CN117997425A - Optical module state monitoring method, device, equipment and medium - Google Patents
Optical module state monitoring method, device, equipment and medium Download PDFInfo
- Publication number
- CN117997425A CN117997425A CN202410225883.XA CN202410225883A CN117997425A CN 117997425 A CN117997425 A CN 117997425A CN 202410225883 A CN202410225883 A CN 202410225883A CN 117997425 A CN117997425 A CN 117997425A
- Authority
- CN
- China
- Prior art keywords
- optical module
- spectrogram
- low
- frequency analysis
- recognition
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 230000003287 optical effect Effects 0.000 title claims abstract description 163
- 238000000034 method Methods 0.000 title claims abstract description 100
- 238000012544 monitoring process Methods 0.000 title claims abstract description 56
- 238000004458 analytical method Methods 0.000 claims abstract description 115
- 238000012549 training Methods 0.000 claims abstract description 50
- 238000012806 monitoring device Methods 0.000 claims abstract description 22
- 230000006870 function Effects 0.000 claims description 27
- 238000012545 processing Methods 0.000 claims description 24
- 230000004913 activation Effects 0.000 claims description 22
- 230000008569 process Effects 0.000 claims description 21
- 208000037170 Delayed Emergence from Anesthesia Diseases 0.000 claims description 16
- 238000010606 normalization Methods 0.000 claims description 15
- 230000007246 mechanism Effects 0.000 claims description 14
- 238000004590 computer program Methods 0.000 claims description 10
- 238000011176 pooling Methods 0.000 claims description 9
- 238000010276 construction Methods 0.000 claims description 6
- 238000001035 drying Methods 0.000 claims description 6
- 238000005070 sampling Methods 0.000 claims description 5
- 238000007670 refining Methods 0.000 claims description 4
- 230000009466 transformation Effects 0.000 claims description 4
- 238000001228 spectrum Methods 0.000 claims description 3
- 238000004891 communication Methods 0.000 abstract description 9
- 238000001514 detection method Methods 0.000 abstract description 7
- 238000010586 diagram Methods 0.000 description 12
- 230000000694 effects Effects 0.000 description 5
- 230000009471 action Effects 0.000 description 3
- 238000007781 pre-processing Methods 0.000 description 3
- 238000012935 Averaging Methods 0.000 description 2
- 230000002159 abnormal effect Effects 0.000 description 2
- 230000032683 aging Effects 0.000 description 2
- 238000013473 artificial intelligence Methods 0.000 description 2
- 238000003745 diagnosis Methods 0.000 description 2
- 230000002708 enhancing effect Effects 0.000 description 2
- 239000000284 extract Substances 0.000 description 2
- 238000000605 extraction Methods 0.000 description 2
- 230000010354 integration Effects 0.000 description 2
- 238000010801 machine learning Methods 0.000 description 2
- 230000000750 progressive effect Effects 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000015556 catabolic process Effects 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 238000006731 degradation reaction Methods 0.000 description 1
- 238000013154 diagnostic monitoring Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 230000007774 longterm Effects 0.000 description 1
- 230000007257 malfunction Effects 0.000 description 1
- 230000005055 memory storage Effects 0.000 description 1
- 230000008447 perception Effects 0.000 description 1
- 230000008707 rearrangement Effects 0.000 description 1
- 238000009877 rendering Methods 0.000 description 1
- 238000010183 spectrum analysis Methods 0.000 description 1
- 230000001052 transient effect Effects 0.000 description 1
Landscapes
- Image Analysis (AREA)
Abstract
The invention relates to the field of optical communication, and discloses an optical module state monitoring method, an optical module state monitoring device, optical module state monitoring equipment and a medium, wherein the optical module state monitoring method comprises the following steps: acquiring a data sample set of the optical module during working; the data sample set comprises a non-fault state light intensity distribution image dataset and a fault state light intensity distribution image dataset; constructing a first low-frequency analysis recognition spectrogram by using the non-fault state light intensity distribution image data set, and constructing a second low-frequency analysis recognition spectrogram by using the fault state light intensity distribution image data set; training classification recognition models for distinguishing different states of the optical module by adopting a first low-frequency analysis recognition spectrogram and a second low-frequency analysis recognition spectrogram; and inputting the low-frequency analysis recognition spectrogram corresponding to the optical module to be tested into the trained classification recognition model, and outputting the current state result of the optical module to be tested. Therefore, the current working state of the optical module can be monitored in real time, whether the optical module fails or not is effectively detected, and the accuracy and the efficiency of the fault detection of the optical module are improved.
Description
Technical Field
The present invention relates to the field of optical communications, and in particular, to a method, an apparatus, a device, and a medium for monitoring an optical module state.
Background
In an optical communication system, an optical module is one of key components, and the optical module is mainly responsible for performing photoelectric conversion on a received signal. However, due to long-term use or environmental factors, the optical module may malfunction, resulting in degradation or even interruption of communication quality. The current monitoring method for the faults of the optical module is mainly based on rules or experience, the monitoring result is inaccurate, false alarm or missing alarm is easy to generate, a large amount of manual intervention is needed, time and effort are consumed in practice, and complex faults cannot be accurately identified.
Therefore, how to provide a method for effectively monitoring the status of the optical module is a technical problem that needs to be solved by those skilled in the art.
Disclosure of Invention
The invention aims to provide an optical module state monitoring method, an optical module state monitoring device, optical module state monitoring equipment and an optical module state monitoring medium, which can monitor the current working state of an optical module in real time, effectively detect whether the optical module fails or not and improve the accuracy and the efficiency of optical module failure detection.
In order to solve the technical problems, the invention provides an optical module state monitoring method, which comprises the following steps:
Acquiring a data sample set of the optical module during working; the data sample set comprises a non-fault state light intensity distribution image dataset and a fault state light intensity distribution image dataset;
Constructing a first low-frequency analysis recognition spectrogram by using the non-fault state light intensity distribution image data set, and constructing a second low-frequency analysis recognition spectrogram by using the fault state light intensity distribution image data set;
training a classification recognition model for distinguishing different states of the optical module by adopting the first low-frequency analysis recognition spectrogram and the second low-frequency analysis recognition spectrogram;
and inputting a low-frequency analysis recognition spectrogram corresponding to the optical module to be tested into the trained classification recognition model, and outputting the current state result of the optical module to be tested.
