CN110210536A - A kind of the physical damnification diagnostic method and device of optical interconnection system - Google Patents
A kind of the physical damnification diagnostic method and device of optical interconnection system Download PDFInfo
- Publication number
- CN110210536A CN110210536A CN201910427762.2A CN201910427762A CN110210536A CN 110210536 A CN110210536 A CN 110210536A CN 201910427762 A CN201910427762 A CN 201910427762A CN 110210536 A CN110210536 A CN 110210536A
- Authority
- CN
- China
- Prior art keywords
- training
- sample
- nervus opticus
- network
- physical damnification
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Biology (AREA)
- General Health & Medical Sciences (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Molecular Biology (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Health & Medical Sciences (AREA)
- Image Analysis (AREA)
Abstract
The embodiment of the present invention provides the physical damnification diagnostic method and device of a kind of optical interconnection system.Wherein, method includes: the follow-up power-off signal for obtaining the receiving end of optical interconnection system and being demodulated to modulated signal；By the corresponding nervus opticus network after diagnostic image is input to training of follow-up power-off signal, according to the output of the nervus opticus network after training as a result, obtaining the physical damnification diagnostic result of optical interconnection system.Method and device provided in an embodiment of the present invention improves diagnosis efficiency to a certain extent and reduces error.The neural network for from the beginning constructing a robust with generalization ability is not needed, network complexity is reduced, it is time-consuming shorter when so that carrying out physical damnification diagnosis to optical interconnection system by the nervus opticus network after training, diagnostic result can be obtained in time；Also, the convergence rate in nervus opticus network in training is faster, the nervus opticus network after capable of quickly being trained.
Description
Technical field
The present invention relates to technical field of photo communication more particularly to the physical damnification diagnostic methods and dress of a kind of optical interconnection system
It sets.
Background technique
Optical interconnection system refers to the system that signal is transmitted using optical fiber or other optical transmission mediums, which includes hair
Sending end and receiving end, wherein electric signal to be sent is modulated on light carrier by transmitting terminal, and it is concurrent to obtain modulated optical signal
Out, modulated signal is sent to receiving end by optical fiber or other optical transmission mediums, and receiving end carries out the modulated signal received
Demodulation obtains electric signal.Due to optical fiber or other optical transmission mediums physical characteristic and transmitting terminal and receiving device it is undesirable
Property etc., signal will receive different type and different degrees of physics damage when being transmitted in optical fiber or other optical transmission mediums
Wound leads to system transmission stability and can so that the electric signal of electric signal and transmission that receiving end demodulates is inconsistent
By property decline.
In order to guarantee the transmission quality of optical interconnection system, after signal is damaged, need to carry out physics to impairment signal
Damage diagnosis.The method that the physical damnification of optical interconnection system is diagnosed in the prior art are as follows: experienced expert engineer
Physical damnification diagnosis is carried out to optical interconnection system by artificial experience, it is low that this has resulted in diagnosis efficiency, error Yamato work at
The disadvantages of this is high.
Summary of the invention
The embodiment of the present invention provides the physical damnification diagnostic method and device of a kind of optical interconnection system, existing to solve
The physical damnification diagnostic method of optical interconnection system cannot obtain the problem of diagnostic result in time.
In a first aspect, the embodiment of the present invention provides a kind of physical damnification diagnostic method of optical interconnection system, comprising:
Obtain the follow-up power-off signal that the receiving end of optical interconnection system demodulates modulated signal；
By the corresponding nervus opticus network after diagnostic image is input to training of the follow-up power-off signal, according to described
The output of nervus opticus network after training is as a result, obtain the physical damnification diagnostic result of the optical interconnection system；
Wherein, the physical damnification diagnostic result includes physical damnification type and degree；
The nervus opticus network is according to the first nerves network struction after transfer learning algorithm and training, after the training
Nervus opticus network according to the second training set training obtain；
First nerves network after the training is obtained according to the training of the first training set；
First training set includes multiple sample images and the corresponding sample physics type of impairment of each sample image
Label；
Second training set includes multiple sample images and the corresponding sample physics type of impairment of each sample image
And degree label.
Second aspect, the embodiment of the present invention provide a kind of physical damnification diagnostic device of optical interconnection system, comprising:
Follow-up power-off signal obtains module, and the receiving end for obtaining optical interconnection system is demodulated to obtain to modulated signal
Follow-up power-off signal；
Diagnostic module, for by the follow-up power-off signal it is corresponding after diagnostic image be input to training after nervus opticus
Network, according to the output of the nervus opticus network after the training as a result, the physical damnification for obtaining the optical interconnection system diagnoses
As a result；
Wherein, the physical damnification diagnostic result includes physical damnification type and degree；
The nervus opticus network is according to the first nerves network struction after transfer learning algorithm and training, after the training
Nervus opticus network according to the second training set training obtain；
First nerves network after the training is obtained according to the training of the first training set；
First training set includes multiple sample images and the corresponding sample physics type of impairment of each sample image
Label；
Second training set includes multiple sample images and the corresponding sample physics type of impairment of each sample image
And degree label.
The third aspect, the embodiment of the present invention provides a kind of electronic equipment, including memory, processor and is stored in memory
Computer program that is upper and can running on a processor, is realized when the processor executes described program as first aspect provides
Method the step of.
Fourth aspect, the embodiment of the present invention provide a kind of non-transient computer readable storage medium, are stored thereon with calculating
Machine program is realized as provided by first aspect when the computer program is executed by processor the step of method.
