A method for screening massive target data from distributed optical fiber sensing based on image processing

Abstract

A method for screening massive target data of distributed fiber optic sensing based on image processing. This method maps the spatiotemporal two-dimensional vibration signal into a grayscale image, utilizes multimodal fusion technology, and combines a lightweight convolutional neural network to achieve efficient classification. It also introduces a dynamic incremental learning mechanism and a spatial propagation attenuation correction algorithm to solve the problems of single feature representation, delayed model update, and insufficient positioning accuracy in traditional methods. This method is applicable to fields such as perimeter security, pipeline monitoring, and geological disaster warning, and can perform real-time screening and target identification of massive vibration data under complex environmental interference. The present invention breaks through the limitations of traditional time-frequency domain analysis through signal imaging representation, significantly improving the ability to distinguish similar events. At the same time, it relies on dynamic model optimization technology to reduce false alarm rates and operation and maintenance costs, and combines the spatial correction algorithm to effectively suppress the impact of fiber optic attenuation on positioning accuracy, providing reliable technical support for the intelligent upgrade of distributed fiber optic sensing systems.

Description

Technical Field

[0001] The present invention belongs to the technical field of intersection of optical fiber sensing and image processing, and specifically designs a distributed optical fiber sensing massive data screening method based on a lightweight convolutional neural network and dynamic incremental learning that integrates image processing and multimodal features. Background Art

[0002] Distributed fiber optic sensing technology, based on the sensing principle of phase-sensitive optical time-domain reflectometry (Φ-OTDR), enables long-distance vibration monitoring in areas such as perimeter security and pipeline monitoring. However, the massive amounts of data generated still face real-time processing bottlenecks. Traditional methods rely on single-dimensional analysis in the time or frequency domain, making it difficult to distinguish the spatiotemporal coupling characteristics of environmental interference from real-world events, resulting in high false alarm rates. Existing shallow machine learning algorithms suffer from feature confusion in multi-target scenarios, while cloud-based offline processing methods incur minute-level delays, making them unable to meet the real-time warning requirements in high-risk scenarios.

[0003] Current deep learning applications face significant limitations: conventional convolutional neural network (CNN) models have excessively high computational loads, making them difficult to adapt to edge devices; static models cannot cope with data distribution shifts caused by environmental changes, requiring frequent manual iterations; and existing signal visualization methods are limited to time-frequency spectrum conversion and lack multimodal fusion of texture, spatial topology, and dynamic behavior, making it difficult to identify similar signals (such as construction vibration and geological hazards). To address these shortcomings, a technical solution integrating lightweight networks, dynamic incremental learning, and collaborative analysis of multimodal features is urgently needed to overcome the technical challenges of real-time screening of massive data, adaptive modeling of dynamic environments, and high-precision target recognition, thereby improving the reliability and timeliness of fiber-optic sensing systems in critical infrastructure monitoring. Summary of the Invention

[0004] Based on the above problems, the present invention provides a method for screening massive target data of distributed fiber optic sensing based on image processing, which solves the technical difficulties of traditional methods such as high false alarm rate, poor real-time performance and weak environmental adaptability caused by single feature representation, delayed model update and insufficient coordination of multimodal information.

[0005] The technical solution adopted in the present invention is as follows:

[0006] A method for screening massive target data of distributed optical fiber sensing based on image processing, comprising the following steps:

[0007] Step 1: Collect spatiotemporal two-dimensional vibration signals through a distributed optical fiber sensing system, normalize the signal amplitude, and convert it into a grayscale image;

[0008] Step 2: Extract multimodal features from the grayscale image in parallel. Specifically, the local binary pattern algorithm is used to extract texture features from the grayscale image, and the shape features of the event region are extracted based on the Hu invariant moment calculation. The motion vector field of adjacent frames is calculated using the Lucas-Kanade optical flow method to extract spatiotemporal dynamic features.