In a first aspect, in the above optical module state monitoring method provided by the present invention, before training a classification recognition model for distinguishing different states of an optical module by using the first low-frequency analysis recognition spectrogram and the second low-frequency analysis recognition spectrogram, the method further includes:
Constructing a super-resolution image reconstruction model, and training the super-resolution image reconstruction model by adopting a low-frequency analysis recognition spectrogram training set;
performing enhancement processing on the first low-frequency analysis recognition spectrogram by adopting the trained super-resolution image reconstruction model;
or, the trained super-resolution image reconstruction model is adopted to carry out enhancement processing on the first low-frequency analysis recognition spectrogram and the second low-frequency analysis recognition spectrogram.
On the other hand, in the above optical module state monitoring method provided by the invention, in the process of constructing a super-resolution image reconstruction model and training the super-resolution image reconstruction model by adopting a low-frequency analysis recognition spectrogram training set, the method comprises the following steps:
constructing the super-resolution image reconstruction model by using a generator and a discriminator;
In the structure of the generator, a convolution layer is utilized to extract the characteristics of the original image in the low-frequency analysis and identification spectrogram training set;
Continuously extracting features through a plurality of residual blocks; the residual block comprises a convolution layer, a batch normalization and activation function;
performing up-sampling operation after extracting features;
Refining the upsampling result by using a plurality of convolution layers after the upsampling operation, and outputting a first image;
in the structure of the arbiter, inputting the first image and the corresponding real high resolution image output by the generator into the arbiter;
extracting basic features of an input image by using a convolution layer; the convolution layer is followed by a first nonlinear activation function;
after the basic features are extracted, a plurality of structures consisting of a convolution layer, batch normalization and a second nonlinear activation function are adopted to process the basic features, so that high-dimensional features are obtained;
And carrying out downsampling operation on the high-dimensional features, carrying out global average pooling or flattening, processing by at least one dense connecting layer, and finally outputting a scalar value for representing the probability that the first image is a real image by the dense layer of a single node.
On the other hand, in the above optical module state monitoring method provided by the invention, in the process of constructing a super-resolution image reconstruction model and training the super-resolution image reconstruction model by adopting a low-frequency analysis recognition spectrogram training set, the method further comprises:
The channel attention mechanism is integrated in a generator of the super resolution image reconstruction model.
On the other hand, in the above optical module state monitoring method provided by the present invention, the integration of the channel attention mechanism in the generator of the super-resolution image reconstruction model includes:
In the structure of the generator, a channel attention mechanism is integrated behind a convolution layer or within a residual block of the generator;
Extracting features of an original image in the low-frequency analysis recognition spectrogram training set by using a convolution layer, compressing spatial information of each channel by using global average pooling after obtaining an original feature image, and generating channel-level description;
through a plurality of full connection layers and activation functions, a complex nonlinear relation among channels is learned, and a weight coefficient of each channel is output;
multiplying the weight coefficient with the original feature map, and recalibrating each channel.
On the other hand, in the above optical module state monitoring method provided by the present invention, in the process of constructing the first low-frequency analysis and identification spectrogram, the method includes:
Converting a time domain signal of the non-fault state light intensity distribution image data into a frequency domain signal by utilizing short-time Fourier transformation, and extracting characteristic frequency components in the frequency domain signal;
after extracting the characteristic frequency components in the frequency domain signal, calculating energy of each characteristic frequency component;
After calculating the energy of each characteristic frequency component, obtaining the energy difference between the characteristic frequency components; the obtained energy difference represents the energy distribution situation among different frequencies;
And converting the energy difference between the obtained characteristic frequency components into corresponding gray values to obtain the first low-frequency analysis identification spectrogram.
On the other hand, in the above optical module state monitoring method provided by the present invention, after acquiring the data sample set during the operation of the optical module, the method further includes:
And sequentially performing de-drying processing and normalization processing on the non-fault state light intensity distribution image data set and the fault state light intensity distribution image data set so as to map the light intensity range into a set numerical range.
In order to solve the technical problem, the invention further provides an optical module state monitoring device, which comprises:
the sample set acquisition module is used for acquiring a data sample set when the optical module works; the data sample set comprises a non-fault state light intensity distribution image dataset and a fault state light intensity distribution image dataset;
the spectrogram construction module is used for constructing a first low-frequency analysis recognition spectrogram by utilizing the non-fault state light intensity distribution image data set and constructing a second low-frequency analysis recognition spectrogram by utilizing the fault state light intensity distribution image data set;
The classification recognition model training module is used for training classification recognition models for distinguishing different states of the optical module by adopting the first low-frequency analysis recognition spectrogram and the second low-frequency analysis recognition spectrogram;
And the classification recognition model reasoning module is used for inputting the low-frequency analysis recognition spectrogram corresponding to the optical module to be detected into the trained classification recognition model and outputting the current state result of the optical module to be detected.
In order to solve the technical problem, the present invention further provides an optical module state monitoring device, including:
A memory for storing a computer program;
And the processor is used for realizing the steps of the optical module state monitoring method when executing the computer program.
In order to solve the above technical problem, the present invention further provides a computer readable storage medium, on which a computer program is stored, which when executed by a processor, implements the steps of the optical module state monitoring method described above.
According to the technical scheme, the optical module state monitoring method provided by the invention comprises the following steps of: acquiring a data sample set of the optical module during working; the data sample set comprises a non-fault state light intensity distribution image dataset and a fault state light intensity distribution image dataset; constructing a first low-frequency analysis recognition spectrogram by using the non-fault state light intensity distribution image data set, and constructing a second low-frequency analysis recognition spectrogram by using the fault state light intensity distribution image data set; training classification recognition models for distinguishing different states of the optical module by adopting a first low-frequency analysis recognition spectrogram and a second low-frequency analysis recognition spectrogram; and inputting the low-frequency analysis recognition spectrogram corresponding to the optical module to be tested into the trained classification recognition model, and outputting the current state result of the optical module to be tested.
The optical module state monitoring method has the beneficial effects that the optical module state monitoring method firstly utilizes the light intensity distribution image data sets of the optical modules in different states to construct the low-frequency analysis recognition spectrogram, and then uses the low-frequency analysis recognition spectrogram corresponding to the different states of the optical modules to train the classification recognition model, so that the current working state of the optical modules can be monitored in real time by using the trained classification recognition model, further, whether the optical modules are in failure can be effectively detected, the accuracy and the efficiency of the optical module failure detection are improved, the failed optical modules can be repaired or replaced by adopting measures in time, and the stability and the reliability of an optical communication system are improved.