The physical damnification diagnostic method and device of a kind of optical interconnection system provided in an embodiment of the present invention, pass through machine learning
Method to optical interconnection system carry out physical damnification diagnosis, compared with the existing technology in Artificial Diagnosis, mention to a certain extent
High diagnosis efficiency simultaneously reduces error.Further, nervus opticus network is according to first after transfer learning algorithm and training
Neural network building, compared with the prior art in neural network, do not need from the beginning to construct the robust with generalization ability
Neural network, reduce network complexity so that by training after nervus opticus network to optical interconnection system carry out physics
It is time-consuming shorter when damage diagnosis, diagnostic result can be obtained in time；Also, the neural network in compared with the prior art, the
Convergence rate in two neural networks in training faster, nervus opticus network after capable of quickly being trained.
Detailed description of the invention
In order to more clearly explain the embodiment of the invention or the technical proposal in the existing technology, to embodiment or will show below
There is attached drawing needed in technical description to be briefly described, it should be apparent that, the accompanying drawings in the following description is this hair
Bright some embodiments for those of ordinary skill in the art without creative efforts, can be with root
Other attached drawings are obtained according to these attached drawings.
Fig. 1 is a kind of physical damnification diagnostic method flow chart of optical interconnection system provided in an embodiment of the present invention；
Fig. 2 is a kind of structural schematic diagram of the physical damnification diagnostic device of optical interconnection system provided in an embodiment of the present invention；
Fig. 3 is the entity structure schematic diagram of a kind of electronic equipment provided in an embodiment of the present invention.
Specific embodiment
In order to make the object, technical scheme and advantages of the embodiment of the invention clearer, below in conjunction with the embodiment of the present invention
In attached drawing, technical scheme in the embodiment of the invention is clearly and completely described, it is clear that described embodiment is
A part of the embodiment of the present invention, instead of all the embodiments.Based on the embodiments of the present invention, those of ordinary skill in the art
Every other embodiment obtained without creative efforts, shall fall within the protection scope of the present invention.
Fig. 1 is a kind of physical damnification diagnostic method flow chart of optical interconnection system provided in an embodiment of the present invention, such as Fig. 1 institute
Show, this method comprises:
Step 101, the follow-up power-off signal that the receiving end of optical interconnection system demodulates modulated signal is obtained.
Specifically, optical interconnection system refers to the system that signal is transmitted using optical fiber or other optical transmission mediums, this is
System includes transmitting terminal and receiving end, wherein electric signal to be sent is modulated on light carrier by transmitting terminal, obtains modulated signal simultaneously
Issue, modulated signal is transmitted to receiving end by optical fiber or other optical transmission mediums, receiving end to the modulated signal received into
Row demodulation obtains electric signal.Due to the physical characteristic and transmitting terminal of optical fiber or other optical transmission mediums and paying no attention to for receiving device
Property thought etc., signal will receive different type and different degrees of physics damage when being transmitted in optical fiber or other optical transmission mediums
Wound, so that the electric signal that receiving end demodulates differs larger with the electric signal of transmission, reduces system transmission stability
And reliability.
In order to guarantee the transmission quality of optical interconnection system, after signal is damaged, object is carried out to impairment signal with needing
Reason damage diagnosis.It should be noted that carrying out physical damnification diagnosis to optical interconnection system refers to the electricity detected to receiving end
The diagnosis of physical damnification type and degree suffered by signal, and as the physical damnification diagnostic result of optical interconnection system.
It should be noted that physical damnification type and degree include physical damnification type and the damage journey under the type
Degree.Wherein, physical damnification type includes: limiting effect, pattern effect, overshoot, " eyes " crooked effect, extinction ratio is insufficient and light
Power mismatch etc., the present invention is not especially limit this.For each physical damnification type, under the type
Degree of injury may each comprise: degree of injury severe, degree of injury moderate and degree of injury are slight etc., the embodiment of the present invention to this not
Make specific limit.
In order to carry out physical damnification diagnosis to optical interconnection system, needs to obtain receiving end and modulated signal is demodulated to obtain
Electric signal, at this point, for ease of description, which is known as follow-up power-off signal.
Step 102, by the follow-up power-off signal it is corresponding after diagnostic image be input to training after nervus opticus network,
According to the output of the nervus opticus network after the training as a result, obtaining the physical damnification diagnostic result of the optical interconnection system.
Specifically, it is corresponding to diagnostic image that follow-up power-off signal is obtained first.It wherein, is that can embody to diagnostic image
The image of follow-up power-off signal Global Information, for example, the eye figure of follow-up power-off signal, planisphere, signal pattern figure, spectrogram or
Spectrogram etc..
Then by the nervus opticus network after diagnostic image is input to training, according to the nervus opticus network after training
Output is as a result, obtain the physical damnification diagnostic result of optical interconnection system.
Wherein, nervus opticus network is according to the first nerves network struction after transfer learning algorithm and training.That is, first will
First nerves network is by including multiple sample images and the corresponding sample physics type of impairment label of each sample image
First training set is trained, the first nerves network after being trained；It is then based on transfer learning algorithm, by after training
Part-structure and parameter in one neural network, or, entire infrastructure and parameter are replicated；The structure finally obtained based on duplication
Nervus opticus network is constructed with parameter, that is, the structure and parameter for obtaining duplication is as a part of nervus opticus network, building
Nervus opticus network.Wherein, transfer learning algorithm includes: to freeze algorithm and fine tuning algorithm etc., and the embodiment of the present invention does not make this
It is specific to limit.
After nervus opticus network struction is good, according to including multiple sample images and the corresponding sample contents of each sample image
Second training set of reason type of impairment and degree label is trained nervus opticus network, the nervus opticus net after being trained
Network.It should be noted that if transfer learning algorithm is to freeze algorithm, then the ginseng of specific structure in nervus opticus network is only trained
Number, wherein specific structure refers to rejecting the other structures except the structure that duplication obtains in nervus opticus network；If migration is learned
Practising algorithm is fine tuning algorithm, then trains the parameter of the entire infrastructure of nervus opticus network.
It is understood that the nervus opticus network after training can be exported to the corresponding physical damnification type of diagnostic image
And degree label, and then according to the physical damnification type and degree of label acquisition optical interconnection system.