[0009] Step 3: Combine texture features, shape features, and dynamic features to generate a joint feature vector;

[0010] Step 4: Use the MobileNetV3 network architecture to process the joint feature vector and obtain the classification confidence;

[0011] Step 5: Set the dynamic incremental learning trigger. When the classification confidence is less than θ, the online hard example mining strategy is triggered and the model parameters are updated using the hybrid loss function.

[0012] Step 6: Perform spatial propagation attenuation correction and perform optical fiber attenuation compensation on the classification results;

[0013] Step 7: When incremental learning is triggered continuously for K ≥ 3 times, pseudo labels are generated based on historical high-confidence samples and expanded to the training dataset to achieve autonomous iterative optimization of the model.

[0014] Furthermore, the signal amplitude is normalized and converted into a grayscale image as follows:

[0015] G (x,t) =floor(255*(S max -S min ) / (S (x,t) -S min ))

[0016] Among them, G (x,t) represents a grayscale image, x is the spatial coordinate of the sensing fiber, t is the time variable; floor refers to the integer downward; S min 、S max are the minimum and maximum values ​​of the signal amplitude respectively; S (x,t) It is a two-dimensional vibration signal in space and time.

[0017] Furthermore, the local binary pattern algorithm is used to extract the grayscale image texture features as follows:

[0018]

[0019] Among them, F texture is the texture feature, g c is the pixel value of the center pixel, g p is the pixel value of the pth sampling point in the neighborhood, and s(·) is a step function.

[0020] Furthermore, the calculation of Hu invariant moments is as follows:

[0021] The main invariant moment combination M1, M2, M3, M4, M5, M6, and M7 calculates the shape feature F of the event area shape :

[0022] M1=(η 20 +η 02 ) 2

[0023] M2=(η 20 +η 02 ) 2 +4η 11 2

[0024] Among them, the first-order moment M1 is used to describe the pixel distribution range, and the second-order moment M2 is used to capture the directional characteristics; the calculation formula of Hu invariant moments M3~M7 is as follows:

[0025] M3=(η 30 -3η 12 ) 2 +(3η 21 -η 03 ) 2

[0026] M4=(η 30 +η 12 ) 2 +(η 21 +η 03 ) 2

[0027] M5=(η 30 +3η 12 )[(η 21 +η 03 ) 2 -3(η 21 +η 03 )] 2

[0028] M6=(η 20 -η 02 )[(η 30 +η 12 ) 2 -(η 21 +η 03 ) 2 ]+4η 11 (η 30 +η 12 )(η 21 +η 03 )]

[0029] M7=(3η21 +η 03 )(3η 21 +η 03 )[(η 30 +η 12 ) 2 -3(η 21 +η 03 ) 2 ]-(η 30 +3η 12 )(η 21

[0030] +η 03 )[3(η 30 +η 12 ) 2 -(η 21 +η 03 ) 2 ]

[0031] Among them, M3 describes the symmetry, M4, M5, M6 and M7 describe the curvature; pq is the normalized central moment, which is calculated as follows:

[0032]

[0033] Among them, u pq Represents the geometric moment of the image relative to the center of mass, which is used to eliminate the effect of translation; u 00 represents the zero-order central moment.

[0034] Furthermore, the dynamic feature satisfies the brightness constant constraint equation:

[0035]

[0036] Among them, x, y, t represent the spatial coordinates and timestamps, Gu, Gv, G represent the horizontal displacement, vertical displacement and start-end linear displacement of the pixel, It means to find partial derivatives of variables x, y, t, Gu, Gv, and G.

[0037] Furthermore, step 3 adopts an attention weighting mechanism, and the weight coefficient is calculated as follows:

[0038]

[0039] Among them, w ij represents the similarity between the features of each modality, q is the learnable query vector, F i With F j is the feature vector of each modality. Furthermore, the process of processing the joint feature vector by the MobileNetV3 network architecture is expressed as follows:

[0040] FCNN=DWConv(ReLU(BN(F joint )))

[0041] Among them, DWConv is a depth-separable convolution, F joint ReLU represents the joint feature; ReLU represents the modified activation function, BN represents the batch normalization function, and the output feature map FCNN; finally, the feature map FCNN uses the normalization function to obtain the maximum probability score of the target object, which is used as the classification confidence P c Output.