In addition, the invention also provides a corresponding optical module state monitoring device, optical module state monitoring equipment and a computer readable storage medium aiming at the optical module state monitoring method, and the optical module state monitoring device and the optical module state monitoring method have the same or corresponding technical characteristics as the optical module state monitoring method, and have the same effects.
Drawings
For a clearer description of embodiments of the present invention, the drawings that are required to be used in the embodiments will be briefly described, it being apparent that the drawings in the following description are only some embodiments of the present invention, and other drawings may be obtained according to the drawings without inventive effort for those skilled in the art.
Fig. 1 is a flowchart of an optical module state monitoring method according to an embodiment of the present invention;
FIG. 2 is a schematic diagram of a construction process of a low-frequency analysis recognition spectrogram provided by an embodiment of the invention;
FIG. 3 is a schematic diagram of a generator according to an embodiment of the present invention;
FIG. 4 is a schematic diagram of a discriminator according to the embodiment of the invention;
Fig. 5 is a schematic structural diagram of an optical module state monitoring device according to an embodiment of the present invention;
fig. 6 is a schematic structural diagram of an optical module state monitoring device according to an embodiment of the present invention.
Detailed Description
Most of the monitoring methods for the faults of the optical module are judged based on rules or experience, false alarm or missing alarm is easy to generate, a large amount of manual intervention is needed, time and effort are consumed in practice, and the faults cannot be accurately identified. In the related technical scheme, most optical modules support digital diagnosis monitoring (Digital Diagnostic Monitoring, DDM) functions, DDM information of the optical modules can be checked by using related instructions, and the warning states of the emitted light power, the received light power, the current, the voltage and the temperature are comprehensively analyzed by the digital diagnosis functions, so that faults are rapidly positioned and solved. However, the DDM technology is used for judging and predicting faults by checking and comprehensively analyzing a plurality of data. For optical modules, different types of faults may result in variations of multiple characteristic parameters, which may increase the complexity of the threshold adjustment. And DDM is mainly suitable for progressive drift detection, while predictive power for sudden faults may be weak.
Based on the above, the invention provides a method, a device, equipment and a medium for monitoring the optical module state, so as to solve the technical problem that the current state of the optical module cannot be accurately identified.
The following description of the embodiments of the present invention will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present invention, but not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by a person of ordinary skill in the art without making any inventive effort are within the scope of the present invention.
Fig. 1 is a flowchart of an optical module state monitoring method according to an embodiment of the present invention, as shown in fig. 1, where the method includes:
S101, acquiring a data sample set of an optical module during working; the data sample set includes a non-fault state light intensity distribution image dataset and a fault state light intensity distribution image dataset.
It should be noted that, the light intensity distribution image of the optical module during operation may be a light intensity distribution image of the optical module at the input port, and is mainly used for subsequent fault monitoring. The non-fault state light intensity distribution image dataset may be understood as a corresponding light intensity distribution image dataset in normal operation of the light module. The faulty state light intensity distribution image dataset may be understood as a corresponding light intensity distribution image dataset in the various faulty states (i.e. abnormal states) in which the light module is operating. Wherein, each fault in the fault state light intensity distribution image dataset can be labeled, and each fault has a corresponding label.
In a specific implementation, after the step S101 is executed to obtain the data sample set during the operation of the optical module, the method may further include: the non-fault state light intensity distribution image data set and the fault state light intensity distribution image data set are subjected to a de-drying process and a normalization process in sequence so as to map the light intensity range into a set numerical range.
In implementations, after the data sample set is acquired, the data sample set may be preprocessed. Illustratively, the non-fault state light intensity distribution image data set and the fault state light intensity distribution image data set are subjected to a de-drying process to eliminate possible interference information, and then the non-fault state light intensity distribution image data set and the fault state light intensity distribution image data set subjected to the de-drying process are subjected to a normalization process to map the light intensity range into a set numerical range, so that subsequent processing is facilitated.
S102, constructing a first low-frequency analysis recognition spectrogram by using the non-fault state light intensity distribution image data set, and constructing a second low-frequency analysis recognition spectrogram by using the fault state light intensity distribution image data set.
It should be noted that a Low-Frequency analysis Recognition (LOFAR) is a framework based on Low-Frequency feature analysis and Recognition, and fault monitoring and Recognition is performed by analyzing a Low-Frequency part of a signal. In the invention, the light intensity distribution image data set of the light module is provided with low-frequency characteristic data, and a low-frequency analysis recognition spectrogram can be constructed by utilizing the light intensity distribution image data set. The spectrogram can be identified by using low-frequency analysis, and the characteristics of the optical module in normal operation can be found out by calculating the abnormal degree of each sample in the spectrogram.
In specific implementation, in the process of constructing the first low-frequency analysis identification spectrogram, the method specifically may include: firstly, converting a time domain signal of non-fault state light intensity distribution image data into a frequency domain signal by utilizing short-time Fourier transformation, and extracting characteristic frequency components in the frequency domain signal; then, calculating the energy of each characteristic frequency component; then, obtaining energy difference between each characteristic frequency component; the obtained energy difference represents the energy distribution situation among different frequencies; and finally, converting the energy difference between the obtained characteristic frequency components into corresponding gray values to obtain a first low-frequency analysis identification spectrogram.
Similarly, in the implementation, in the process of constructing the second low-frequency analysis identification spectrogram, the method specifically may include: firstly, converting a time domain signal of fault state light intensity distribution image data into a frequency domain signal by utilizing short-time Fourier transform, and extracting characteristic frequency components in the frequency domain signal; then, calculating the energy of each characteristic frequency component; then, obtaining energy difference between each characteristic frequency component; the obtained energy difference represents the energy distribution situation among different frequencies; and finally, converting the energy difference between the obtained characteristic frequency components into corresponding gray values to obtain a second low-frequency analysis identification spectrogram.
Fig. 2 is a schematic diagram of a construction process of a low-frequency analysis recognition spectrogram provided by an embodiment of the invention. As shown in fig. 2, after inputting the light intensity distribution image data, preprocessing, extracting the characteristic frequency, calculating the energy difference, and obtaining the gray sample, the low-frequency analysis recognition spectrogram is obtained.