The physical damnification diagnostic method of optical interconnection system provided in an embodiment of the present invention, by the method for machine learning to light
Interconnection system carry out physical damnification diagnosis, compared with the existing technology in Artificial Diagnosis, improve to a certain extent diagnosis effect
Rate simultaneously reduces error.Further, nervus opticus network is according to the first nerves network structure after transfer learning algorithm and training
Build, compared with the prior art in neural network, do not need from the beginning to construct the neural network of a robust with generalization ability,
Network complexity is reduced, when so that carrying out physical damnification diagnosis to optical interconnection system by the nervus opticus network after training,
It is time-consuming shorter, diagnostic result can be obtained in time；Also, the neural network in compared with the prior art, nervus opticus network
In convergence rate in training faster, nervus opticus network after capable of quickly being trained.
Based on any of the above-described embodiment, the embodiment of the present invention is illustrated the acquisition of the nervus opticus network after training,
That is, by the corresponding nervus opticus network after diagnostic image is input to training of the follow-up power-off signal, before further include:
Step 001, multiple sample images and the corresponding sample physics type of impairment of each sample image and degree are obtained
Label；
Step 002, using each sample image and corresponding sample physics type of impairment and degree label as a training
Sample；
Step 003, multiple training samples are obtained, using the multiple training sample as the second training set, pass through described
Two training sets are trained nervus opticus network, the nervus opticus network after being trained.
Specifically, it will also need before the nervus opticus network that diagnostic image is input to after training to nervus opticus net
Network is trained, and specific training process is as follows:
Firstly, obtaining multiple sample images and the corresponding sample physics type of impairment of each sample image and degree mark
Label.Wherein, each sample image can obtain in the following way:
Analogue system is established based on VPI Transmission Maker9.0, certain is generated by pseudo-random binary sequence
The optical signal of certain degree of injury under physical damnification type, in order to simulate true optical signal as far as possible, in the analogue system
It is added to dispersion, by photoelectric detector, converts optical signals to electric signal, then obtains the corresponding sample image of electric signal.And
And a corresponding sample physics type of impairment and degree label are set for sample image.For example, sample physics type of impairment and
Degree label can be a string of vectors being made of 29 bits, and the present invention is not especially limit this.
Then, using a sample image and the corresponding sample physics type of impairment of the sample image and degree label as
One training sample, so that multiple training samples can be obtained.
Finally, multiple training samples are sequentially input to nervus opticus network, according to the defeated each time of nervus opticus network
Result is adjusted the parameter of nervus opticus network out, is finally completed the training of nervus opticus network, and after being trained
Two neural networks.
Based on the acquisition modes for each sample image mentioned in the embodiment of the present invention, to the acquisition process of the second training set
It is further illustrated by:
If sample image is eye figure, physical damnification type is following 6 kinds: limiting effect, pattern effect, overshoot, " eyes "
Crooked effect, extinction ratio are insufficient, optical power mismatches.And limiting effect is divided into 9 kinds of degree of injury, pattern effect is divided into 4 kinds of damages
Hurt degree, overshoot is divided into 4 kinds of degree of injury, effect is divided into 4 kinds of degree of injury to " eyes " skew, extinction ratio deficiency is divided into 4 kinds of damages
Hurt degree, optical power mismatch is divided into 4 kinds of degree of injury.Acquisition modes then based on above-mentioned each sample image, to every kind of object
Each degree of injury under reason type of impairment collects the eye figure of " jpg " format that 100 pixel sizes are 999 × 680 respectively, this
Place, using every figure and corresponding physical damnification type at a glance and degree label as a training sample, therefore second trains lump
It altogether include 2900 training samples.
Based on any of the above-described embodiment, the embodiment of the present invention carries out the training process of nervus opticus network further specific
Explanation, that is, nervus opticus network is trained by second training set, comprising:
Step 0031, for any one training sample, the sample image in the training sample is input to described
Two neural networks obtain the corresponding prediction physical damnification type of the sample image and degree of the nervus opticus network output
Label；
Step 0032, according to the sample image in the prediction physical damnification type and degree label and the training sample
Corresponding sample physics type of impairment and degree label obtain the penalty values of the nervus opticus network；
Step 0033, if the penalty values are less than preset threshold, the nervus opticus network training is completed.
Specifically, after obtaining the second training set, for any one training sample in the second training set, by the instruction
Practice the sample image in sample and be input to nervus opticus network, the sample image for obtaining the output of nervus opticus network is corresponding pre-
Survey physical damnification type and degree label.
On this basis, by preset loss function according to the corresponding prediction physical damnification type of the sample image and journey
Scale label sample physics type of impairment corresponding with the sample image in the training sample and degree label calculate penalty values.
Wherein, preset loss function can be cross entropy loss function, and the embodiment of the present invention is not specifically limited herein.
After calculating acquisition penalty values, this training process terminates, and error backpropagation algorithm is recycled to update second
Parameter in neural network, is trained next time again later.During training, if being obtained for the calculating of some training sample
The penalty values obtained are less than preset threshold, then nervus opticus network training is completed.It should be noted that preset threshold is according to specific feelings
Condition is selected, and the present invention is not especially limit this.
It is reversed based on error if the penalty values are greater than or equal to the preset threshold based on any of the above-described embodiment
Propagation algorithm updates the parameter of the nervus opticus network.
Specifically, if migration algorithm when constructing nervus opticus network is to freeze algorithm, it is based on error back propagation
Algorithm only updates the parameter of specific structure in nervus opticus network, wherein specific structure refers to rejecting in nervus opticus network
Replicate the other structures except obtained structure；If transfer learning algorithm is fine tuning algorithm, it is based on error backpropagation algorithm
Update the parameter of the entire infrastructure of nervus opticus network.