[0042] Furthermore, the hybrid loss function calculation process is as follows:

[0043] L=λ·LCE+(1-λ)·L triplet

[0044] Among them, L represents the regularization loss, LCE is the cross entropy loss, and L triplet is the triplet loss and λ is the regularization parameter.

[0045] Furthermore, the correction calculation formula of step 6 is as follows:

[0046]

[0047] Among them, P final (x) represents the final power at position x, P c (x) represents the initial power at the reference point x0, α ​​is the fiber attenuation coefficient, x0 is the initial position of the event, and |x-x0| represents the absolute distance from the reference point x0 to the fiber sensor position x.

[0048] The advantages and technical effects of the present invention are as follows:

[0049] This invention is applicable to perimeter security, oil and gas pipeline monitoring, rail transit status perception, and geological disaster early warning. Based on image signal representation and multimodal feature fusion technology, it can efficiently screen massive amounts of vibration data in complex scenarios such as wind and rain interference, mechanical vibration, and geological deformation, increasing the accuracy of target event recognition to over 95%.

[0050] The present invention does not require the use of high-cost hardware acceleration equipment or frequent manual labeling, and is suitable for the deployment of edge computing nodes. It relies on lightweight convolutional neural networks (CPU computing load is reduced by 70%) and dynamic incremental learning mechanisms to improve the generalization ability of traditional static models. Through the collaborative analysis of spatiotemporal-texture-shape multimodal features, an environmentally adaptive data screening system is established to achieve millisecond-level real-time response at a sampling frequency of 10kHz. At the same time, it supports autonomous model evolution of long-period scenarios (such as seasonal changes and geological subsidence), and the false alarm rate is stably below 5% for a long time. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 It is a schematic diagram of the composition of each functional module provided by the present invention.

[0052] Figure 2 It is a flow chart of the method of the present invention. DETAILED DESCRIPTION

[0053] See also Figure 1 , Figure 1 This is a schematic diagram of signal image conversion provided by an embodiment of the present invention; please refer to Figure 2 , Figure 2 Flowchart of an embodiment of the present invention. A method for screening massive target data of distributed optical fiber sensing based on image processing, comprising the following steps:

[0054] Step 1: Use the signal conversion imaging module to realize the spatial-temporal vibration signal imaging conversion, and collect the spatial-temporal two-dimensional vibration signal S through the distributed optical fiber sensing system. (x,t) (such as ground vibration caused by vehicle driving), the signal amplitude is normalized and converted into a grayscale image G (x,t) , where x is the spatial position coordinate of the sensing fiber, and t is the time variable. The conversion formula is as follows:

[0055] G (x,t) =floor(255*(S max -S min ) / (S (x,t) -s min ))

[0056] Among them, floor refers to rounding down; S min 、S max are the minimum and maximum values ​​of the signal amplitude respectively; the sampling frequency of the spatiotemporal two-dimensional vibration signal fs≥10kHz, and the spatial resolution Δx≤1m;

[0057] Step 2: Use the multimodal feature extraction module to extract multimodal features from the grayscale image in parallel. Specifically, the local binary pattern (LBP) algorithm is used to extract the texture feature F of the grayscale image G(x,t). texture ; Calculate the event area based on Hu invariant moment to extract shape features F shape ; The Lucas-Kanade optical flow method (the relevant algorithm module can be directly called) is used to calculate the motion vector field of adjacent frames and then extract the spatiotemporal dynamic features F motion ;

[0058] Specifically, the spatial texture distribution is extracted by the LBP algorithm to obtain the texture feature F texture The calculation process is as follows:

[0059]

[0060] where g c is the pixel value of the center pixel, g p is the pixel value of the pth sampling point in the neighborhood, and s(·) is the step function;

[0061] Hu invariant moment geometric transformation (translation, rotation, scaling), main invariant moment combination M1, M2, M3, M4, M5, M6, M7 calculates the shape feature F of the event area shape , the calculation formula is as follows:

[0062] M1=(η 20 +η 02 ) 2

[0063] M2=(η 20 +η 02 ) 2 +4η 11 2

[0064] Among them, the first-order moment M1 is used to describe the pixel distribution range, and the second-order moment M2 is used to capture directional features (such as the tilt of the shape or the direction of the main axis). Furthermore, the calculation formula of Hu invariant moments M3 to M7 is as follows:

[0065] M3=(η 30 -3η 12 ) 2 +(3η 21 -η 03 ) 2

[0066] M4=(η 30 +η 12 ) 2 +(η 21 +η 03 ) 2

[0067] M5=(η 30 +3η 12 )[(η 21 +η 03 ) 2 -3(η 21 +η 03 )] 2

[0068] M6=(η 20 -η 02 )[(η 30 +η 12 ) 2 -(η 21 +η03 ) 2 ]+4η 11 (η 30 +η 12 )(η 21 +η 03 )]

[0069] M7=(3η 21 +η 03 )(3η 21 +η 03 )[(η 30 +η 12 ) 2 -3(η 21 +η 03 ) 2 ]-(η 30 +3η 12 )(η 21

[0070] +η 03 )[3(η 30 +η 12 ) 2 -(η 21 +η 03 ) 2 ]

[0071] Among them, M3 describes the symmetry, M4, M5, M6 and M7 describe the curvature. In addition, in the formula pq is the normalized central moment, which is calculated as follows:

[0072]

[0073] Among them, u pq Represents the geometric moment of the image relative to the center of mass, which is used to eliminate the effect of translation. 00 represents the zero-order central moment (i.e., the total number of pixels), which is used as a normalization criterion to eliminate the scale effect.

[0074] Based on the optical flow method, the motion vector field between adjacent frames is calculated to obtain the dynamic feature F motion , this feature satisfies the brightness constant constraint equation:

[0075]

[0076] Among them, x, y, t represent the spatial coordinates and timestamps, Gu, Gv, G represent the horizontal displacement, vertical displacement and start-end linear displacement of the pixel, It means to find partial derivatives of variables x, y, t, Gu, Gv, and G.

[0077] Step 3: Transform the texture feature F texture , shape feature Fshape and dynamic characteristics F motion Splice and generate joint feature vector F joint , to achieve feature fusion; multimodal feature fusion adopts attention weighting mechanism, and the weight coefficient calculation formula is as follows:

[0078]

[0079] Among them, w ij represents the similarity between the features of each modality, q is the learnable query vector, F i With F j is the feature vector of each modality; the dimension of the query vector q is d=128 and is dynamically updated through back propagation.

[0080] Step 4: Use the MobileNetV3 lightweight CNN network architecture to process the joint features, and use the normalization function in its fully connected layer to obtain the maximum probability score of the target object, which is used as the classification confidence P c Output. The multimodal features are input into a lightweight convolutional neural network (CNN) classification model. The CNN adopts the MobileNetV3 architecture, with a depthwise separable convolution kernel size of 3×3, a channel compression ratio of 1 / 4, and the number of neurons in the last fully connected layer is N. fc =256; This architecture uses a depthwise separable convolutional layer to output a feature map FCNN, which is calculated as follows:

[0081] FCNN=DWConv(ReLU(BN(F joint )))

[0082] Among them, DWConv is a depth-separable convolution, F joint ReLU represents the corrected activation function, and BN represents the batch normalization function.

[0083] Step 5: Set dynamic incremental learning trigger, when the classification confidence P c When θ<θ, the online hard example mining strategy is triggered, and the model parameters W are updated using the hybrid loss function, where θ∈[0.6,0.8]. The hybrid loss function calculation process is as follows:

[0084] L=λ·LCE+(1-λ)·L triplet

[0085] Among them, L represents the regularization loss, LCE is the cross entropy loss, and L triplet is the triplet loss and λ is the regularization parameter.

