Optical fiber network monitoring method and system

Abstract

The invention discloses an optical fiber network monitoring method and system, and relates to the technical field of optical fiber network monitoring, and the method comprises the steps: obtaining a plurality of training optical fiber data, carrying out the preprocessing of each training optical fiber data, obtaining a corresponding optical fiber feature set, training a preset initial optical fiber fault detection model through employing the optical fiber feature set, and obtaining an optical fiber fault detection model; and obtaining a corresponding optical fiber fault detection model, obtaining a light leakage signal and a terminal low-light image of the optical fiber network, and carrying out fault detection on the light leakage signal and the terminal low-light image through the optical fiber fault detection model to obtain a corresponding monitoring result. The technical problems that in the prior art, the optical fiber network is monitored mainly through a pre-trained graph neural network, but associated information of multi-dimensional data such as light leakage signals and terminal appearance characteristics cannot be excavated, so that the fault early warning accuracy is low, and the operation reliability of the optical fiber network is reduced are solved.

Description

Technical Field

[0001] This invention relates to the field of fiber optic network monitoring technology, and in particular to a fiber optic network monitoring method and system. Background Technology

[0002] With the large-scale deployment of 5G base stations and the surge in the number of cloud computing and IoT terminals, the total length of global fiber optic links has experienced explosive growth. The network topology has evolved from the traditional two-layer "backbone-access network" structure to a complex mesh structure interwoven with multiple nodes and links. Furthermore, a large number of links need to be deployed in complex environments such as underground pipelines, overhead cables, and remote fields, significantly increasing the difficulty of operation and maintenance. During long-term operation, fiber optic networks are susceptible to external interference and internal losses, leading to faults. If not detected in time, this can result in minor network outages or, in severe cases, major incidents such as financial transaction delays and energy dispatch failures. Therefore, accurate monitoring of fiber optic networks is crucial.

[0003] Currently, existing technologies mainly monitor fiber optic networks through pre-trained graph neural networks, but they cannot extract correlation information between leakage signals and multi-dimensional data such as terminal appearance features, resulting in low accuracy of fault warning and reduced reliability of fiber optic network operation. Summary of the Invention

[0004] This invention provides a fiber optic network monitoring method and system, which solves the technical problem that existing technologies mainly monitor fiber optic networks through pre-trained graph neural networks, but cannot mine the correlation information between leakage light signals and multi-dimensional data such as terminal appearance characteristics, resulting in low fault warning accuracy and reduced reliability of fiber optic network operation.

[0005] The first aspect of this invention provides a fiber optic network monitoring method, comprising:

[0006] Acquire multiple training fiber data sets, preprocess each training fiber data set, and obtain the corresponding fiber feature set;

[0007] The optical fiber feature set is used to train the preset initial optical fiber fault detection model to obtain the corresponding optical fiber fault detection model, wherein the optical fiber fault detection model includes a multi-mode feature extraction network and a fault detection network.

[0008] The leakage light signal and terminal low-light image of the optical fiber network are acquired. The leakage light signal and terminal low-light image of the optical fiber network are acquired. The multimodal feature extraction network is used to extract multimodal features from the leakage light signal and the terminal low-light image to obtain the corresponding cross-modal fusion feature vector.

[0009] The fault detection network is used to detect faults in the cross-modal fusion feature vector to obtain the corresponding monitoring results.

[0010] Optionally, the step of training a preset initial fiber optic fault detection model using the fiber optic feature set to obtain a corresponding fiber optic fault detection model includes:

[0011] The optical fiber feature set is used as input to train the preset initial optical fiber fault detection model to obtain the corresponding training optical fiber detection data.

[0012] Based on the preset cross-entropy loss function, the first loss value of the fiber feature set is calculated according to the training fiber detection data;

[0013] Based on a preset calibration loss function, the second loss value of the fiber feature set is calculated according to the training fiber detection data;

[0014] Based on preset loss weights, the first loss value and the second loss value are weighted and calculated to obtain the corresponding training loss function value;

[0015] When the training loss function value is greater than or equal to the preset standard loss function value, the network parameters of the initial fiber optic fault detection model are adjusted, and the process jumps to the step of training the preset initial fiber optic fault detection model using the fiber optic feature set to obtain the corresponding training fiber optic detection data, until the training loss function value is less than the standard loss function value.

[0016] When the training loss function value is less than the standard loss function value, an optical fiber fault detection model is generated.

[0017] Optionally, the multimodal feature extraction network includes a temporal branch, a visual branch, and a cross-modal attention fusion network. The step of acquiring the leakage light signal and terminal low-light image of the optical fiber network, and performing multimodal feature extraction on the leakage light signal and the terminal low-light image through the multimodal feature extraction network to obtain the corresponding cross-modal fusion feature vector includes:

[0018] The time-series branch is used to extract time-series features from the light leakage signal to obtain the corresponding time-series feature vector. The time-series branch includes a one-dimensional convolutional layer, a one-dimensional convolutional layer, a bidirectional long short-term memory network, and an attention pooling layer connected in sequence.

[0019] Visual features are extracted from the terminal low-light image through the visual branch to obtain the corresponding visual feature vector. The visual branch includes a first convolutional layer, a first pooling layer, a basic residual block, a morphological perception residual block, a basic residual block, a basic residual block, a spatial attention pooling layer, and a global average pooling layer connected in sequence.

[0020] The temporal feature vector and the visual feature vector are fused using the cross-modal attention fusion network to obtain the corresponding cross-modal fused feature vector.

[0021] Optionally, the shape-aware residual block includes three 1×1 convolutional layers, a feature fusion layer, two ReLU activation function layers, and a shape-aware convolutional module. The specific processing procedure of the shape-aware residual block is as follows:

[0022] The input intermediate feature map is reduced in dimensionality by a 1×1 convolutional layer to obtain the corresponding first intermediate feature map.

[0023] The morphological features of the first intermediate feature map are extracted by the morphological awareness convolution module to obtain the corresponding morphological feature map.