Based on any of the above-described embodiment, the embodiment of the present invention is illustrated the acquisition of the first nerves network after training,
That is, by the corresponding nervus opticus network after diagnostic image is input to training of the follow-up power-off signal, before further include:
Obtain multiple sample images and the corresponding sample physics type of impairment label of each sample image；
Using each sample image and corresponding sample physics type of impairment label as a training sample；
Multiple training samples are obtained, using the multiple training sample as the first training set, pass through first training set
First nerves network is trained, the first nerves network after being trained.
Specifically, it will also need before the nervus opticus network that diagnostic image is input to after training to first nerves net
Network is trained, and specific training process is as follows:
Firstly, obtaining multiple sample images and the corresponding sample physics type of impairment label of each sample image.Wherein,
Each sample image can obtain in the following way:
Analogue system is established based on VPI Transmission Maker9.0, number is generated by pseudo-random binary sequence
Signal is added to dispersion in the analogue system to simulate true optical signal as far as possible, by photoelectric detector, by light
Signal is converted to electric signal, then obtains the corresponding sample image of electric signal.Also, a corresponding sample is set for the sample image
This physical damnification type label.For example, sample physics type of impairment label can be a string of vectors being made of 6 bits, this
Inventive embodiments are not especially limited this.
Then, it is instructed using a sample image and the corresponding sample physics type of impairment label of the sample image as one
Practice sample, so that multiple training samples can be obtained.
Finally, multiple training samples are sequentially input to first nerves network, according to the defeated each time of first nerves network
Result is adjusted first nerves network parameter out, is finally completed the training of first nerves network, and first after being trained
Neural network.
Based on the acquisition modes for each sample image mentioned in the embodiment of the present invention, to the acquisition process of the first training set
It is further illustrated by:
If sample image is eye figure, physical damnification type is following 6 kinds: limiting effect, pattern effect, overshoot, " eyes "
Crooked effect, extinction ratio are insufficient, optical power mismatches.Acquisition modes then based on above-mentioned each sample image, to every kind of physics
Type of impairment collects the eye figure of " jpg " format that 100 pixel sizes are 999 × 680 respectively, herein, with per figure at a glance and right
The physical damnification type label answered is as a training sample, therefore the first training set includes 600 training samples in total.
First nerves network is trained by first training set, is specifically included:
For any one training sample, the sample image in the training sample is input to the first nerves net
Network obtains the corresponding prediction physical damnification type label of the sample image of the first nerves network output；
According to the corresponding sample physics of sample image in the prediction physical damnification type label and the training sample
Type of impairment label obtains the penalty values of the first nerves network；
If the penalty values are completed less than the second preset threshold, the first nerves network training；
If the penalty values are greater than or equal to second preset threshold, based on described in error backpropagation algorithm update
The parameter of first nerves network.It should be noted that the second preset threshold is selected as the case may be, the embodiment of the present invention
This is not especially limited.
Based on any of the above-described embodiment, the preprocessing process that the embodiment of the present invention treats diagnostic image is illustrated, that is, will
The corresponding nervus opticus network after diagnostic image is input to training of the follow-up power-off signal, before further include:
Described down-sampling will be carried out to diagnostic image, and obtain down-sampled images；
Correspondingly, by the follow-up power-off signal it is corresponding after diagnostic image be input to training after nervus opticus network,
Specifically:
The down-sampled images are input to the nervus opticus network after training.
Specifically, in embodiments of the present invention, by after diagnostic image be input to training after nervus opticus network before, first
It treats diagnostic image and carries out down-sampling, obtain down-sampled images.Wherein, the down-sampling of image refers to downscaled images, so that picture
Element is biggish to be converted to the lesser down-sampled images of pixel to diagnostic image.This have the advantage that: reduce nervus opticus
The calculation amount of network saves computing resource expense.After obtaining down-sampled images, then by down-sampled images be input to training after
Nervus opticus network.
It is understood that in the training process of nervus opticus network, also by each sample image in the second training set
Nervus opticus network is trained again after having done above-mentioned pretreatment, details are not described herein again for preprocessing process.
It should be noted that in the training process of first nerves network, also by each sample image in the first training set
First nerves network is trained again after having done above-mentioned pretreatment, details are not described herein again for preprocessing process.
Based on any of the above-described embodiment, the embodiment of the present invention to above-described embodiment treat the preprocessing process of diagnostic image into
Row explanation, that is, described will carry out down-sampling to diagnostic image, and obtain down-sampled images, before further include:
Described gray processing will be carried out to diagnostic image, and obtain gray level image；
Correspondingly, described down-sampling will be carried out to diagnostic image, and will obtain down-sampled images, specifically:
The gray level image is subjected to down-sampling, obtains down-sampled images.
Specifically, in embodiments of the present invention, before it will carry out down-sampling to diagnostic image, first treat diagnostic image into
Row gray processing, obtains gray level image.Wherein, the gray processing of image refers to the colored ash for being converted to gray scale to diagnostic image
Degreeization image.This have the advantage that: the original data volume of image is reduced, to reduce the calculating of nervus opticus network
Amount, saves computing resource expense.After obtaining gray level image, then gray level image is subjected to down-sampling, obtains down-sampling figure
Picture.
It is understood that in the training process of nervus opticus network, also by each sample image in the second training set
Nervus opticus network is trained again after having done above-mentioned pretreatment, details are not described herein again for preprocessing process.
It should be noted that in the training process of first nerves network, also by each sample image in the first training set
First nerves network is trained again after having done above-mentioned pretreatment, details are not described herein again for preprocessing process.