[0086] Step 6: Perform spatial propagation attenuation correction and perform optical fiber attenuation compensation on the classification results. The correction calculation formula is as follows:

[0087]

[0088] Among them, P final (x) represents the final power at position x, P c (x) represents the initial power at the reference point x0, α ​​is the fiber attenuation coefficient (obtained through calibration experiments), x0 is the initial position of the event (the event that causes optical signal attenuation, such as a physical disturbance), and |x-x0| represents the absolute distance from the reference point x0 to the fiber sensor position x.

[0089] Specifically, the spatial correction model can be established as follows:

[0090] Based on the attenuation compensation results of step 6 and combined with the optical fiber layout topology, the event propagation attenuation mapping matrix A is constructed to achieve collaborative correction of positioning errors in multi-event concurrent scenarios. On this basis, the event propagation attenuation mapping matrix is ​​constructed. The calculation formula of the event propagation attenuation mapping matrix is ​​as follows:

[0091] A ij =α·L ij =α·|x i -d j |

[0092] α represents the power consumption per unit distance of optical fiber, x i represents the location of the physical disturbance event, d j Indicates the signal receiving point, L ij Indicates the path length. Furthermore, when multiple events occur simultaneously, the event types are distinguished by separating the source signals, PCA analysis, and time-frequency analysis. Furthermore, a collaborative localization algorithm is used to solve the event location through an optimization function. The optimization function calculation formula is as follows:

[0093]

[0094] Among them, min represents the minimum value, S j A represents the jth detection point (sensor) signal, kj represents the signal attenuation coefficient from the jth event source to the kth detection point, Represents the kth event source signal to be estimated (such as vibration amplitude, position parameters, etc.). It represents the mean square error between the predicted signal and the actual signal.

[0095] Step 7: When the number of consecutive incremental learning triggers K ≥ 3, based on the historical high confidence samples (P c ≥0.9) to generate pseudo labels and expand them to the training dataset to achieve autonomous iterative optimization of the model.

[0096] It will be understood that the present invention is described by way of some embodiments, and it will be appreciated by those skilled in the art that various changes or equivalent substitutions may be made to these features and embodiments without departing from the spirit and scope of the present invention. In addition, under the teachings of the present invention, these features and embodiments may be modified to adapt to specific circumstances and materials without departing from the spirit and scope of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed herein, and all embodiments falling within the scope of the claims of this application are intended to be protected by the present invention.

Claims

1. A method for screening massive target data from distributed optical fiber sensing based on image processing, characterized in that: The following steps are involved: Step 1: Collect spatiotemporal two-dimensional vibration signals through a distributed optical fiber sensing system, normalize the signal amplitude, and convert it into a grayscale image; Step 2: Extract multimodal features from the grayscale image in parallel. Specifically, the local binary pattern algorithm is used to extract texture features from the grayscale image, and the shape features of the event region are extracted based on the Hu invariant moment calculation. The motion vector field of adjacent frames is calculated using the Lucas-Kanade optical flow method to extract spatiotemporal dynamic features. Step 3: Combine texture features, shape features, and dynamic features to generate a joint feature vector; Step 4: Use the MobileNetV3 network architecture to process the joint feature vector and obtain the classification confidence; Step 5: Set the dynamic incremental learning trigger. When the classification confidence is less than θ, the online hard example mining strategy is triggered and the model parameters are updated using the hybrid loss function. Step 6: Perform spatial propagation attenuation correction and perform optical fiber attenuation compensation on the classification results; Step 7: When incremental learning is triggered continuously for K ≥ 3 times, pseudo labels are generated based on historical high-confidence samples and expanded to the training dataset to achieve autonomous iterative optimization of the model.

2. The method for screening massive target data using distributed optical fiber sensing based on image processing according to claim 1, characterized in that: The signal amplitude is normalized and converted into a grayscale image as follows: G (x,t) =floor(255*(S max -S min ) / (S (x,t) -S min )) Among them, G (x,t) represents a grayscale image, x is the spatial coordinate of the sensing fiber, t is the time variable; floor refers to the integer downward; S min 、S max are the minimum and maximum values ​​of the signal amplitude respectively; S (x,t) It is a two-dimensional vibration signal in space and time.