First, after acquiring the light intensity distribution image data, which typically contains possible disturbances, the preprocessing step includes filtering noise, signal enhancement and removing extraneous signals, whereas the preprocessed signals are more suitable for feature extraction. In implementations, a first low frequency analysis recognition spectrum may be constructed using the pre-processed non-fault state light intensity distribution image dataset and a second low frequency analysis recognition spectrum may be constructed using the pre-processed fault state light intensity distribution image dataset. The first low-frequency analysis and identification spectrogram has corresponding labels aiming at the non-fault state of the optical module, and the second low-frequency analysis and identification spectrogram has different labels aiming at different faults of the optical module.
And extracting characteristic frequencies in the characteristic frequency signals is a key for constructing a low-frequency analysis identification spectrogram. Extracting the characteristic frequencies typically involves spectral analysis, such as a fast fourier transform or wavelet transform, etc. By means of the techniques, the time domain signal can be converted into the frequency domain signal, and the main frequency component of the signal is extracted, and the method mainly adopts short-time Fourier transform, so that the method is a method with high practicability.
The energy difference is then calculated, and after the characteristic frequency is acquired, the energy of each frequency component is calculated, which helps determine the importance of the individual frequency components. The energy difference refers to the energy distribution between different frequencies. In some cases, a particular energy distribution pattern may be associated with a particular signal source or event.
In acquiring a gray scale sample, the low frequency analysis identifies a spectrogram similar to a gray scale image, where each characteristic frequency corresponds to a horizontal line of the image and its energy value corresponds to the gray scale level of that line. Acquiring a gray sample refers to converting an energy value into a corresponding gray value, with higher gray values representing stronger energy. Finally, by drawing all the characteristic frequencies and the corresponding gray values on a two-dimensional plane, a low-frequency analysis recognition spectrogram can be obtained. The horizontal axis generally represents time, the vertical axis represents frequency, and the gray value of the image represents the energy level of the signal at a particular time and frequency.
For example, in the process of obtaining the second low-frequency analysis recognition spectrogram, the distance between each gray sample and the nearest neighbor gray sample can be calculated, and then the local anomaly factor of each gray sample is calculated according to the density and distance information of the nearest neighbor gray sample. Where density refers to the number of adjacent gray samples around each gray sample. If there are many adjacent gray samples around a gray sample, the density of the gray sample is higher; conversely, the density is lower. This may help identify gray samples of relatively few neighbors around it, which may be considered outliers. The distance information refers to the distance between each gray sample and its nearest neighbor gray sample. By comparing the distance between a gray sample and its nearest neighbor gray sample, the degree of tightness between the gray sample and its neighbors can be measured. If a gray sample is far farther from its nearest neighbor than the other gray samples, it may have a higher outlier score. And finally, sorting according to the values of the local anomaly factors to form a final second low-frequency analysis recognition spectrogram.
S103, training classification recognition models for distinguishing different states of the optical module by adopting the first low-frequency analysis recognition spectrogram and the second low-frequency analysis recognition spectrogram.
In the invention, a machine learning network model, namely a classification recognition model, is trained by adopting a low-frequency analysis recognition spectrogram corresponding to a non-fault state of the optical module and a low-frequency analysis recognition spectrogram corresponding to a fault state so as to learn the characteristic mode of each state of the optical module. The classification recognition model can be used for distinguishing different states of the optical module, including non-fault states (namely normal states), fault states (comprising serious aging, damage and the like), namely the classification recognition model aims at classifying an input low-frequency analysis recognition spectrogram into a plurality of predefined optical module states, such as normal states, serious aging, damage and the like, and each state can be marked after classification.
S104, inputting the low-frequency analysis recognition spectrogram corresponding to the optical module to be tested into the trained classification recognition model, and outputting the current state result of the optical module to be tested.
Step S104 is an inference stage of the classification recognition model, and the current state of the optical module can be detected in real time by using the trained classification recognition model. When the low-frequency analysis recognition spectrogram corresponding to the optical module to be detected is input into the trained classification recognition model, the classification recognition model can classify according to the classification labels, and whether the optical module to be detected is non-fault data or not can be output.
In the optical module state monitoring method provided by the embodiment of the invention, the low-frequency analysis recognition spectrogram is firstly constructed by utilizing the light intensity distribution image data sets of the optical module in different states, and then the classification recognition model is trained by utilizing the low-frequency analysis recognition spectrogram corresponding to the different states of the optical module, so that the current working state of the optical module can be monitored in real time by using the trained classification recognition model, further, whether the optical module fails or not can be effectively detected, the accuracy and the efficiency of the optical module failure detection are improved, and the optical module with failure can be repaired or replaced by adopting measures in time, thereby improving the stability and the reliability of an optical communication system.
Further, in a specific implementation, in the above optical module state monitoring method provided by the embodiment of the present invention, before executing step S103 to train the classification recognition model for distinguishing different states of the optical module by using the first low-frequency analysis recognition spectrogram and the second low-frequency analysis recognition spectrogram, the method may further include: firstly, constructing a super-resolution image reconstruction model, and adopting a low-frequency analysis recognition spectrogram training set to train the super-resolution image reconstruction model; then, a trained super-resolution image reconstruction model is adopted to carry out enhancement treatment on the first low-frequency analysis recognition spectrogram; or, adopting a trained super-resolution image reconstruction model to carry out enhancement treatment on the first low-frequency analysis recognition spectrogram and the second low-frequency analysis recognition spectrogram.
It should be noted that, the Super Resolution image reconstruction model (Super Resolution GENERATIVE ADVERSARIAL Network, SRGAN) is a Network of generating countermeasure structures for improving the Resolution and quality of the image. The input blurred image is converted into a more real high-resolution image with rich details by using a generator in a super-resolution image reconstruction model, and a discriminator is used for distinguishing a false high-resolution image and a real high-resolution image generated by the generator. The generator and the arbiter are used for generating super-resolution images with high quality and authenticity through countermeasure training and continuously adjusting network parameters.
In the invention, a first low-frequency analysis and identification spectrogram corresponding to a non-fault state of an optical module or all low-frequency analysis and identification spectrograms corresponding to the optical module when in operation are used as input of a generator of a super-resolution image reconstruction model, and a generated super-resolution image is used as output. And a super-resolution image reconstruction model capable of effectively improving the image resolution is trained by continuously optimizing the countermeasure process between the generator and the discriminator.