Based on any of the above-described embodiment, as a preferred embodiment, the embodiment of the present invention is to the second mind after training
Structure and its use process through network are illustrated:
Nervus opticus network after training includes:
One input layer, four convolutional layers (C1, C2, C3, C4), four pond layers (P1, P2, P3, P4), a full connection
Layer (F1) and an output layer.Wherein, input layer, convolutional layer C1, pond layer P1, convolutional layer C2, pond layer P2, convolutional layer C3,
Pond layer P3, convolutional layer C4, pond layer C4, full articulamentum F1 and output layer are sequentially connected.
It is connected from as input layer with convolutional layer C1 by the eye figure that pretreated pixel size is 64 × 64；Input
Eye figure is by the convolutional layer C1 for the convolution kernel for being 5 × 5 containing 32 sizes, and obtaining 32 sizes is 64 × 64 characteristic patterns, in turn
Obtained characteristic pattern is sent to pond layer P1；Pond layer P1 carries out maximum pond to 32 characteristic patterns with 2 × 2 sample size
Change, the characteristic pattern after obtaining the sampling that corresponding 32 sizes are 32 × 32, and then obtained characteristic pattern is sent to convolutional layer
C2；Convolutional layer C2 contains the convolution kernel that 64 sizes are 5 × 5, and layer P1 resulting 32 characteristic patterns in pond are obtained through convolutional layer C2
64 sizes are 32 × 32 characteristic pattern, and then obtained characteristic pattern is sent to pond layer P2；Pond layer P2 is equally with 2 × 2
Sample size 64 sizes that convolutional layer C2 is generated be 32 × 32 characteristic pattern carry out maximum pond, obtain corresponding 64
Characteristic pattern after the sampling that size is 16 × 16, and then obtained characteristic pattern is sent to convolutional layer C3；Convolutional layer C3 contains 128
The convolution kernel that a size is 5 × 5, it is 16 × 16 that layer P2 resulting 64 characteristic patterns in pond, which obtain 128 sizes through convolutional layer C3,
Characteristic pattern, and then obtained characteristic pattern is sent to pond layer P3；Pond layer P3 is equally with 2 × 2 sample size to convolution
The characteristic pattern that 128 sizes that layer C2 is generated are 16 × 16 carries out maximum pond, obtains corresponding 128 sizes and adopts for 8 × 8
Characteristic pattern after sample, and then obtained characteristic pattern is sent to convolutional layer C4；Convolutional layer C4 contains the volume that 256 sizes are 5 × 5
Product core, resulting 128 characteristic patterns of pond layer P3 obtain 256 sizes through convolutional layer C3 as 8 × 8 characteristic pattern, and then will
To characteristic pattern be sent to pond layer P4；Pond layer P4 equally generates convolutional layer C2 with 2 × 2 sample size 256 big
The small characteristic pattern for being 8 × 8 carries out maximum pond, and the characteristic pattern after obtaining the sampling that corresponding 256 sizes are 4 × 4 then will
Obtained characteristic pattern is sent to full articulamentum F1；The pixel of the resulting all characteristic patterns of pond layer P4 is mapped as one-dimensional connecting entirely
Meet a layer F1, each pixel represents a neuron node of full articulamentum F1, each neuron node of full articulamentum F1 with it is defeated
Layer is connected entirely out；Last output layer exports the corresponding physical damnification type of eye figure and degree label, can be obtained according to the label
Take the physical damnification diagnostic result of optical interconnection system.
Based on any of the above-described embodiment, Fig. 2 is that a kind of physical damnification of optical interconnection system provided in an embodiment of the present invention is examined
The structural schematic diagram of disconnected device, as shown in Fig. 2, the device includes:
Follow-up power-off signal obtains module 201, and the receiving end for obtaining optical interconnection system demodulates modulated signal
Obtained follow-up power-off signal；
Diagnostic module 202, for by the follow-up power-off signal it is corresponding after diagnostic image be input to training after second
Neural network, according to the output of the nervus opticus network after the training as a result, obtaining the physical damnification of the optical interconnection system
Diagnostic result；
Wherein, the physical damnification diagnostic result includes physical damnification type and degree；
The nervus opticus network is according to the first nerves network struction after transfer learning algorithm and training, after the training
Nervus opticus network according to the second training set training obtain；
First nerves network after the training is obtained according to the training of the first training set；
First training set includes multiple sample images and the corresponding sample physics type of impairment of each sample image
Label；
Second training set includes multiple sample images and the corresponding sample physics type of impairment of each sample image
And degree label.
Device provided in an embodiment of the present invention, specifically executes above-mentioned each method embodiment process, please specifically be detailed in above-mentioned each
The content of embodiment of the method, details are not described herein again.Device provided in an embodiment of the present invention, it is mutual to light by the method for machine learning
Link system carry out physical damnification diagnosis, compared with the existing technology in Artificial Diagnosis, improve diagnosis efficiency to a certain extent
And reduce error.Further, first nerves network struction of the nervus opticus network according to transfer learning algorithm and after training,
Neural network in compared with the prior art does not need the neural network for from the beginning constructing a robust with generalization ability, drop
Low network complexity, when so that carrying out physical damnification diagnosis to optical interconnection system by the nervus opticus network after training, consumption
When it is shorter, diagnostic result can be obtained in time；Also, the neural network in compared with the prior art, in nervus opticus network
Convergence rate in training faster, nervus opticus network after capable of quickly being trained.
Fig. 3 is the entity structure schematic diagram of a kind of electronic equipment provided in an embodiment of the present invention, as shown in figure 3, the electronics
Equipment may include: processor (processor) 301, communication interface (Communications Interface) 302, storage
Device (memory) 303 and communication bus 304, wherein processor 301, communication interface 302, memory 303 pass through communication bus
304 complete mutual communication.Processor 301, which can call, to be stored on memory 303 and can run on processor 301
Computer program, the method to execute the various embodiments described above offer, for example, obtain the receiving end of optical interconnection system to modulation
The follow-up power-off signal that signal is demodulated；The follow-up power-off signal is corresponding after diagnostic image is input to training
Nervus opticus network, according to the nervus opticus network after the training output as a result, obtaining the object of the optical interconnection system
Reason damage diagnostic result；Wherein, the physical damnification diagnostic result includes physical damnification type and degree；The nervus opticus net
Network is according to the first nerves network struction after transfer learning algorithm and training, and the nervus opticus network after the training is according to second
Training set training obtains；First nerves network after the training is obtained according to the training of the first training set；First training set
Including multiple sample images and the corresponding sample physics type of impairment label of each sample image；Second training set includes
Multiple sample images and the corresponding sample physics type of impairment of each sample image and degree label.
In addition, the logical order in above-mentioned memory 303 can be realized by way of SFU software functional unit and conduct
Independent product when selling or using, can store in a computer readable storage medium.Based on this understanding, originally
The technical solution of the inventive embodiments substantially part of the part that contributes to existing technology or the technical solution in other words
It can be embodied in the form of software products, which is stored in a storage medium, including several fingers
It enables and using so that a computer equipment (can be personal computer, server or the network equipment etc.) executes the present invention respectively
The all or part of the steps of a embodiment the method.And storage medium above-mentioned includes: USB flash disk, mobile hard disk, read-only memory
(ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), magnetic or disk
Etc. the various media that can store program code.
The embodiment of the present invention also provides a kind of non-transient computer readable storage medium, is stored thereon with computer program,
The computer program is implemented to carry out the transmission method of the various embodiments described above offer when being executed by processor, for example, obtain
The follow-up power-off signal that the receiving end of optical interconnection system demodulates modulated signal；The follow-up power-off signal is corresponding
After diagnostic image be input to training after nervus opticus network, according to the output knot of the nervus opticus network after the training
Fruit obtains the physical damnification diagnostic result of the optical interconnection system；Wherein, the physical damnification diagnostic result includes physical damnification
Type and degree；The nervus opticus network is according to the first nerves network struction after transfer learning algorithm and training, the instruction
Nervus opticus network after white silk is obtained according to the training of the second training set；First nerves network after the training is according to the first training
Training is got；First training set includes multiple sample images and the corresponding sample physical damnification class of each sample image
Type label；Second training set include multiple sample images and the corresponding sample physics type of impairment of each sample image and
Degree label.
The apparatus embodiments described above are merely exemplary, wherein described, unit can as illustrated by the separation member
It is physically separated with being or may not be, component shown as a unit may or may not be physics list
Member, it can it is in one place, or may be distributed over multiple network units.It can be selected according to the actual needs
In some or all of the modules achieve the purpose of the solution of this embodiment.Those of ordinary skill in the art are not paying creativeness
Labour in the case where, it can understand and implement.
Through the above description of the embodiments, those skilled in the art can be understood that each embodiment can
It realizes by means of software and necessary general hardware platform, naturally it is also possible to pass through hardware.Based on this understanding, on
Stating technical solution, substantially the part that contributes to existing technology can be embodied in the form of software products in other words, should
Computer software product may be stored in a computer readable storage medium, such as ROM/RAM, magnetic disk, CD, including several fingers
It enables and using so that a computer equipment (can be personal computer, server or the network equipment etc.) executes each implementation
Method described in certain parts of example or embodiment.
Finally, it should be noted that the above embodiments are merely illustrative of the technical solutions of the present invention, rather than its limitations；Although
Present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that: it still may be used
To modify the technical solutions described in the foregoing embodiments or equivalent replacement of some of the technical features；
And these are modified or replaceed, technical solution of various embodiments of the present invention that it does not separate the essence of the corresponding technical solution spirit and
Range.
Claims (10)
1. a kind of physical damnification diagnostic method of optical interconnection system characterized by comprising
Obtain the follow-up power-off signal that the receiving end of optical interconnection system demodulates modulated signal；
By the corresponding nervus opticus network after diagnostic image is input to training of the follow-up power-off signal, according to the training
The output of nervus opticus network afterwards is as a result, obtain the physical damnification diagnostic result of the optical interconnection system；
Wherein, the physical damnification diagnostic result includes physical damnification type and degree；
The nervus opticus network according to the first nerves network struction after transfer learning algorithm and training, after the training the
Two neural networks are obtained according to the training of the second training set；
First nerves network after the training is obtained according to the training of the first training set；
First training set includes multiple sample images and the corresponding sample physics type of impairment label of each sample image；
Second training set includes multiple sample images and the corresponding sample physics type of impairment of each sample image and journey
Scale label.
2. the physical damnification diagnostic method of optical interconnection system according to claim 1, which is characterized in that by described wait diagnose
The corresponding nervus opticus network after diagnostic image is input to training of electric signal, before further include:
Obtain multiple sample images and the corresponding sample physics type of impairment of each sample image and degree label；
Using each sample image and corresponding sample physics type of impairment and degree label as a training sample；
Multiple training samples are obtained, using the multiple training sample as the second training set, by second training set to
Two neural networks are trained, the nervus opticus network after being trained.
3. the physical damnification diagnostic method of optical interconnection system according to claim 2, which is characterized in that pass through described second
Training set is trained nervus opticus network, comprising:
For any one training sample, the sample image in the training sample is input to the nervus opticus network, is obtained
Take the corresponding prediction physical damnification type of the sample image and degree label of the nervus opticus network output；
According to the corresponding sample contents of sample image in the prediction physical damnification type and degree label and the training sample
Type of impairment and degree label are managed, the penalty values of the nervus opticus network are obtained；
If the penalty values are less than preset threshold, the nervus opticus network training is completed.
4. the physical damnification diagnostic method of optical interconnection system according to claim 3, which is characterized in that if the penalty values
More than or equal to the preset threshold, then the parameter of the nervus opticus network is updated based on error backpropagation algorithm.
5. the physical damnification diagnostic method of optical interconnection system according to claim 1, which is characterized in that by described wait diagnose
The corresponding nervus opticus network after diagnostic image is input to training of electric signal, before further include:
Obtain multiple sample images and the corresponding sample physics type of impairment label of each sample image；
Using each sample image and corresponding sample physics type of impairment label as a training sample；
Multiple training samples are obtained, using the multiple training sample as the first training set, by first training set to
One neural network is trained, the first nerves network after being trained.
6. the physical damnification diagnostic method of optical interconnection system according to claim 1, which is characterized in that by described wait diagnose
The corresponding nervus opticus network after diagnostic image is input to training of electric signal, before further include:
Described down-sampling will be carried out to diagnostic image, and obtain down-sampled images；
Correspondingly, by the corresponding nervus opticus network after diagnostic image is input to training of the follow-up power-off signal, specifically
Are as follows:
The down-sampled images are input to the nervus opticus network after training.
7. the physical damnification diagnostic method of optical interconnection system according to claim 6, which is characterized in that by described wait diagnose
Image carries out down-sampling, obtains down-sampled images, before further include:
Described gray processing will be carried out to diagnostic image, and obtain gray level image；
Correspondingly, described down-sampling will be carried out to diagnostic image, and will obtain down-sampled images, specifically:
The gray level image is subjected to down-sampling, obtains down-sampled images.
8. a kind of physical damnification diagnostic device of optical interconnection system characterized by comprising
Follow-up power-off signal obtains module, for obtain the receiving end of optical interconnection system to modulated signal demodulated to
Diagnose electric signal；
Diagnostic module, for by the follow-up power-off signal it is corresponding after diagnostic image be input to training after nervus opticus net
Network, according to the output of the nervus opticus network after the training as a result, obtaining the physical damnification diagnosis knot of the optical interconnection system
Fruit；
Wherein, the physical damnification diagnostic result includes physical damnification type and degree；
The nervus opticus network according to the first nerves network struction after transfer learning algorithm and training, after the training the
Two neural networks are obtained according to the training of the second training set；
First nerves network after the training is obtained according to the training of the first training set；
First training set includes multiple sample images and the corresponding sample physics type of impairment label of each sample image；
Second training set includes multiple sample images and the corresponding sample physics type of impairment of each sample image and journey
Scale label.
9. a kind of electronic equipment including memory, processor and stores the calculating that can be run on a memory and on a processor
Machine program, which is characterized in that the processor realizes the light network as described in any one of claim 1 to 7 when executing described program
The step of physical damnification diagnostic method of system.
10. a kind of non-transient computer readable storage medium, is stored thereon with computer program, which is characterized in that the computer
Realizing the physical damnification diagnostic method of optical interconnection system as described in any one of claim 1 to 7 when program is executed by processor
Step.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910427762.2A CN110210536A (en) | 2019-05-22 | 2019-05-22 | A kind of the physical damnification diagnostic method and device of optical interconnection system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910427762.2A CN110210536A (en) | 2019-05-22 | 2019-05-22 | A kind of the physical damnification diagnostic method and device of optical interconnection system |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110210536A true CN110210536A (en) | 2019-09-06 |
Family
ID=67788183
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910427762.2A Pending CN110210536A (en) | 2019-05-22 | 2019-05-22 | A kind of the physical damnification diagnostic method and device of optical interconnection system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110210536A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111444864A (en) * | 2020-03-30 | 2020-07-24 | 北京旷视科技有限公司 | Target object wear degree detection method and device |
| CN113627541A (en) * | 2021-08-13 | 2021-11-09 | 北京邮电大学 | Light path transmission quality prediction method based on sample migration screening |
Citations (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102238060A (en) * | 2010-04-21 | 2011-11-09 | 华为技术有限公司 | Routing method and apparatus thereof |
| CN103944638A (en) * | 2014-04-18 | 2014-07-23 | 华中科技大学 | Optical signal modulation format recognition method and system based on nonlinear digital processing |
| CN105515650A (en) * | 2015-11-24 | 2016-04-20 | 华中科技大学 | Optical signal modulation format identification method based on spectral characteristics |
| CN107342962A (en) * | 2017-07-03 | 2017-11-10 | 北京邮电大学 | Deep learning intelligence Analysis On Constellation Map method based on convolutional neural networks |
| CN107358596A (en) * | 2017-04-11 | 2017-11-17 | 阿里巴巴集团控股有限公司 | A kind of car damage identification method based on image, device, electronic equipment and system |
| CN108053030A (en) * | 2017-12-15 | 2018-05-18 | 清华大学 | A kind of transfer learning method and system of Opening field |
| CN108182427A (en) * | 2018-01-30 | 2018-06-19 | 电子科技大学 | A kind of face identification method based on deep learning model and transfer learning |
| CN108229549A (en) * | 2017-12-28 | 2018-06-29 | 杭州大搜车汽车服务有限公司 | A kind of intelligent recognition car trader fits up method, electronic equipment and the storage medium of degree |
| CN108446631A (en) * | 2018-03-20 | 2018-08-24 | 北京邮电大学 | The smart frequency spectrum figure analysis method of deep learning based on convolutional neural networks |
| CN108492873A (en) * | 2018-03-13 | 2018-09-04 | 山东大学 | A kind of knowledge migration learning method for auxiliary diagnosis Alzheimer's disease |
| CN108564122A (en) * | 2018-04-10 | 2018-09-21 | 北京工业大学 | The integrated training method of image description model based on convolution-cycle hybrid network |
| CN108829684A (en) * | 2018-05-07 | 2018-11-16 | 内蒙古工业大学 | A kind of illiteracy Chinese nerve machine translation method based on transfer learning strategy |
| CN108874900A (en) * | 2018-05-24 | 2018-11-23 | 四川斐讯信息技术有限公司 | A kind of acquisition methods and system of samples pictures data acquisition system |
| CN109167627A (en) * | 2018-10-09 | 2019-01-08 | 北京邮电大学 | A kind of modulation format and OSNR Monitoring Method and device |
| CN109508634A (en) * | 2018-09-30 | 2019-03-22 | 上海鹰觉科技有限公司 | Ship Types recognition methods and system based on transfer learning |
-
2019
- 2019-05-22 CN CN201910427762.2A patent/CN110210536A/en active Pending
Patent Citations (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102238060A (en) * | 2010-04-21 | 2011-11-09 | 华为技术有限公司 | Routing method and apparatus thereof |
| CN103944638A (en) * | 2014-04-18 | 2014-07-23 | 华中科技大学 | Optical signal modulation format recognition method and system based on nonlinear digital processing |
| CN105515650A (en) * | 2015-11-24 | 2016-04-20 | 华中科技大学 | Optical signal modulation format identification method based on spectral characteristics |
| CN107358596A (en) * | 2017-04-11 | 2017-11-17 | 阿里巴巴集团控股有限公司 | A kind of car damage identification method based on image, device, electronic equipment and system |
| CN107342962A (en) * | 2017-07-03 | 2017-11-10 | 北京邮电大学 | Deep learning intelligence Analysis On Constellation Map method based on convolutional neural networks |
| CN108053030A (en) * | 2017-12-15 | 2018-05-18 | 清华大学 | A kind of transfer learning method and system of Opening field |
| CN108229549A (en) * | 2017-12-28 | 2018-06-29 | 杭州大搜车汽车服务有限公司 | A kind of intelligent recognition car trader fits up method, electronic equipment and the storage medium of degree |
| CN108182427A (en) * | 2018-01-30 | 2018-06-19 | 电子科技大学 | A kind of face identification method based on deep learning model and transfer learning |
| CN108492873A (en) * | 2018-03-13 | 2018-09-04 | 山东大学 | A kind of knowledge migration learning method for auxiliary diagnosis Alzheimer's disease |
| CN108446631A (en) * | 2018-03-20 | 2018-08-24 | 北京邮电大学 | The smart frequency spectrum figure analysis method of deep learning based on convolutional neural networks |
| CN108564122A (en) * | 2018-04-10 | 2018-09-21 | 北京工业大学 | The integrated training method of image description model based on convolution-cycle hybrid network |
| CN108829684A (en) * | 2018-05-07 | 2018-11-16 | 内蒙古工业大学 | A kind of illiteracy Chinese nerve machine translation method based on transfer learning strategy |
| CN108874900A (en) * | 2018-05-24 | 2018-11-23 | 四川斐讯信息技术有限公司 | A kind of acquisition methods and system of samples pictures data acquisition system |
| CN109508634A (en) * | 2018-09-30 | 2019-03-22 | 上海鹰觉科技有限公司 | Ship Types recognition methods and system based on transfer learning |
| CN109167627A (en) * | 2018-10-09 | 2019-01-08 | 北京邮电大学 | A kind of modulation format and OSNR Monitoring Method and device |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111444864A (en) * | 2020-03-30 | 2020-07-24 | 北京旷视科技有限公司 | Target object wear degree detection method and device |
| CN111444864B (en) * | 2020-03-30 | 2023-11-17 | 北京旷视科技有限公司 | Method and device for detecting abrasion degree of target object |
| CN113627541A (en) * | 2021-08-13 | 2021-11-09 | 北京邮电大学 | Light path transmission quality prediction method based on sample migration screening |
| CN113627541B (en) * | 2021-08-13 | 2023-07-21 | 北京邮电大学 | Optical path transmission quality prediction method based on sample migration screening |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| KR102545128B1 (en) | Client device with neural network and system including the same | |
| US11586911B2 (en) | Pre-training system for self-learning agent in virtualized environment | |
| CN107342962B (en) | deep learning intelligent constellation diagram analysis method based on convolutional neural network | |
| CN112651978B (en) | Sublingual microcirculation image segmentation method and device, electronic equipment and storage medium | |
| Robert et al. | Hybridnet: Classification and reconstruction cooperation for semi-supervised learning | |
| CN110932809B (en) | Fiber channel model simulation method, device, electronic equipment and storage medium | |
| CN110349082A (en) | Method of cutting out and device, the storage medium and electronic device of image-region | |
| US20220284349A1 (en) | Techniques to generate network simulation scenarios | |
| CN110210536A (en) | A kind of the physical damnification diagnostic method and device of optical interconnection system | |
| CN110324080A (en) | A kind of method, apparatus of optical information networks, electronic equipment and medium | |
| CN110190909A (en) | A kind of signal equalizing method and device for optic communication | |
| CN109995449A (en) | A kind of millimeter-wave signal detection method based on deep learning | |
| Wang et al. | Convolutional neural network-based deep learning for intelligent OSNR estimation on eye diagrams | |
| Liu et al. | RB-Net: Training highly accurate and efficient binary neural networks with reshaped point-wise convolution and balanced activation | |
| CN106778910A (en) | Deep learning system and method based on local training | |
| CN114820871A (en) | Font generation method, model training method, device, equipment and medium | |
| CN115358931A (en) | Image reconstruction method and device for warehouse logistics system | |
| CN113987236A (en) | Unsupervised training method and unsupervised training device for visual retrieval model based on graph convolution network | |
| CN112748941A (en) | Feedback information-based target application program updating method and device | |
| CN110490876B (en) | Image segmentation method based on lightweight neural network | |
| CN107633219A (en) | Integrated optical character identifying method and system | |
| CN112926691A (en) | Convolutional dendrite method for extracting feature logic for classification | |
| CN113012072A (en) | Image motion deblurring method based on attention network | |
| CN116310391B (en) | Identification method for tea diseases | |
| Fakhari et al. | An image restoration architecture using abstract features and generative models |
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 | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190906 |