[0024] The morphological feature map is subjected to feature extraction by sequentially connecting 1×1 convolutional layers and ReLU activation function layers to obtain the corresponding second intermediate feature map;

[0025] The intermediate feature map is dimension-matched by a 1×1 convolutional layer to obtain the corresponding third intermediate feature map;

[0026] The second intermediate feature map and the third intermediate feature map are spliced ​​together by the feature fusion layer to obtain the corresponding fused intermediate feature map.

[0027] The fused intermediate feature map is nonlinearly transformed by a ReLU activation function layer to obtain the corresponding target morphological feature map.

[0028] Optionally, the morphology-aware convolutional module includes a 3×3 convolutional layer, a 1×5 convolutional layer, a 5×1 convolutional layer, and a feature fusion layer. The step of extracting morphological features from the first intermediate feature map using the morphology-aware convolutional module to obtain the corresponding morphological feature map includes:

[0029] Global morphological features are extracted from the first intermediate feature map by using a 3×3 convolutional layer to obtain the corresponding global morphological feature map.

[0030] The first intermediate feature map is processed by a 1×5 convolutional layer to extract horizontal crack features, and the corresponding horizontal crack feature map is obtained.

[0031] The first intermediate feature map is subjected to vertical offset feature extraction by a 5×1 convolutional layer to obtain the corresponding vertical offset feature map.

[0032] The global morphological feature map, the horizontal crack feature map, and the vertical offset feature map are stitched together by a feature fusion layer to obtain the corresponding morphological feature map.

[0033] Optionally, the fault detection network includes a defect calibration network, a first fully connected layer, and a second fully connected layer. The step of performing fault detection on the cross-modal fused feature vector through the fault detection network to obtain the corresponding monitoring result includes:

[0034] The cross-modal fusion feature vector is calibrated using the defect calibration network to obtain the corresponding calibration feature vector;

[0035] The calibration feature vector is classified into faults by sequentially connecting a first fully connected layer and a second fully connected layer to obtain the corresponding monitoring results.

[0036] Optionally, the defect calibration network includes a cosine similarity module, a weighted summation module, and a multiplication layer. The step of performing feature calibration on the cross-modal fused feature vector through the defect calibration network to obtain the corresponding calibration feature vector includes:

[0037] The cosine similarity module calculates the cosine similarity between the cross-modal fusion feature vector and each fault feature vector in the preset prior matrix, and obtains multiple correlation values.

[0038] The associated values ​​are normalized to obtain multiple target associated values;

[0039] Based on the weighted summation module, a weighted summation operation is performed on each of the fault feature vectors according to each of the target correlation values ​​to obtain the corresponding weight vector;

[0040] The corresponding calibration feature vector is obtained by performing element-wise multiplication of the weight vector and the cross-modal fusion feature vector through the multiplication layer.

[0041] A second aspect of the present invention provides a fiber optic network monitoring system, comprising:

[0042] The preprocessing module is used to acquire multiple training fiber data, preprocess each training fiber data, and obtain the corresponding fiber feature set.

[0043] The training module is used to train the preset initial optical fiber fault detection model using the optical fiber feature set to obtain the corresponding optical fiber fault detection model, wherein the optical fiber fault detection model includes a multi-mode feature extraction network and a fault detection network.

[0044] The acquisition module is used to acquire the leakage light signal and terminal low light image of the optical fiber network, and to extract multimodal features from the leakage light signal and terminal low light image through the multimodal feature extraction network to obtain the corresponding cross-modal fusion feature vector.

[0045] The detection module is used to perform fault detection on the cross-modal fusion feature vector through the fault detection network to obtain the corresponding monitoring results.

[0046] A third aspect of the present invention provides an electronic device, including a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor performs the steps of the fiber optic network monitoring method as described in any of the preceding claims.

[0047] The fourth aspect of the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed, implements the fiber optic network monitoring method as described in any of the preceding claims.

[0048] The fifth aspect of the present invention provides a computer program product, the computer program product comprising a computer program stored on a non-transitory computer-readable storage medium, the computer program comprising program instructions, wherein, when the program instructions are executed by a computer, the computer performs the fiber optic network monitoring method as described in any of the preceding claims.

[0049] As can be seen from the above technical solutions, the present invention has the following advantages:

[0050] This invention acquires multiple training fiber data sets, preprocesses each set to obtain corresponding fiber feature sets, and uses these feature sets to train a pre-set initial fiber fault detection model. This results in a corresponding fiber fault detection model, which acquires leakage signals and terminal low-light images of the fiber network. The fiber fault detection model then performs fault detection on these leak signals and terminal low-light images to obtain corresponding monitoring results. This overcomes the technical problem of existing technologies that primarily monitor fiber networks using pre-trained graph neural networks, but fail to mine the correlation information between leakage signals and multi-dimensional data such as terminal appearance features, leading to low fault warning accuracy and reduced reliability of fiber network operation. Compared with traditional fiber optic network monitoring methods, this invention uses a fiber optic fault detection model to detect leaked light signals and terminal low-light images, obtaining corresponding monitoring results. The fiber optic fault detection model includes a multi-mode feature extraction network and a fault detection network. The multi-mode feature extraction network extracts multi-mode features from the leaked light signals and terminal low-light images, thereby fully exploring the correlation information between the leaked light signals and terminal low-light images to obtain corresponding cross-mode fusion feature vectors. Then, the fault detection network performs fault detection on the cross-mode fusion feature vectors to obtain the corresponding monitoring results, significantly improving the accuracy of fiber optic network fault prediction and thus enhancing the reliability of fiber optic network operation. Attached Figure Description

[0051] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0052] Figure 1 This is a flowchart of the steps of a fiber optic network monitoring method provided in Embodiment 1 of the present invention;

[0053] Figure 2 This is a flowchart illustrating the steps of a fiber optic network monitoring method provided in Embodiment 2 of the present invention.