In the implementation, the super-resolution image reconstruction model is utilized to carry out enhancement processing on the low-frequency analysis recognition spectrogram, so that the resolution and quality of the low-frequency analysis recognition spectrogram can be improved. The enhanced low-frequency analysis recognition spectrogram is utilized to train the classification recognition model, so that parameters of the classification recognition model are more accurate, the accuracy of the reasoning result of the classification recognition model is further improved, the characteristics are more obvious during classification, the effect obtained by classification is clearer, and the accuracy and the efficiency of optical module fault monitoring are ensured.
Based on this, in implementation, in the process of constructing the super-resolution image reconstruction model in the above steps and training the super-resolution image reconstruction model by using the low-frequency analysis recognition spectrogram training set, the method specifically may include the following steps:
Constructing a super-resolution image reconstruction model by using a generator and a discriminator;
in the structure of the generator, a convolution layer is utilized to extract the characteristics of the original image in the low-frequency analysis recognition spectrogram training set;
continuously extracting features through a plurality of residual blocks; the residual block comprises a convolution layer and a batch normalization and activation function;
performing up-sampling operation after extracting features;
Refining the upsampling result by using a plurality of convolution layers after the upsampling operation, and outputting a first image;
In the structure of the discriminator, inputting the first image output by the generator and the corresponding real high-resolution image into the discriminator;
extracting basic features of an input image by using a convolution layer; the convolution layer is followed by a first nonlinear activation function;
After extracting the basic features, processing the basic features by adopting a plurality of structures consisting of a convolution layer, batch normalization and a second nonlinear activation function to obtain high-dimensional features;
And carrying out downsampling operation on the high-dimensional features, carrying out global average pooling or flattening, processing by at least one dense connecting layer, and finally outputting a scalar value for representing the probability that the first image is a real image by the dense layer of one single node.
Fig. 3 is a schematic structural diagram of a generator according to an embodiment of the present invention. As shown in fig. 3, in the generator, feature extraction is first performed on an input low-resolution original image using one convolution layer. Then, continuously extracting features through a plurality of residual blocks and enhancing the learning capacity of the model; the skipped connections contained in these residual blocks help learn image details and alleviate the gradient vanishing problem. The residual block consists of a convolution layer, batch normalization and activation functions. After extracting sufficient features, the network increases the spatial resolution of the image through an upsampling layer, which may be done by transpose convolution or pixel rearrangement. Finally, the up-sampling result is refined by a series of convolution layers, and a final high-resolution image, namely a first image, is output. Overall, this architecture of the generator aims to effectively reconstruct high frequency details in the high resolution image to produce a high quality output without significantly increasing computational complexity. During the course of the countermeasure training, the generator will constantly learn how to produce more realistic high resolution images to better fool the discriminator.
Fig. 4 is a schematic structural diagram of a discriminator according to the embodiment of the invention. As shown in fig. 4, in the arbiter the input is a pair of images, namely the first image produced by the generator and the corresponding real high resolution image. First, a convolution layer is responsible for extracting the basic features of the image, and a first nonlinear activation function, such as a Leaky ReLU (a nonlinear function related to Sigmoid and ReLU functions) activation function, is used. Subsequently, a series of deeper convolutional layers, possibly in combination with a batch normalization and a second nonlinear activation function, such as the leak ReLU activation function, increase the number of convolutional kernels layer by layer to capture more complex features. In deep networks, a step-wise convolutional layer is used to reduce the size and increase the depth of the feature map, acting as a downsampling to reduce the computational effort and widen the receptive field. Next, the feature map is pooled or flattened by global averaging, and then processed by one or more dense connection layers to transform the high-dimensional features into a final classification result. The final output of the arbiter is given by a dense layer of single nodes that activate the function using Sigmoid (a nonlinear function that can map the real value of the input to a value between 0-1 and has saturation), outputting a scalar value representing the probability that the image is a true image, 0 representing a "false image" and 1 representing a "true image". The thus constructed discriminators are intended to distinguish the generated image from the actual image, providing discrimination capability for countermeasure training.
Further, in a specific implementation, in the above optical module state monitoring method provided by the embodiment of the present invention, in a process of constructing a super-resolution image reconstruction model and training the super-resolution image reconstruction model by using a low-frequency analysis recognition spectrogram training set, the method may further include: the channel attention mechanism is integrated in the generator of the super resolution image reconstruction model.
In order to further improve the accuracy of the fault detection of the optical module, the invention introduces a attention mechanism into the generator to make the generator pay more attention to the mutual influence of the characteristics among different channels, so that the perception and utilization capability of the super-resolution image reconstruction model on key information corresponding to different faults of the optical module can be enhanced by changing the attention degree of the super-resolution image reconstruction model to input data, the interference of irrelevant information is reduced, and the prediction precision of the super-resolution image reconstruction model is improved. Meanwhile, as the learning time is longer, the model is more complete to construct, and the training time of the model after the attention mechanism is introduced is also reduced.
More or less attention on different areas may be achieved by weighting the data to be processed, for example. The mechanism of distraction means that the model is designed to more focus on identifying and processing the most important parts of the information.
In a specific implementation, the integration of the channel attention mechanism in the generator of the super-resolution image reconstruction model in the above steps may specifically include:
in the structure of the generator, the channel attention mechanism is integrated behind the convolution layer of the generator or in the residual block;
Extracting features of an original image in a low-frequency analysis recognition spectrogram training set by using a convolution layer, compressing spatial information of each channel by using global average pooling after obtaining the original feature image, and generating channel-level description;
through a plurality of full connection layers and activation functions, a complex nonlinear relation among channels is learned, and a weight coefficient of each channel is output;
multiplying the weight coefficient with the original characteristic diagram, and carrying out recalibration on each channel.
In practice, first, the generator extracts feature maps from the input image through the convolutional layer, and then compresses the spatial information for each channel using global averaging pooling, resulting in a description of the channel level. Then, through a series of full connection layers and ReLU and Sigmoid activation functions, the network learns complex nonlinear relations among channels, and outputs the weight coefficient of each channel. Finally, the coefficients are multiplied by the original feature map, and the channels are recalibrated, so that important features are strengthened and unimportant is restrained. By embedding this attention module in a key location of the generator, such as after a convolution layer or within a residual block, the generator can focus more on key channel features during reconstruction of the high resolution image, thereby generating a higher quality image during the countermeasure training.
