness value of the updated individual and record the optimal fitness value of the current iteration as ,if , for The fitness value of the wolf is Updated to , and record the optimal position as The wolf's position; if , for The fitness value of the wolf is Assign to Wolf, and record the optimal position as The position of the wolf; if , for The fitness value of the wolf is Assign to Wolf, and record the optimal position as The location of the wolf; Determine whether the number of iterations reaches the maximum number of iterations. If so, the loop terminates and the optimal penalty parameter is obtained. and the optimal kernel function parameters , and use the optimal penalty parameter and the optimal kernel function parameters Get the SVM classifier optimized by the gray wolf optimization algorithm; if it is not reached, return to continue iterating.

Images