[0054] Figure 3 This is a schematic diagram of the structure of the optical fiber fault detection model provided in Embodiment 2 of the present invention;

[0055] Figure 4 This is a structural block diagram of an optical fiber network monitoring system provided in Embodiment 3 of the present invention;

[0056] Figure 5 This is a structural block diagram of an electronic device provided in Embodiment 4 of the present invention. Detailed Implementation

[0057] This invention provides a fiber optic network monitoring method and system to address the technical problem that existing technologies mainly monitor fiber optic networks through pre-trained graph neural networks, but cannot extract correlation information between leakage signals and multi-dimensional data such as terminal appearance features, resulting in low accuracy of fault warnings and reduced reliability of fiber optic network operation.

[0058] To make the objectives, features, and advantages of this invention more apparent and understandable, the technical solutions of the embodiments of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the embodiments described below are only some embodiments of this invention, and not all embodiments. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0059] Please see Figure 1 , Figure 1 The flowchart illustrates the steps of a fiber optic network monitoring method provided in Embodiment 1 of the present invention.

[0060] The present invention provides a fiber optic network monitoring method, comprising:

[0061] Step 101: Obtain multiple training fiber data sets, preprocess each training fiber data set, and obtain the corresponding fiber feature set.

[0062] Training fiber data refers to multi-dimensional raw data collected for training fiber fault detection models, reflecting the normal and fault states of the fiber, including leakage signals, terminal low-light images, etc.

[0063] The fiber feature set refers to the preprocessed training fiber data.

[0064] In this embodiment of the invention, multiple training fiber data are acquired, and each training fiber data is preprocessed (preprocessing includes data cleaning and standardization) to obtain the corresponding fiber feature set.

[0065] Step 102: Train the preset initial fiber optic fault detection model using fiber optic feature set to obtain the corresponding fiber optic fault detection model, wherein the fiber optic fault detection model includes a multi-mode feature extraction network and a fault detection network.

[0066] In this embodiment of the invention, a preset initial fiber optic fault detection model is trained using a fiber optic feature set as input to obtain corresponding training fiber optic detection data. A first loss value for the fiber optic feature set is calculated based on a preset cross-entropy loss function and the training fiber optic detection data. A second loss value for the fiber optic feature set is calculated based on a preset calibration loss function and the training fiber optic detection data. The first and second loss values ​​are weighted based on preset loss weights to obtain the corresponding training loss function value. When the training loss function value is greater than or equal to a preset standard loss function value, the network parameters of the initial fiber optic fault detection model are adjusted, and the process jumps to the step of training the preset initial fiber optic fault detection model using a fiber optic feature set as input to obtain corresponding training fiber optic detection data, until the training loss function value is less than the standard loss function value. When the training loss function value is less than the standard loss function value, a fiber optic fault detection model is generated, wherein the fiber optic fault detection model includes a multimodal feature extraction network and a fault detection network.

[0067] Step 103: Obtain the leakage light signal and terminal low light image of the optical fiber network, and extract multimodal features from the leakage light signal and terminal low light image through a multimodal feature extraction network to obtain the corresponding cross-modal fusion feature vector.

[0068] In this embodiment of the invention, the leakage light signal and terminal low light image of the optical fiber network are acquired, and the leakage light signal and terminal low light image are subjected to multimodal feature extraction through a multimodal feature extraction network to obtain the corresponding cross-modal fusion feature vector. The multimodal feature extraction network includes a temporal branch, a visual branch and a cross-modal attention fusion network.

[0069] Step 104: Perform fault detection on the cross-modal fused feature vector through a fault detection network to obtain the corresponding monitoring results.

[0070] In this embodiment of the invention, a fault detection network is used to detect faults in cross-modal fused feature vectors to obtain corresponding monitoring results. The fault detection network includes a defect calibration network, a first fully connected layer, and a second fully connected layer.

[0071] In this embodiment of the invention, multiple training fiber data sets are acquired, and each set is preprocessed to obtain a corresponding fiber feature set. This fiber feature set is then used to train a pre-set initial fiber fault detection model, resulting in a corresponding fiber fault detection model. Leakage signals and terminal low-light images of the fiber network are acquired, and the fiber fault detection model is used to detect faults in the leakage signals and terminal low-light images, yielding corresponding monitoring results. This overcomes the technical problem of existing technologies that primarily monitor fiber networks using pre-trained graph neural networks, but fail to mine the correlation information between leakage signals and multi-dimensional data such as terminal appearance features, leading to low fault warning accuracy and reduced reliability of fiber network operation. Compared with traditional fiber optic network monitoring methods, this invention uses a fiber optic fault detection model to detect leaked light signals and terminal low-light images, obtaining corresponding monitoring results. The fiber optic fault detection model includes a multi-mode feature extraction network and a fault detection network. The multi-mode feature extraction network extracts multi-mode features from the leaked light signals and terminal low-light images, thereby fully exploring the correlation information between the leaked light signals and terminal low-light images to obtain corresponding cross-mode fusion feature vectors. Then, the fault detection network performs fault detection on the cross-mode fusion feature vectors to obtain the corresponding monitoring results, significantly improving the accuracy of fiber optic network fault prediction and thus enhancing the reliability of fiber optic network operation.

[0072] Please see Figure 2 , Figure 2 This is a flowchart illustrating the steps of a fiber optic network monitoring method provided in Embodiment 2 of the present invention.

[0073] The present invention provides a fiber optic network monitoring method, comprising:

[0074] Step 201: Obtain multiple training fiber data sets, preprocess each training fiber data set, and obtain the corresponding fiber feature set.

[0075] In this embodiment of the invention, multiple training fiber data are acquired, and each training fiber data is cleaned and standardized to obtain a corresponding fiber feature set.

[0076] Step 202: Train the preset initial fiber optic fault detection model using the fiber optic feature set input to obtain the corresponding training fiber optic detection data.

[0077] Training fiber optic detection data refers to the prediction results output by the initial fiber optic fault detection model after receiving the fiber optic feature set.

[0078] In this embodiment of the invention, the fiber feature set is input into a preset initial fiber fault detection model for training to obtain corresponding training fiber detection data.

[0079] Step 203: Based on the preset cross-entropy loss function, calculate the first loss value of the fiber feature set according to the training fiber detection data.