It should be added that in the process of training the super-resolution image reconstruction model or the classification recognition model, the convergence speed and performance of the model may be affected to a certain extent by the selection of the learning rate. If the learning rate is chosen too high, the step size of the gradient decrease may be too large, resulting in an incorrect convergence of the algorithm and possibly even an unstable result. If the learning rate of choice is too low, the training speed of the model can become very slow and may cause the algorithm to sink to local minima. Therefore, when the model is found to be unstable during training or the convergence speed is slow, an attempt may be made to reduce the learning rate.
In the above embodiments, the present invention further provides an optical module state monitoring device and an embodiment corresponding to the optical module state monitoring device. It should be noted that the present invention describes an embodiment of the device portion from two angles, one based on the angle of the functional module and the other based on the angle of the hardware.
Fig. 5 is a schematic structural diagram of an optical module state monitoring device according to an embodiment of the present invention. The embodiment is based on the angle of the functional module, as shown in fig. 5, and the device includes:
the sample set acquisition module 10 is used for acquiring a data sample set when the optical module works; the data sample set comprises a non-fault state light intensity distribution image dataset and a fault state light intensity distribution image dataset;
A spectrogram construction module 11, configured to construct a first low-frequency analysis and identification spectrogram using the non-fault state light intensity distribution image dataset, and construct a second low-frequency analysis and identification spectrogram using the fault state light intensity distribution image dataset;
The classification recognition model training module 12 is configured to train a classification recognition model for distinguishing different states of the optical module by using the first low-frequency analysis recognition spectrogram and the second low-frequency analysis recognition spectrogram;
and the classification recognition model reasoning module 13 is used for inputting the low-frequency analysis recognition spectrogram corresponding to the optical module to be detected into the trained classification recognition model and outputting the current state result of the optical module to be detected.
In the optical module state monitoring device provided by the embodiment of the invention, the current working state of the optical module can be monitored in real time through the interaction of the four modules, so that whether the optical module fails or not is effectively detected, the accuracy and the efficiency of the optical module failure detection are improved, the failed optical module can be repaired or replaced by adopting measures in time, and the stability and the reliability of an optical communication system are improved.
Since the embodiments of the apparatus portion and the embodiments of the method portion correspond to each other, the embodiments of the apparatus portion are referred to the description of the embodiments of the method portion, and are not repeated herein. And has the same advantageous effects as the above-mentioned optical mode block state monitoring method.
Further, in a specific implementation, the optical module state monitoring device provided by the embodiment of the present invention may further include:
The preprocessing module is used for sequentially performing the de-drying processing and the normalization processing on the non-fault state light intensity distribution image data set and the fault state light intensity distribution image data set so as to map the light intensity range into a set numerical range.
Further, in the optical module state monitoring device provided by the embodiment of the present invention, the spectrogram construction module 11 may be specifically configured to convert a time domain signal of the non-fault state light intensity distribution image data into a frequency domain signal by using short-time fourier transform, and extract a characteristic frequency component in the frequency domain signal; then, calculating the energy of each characteristic frequency component; then, obtaining energy difference between each characteristic frequency component; the obtained energy difference represents the energy distribution situation among different frequencies; finally, converting the energy difference between the obtained characteristic frequency components into corresponding gray values to obtain a first low-frequency analysis identification spectrogram; the method can be used for converting time domain signals of fault state light intensity distribution image data into frequency domain signals by short-time Fourier transformation, and extracting characteristic frequency components in the frequency domain signals; then, calculating the energy of each characteristic frequency component; then, obtaining energy difference between each characteristic frequency component; the obtained energy difference represents the energy distribution situation among different frequencies; and finally, converting the energy difference between the obtained characteristic frequency components into corresponding gray values to obtain a second low-frequency analysis identification spectrogram.
Further, in a specific implementation, the optical module state monitoring device provided by the embodiment of the present invention may further include:
The image reconstruction model training module is used for constructing a super-resolution image reconstruction model and training the super-resolution image reconstruction model by adopting a low-frequency analysis recognition spectrogram training set; the method is also used for enhancing the first low-frequency analysis recognition spectrogram by adopting a trained super-resolution image reconstruction model; or, adopting a trained super-resolution image reconstruction model to carry out enhancement treatment on the first low-frequency analysis recognition spectrogram and the second low-frequency analysis recognition spectrogram.
When in specific implementation, the image reconstruction model training module can be specifically used for constructing a super-resolution image reconstruction model by utilizing the generator and the discriminator; in the structure of the generator, a convolution layer is utilized to extract the characteristics of the original image in the low-frequency analysis recognition spectrogram training set; continuously extracting features through a plurality of residual blocks; the residual block comprises a convolution layer and a batch normalization and activation function; performing up-sampling operation after extracting features; refining the upsampling result by using a plurality of convolution layers after the upsampling operation, and outputting a first image; in the structure of the discriminator, inputting the first image output by the generator and the corresponding real high-resolution image into the discriminator; extracting basic features of an input image by using a convolution layer; the convolution layer is followed by a first nonlinear activation function; after extracting the basic features, processing the basic features by adopting a plurality of structures consisting of a convolution layer, batch normalization and a second nonlinear activation function to obtain high-dimensional features; and carrying out downsampling operation on the high-dimensional features, carrying out global average pooling or flattening, processing by at least one dense connecting layer, and finally outputting a scalar value for representing the probability that the first image is a real image by the dense layer of one single node.
In particular implementations, the image reconstruction model training module may be further configured to integrate a channel attention mechanism into a generator of the super-resolution image reconstruction model, and may include: in the structure of the generator, the channel attention mechanism is integrated behind the convolution layer of the generator or in the residual block; extracting features of an original image in a low-frequency analysis recognition spectrogram training set by using a convolution layer, compressing spatial information of each channel by using global average pooling after obtaining the original feature image, and generating channel-level description; through a plurality of full connection layers and activation functions, a complex nonlinear relation among channels is learned, and a weight coefficient of each channel is output; multiplying the weight coefficient with the original characteristic diagram, and carrying out recalibration on each channel.