3. The method for screening massive target data using distributed optical fiber sensing based on image processing according to claim 2, characterized in that: The local binary pattern algorithm is used to extract the texture features of the grayscale image as follows: Among them, F texture is the texture feature, g c is the pixel value of the center pixel, g p is the pixel value of the pth sampling point in the neighborhood, and s(·) is a step function.

4. The method for screening massive target data using distributed optical fiber sensing based on image processing according to claim 3, characterized in that: The calculation of Hu invariant moments is as follows: The main invariant moment combination M1, M2, M3, M4, M5, M6, and M7 calculates the shape feature F of the event area shape : M1=(η 20 +n 02 ) 2 M2=(η 20 +n 02 ) 2 +4th 11 2 Among them, the first-order moment M1 is used to describe the pixel distribution range, and the second-order moment M2 is used to capture the directional characteristics; the calculation formula of Hu invariant moments M3~M7 is as follows: M3=(η 30 -3rd 12 ) 2 +(3rd 21 -or 03 ) 2 M4=(η 30 +n 12 ) 2 +(the 21 +n 03 ) 2 M5=(η 30 +3rd 12 )[(or 21 +n 03 ) 2 -3(h 21 +n 03 )] 2 M6=(η 20 -or 02 )[(or 30 +n 12 ) 2 -(or 21 +n 03 ) 2 ]+4th 11 (or 30 +n 12 )(or 21 +n 03 )] M7=(3rd 21 +n 03 (3rd) 21 +n 03 )[(or 30 +n 12 ) 2 -3(h 21 +n 03 ) 2 ]-(or 30 +3rd 12 )(or 21 +n 03 )[3(h 30 +n 12 ) 2 -(or 21 +n 03 ) 2 ] Among them, M3 describes the symmetry, M4, M5, M6 and M7 describe the curvature; pq is the normalized central moment, which is calculated as follows: Among them, u pq Represents the geometric moment of the image relative to the center of mass, which is used to eliminate the translation effect; u 00 represents the zero-order central moment.

5. The method for screening massive target data using distributed optical fiber sensing based on image processing according to claim 4, characterized in that: The dynamic characteristics satisfy the brightness constant constraint equation: Among them, x, y, t represent the spatial coordinates and timestamps, Gu, Gv, G represent the horizontal displacement, vertical displacement and start-end linear displacement of the pixel, It means to find partial derivatives of variables x, y, t, Gu, Gv, and G.

6. The method for screening massive target data using distributed optical fiber sensing based on image processing according to claim 5, characterized in that: Step 3 adopts the attention weighting mechanism, and the weight coefficient is calculated as follows: Among them, w ij represents the similarity between the features of each modality, q is the learnable query vector, F i With F j is the eigenvector of each mode.

7. The method for screening massive target data using distributed optical fiber sensing based on image processing according to claim 6, characterized in that: The process of processing the joint feature vector in the MobileNetV3 network architecture is expressed as follows: FCNN=DWConv(ReLU(BN(F joint ))) Among them, DWConv is a depth-separable convolution, F joint ReLU represents the joint feature; ReLU represents the modified activation function, BN represents the batch normalization function, and the output feature map FCNN; finally, the feature map FCNN uses the normalization function to obtain the maximum probability score of the target object, which is used as the classification confidence P c Output.

8. The method for screening massive target data using distributed optical fiber sensing based on image processing according to claim 7, characterized in that: The hybrid loss function calculation process is as follows: L=λ·LCE+(1-λ)·L triplet Among them, L represents the regularization loss, LCE is the cross entropy loss, and L triplet is the triplet loss and λ is the regularization parameter.

9. The method for screening massive target data using distributed optical fiber sensing based on image processing according to claim 8, characterized in that: The correction calculation formula of step 6 is as follows: Among them, P final (x) represents the final power at position x, P c (x) represents the initial power at the reference point x0, α ​​is the fiber attenuation coefficient, x0 is the initial position of the event, and |x-x0| represents the absolute distance from the reference point x0 to the fiber sensor position x.

Images