[0080] In this embodiment of the invention, the training fiber detection data and fiber feature set are input into a preset cross-entropy loss function to obtain the corresponding first loss value.

[0081] It should be noted that the cross-entropy loss function is as follows:

[0082]

[0083] in, The first loss value, The weight of category c, The true label of sample i in category c. Let be the predicted probability of sample i in category c, where c is the fault category index, i is the index of the sample, and N is the total number of samples.

[0084] Step 204: Based on the preset calibration loss function, calculate the second loss value of the fiber feature set according to the training fiber detection data.

[0085] In this embodiment of the invention, the training fiber detection data and fiber feature set are input into a preset calibration loss function to obtain the corresponding second loss value.

[0086] It should be noted that the calibration loss function is as follows:

[0087]

[0088] in, This is the second loss value. Let be the calibrated feature value of sample i in dimension d. This represents the association strength of category c in dimension d within the prior matrix. d represents the ideal feature strength, and d is the dimension index.

[0089] Step 205: Based on the preset loss weights, perform a weighted operation on the first loss value and the second loss value to obtain the corresponding training loss function value.

[0090] In this embodiment of the invention, the first loss value and the second loss value are weighted based on a preset loss weight to obtain the corresponding training loss function value. For example, if the loss weight coefficient is 0.6, then the training loss function value = 0.6 * first loss value + (1 - 0.6) * second loss value.

[0091] Step 206: When the training loss function value is greater than or equal to the preset standard loss function value, adjust the network parameters of the initial fiber optic fault detection model, and jump to the step of training the preset initial fiber optic fault detection model by inputting the fiber optic feature set to obtain the corresponding training fiber optic detection data, until the training loss function value is less than the standard loss function value.

[0092] The standard loss function value refers to the loss threshold that is set in advance according to the accuracy requirements of the fiber optic fault detection scenario. It is the core standard for judging whether the model training has met the requirements. When the training loss function value is less than the threshold, it indicates that the model prediction error has met the requirements of practical application.

[0093] In this embodiment of the invention, when the training loss function value is greater than or equal to the preset standard loss function value, the network parameters of the initial fiber optic fault detection model (such as the weights and biases of the convolutional layer, and the weights and biases of the fully connected layer) are adjusted, and steps 202-206 are executed until the training loss function value is less than the standard loss function value.

[0094] Step 207: When the training loss function value is less than the standard loss function value, an optical fiber fault detection model is generated. The optical fiber fault detection model includes a multimodal feature extraction network and a fault detection network.

[0095] In this embodiment of the invention, when the training loss function value is less than the standard loss function value, it indicates that the model prediction error has met the requirements of practical application, and an optical fiber fault detection model is generated. The optical fiber fault detection model includes a multimodal feature extraction network and a fault detection network.

[0096] Step 208: Obtain the leakage light signal and terminal low light image of the optical fiber network, and extract multimodal features from the leakage light signal and terminal low light image through a multimodal feature extraction network to obtain the corresponding cross-modal fusion feature vector.

[0097] Further, see Figure 3 As shown, the multimodal feature extraction network includes a temporal branch, a visual branch, and a cross-modal attention fusion network. Step 208 includes the following sub-steps:

[0098] S11. Temporal features are extracted from the light leakage signal through temporal branches to obtain the corresponding temporal feature vector. The temporal branches include a one-dimensional convolutional layer, a one-dimensional convolutional layer, a bidirectional long short-term memory network, and an attention pooling layer connected in sequence.

[0099] Leak light signal refers to a continuous data sequence (i.e., Leak Light signal) in which the light intensity changes over time due to partial optical signal leakage caused by link damage (such as microcracks or poor terminal contact) during signal transmission in optical fiber.

[0100] In this embodiment of the invention, the time-series features of the light leakage signal are extracted by sequentially connecting a one-dimensional convolutional layer, a one-dimensional convolutional layer, a bidirectional long short-term memory network, and an attention pooling layer to obtain the corresponding time-series feature vector.

[0101] It should be noted that the first one-dimensional convolutional layer is used to capture local intensity abrupt changes in the light leakage signal (such as a sudden increase in light leakage caused by a crack). The second one-dimensional convolutional layer is used to extract mesoscale fluctuation features in the light leakage signal (such as periodic light leakage due to loose terminals). A bidirectional long short-term memory network is used to distinguish between "persistent damage" and "temporary interference" in the light leakage signal. An attention pooling layer is used to extract key time points in the light leakage signal (such as the signal peak at the moment of the fault).

[0102] S12. Visual features are extracted from the terminal low-light image through the visual branch to obtain the corresponding visual feature vector. The visual branch includes a first convolutional layer, a first pooling layer, a basic residual block, a morphological perception residual block, a basic residual block, a basic residual block, a spatial attention pooling layer, and a global average pooling layer connected in sequence.

[0103] Terminal low-light images refer to images of fiber optic terminals (components used to connect fibers to each other or between fibers and equipment) acquired by low-light imaging equipment.

[0104] The first convolutional layer refers to a 3×3 convolutional layer.

[0105] The first pooling layer refers to the 3×3 pooling layer.

[0106] In this embodiment of the invention, visual features are extracted from the terminal low-light image by sequentially connecting a first convolutional layer, a first pooling layer, a basic residual block, a morphological perception residual block, a basic residual block, a basic residual block, a spatial attention pooling layer, and a global average pooling layer, to obtain the corresponding visual feature vector.

[0107] It is worth mentioning that the spatial attention pooling layer is used to suppress background redundancy (such as equipment housings and cables) and enhance the defect features of the core areas of the terminals (weld joints and snap-fits).

[0108] It should be noted that the difference between the basic residual block and the morphology-aware residual block is that the morphology-aware convolutional module in the morphology-aware residual block is replaced by a 3×3 convolutional layer.

[0109] It should be noted that the morphology-aware residual block consists of three 1×1 convolutional layers, a feature fusion layer, two ReLU activation function layers, and a morphology-aware convolutional module. The specific processing procedure of the morphology-aware residual block is as follows:

[0110] A1. The input intermediate feature map is reduced in dimensionality by a 1×1 convolutional layer to obtain the corresponding first intermediate feature map.

[0111] In this embodiment of the invention, the input intermediate feature map is dimensionality reduced by a 1×1 convolutional layer (i.e., the number of feature map channels is reduced, the subsequent computation is reduced, and redundant information is compressed) to obtain the corresponding first intermediate feature map.

[0112] A2. The morphological features of the first intermediate feature map are extracted by the morphological awareness convolution module to obtain the corresponding morphological feature map.

[0113] Furthermore, the morphology-aware convolutional module includes a 3×3 convolutional layer, a 1×5 convolutional layer, a 5×1 convolutional layer, and a feature fusion layer. A2 includes the following sub-steps:

[0114] A21. Global morphological features are extracted from the first intermediate feature map using a 3×3 convolutional layer to obtain the corresponding global morphological feature map.

[0115] In this embodiment of the invention, a 3×3 convolutional layer is used to extract global morphological features from the first intermediate feature map (i.e., extract global morphological features of the terminal (such as whether the whole is tilted or whether there is large-area deformation on the surface)) to obtain the corresponding global morphological feature map. The 3×3 convolutional layer has 128 convolutional kernels, a stride of 1, and uses the ReLU activation function.

[0116] A22. Horizontal crack features are extracted from the first intermediate feature map using a 1×5 convolutional layer to obtain the corresponding horizontal crack feature map.

[0117] In this embodiment of the invention, a 1×5 convolutional layer is used to extract horizontal crack features from the first intermediate feature map (i.e., extract horizontal features (horizontal cracks at the terminal weld)) to obtain the corresponding horizontal crack feature map. The 1×5 convolutional layer has 128 kernels, a stride of 1, and uses the ReLU activation function.

[0118] A23. Vertical offset feature extraction is performed on the first intermediate feature map through a 5×1 convolutional layer to obtain the corresponding vertical offset feature map.

[0119] In this embodiment of the invention, a 5×1 convolutional layer is used to extract vertical offset features from the first intermediate feature map (i.e., extract vertical direction features (vertical offset between the terminal and the optical fiber)) to obtain the corresponding vertical offset feature map. The 5×1 convolutional layer has 128 convolutional kernels, a stride of 1, and uses the ReLU activation function.

[0120] A24. The global morphological feature map, horizontal crack feature map, and vertical offset feature map are spliced ​​together by the feature fusion layer to obtain the corresponding morphological feature map.

[0121] In this embodiment of the invention, a feature fusion layer is used to fuse the global morphological feature map, the horizontal crack feature map, and the vertical offset feature map to obtain the corresponding morphological feature map.

[0122] A3. By sequentially connecting 1×1 convolutional layers and ReLU activation function layers, feature extraction is performed on the morphological feature map to obtain the corresponding second intermediate feature map.

[0123] In this embodiment of the invention, a 1×1 convolutional layer and a ReLU activation function layer connected in sequence are used to extract features from the morphological feature map to obtain the corresponding second intermediate feature map.

[0124] A4. Perform dimension matching on the intermediate feature map using a 1×1 convolutional layer to obtain the corresponding third intermediate feature map.

[0125] In this embodiment of the invention, a 1×1 convolutional layer is used to extract features from the intermediate feature map to obtain the corresponding third intermediate feature map.

[0126] A5. The second and third intermediate feature maps are spliced ​​together by the feature fusion layer to obtain the corresponding fused intermediate feature map.

[0127] In this embodiment of the invention, the second intermediate feature map and the third intermediate feature map are fused by a feature fusion layer to obtain the corresponding fused intermediate feature map.

[0128] A6. The intermediate feature map is transformed nonlinearly by the ReLU activation function layer to obtain the corresponding target morphological feature map.

[0129] In this embodiment of the invention, a ReLU activation function layer is used to perform a nonlinear transformation on the fused intermediate feature map to obtain the corresponding target morphological feature map.

[0130] It is worth mentioning that the visual branch is embedded with a morphology-aware convolution module, which specifically enhances the visual branch's ability to extract features of morphological defects (cracks, offsets, and buckle damage) in low-light images of terminals.

[0131] S13. The temporal feature vector and the visual feature vector are fused through a cross-modal attention fusion network to obtain the corresponding cross-modal fused feature vector.

[0132] In this embodiment of the invention, attention weights for temporal and visual feature vectors are calculated using intra-modal attention and inter-modal attention within the cross-modal attention fusion network, resulting in corresponding visual and temporal attention weights. Based on these visual and temporal attention weights, the visual and temporal feature vectors are fused to obtain the corresponding cross-modal fused feature vector (i.e., cross-modal fused feature vector = visual attention weight * visual feature vector + temporal feature vector * temporal attention weight).

[0133] Step 209: Perform fault detection on the cross-modal fusion feature vector through a fault detection network to obtain the corresponding monitoring results.

[0134] Further, see Figure 3 As shown, the fault detection network includes a defect calibration network, a first fully connected layer, and a second fully connected layer. Step 209 includes the following sub-steps:

[0135] S21. The cross-modal fusion feature vector is calibrated using a defect calibration network to obtain the corresponding calibration feature vector.

[0136] Furthermore, the defect calibration network includes a cosine similarity module, a weighted summation module, and a multiplication layer. S21 includes the following sub-steps:

[0137] S211. The cosine similarity between the cross-modal fusion feature vector and each fault feature vector in the preset prior matrix is ​​calculated using the cosine similarity module to obtain multiple correlation values.

[0138] The prior matrix refers to the fault feature benchmark matrix constructed based on historical fault sample data. Each row of the matrix corresponds to a fault feature vector of a typical fault category. These vectors are obtained through cross-modal fusion feature statistical optimization of historical fault samples.

[0139] The correlation value refers to the cosine similarity between the cross-modal fused feature vector and a single fault feature vector.

[0140] In this embodiment of the invention, the cosine similarity between the cross-modal fusion feature vector and each fault feature vector in the preset prior matrix is ​​calculated by the cosine similarity module to obtain multiple correlation values.

[0141] S212. Normalize each correlation value to obtain multiple target correlation values.

[0142] In this embodiment of the invention, each associated value is normalized to obtain multiple target associated values.

[0143] S213. Based on the weighted summation module, perform a weighted operation (i.e., target association value * fault feature vector) on each target association value and the corresponding fault feature vector, and then sum the weighted fault feature vectors according to the same dimension to obtain the corresponding weight vector.

[0144] In this embodiment of the invention, the weighted summation module performs a weighted summation operation on each target correlation value and each fault feature vector to obtain the corresponding weight vector.

[0145] S214. The corresponding calibration feature vector is obtained by performing element-wise multiplication of the weight vector and the cross-modal fusion feature vector through the multiplication layer.

[0146] In this embodiment of the invention, a multiplication layer is used to perform element-wise multiplication of the weight vector and the cross-modal fusion feature vector to obtain the corresponding calibration feature vector.

[0147] S22. The calibration feature vector is classified for faults by sequentially connecting the first fully connected layer and the second fully connected layer to obtain the corresponding monitoring results.

[0148] It should be noted that the first fully connected layer contains a fully connected layer with a ReLU activation function. The second fully connected layer contains a fully connected layer with a Softmax activation function.

[0149] In this embodiment of the invention, the calibration feature vector is classified for faults by sequentially connecting a first fully connected layer and a second fully connected layer to obtain the corresponding monitoring results. The monitoring results include one of the following: normal, material defect, morphological defect, and mixed fault.

[0150] In this embodiment of the invention, multiple training fiber data sets are acquired, and each set is preprocessed to obtain a corresponding fiber feature set. This fiber feature set is then used to train a pre-set initial fiber fault detection model, resulting in a corresponding fiber fault detection model. Leakage signals and terminal low-light images of the fiber network are acquired, and the fiber fault detection model is used to detect faults in the leakage signals and terminal low-light images, yielding corresponding monitoring results. This overcomes the technical problem of existing technologies that primarily monitor fiber networks using pre-trained graph neural networks, but fail to mine the correlation information between leakage signals and multi-dimensional data such as terminal appearance features, leading to low fault warning accuracy and reduced reliability of fiber network operation. Compared with traditional fiber optic network monitoring methods, this invention uses a fiber optic fault detection model to detect leaked light signals and terminal low-light images, obtaining corresponding monitoring results. The fiber optic fault detection model includes a multi-mode feature extraction network and a fault detection network. The multi-mode feature extraction network extracts multi-mode features from the leaked light signals and terminal low-light images, thereby fully exploring the correlation information between the leaked light signals and terminal low-light images to obtain corresponding cross-mode fusion feature vectors. Then, the fault detection network performs fault detection on the cross-mode fusion feature vectors to obtain the corresponding monitoring results, significantly improving the accuracy of fiber optic network fault prediction and thus enhancing the reliability of fiber optic network operation.

[0151] Please see Figure 4 , Figure 4 This is a structural block diagram of an optical fiber network monitoring system provided in Embodiment 3 of the present invention.

[0152] The present invention provides a fiber optic network monitoring system, comprising:

[0153] The preprocessing module 301 is used to acquire multiple training fiber data, preprocess each training fiber data, and obtain the corresponding fiber feature set.

[0154] The training module 302 is used to train the preset initial fiber optic fault detection model using fiber optic feature set to obtain the corresponding fiber optic fault detection model, wherein the fiber optic fault detection model includes a multimodal feature extraction network and a fault detection network.

[0155] The acquisition module 303 is used to acquire the leakage light signal and terminal low light image of the optical fiber network, and to extract multimodal features from the leakage light signal and terminal low light image through a multimodal feature extraction network to obtain the corresponding cross-modal fusion feature vector.

[0156] The detection module 304 is used to perform fault detection on the cross-modal fused feature vector through the fault detection network and obtain the corresponding monitoring results.

[0157] Furthermore, training module 302 includes:

[0158] The training submodule is used to train the preset initial fiber optic fault detection model using the fiber optic feature set as input, and obtain the corresponding training fiber optic detection data.

[0159] The loss submodule is used to calculate the first loss value of the fiber feature set based on the training fiber detection data, using a preset cross-entropy loss function.

[0160] Based on the preset calibration loss function, the second loss value of the fiber feature set is calculated according to the training fiber detection data;

[0161] Based on the preset loss weights, the first loss value and the second loss value are weighted and calculated to obtain the corresponding training loss function value;

[0162] The adjustment submodule is used to adjust the network parameters of the initial fiber optic fault detection model when the training loss function value is greater than or equal to the preset standard loss function value, and then jump to execute the step of training the preset initial fiber optic fault detection model by inputting the fiber optic feature set to obtain the corresponding training fiber optic detection data, until the training loss function value is less than the standard loss function value.

[0163] When the training loss function value is less than the standard loss function value, a fiber optic fault detection model is generated.

[0164] Furthermore, the multimodal feature extraction network includes a temporal branch, a visual branch, and a cross-modal attention fusion network. The acquisition module 303 includes:

[0165] The temporal feature submodule is used to extract temporal features from the light leakage signal through temporal branches to obtain the corresponding temporal feature vector. The temporal branches include a one-dimensional convolutional layer, a one-dimensional convolutional layer, a bidirectional long short-term memory network, and an attention pooling layer connected in sequence.

[0166] The visual feature submodule is used to extract visual features from the terminal low-light image through the visual branch to obtain the corresponding visual feature vector. The visual branch includes a first convolutional layer, a first pooling layer, a basic residual block, a morphological perception residual block, a basic residual block, a basic residual block, a spatial attention pooling layer, and a global average pooling layer connected in sequence.

[0167] The cross-modal fusion submodule is used to fuse temporal feature vectors and visual feature vectors through a cross-modal attention fusion network to obtain the corresponding cross-modal fused feature vector.

[0168] Furthermore, the morphology-aware residual block comprises three 1×1 convolutional layers, a feature fusion layer, two ReLU activation function layers, and a morphology-aware convolutional module. The specific processing procedure of the morphology-aware residual block is as follows:

[0169] The input intermediate feature map is reduced in dimensionality by a 1×1 convolutional layer to obtain the corresponding first intermediate feature map.

[0170] The morphological features of the first intermediate feature map are extracted by the morphological awareness convolution module to obtain the corresponding morphological feature map.

[0171] The morphological feature map is extracted by sequentially connecting 1×1 convolutional layers and ReLU activation function layers to obtain the corresponding second intermediate feature map.

[0172] The intermediate feature map is dimension-matched by a 1×1 convolutional layer to obtain the corresponding third intermediate feature map;

[0173] The second and third intermediate feature maps are concatenated by a feature fusion layer to obtain the corresponding fused intermediate feature map.

[0174] The intermediate feature map is nonlinearly transformed by the ReLU activation function layer to obtain the corresponding target morphological feature map.

[0175] Furthermore, the morphology-aware convolutional module includes a 3×3 convolutional layer, a 1×5 convolutional layer, a 5×1 convolutional layer, and a feature fusion layer. The step of extracting morphological features from the first intermediate feature map using the morphology-aware convolutional module to obtain the corresponding morphological feature map includes:

[0176] Global morphological features are extracted from the first intermediate feature map using a 3×3 convolutional layer to obtain the corresponding global morphological feature map.

[0177] The horizontal crack feature is extracted from the first intermediate feature map by using a 1×5 convolutional layer to obtain the corresponding horizontal crack feature map.

[0178] The first intermediate feature map is vertically offset feature extracted by a 5×1 convolutional layer to obtain the corresponding vertically offset feature map.

[0179] The global morphological feature map, horizontal crack feature map, and vertical offset feature map are stitched together by a feature fusion layer to obtain the corresponding morphological feature map.

[0180] Furthermore, the fault detection network includes a defect calibration network, a first fully connected layer, and a second fully connected layer. The detection module 304 includes:

[0181] The calibration submodule is used to perform feature calibration on the cross-modal fusion feature vector through the defect calibration network to obtain the corresponding calibration feature vector.

[0182] The fault classification submodule is used to classify the calibration feature vector for faults through a first fully connected layer and a second fully connected layer connected in sequence, so as to obtain the corresponding monitoring results.

[0183] Furthermore, the defect calibration network includes a cosine similarity module, a weighted summation module, and a multiplication layer. The calibration submodule includes:

[0184] The similarity unit is used to calculate the cosine similarity between the cross-modal fusion feature vector and each fault feature vector in the preset prior matrix through the cosine similarity module, and obtain multiple association values.

[0185] The normalization unit is used to normalize each correlation value to obtain multiple target correlation values.

[0186] The weighted summation unit is used to perform a weighted summation operation on each fault feature vector based on the weighted summation module according to the correlation values ​​of each target, so as to obtain the corresponding weight vector.

[0187] The multiplication unit is used to perform element-wise multiplication of the weight vector and the cross-modal fusion feature vector through the multiplication layer to obtain the corresponding calibration feature vector.

[0188] Please see Figure 5 , Figure 5 This is a structural block diagram of an electronic device provided in Embodiment 4 of the present invention.

[0189] An electronic device according to an embodiment of the present invention includes: a memory 401 and a processor 402. The memory 401 stores a computer program. When the computer program is executed by the processor 402, the processor 402 performs the fiber optic network monitoring method as described in any of the above embodiments.

[0190] Memory 401 may be an electronic memory such as flash memory, EEPROM (Electrically Erasable Programmable Read-Only Memory), EPROM, hard disk, or ROM. Memory 401 has storage space 403 for program code 413 for performing any of the method steps described above. For example, storage space 403 for program code may include individual program codes 413 for implementing the various steps in the methods described above. This program code may be read from or written to one or more computer program products. These computer program products include program code carriers such as hard disks, CDs, memory cards, or floppy disks. The program code may be compressed, for example, in a suitable form. When run by a computing processing device, this code causes the computing processing device to perform the various steps in the methods described above. This program code may be read from or written to one or more computer program products. These computer program products include program code carriers such as hard disks, CDs, memory cards, or floppy disks. The program code may be compressed, for example, in a suitable form. When this code is run by a computing device, it causes the computing device to perform the various steps in the fiber optic network monitoring method described above.

[0191] Embodiment 5 of the present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the fiber optic network monitoring method as described in any of the above embodiments.

[0192] Embodiment 6 of the present invention also provides a computer program product, which includes a computer program stored on a non-transitory computer-readable storage medium. The computer program includes program instructions, wherein when the program instructions are executed by a computer, the computer performs the fiber optic network monitoring method as described in any of the above embodiments.

[0193] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0194] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces, or indirect coupling or communication connection between apparatuses or units, and may be electrical, mechanical, or other forms.

[0195] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0196] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0197] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0198] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for monitoring fiber optic networks, characterized in that, include: Acquire multiple training fiber data sets, preprocess each training fiber data set, and obtain the corresponding fiber feature set; The optical fiber feature set is used to train the preset initial optical fiber fault detection model to obtain the corresponding optical fiber fault detection model, wherein the optical fiber fault detection model includes a multi-mode feature extraction network and a fault detection network. The leakage light signal and terminal low-light image of the optical fiber network are acquired, and the multimodal feature extraction network is used to extract multimodal features from the leakage light signal and the terminal low-light image to obtain the corresponding cross-modal fusion feature vector. The fault detection network is used to detect faults in the cross-modal fusion feature vector to obtain the corresponding monitoring results.

2. The fiber optic network monitoring method according to claim 1, characterized in that, The step of training a preset initial fiber optic fault detection model using the fiber optic feature set to obtain a corresponding fiber optic fault detection model includes: The optical fiber feature set is used as input to train the preset initial optical fiber fault detection model to obtain the corresponding training optical fiber detection data. Based on the preset cross-entropy loss function, the first loss value of the fiber feature set is calculated according to the training fiber detection data; Based on a preset calibration loss function, the second loss value of the fiber feature set is calculated according to the training fiber detection data; Based on preset loss weights, the first loss value and the second loss value are weighted and calculated to obtain the corresponding training loss function value; When the training loss function value is greater than or equal to the preset standard loss function value, the network parameters of the initial fiber optic fault detection model are adjusted, and the process jumps to the step of training the preset initial fiber optic fault detection model using the fiber optic feature set to obtain the corresponding training fiber optic detection data, until the training loss function value is less than the standard loss function value. When the training loss function value is less than the standard loss function value, an optical fiber fault detection model is generated.

3. The fiber optic network monitoring method according to claim 1, characterized in that, The multimodal feature extraction network includes a temporal branch, a visual branch, and a cross-modal attention fusion network. The step of acquiring the leakage light signal and terminal low-light image of the optical fiber network, and performing multimodal feature extraction on the leakage light signal and the terminal low-light image through the multimodal feature extraction network to obtain the corresponding cross-modal fusion feature vector includes: The time-series branch is used to extract time-series features from the light leakage signal to obtain the corresponding time-series feature vector. The time-series branch includes a one-dimensional convolutional layer, a one-dimensional convolutional layer, a bidirectional long short-term memory network, and an attention pooling layer connected in sequence. Visual features are extracted from the terminal low-light image through the visual branch to obtain the corresponding visual feature vector. The visual branch includes a first convolutional layer, a first pooling layer, a basic residual block, a morphological perception residual block, a basic residual block, a basic residual block, a spatial attention pooling layer, and a global average pooling layer connected in sequence. The temporal feature vector and the visual feature vector are fused using the cross-modal attention fusion network to obtain the corresponding cross-modal fused feature vector.

4. The fiber optic network monitoring method according to claim 3, characterized in that, The morphology-aware residual block comprises three 1×1 convolutional layers, a feature fusion layer, two ReLU activation function layers, and a morphology-aware convolutional module. The specific processing procedure of the morphology-aware residual block is as follows: The input intermediate feature map is reduced in dimensionality by a 1×1 convolutional layer to obtain the corresponding first intermediate feature map. The morphological features of the first intermediate feature map are extracted by the morphological awareness convolution module to obtain the corresponding morphological feature map. The morphological feature map is subjected to feature extraction by sequentially connecting 1×1 convolutional layers and ReLU activation function layers to obtain the corresponding second intermediate feature map; The intermediate feature map is dimension-matched by a 1×1 convolutional layer to obtain the corresponding third intermediate feature map; The second intermediate feature map and the third intermediate feature map are spliced ​​together by the feature fusion layer to obtain the corresponding fused intermediate feature map. The fused intermediate feature map is nonlinearly transformed by a ReLU activation function layer to obtain the corresponding target morphological feature map.

5. The fiber optic network monitoring method according to claim 4, characterized in that, The morphology-aware convolutional module includes a 3×3 convolutional layer, a 1×5 convolutional layer, a 5×1 convolutional layer, and a feature fusion layer. The step of extracting morphological features from the first intermediate feature map using the morphology-aware convolutional module to obtain the corresponding morphological feature map includes: Global morphological features are extracted from the first intermediate feature map by a 3×3 convolutional layer to obtain the corresponding global morphological feature map. The first intermediate feature map is processed by a 1×5 convolutional layer to extract horizontal crack features, and the corresponding horizontal crack feature map is obtained. The first intermediate feature map is subjected to vertical offset feature extraction by a 5×1 convolutional layer to obtain the corresponding vertical offset feature map. The global morphological feature map, the horizontal crack feature map, and the vertical offset feature map are stitched together by a feature fusion layer to obtain the corresponding morphological feature map.

6. The fiber optic network monitoring method according to claim 1, characterized in that, The fault detection network includes a defect calibration network, a first fully connected layer, and a second fully connected layer. The step of performing fault detection on the cross-modal fused feature vector through the fault detection network to obtain the corresponding monitoring result includes: The cross-modal fusion feature vector is calibrated using the defect calibration network to obtain the corresponding calibration feature vector; The calibration feature vector is classified for faults by sequentially connecting a first fully connected layer and a second fully connected layer, and the corresponding monitoring results are obtained.

7. The fiber optic network monitoring method according to claim 6, characterized in that, The defect calibration network includes a cosine similarity module, a weighted summation module, and a multiplication layer. The step of performing feature calibration on the cross-modal fused feature vector through the defect calibration network to obtain the corresponding calibrated feature vector includes: The cosine similarity module calculates the cosine similarity between the cross-modal fusion feature vector and each fault feature vector in the preset prior matrix, and obtains multiple correlation values. The associated values ​​are normalized to obtain multiple target associated values; Based on the weighted summation module, a weighted summation operation is performed on each of the fault feature vectors according to each of the target correlation values ​​to obtain the corresponding weight vector; The corresponding calibration feature vector is obtained by performing element-wise multiplication of the weight vector and the cross-modal fusion feature vector through the multiplication layer.

8. A fiber optic network monitoring system, characterized in that, include: The preprocessing module is used to acquire multiple training fiber data, preprocess each training fiber data, and obtain the corresponding fiber feature set. The training module is used to train the preset initial optical fiber fault detection model using the optical fiber feature set to obtain the corresponding optical fiber fault detection model, wherein the optical fiber fault detection model includes a multi-mode feature extraction network and a fault detection network. The acquisition module is used to acquire the leakage light signal and terminal low light image of the optical fiber network, and to extract multimodal features from the leakage light signal and terminal low light image through the multimodal feature extraction network to obtain the corresponding cross-modal fusion feature vector. The detection module is used to perform fault detection on the cross-modal fusion feature vector through the fault detection network to obtain the corresponding monitoring results.

9. An electronic device, characterized in that, The device includes a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor performs the steps of the fiber optic network monitoring method as described in any one of claims 1-7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed, it implements the fiber optic network monitoring method as described in any one of claims 1-7.

Images