Fig. 6 is a schematic structural diagram of an optical module state monitoring device according to an embodiment of the present invention. The optical module state monitoring device of this embodiment includes, based on the hardware angle, as shown in fig. 6:
a memory 20 for storing a computer program;
A processor 21 for implementing the steps of the optical mode block state monitoring method as mentioned in the above embodiments when executing a computer program.
Processor 21 may include one or more processing cores, such as a 4-core processor, an 8-core processor, etc. The Processor 21 may be implemented in at least one hardware form of a digital signal Processor (DIGITAL SIGNAL Processor, DSP), field-Programmable gate array (Field-Programmable GATE ARRAY, FPGA), programmable logic array (Programmable Logic Array, PLA). The processor 21 may also include a main processor, which is a processor for processing data in an awake state, also called CPU, and a coprocessor; a coprocessor is a low-power processor for processing data in a standby state. In some embodiments, the processor 21 may be integrated with a graphics processor (Graphics Processing Unit, GPU) for rendering and drawing of content required to be displayed by the display screen. In some embodiments, the processor 21 may also include an artificial intelligence (ARTIFICIAL INTELLIGENCE, AI) processor for processing computing operations related to machine learning.
Memory 20 may include one or more computer-readable storage media, which may be non-transitory. Memory 20 may also include high-speed random access memory, as well as non-volatile memory, such as one or more magnetic disk storage devices, flash memory storage devices. In this embodiment, the memory 20 is at least used for storing a computer program 201, which, when loaded and executed by the processor 21, is capable of implementing the relevant steps of the optical mode block state monitoring method disclosed in any of the foregoing embodiments. In addition, the resources stored in the memory 20 may further include an operating system 202, data 203, and the like, where the storage manner may be transient storage or permanent storage. Operating system 202 may include Windows, unix, linux, among other things. The data 203 may include, but is not limited to, the data related to the above-mentioned optical mode state monitoring method, etc.
In some embodiments, the optical module status monitoring device may further include a display 22, an input-output interface 23, a communication interface 24, a power supply 25, and a communication bus 26.
Those skilled in the art will appreciate that the configuration shown in fig. 6 is not limiting of the light module status monitoring device and may include more or fewer components than shown.
The optical module state monitoring device provided by the embodiment of the invention comprises a memory and a processor, wherein the processor can realize the following method when executing a program stored in the memory: the optical module state monitoring method has the same effect.
Finally, the invention also provides a corresponding embodiment of the computer readable storage medium. The computer-readable storage medium has stored thereon a computer program which, when executed by a processor, performs the steps as described in the method embodiments above.
It will be appreciated that the methods of the above embodiments, if implemented in the form of software functional units and sold or used as stand-alone products, may be stored on a computer readable storage medium. Based on this understanding, the technical solution of the present invention may be embodied essentially or in part or all of the technical solution or in part in the form of a software product stored in a storage medium for performing all or part of the steps of the method according to the embodiments of the present invention. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-Only Memory (ROM), a random access Memory (Random Access Memory, RAM), a magnetic disk, or an optical disk, or other various media capable of storing program codes.
The computer readable storage medium provided by the invention comprises the optical module state monitoring method, and the effects are the same as those of the optical module state monitoring method.
It should also be noted that in this specification, relational terms such as first and second, and the like are used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Moreover, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one … …" does not exclude the presence of other like elements in a process, method, article, or apparatus that comprises the element.
The method, the device, the equipment and the medium for monitoring the optical mode state provided by the invention are described in detail. In the description, each embodiment is described in a progressive manner, and each embodiment is mainly described by the differences from other embodiments, so that the same similar parts among the embodiments are mutually referred. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant points refer to the description of the method section. It should be noted that it will be apparent to those skilled in the art that the present invention may be modified and practiced without departing from the spirit of the present invention.
Claims (10)
1. An optical mode block state monitoring method, the method comprising:
Acquiring a data sample set of the optical module during working; the data sample set comprises a non-fault state light intensity distribution image dataset and a fault state light intensity distribution image dataset;
Constructing a first low-frequency analysis recognition spectrogram by using the non-fault state light intensity distribution image data set, and constructing a second low-frequency analysis recognition spectrogram by using the fault state light intensity distribution image data set;
training a classification recognition model for distinguishing different states of the optical module by adopting the first low-frequency analysis recognition spectrogram and the second low-frequency analysis recognition spectrogram;
and inputting a low-frequency analysis recognition spectrogram corresponding to the optical module to be tested into the trained classification recognition model, and outputting the current state result of the optical module to be tested.
2. The optical module state monitoring method of claim 1, further comprising, prior to training a classification recognition model for distinguishing between different states of an optical module using the first low frequency analysis recognition spectrogram and the second low frequency analysis recognition spectrogram:
Constructing a super-resolution image reconstruction model, and training the super-resolution image reconstruction model by adopting a low-frequency analysis recognition spectrogram training set;
performing enhancement processing on the first low-frequency analysis recognition spectrogram by adopting the trained super-resolution image reconstruction model;
or, the trained super-resolution image reconstruction model is adopted to carry out enhancement processing on the first low-frequency analysis recognition spectrogram and the second low-frequency analysis recognition spectrogram.
3. The optical module state monitoring method according to claim 2, wherein in the process of constructing a super-resolution image reconstruction model and training the super-resolution image reconstruction model by using a low-frequency analysis recognition spectrogram training set, the method comprises the steps of:
constructing the super-resolution image reconstruction model by using a generator and a discriminator;
In the structure of the generator, a convolution layer is utilized to extract the characteristics of the original image in the low-frequency analysis and identification spectrogram training set;
Continuously extracting features through a plurality of residual blocks; the residual block comprises a convolution layer, a batch normalization and activation function;
performing up-sampling operation after extracting features;
Refining the upsampling result by using a plurality of convolution layers after the upsampling operation, and outputting a first image;
in the structure of the arbiter, inputting the first image and the corresponding real high resolution image output by the generator into the arbiter;
extracting basic features of an input image by using a convolution layer; the convolution layer is followed by a first nonlinear activation function;
after the basic features are extracted, a plurality of structures consisting of a convolution layer, batch normalization and a second nonlinear activation function are adopted to process the basic features, so that high-dimensional features are obtained;
And carrying out downsampling operation on the high-dimensional features, carrying out global average pooling or flattening, processing by at least one dense connecting layer, and finally outputting a scalar value for representing the probability that the first image is a real image by the dense layer of a single node.
4. The method of optical module state monitoring according to claim 3, wherein in the process of constructing a super-resolution image reconstruction model and training the super-resolution image reconstruction model by using a low-frequency analysis recognition spectrogram training set, the method further comprises:
The channel attention mechanism is integrated in a generator of the super resolution image reconstruction model.
5. The optical module state monitoring method of claim 4, wherein integrating a channel attention mechanism in the generator of the super-resolution image reconstruction model comprises:
In the structure of the generator, a channel attention mechanism is integrated behind a convolution layer or within a residual block of the generator;
Extracting features of an original image in the low-frequency analysis recognition spectrogram training set by using a convolution layer, compressing spatial information of each channel by using global average pooling after obtaining an original feature image, and generating channel-level description;
through a plurality of full connection layers and activation functions, a complex nonlinear relation among channels is learned, and a weight coefficient of each channel is output;
multiplying the weight coefficient with the original feature map, and recalibrating each channel.
6. The optical module state monitoring method of claim 1, wherein in constructing the first low frequency analysis identification spectrum, comprising:
Converting a time domain signal of the non-fault state light intensity distribution image data into a frequency domain signal by utilizing short-time Fourier transformation, and extracting characteristic frequency components in the frequency domain signal;
after extracting the characteristic frequency components in the frequency domain signal, calculating energy of each characteristic frequency component;
After calculating the energy of each characteristic frequency component, obtaining the energy difference between the characteristic frequency components; the obtained energy difference represents the energy distribution situation among different frequencies;
And converting the energy difference between the obtained characteristic frequency components into corresponding gray values to obtain the first low-frequency analysis identification spectrogram.
7. The optical module state monitoring method of claim 1, further comprising, after acquiring the data sample set of the optical module in operation:
And sequentially performing de-drying processing and normalization processing on the non-fault state light intensity distribution image data set and the fault state light intensity distribution image data set so as to map the light intensity range into a set numerical range.
8. An optical mode block state monitoring device, the device comprising:
the sample set acquisition module is used for acquiring a data sample set when the optical module works; the data sample set comprises a non-fault state light intensity distribution image dataset and a fault state light intensity distribution image dataset;
the spectrogram construction module is used for constructing a first low-frequency analysis recognition spectrogram by utilizing the non-fault state light intensity distribution image data set and constructing a second low-frequency analysis recognition spectrogram by utilizing the fault state light intensity distribution image data set;
The classification recognition model training module is used for training classification recognition models for distinguishing different states of the optical module by adopting the first low-frequency analysis recognition spectrogram and the second low-frequency analysis recognition spectrogram;
And the classification recognition model reasoning module is used for inputting the low-frequency analysis recognition spectrogram corresponding to the optical module to be detected into the trained classification recognition model and outputting the current state result of the optical module to be detected.
9. An optical mode block state monitoring device, the device comprising:
A memory for storing a computer program;
a processor for implementing the steps of the optical mode block state monitoring method according to any one of claims 1 to 7 when executing said computer program.
10. A computer readable storage medium, characterized in that the computer readable storage medium has stored thereon a computer program which, when executed by a processor, implements the steps of the optical mode block state monitoring method according to any of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410225883.XA CN117997425A (en) | 2024-02-29 | 2024-02-29 | Optical module state monitoring method, device, equipment and medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410225883.XA CN117997425A (en) | 2024-02-29 | 2024-02-29 | Optical module state monitoring method, device, equipment and medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN117997425A true CN117997425A (en) | 2024-05-07 |
Family
ID=90892992
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202410225883.XA Pending CN117997425A (en) | 2024-02-29 | 2024-02-29 | Optical module state monitoring method, device, equipment and medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117997425A (en) |
-
2024
- 2024-02-29 CN CN202410225883.XA patent/CN117997425A/en active Pending
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112651978B (en) | Sublingual microcirculation image segmentation method and device, electronic equipment and storage medium | |
| CN111914883B (en) | Spindle bearing state evaluation method and device based on deep fusion network | |
| CN111650453A (en) | Power equipment diagnosis method and system based on windowing characteristic Hilbert imaging | |
| KR20210141060A (en) | Machine learning based image anomaly detection system | |
| CN113936132B (en) | Method and system for detecting water pollution of chemical plant based on computer vision | |
| CN114037478A (en) | Advertisement abnormal flow detection method and system, electronic equipment and readable storage medium | |
| CN114782734A (en) | Visual detection method for pipeline leakage of valve cooling system, computer and storage medium | |
| CN115114965B (en) | Wind turbine generator gearbox fault diagnosis method, device, equipment and storage medium | |
| CN115908407B (en) | Power equipment defect detection method and device based on infrared image temperature value | |
| Shajihan et al. | CNN based data anomaly detection using multi-channel imagery for structural health monitoring | |
| CN116680643A (en) | Abnormality detection method combining deep contrast learning and density clustering | |
| CN117115715A (en) | Video anomaly detection method based on combination of stream reconstruction and frame prediction | |
| CN116842379A (en) | Mechanical bearing residual service life prediction method based on DRSN-CS and BiGRU+MLP models | |
| CN111476726A (en) | Unsupervised two-photon calcium imaging denoising method and device based on antagonistic neural network | |
| CN113487223B (en) | Risk assessment method and system based on information fusion | |
| KR102163155B1 (en) | Method and apparatus for converting data by using nested scatter plot | |
| CN114707577A (en) | Anomaly detection method and system based on self-confrontation variational self-encoder | |
| Wang et al. | A structurally re-parameterized convolution neural network-based method for gearbox fault diagnosis in edge computing scenarios | |
| CN112330562B (en) | Heterogeneous remote sensing image transformation method and system | |
| CN117079005A (en) | Optical cable fault monitoring method, system, device and readable storage medium | |
| CN112613494A (en) | Power line monitoring abnormity identification method and system based on deep countermeasure network | |
| CN113052103A (en) | Electrical equipment defect detection method and device based on neural network | |
| CN115356599B (en) | Multi-mode urban power grid fault diagnosis method and system | |
| CN112614094A (en) | Insulator string abnormity positioning and identifying method based on sequence state coding | |
| CN116541771A (en) | Unbalanced sample bearing fault diagnosis method based on multi-scale feature fusion |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |