CN110197212A - Image classification method, system and computer readable storage medium - Google Patents
Image classification method, system and computer readable storage medium Download PDFInfo
- Publication number
- CN110197212A CN110197212A CN201910419415.5A CN201910419415A CN110197212A CN 110197212 A CN110197212 A CN 110197212A CN 201910419415 A CN201910419415 A CN 201910419415A CN 110197212 A CN110197212 A CN 110197212A
- Authority
- CN
- China
- Prior art keywords
- sample
- classification
- radio radiation
- solar radio
- spectrogram
- 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
Abstract
The invention discloses a kind of image classification method, system and computer readable storage mediums, it is related to solar radio radiation detection technique field, including carrying out sample equilibrium treatment to classified multiple original solar radio radiation spectrograms, multiple samples that each classification participates in training opportunity equalization are obtained；It is input with the multiple sample, the multiple sample is trained using convolutional neural networks, obtains solar radio radiation spectrogram disaggregated model；Based on the solar radio radiation spectrogram disaggregated model, classifies to non-classified solar radio radiation spectrogram, obtain classification results.The present invention makes each sample participate in training opportunity equalization by above-mentioned sample equilibrium treatment, so that improving the accuracy of classification results because the unbalance influence of sample causes the inaccurate rate of classification results greatly to reduce.
Description
Technical field
The present invention relates to solar radio radiation detection technique field, a kind of image classification method, system and computer are particularly related to
Readable storage medium storing program for executing.
Background technique
All activity phenomenons are referred to as solar activity in solar atmosphere, and solar radio burst is also the one of solar activity
Kind.Since solar radio burst phenomenon usually occurs very unexpected, its radiation intensity is big, variation is violent, when the arrival earth
When, it will cause a series of geophysical effect, such as magnetic storm, communication interference.In addition, solar radio burst can be divided into five again
Seed type, each type of meaning is again different, and II type radio storm is commonly used for the early warning for diastrous weather, and III type is penetrated
Electricity is conducive to us cruelly and studies solar flare coronal mass ejection.I can be made for the classification work of solar radio burst
Preferably research solar burst when corona in the small small dimensional variation process of energy release process and Coronal Magnetic Field.Therefore,
How to carry out classification to solar radio radiation spectrogram is one of the key technical problem that solar radio radiation detection technique field needs to solve.
Summary of the invention
In view of this, it is an object of the invention to propose a kind of image classification method, system and computer-readable storage medium
Matter promotes the accuracy rate of classification results for carrying out Accurate classification to solar radio radiation spectrogram.
Based on above-mentioned purpose, the present invention provides a kind of image classification methods, comprising the following steps:
Sample equilibrium treatment is carried out to classified multiple original solar radio radiation spectrograms, each classification is obtained and participates in training
Equal-opportunity multiple samples；
It is input with the multiple sample, the multiple sample is trained using convolutional neural networks, obtains the sun
Radio spectra figure disaggregated model；
Based on the solar radio radiation spectrogram disaggregated model, classifies to non-classified solar radio radiation spectrogram, obtain
To classification results.
Optionally, the sample equilibrium treatment includes:
Equilibrium treatment is carried out to sample based on production confrontation network, each classification is obtained and participates in the more of training opportunity equalization
A sample.
Optionally, the production confrontation network includes: to sample progress equilibrium treatment
Obtain the original solar radio radiation frequency spectrum classified known to one, and by the original solar radio radiation frequency spectrum of the known classification
It is sent in the production confrontation network；
Study processing is carried out to the original solar radio radiation frequency spectrum of the known classification, exports exptended sample；
The original solar radio radiation frequency spectrum of the exptended sample and the known classification is received, and to the exptended sample and institute
It states the known original solar radio radiation frequency spectrum classified to be compared, the parameter of production confrontation network is adjusted according to comparison result；
Until comparison result difference is received between the exptended sample and the original solar radio radiation frequency spectrum of the known classification
Until holding back, final exptended sample is obtained.
Optionally, the final exptended sample is combined with solar radio radiation spectrogram, using convolutional neural networks to institute
It states final exptended sample to be trained with the solar radio radiation spectrogram, obtains solar radio radiation spectrogram disaggregated model.
Optionally, solar radio radiation spectrogram is trained the following steps are included:
Feature is extracted from the solar radio radiation spectrogram, obtains eigenmatrix figure；
According to the parameter reduced in network, the size of the eigenmatrix figure is reduced；
Reduced eigenmatrix figure is compared with the eigenmatrix figure of previous step, image instruction is adjusted according to comparison result
Practice the parameter in network；
Until comparison result difference convergence between the eigenmatrix figure of reduced eigenmatrix figure and previous step, obtain
To final eigenmatrix figure.
It is activated based on line rectification function, is resolved the gradient disperse problem in back-propagation process, and
Accelerate the renewal process of parameter in network；
Final eigenmatrix figure is mapped in classifier, the solar radio radiation spectrogram disaggregated model is obtained.
Optionally, the method further includes: by Principal Component Analysis to the original solar radio radiation spectrogram into
Row pretreatment, and dimensionality reduction is carried out to the original solar radio radiation spectrogram.
Optionally, the sample equilibrium treatment uses up-sampling method or down-sampling method, wherein
The up-sampling method includes the image of the less classification of reproduction copies until consistent with the more multi-class sample number of sample；
The down-sampling method, which is included in, chooses image when carrying out handling trained, reduces the more multi-class picture number of sample.
Optionally, the sample equilibrium treatment use aligned sample method, wherein the aligned sample method the following steps are included:
It obtains sample to be trained and is grouped according to classification；
Based on each classification group, corresponding sample list is generated；
A classification is determined at random, and sample is obtained in its corresponding sample list for training.
Based on identical innovation and creation, the present invention also provides a kind of image classification systems, comprising:
Balance processing module is obtained for carrying out sample equilibrium treatment to classified multiple original solar radio radiation spectrograms
Obtain multiple samples that each classification participates in training opportunity equalization；
Image training module, for being input with the multiple sample, using convolutional neural networks to the multiple sample
It is trained, obtains solar radio radiation spectrogram disaggregated model；
Categorization module, for being based on the solar radio radiation spectrogram disaggregated model, to non-classified solar radio radiation frequency spectrum
Figure is classified, and classification results are obtained.
Based on identical innovation and creation, the present invention also provides a kind of computer readable storage mediums, which is characterized in that packet
It includes:
At least one processor；And
The memory being connect at least one described processor communication；Wherein,
The memory is stored with the instruction that can be executed by least one described processor, and described instruction is by described at least one
A processor executes, so that at least one described processor is able to carry out the image as described in claim 1-8 any one point
Class method.
The present invention obtains each class by carrying out sample equilibrium treatment to classified multiple original solar radio radiation spectrograms
Not Can Yu training opportunity equalization multiple samples；It is input with the multiple sample, using convolutional neural networks to the multiple
Sample is trained, and obtains solar radio radiation spectrogram disaggregated model；Based on the solar radio radiation spectrogram disaggregated model, to without
The solar radio radiation spectrogram of classification is classified, and classification results are obtained.
By above-mentioned steps it is found that making each sample participate in training opportunity equalization by above-mentioned sample equilibrium treatment, thus
So that improving the accurate of classification results because the unbalance influence of sample causes the inaccurate rate of classification results greatly to reduce
Property.
Detailed description of the invention
Fig. 1 is a kind of image classification method flow chart of the embodiment of the present invention；
Fig. 2 is that production of embodiment of the present invention confrontation network carries out expansion flow chart to sample size；
Fig. 3 is that the embodiment of the present invention utilizes convolutional neural networks to the final exptended sample and the solar radio radiation frequency spectrum
Figure is trained flow chart；
Fig. 4 is the structural block diagram of image classification system of the embodiment of the present invention；
Fig. 5 is computer readable storage medium of embodiment of the present invention hardware structural diagram；
Fig. 6 is the generation schematic diagram of exptended sample of the embodiment of the present invention；
Fig. 7 is the training process schematic diagram of solar radio radiation of embodiment of the present invention spectrogram.
Specific embodiment
To make the objectives, technical solutions, and advantages of the present invention clearer, below in conjunction with specific embodiment, and reference
Attached drawing, the present invention is described in more detail.
With the appearance of big data and the raising of machine processing ability, depth learning technology had obtained widely answering in recent years
With, as shown in Figure 1, can be used for classifying to solar radio radiation spectrogram the invention proposes a kind of image classification method, including
Following steps:
S11: sample equilibrium treatment is carried out to classified multiple original solar radio radiation spectrograms, each classification is obtained and participates in
Multiple samples of training opportunity equalization；
S12: it is input with the multiple sample, the multiple sample is trained using convolutional neural networks, is obtained
Solar radio radiation spectrogram disaggregated model；
S13: it is based on the solar radio radiation spectrogram disaggregated model, non-classified solar radio radiation spectrogram is divided
Class obtains classification results.
By above-mentioned steps it is found that making the sample size in less sample size classification be increased by sample equalization methods,
So that the sample size of each classification tends to balance, or the sample size for making each classification participate in training opportunity equalization, from
And to improve the standard of classification results because the unbalance influence of sample causes the inaccurate rate of classification results to greatly reduce
True property.And the above-mentioned classification results for solar radio burst can be applied to energy release small in corona when solar burst
The research of the processes such as process and the small dimensional variation process of Coronal Magnetic Field.
In an embodiment of the present invention, to classified multiple original solar radio radiation spectrograms described in above-mentioned steps S11
Carrying out sample equilibrium treatment may comprise steps of: be expanded based on production confrontation network sample size, so that each
The sample size of classification tends to balance.
Wherein, the core concept of production confrontation network is the nash banlance of game theory, by generator and arbiter two
Part forms.Wherein, the purpose of generator is to try to learn true data distribution, and the purpose of arbiter is to try to correctly sentence
Other input data is from truthful data or to carry out self-generator.The two is continued to optimize during network training, final to instruct
It is experienced the result is that reaching nash banlance between the two.
Specifically, the production confrontation network expands sample size, as shown in Figure 2, comprising the following steps:
S21: obtaining the original solar radio radiation frequency spectrum classified known to one, and by the original solar radio radiation of the known classification
Frequency spectrum is sent in the production confrontation network；
S22: study processing is carried out to the original solar radio radiation frequency spectrum of the known classification, exports exptended sample；
S23: the original solar radio radiation frequency spectrum of the exptended sample and the known classification is received, and to the exptended sample
It is compared with the original solar radio radiation frequency spectrum of the known classification, the ginseng of production confrontation network is adjusted according to comparison result
Number；
S24: until comparison result is poor between the exptended sample and the original solar radio radiation frequency spectrum of the known classification
Not Shou Lian until, obtain final exptended sample.
By above-mentioned steps it is found that having finally obtained available exptended sample by obtaining stochastic variable, to realize pair
The expansion of sample, so that the sample size of each classification tends to balance, to be trained to be subsequent to solar radio radiation spectrogram
When provide sample size tend to balance it is different classes of.
Other than the above-mentioned mode expanded based on production confrontation network sample size, in some embodiments,
Carrying out sample equilibrium treatment to classified multiple original solar radio radiation spectrograms described in above-mentioned steps S11 can also be using such as
Under various other method.
In some embodiments, the sample equilibrium treatment can use up-sampling method.The up-sampling method is duplication sample
The image of this less classification is until consistent with the more multi-class sample number of sample, so that sample size tends to balance, promotion divides
The accuracy of class.
In some embodiments, the sample equilibrium treatment can use down-sampling method.The down-sampling method is to choose
When image carries out handling trained, the more multi-class picture number of sample is reduced, so that sample size tends to balance, promotes classification
Accuracy.
In some embodiments, the sample equilibrium treatment can also use aligned sample method.The aligned sample method tool
Body may comprise steps of:
It obtains sample to be trained and is grouped according to classification；
Based on each classification group, corresponding sample list is generated；And
A classification is determined at random, and sample is obtained in its corresponding sample list for training.
It can guarantee that each classification participates in having equal opportunities for training using method by above-mentioned equilibrium, to promote the standard of classification
True property.
After carrying out sample equilibrium treatment to classified multiple original solar radio radiation spectrograms, the instruction of second part is begun to
Practice.
In an embodiment of the present invention, the multiple sample is carried out using convolutional neural networks described in above-mentioned steps S12
Training may include: to combine the final exptended sample with solar radio radiation spectrogram, using convolutional neural networks to described
Final exptended sample is trained with the solar radio radiation spectrogram, obtains classifier namely solar radio radiation spectrogram classification mould
Type.
Here the training of image is carried out using convolutional neural networks.Engineer is needed compared to traditional network, is mentioned
The step of taking feature avoids this cumbersome process, and, directly using the image data collection being collected into as input, network can be right for it
Each fritter pixel region is handled on picture, and this way strengthens the continuity of pictorial information, and then can deepen net
Understanding of the network for picture.In the present invention, by convolutional neural networks, it can be using sun radio spectra figure as input, net
The front end of network is several alternate convolutional layers and pond layer, for constantly extracting feature on picture, these can be mentioned later
The feature got is input to subsequent full articulamentum, and the classification results for finally entering spectrogram can be obtained.
Specifically, in an embodiment of the present invention, using convolutional neural networks to the final exptended sample and it is described too
Positive radio spectra figure is trained, as shown in figure 3, can specifically include following steps:
S31: extracting feature from the solar radio radiation spectrogram, obtains eigenmatrix figure；
S32: according to the parameter reduced in network, the size of the eigenmatrix figure is reduced；
S33: reduced eigenmatrix figure is compared with the eigenmatrix figure of previous step, is adjusted and is schemed according to comparison result
As the parameter in training network；
S34: until comparison result difference converges between the eigenmatrix figure of reduced eigenmatrix figure and previous step
Only, final eigenmatrix figure is obtained；
S35: being activated based on line rectification function, is resolved the gradient disperse problem in back-propagation process,
And accelerate the renewal process of parameter in network；And
S36: final eigenmatrix figure is mapped in classifier, obtains the solar radio radiation spectrogram disaggregated model.
In conclusion meeting that sample is unbalance is so that during training, model more side due in the application of deep learning
Classification more than those sample sizes of weight, causes the generalization ability of model very impacted.And production confrontation network is being used to carry out
After sample supplement, the influence that sample is unbalance is greatly reduced.Final classification accuracy has reached 84.6%, has been more than it
It is preceding tradition machine learning method 52.7% and depth confidence network 67.4%, the experiment show this method it is effective
Property.
Based on identical innovation and creation, the present invention also provides a kind of image classification system, can be used for solar radio radiation frequency
Spectrogram is classified, as shown in figure 4, the system may include:
Balance processing module is obtained for carrying out sample equilibrium treatment to classified multiple original solar radio radiation spectrograms
Obtain multiple samples that each classification participates in training opportunity equalization；
Image training module, for being input with the multiple sample, using convolutional neural networks to the multiple sample
It is trained, obtains solar radio radiation spectrogram disaggregated model；
Categorization module, for being based on the solar radio radiation spectrogram disaggregated model, to non-classified solar radio radiation frequency spectrum
Figure is classified, and classification results are obtained.
It should be noted that the concrete methods of realizing of above-mentioned modules function can refer to above-mentioned image classification method.
The embodiment of the invention provides a kind of solar radio radiation classification methods, are divided into two parts, first part is sample
Generation work, the second part is the extraction work of picture feature, below will gather with reference to figure, do specific description.
As shown in fig. 6, being the generation schematic diagram of exptended sample.As can be seen from the figure: production fights the defeated of network
Entering stochastic variable is noise, its essence is an one-dimensional variable, and generator handles the noise variance, obtains a void
False pictorial information, and as the input of arbiter.In addition, the importation of arbiter also needs to include true picture
Sample, the output of arbiter are exactly that the picture of judgement input is true or false.It is worth noting that, by true picture
Sample be sent to differentiation its training before, need using Principal Component Analysis (Principal Components Analysis,
PCA) image is pre-processed, this pretreated purpose is that PCA can carry out dimensionality reduction to original image, reduces generator
To the difficulty of original image feature distribution, to guarantee to reduce the need for original sample amount on the basis of guaranteeing validity
It asks.From this figure it can be seen that the structure of generator and arbiter is all convolutional neural networks, such design be also advantageous that in
Abundant study for radio spectra figure feature increases the reliability for generating sample.
It is the training process schematic diagram for solar radio radiation spectrogram as shown in Figure 7, uses network structure as shown in the figure
It is trained.Whole network includes five convolution groups, and each convolution group includes 2-3 convolutional layer, each convolution kernel in convolutional layer
Size be all 3 × 3, its effect is to extract feature from image, obtains eigenmatrix figure, referred to as feature_map；
A maximum pond layer can be followed behind each convolution group, the effect of maximum pond layer is the ruler for reducing previous step eigenmatrix figure
It is very little, reduce the parameter in network；In addition, being activated in network using ReLU (line rectification) function, its advantage is that can
So that the gradient disperse problem in back-propagation process is resolved, and accelerate the renewal process of parameter in network；Network
It is finally three full articulamentums, for the feature_map extracted before to be mapped to final sample space, obtains final
Classification results.
The present invention also provides a kind of computer readable storage mediums, as shown in figure 5, including at least one processor；With
And the memory being connect at least one described processor communication；Wherein, the memory be stored with can by it is described at least one
The instruction that processor executes, described instruction is executed by least one described processor, so that at least one described processor can
Execute any one method as described above.
By taking electronic equipment as shown in the figure as an example, in the electronic equipment include a processor and a memory,
It and can also include: input unit and output device.
Processor, memory, input unit and output device can be connected by bus or other modes, with logical in figure
It crosses for bus connection.
Memory as a kind of non-volatile computer readable storage medium storing program for executing, can be used for storing non-volatile software program,
Non-volatile computer executable program and module, such as the computation migration of the program of mobile terminal in the embodiment of the present application
Corresponding program instruction/the module of method.Processor by run non-volatile software program stored in memory, instruction with
And module realizes the mobile end of above method embodiment thereby executing the various function application and data processing of server
Hold the computation migration method of program.
Memory may include storing program area and storage data area, wherein storing program area can storage program area, extremely
Application program required for a few function；Storage data area can be stored to be made according to the computation migration device of program of mobile terminal
With the data etc. created.In addition, memory may include high-speed random access memory, it can also include non-volatile memories
Device, for example, at least a disk memory, flush memory device or other non-volatile solid state memory parts.In some embodiments
In, optional memory includes the memory remotely located relative to processor.The example of above-mentioned network includes but is not limited to interconnect
Net, intranet, local area network, mobile radio communication and combinations thereof.
Input unit can receive the number or character information of input, and generates and fill with the computation migration of program of mobile terminal
The related key signals input of the user setting and function control set.Output device may include that display screen etc. shows equipment.
One or more of module storages in the memory, when being executed by the processor, execute above-mentioned
The computation migration method of program of mobile terminal in any means embodiment.
Any one embodiment of the electronic equipment of the computation migration method for executing the program of mobile terminal, can be with
Achieve the effect that corresponding aforementioned any means embodiment is identical or similar.
Those of ordinary skill in the art will appreciate that realizing all or part of the process in above-described embodiment method, being can be with
Related hardware is instructed to complete by computer program, the program can be stored in a computer-readable storage medium
In, the program is when being executed, it may include such as the process of the embodiment of above-mentioned each method.Wherein, the storage medium can be magnetic
Dish, CD, read-only memory (Read-OnlyMemory, ROM) or random access memory
(RandomAccessMemory, RAM) etc..The embodiment of the computer program can achieve corresponding aforementioned any
The identical or similar effect of embodiment of the method.
In addition, being also implemented as the computer program executed by CPU, the computer program according to disclosed method
It may be stored in a computer readable storage medium.When the computer program is executed by CPU, executes and limited in disclosed method
Fixed above-mentioned function.
In addition, above method step and system unit also can use controller and for storing so that controller is real
The computer readable storage medium of the computer program of existing above-mentioned steps or Elementary Function is realized.
Those skilled in the art will also understand is that, various illustrative logical blocks, mould in conjunction with described in disclosure herein
Block, circuit and algorithm steps may be implemented as the combination of electronic hardware, computer software or both.It is hard in order to clearly demonstrate
This interchangeability of part and software, with regard to various exemplary components, square, module, circuit and step function to its into
General description is gone.This function is implemented as software and is also implemented as hardware depending on concrete application and application
To the design constraint of whole system.Those skilled in the art can realize described in various ways for every kind of concrete application
Function, but this realization decision should not be interpreted as causing a departure from the scope of the present disclosure.
Various illustrative logical blocks, module and circuit, which can use, in conjunction with described in disclosure herein is designed to
The following component of function described here is executed to realize or execute: general processor, digital signal processor (DSP), dedicated collection
At circuit (ASIC), field programmable gate array (FPGA) or other programmable logic device, discrete gate or transistor logic, divide
Any combination of vertical hardware component or these components.General processor can be microprocessor, but alternatively, processing
Device can be any conventional processors, controller, microcontroller or state machine.Processor also may be implemented as calculating equipment
Combination, for example, the combination of DSP and microprocessor, multi-microprocessor, one or more microprocessors combination DSP core or any
Other this configurations.
It should be understood by those ordinary skilled in the art that: the discussion of any of the above embodiment is exemplary only, not
It is intended to imply that the scope of the present disclosure (including claim) is limited to these examples；Under thinking of the invention, above embodiments
Or can also be combined between the technical characteristic in different embodiments, step can be realized with random order, and be existed such as
Many other variations of the upper different aspect of the invention, for simplicity, they are not provided in details.
Claims (10)
1. a kind of image classification method, which comprises the following steps:
Sample equilibrium treatment is carried out to classified multiple original solar radio radiation spectrograms, each classification is obtained and participates in training opportunity
Impartial multiple samples；
It is input with the multiple sample, the multiple sample is trained using convolutional neural networks, obtains solar radio radiation
Spectrogram disaggregated model；
Based on the solar radio radiation spectrogram disaggregated model, classifies to non-classified solar radio radiation spectrogram, divided
Class result.
2. a kind of image classification method according to claim 1, which is characterized in that the sample equilibrium treatment includes:
Equilibrium treatment is carried out to sample based on production confrontation network, obtains multiple samples that each classification participates in training opportunity equalization
This.
3. a kind of image classification method according to claim 2, which is characterized in that the production confrontation network is to sample
Carrying out equilibrium treatment includes:
The original solar radio radiation frequency spectrum classified known to one is obtained, and the original solar radio radiation frequency spectrum of the known classification is sent
Into production confrontation network；
Study processing is carried out to the original solar radio radiation frequency spectrum of the known classification, exports exptended sample；
Receive the original solar radio radiation frequency spectrum of the exptended sample and the known classification, and to the exptended sample and it is described
Know that the original solar radio radiation frequency spectrum of classification is compared, the parameter of production confrontation network is adjusted according to comparison result；
Until comparison result difference converges between the exptended sample and the original solar radio radiation frequency spectrum of the known classification
Only, final exptended sample is obtained.
4. a kind of image classification method according to claim 3, which is characterized in that by the final exptended sample and the sun
Radio spectra figure combines, and is instructed using convolutional neural networks to the final exptended sample and the solar radio radiation spectrogram
Practice, obtains solar radio radiation spectrogram disaggregated model.
5. a kind of image classification method according to claim 4, which is characterized in that be trained to solar radio radiation spectrogram
The following steps are included:
Feature is extracted from the solar radio radiation spectrogram, obtains eigenmatrix figure；
According to the parameter reduced in network, the size of the eigenmatrix figure is reduced；
Reduced eigenmatrix figure is compared with the eigenmatrix figure of previous step, image training net is adjusted according to comparison result
Parameter in network；
Until comparison result difference convergence between the eigenmatrix figure of reduced eigenmatrix figure and previous step, obtain most
Whole eigenmatrix figure.
It is activated based on line rectification function, is resolved the gradient disperse problem in back-propagation process, and accelerate
The renewal process of parameter in network；
Final eigenmatrix figure is mapped in classifier, the solar radio radiation spectrogram disaggregated model is obtained.
6. a kind of image classification method according to claim 5, which is characterized in that the method further includes: pass through
Principal Component Analysis pre-processes the original solar radio radiation spectrogram, and carries out to the original solar radio radiation spectrogram
Dimensionality reduction.
7. a kind of image classification method according to claim 1, which is characterized in that the sample equilibrium treatment is used and above adopted
Sample method or down-sampling method, wherein
The up-sampling method includes the image of the less classification of reproduction copies until consistent with the more multi-class sample number of sample；
The down-sampling method, which is included in, chooses image when carrying out handling trained, reduces the more multi-class picture number of sample.
8. a kind of image classification method according to claim 1, which is characterized in that the sample equilibrium treatment is using balanced
Sampling method, wherein the aligned sample method the following steps are included:
It obtains sample to be trained and is grouped according to classification；
Based on each classification group, corresponding sample list is generated；
A classification is determined at random, and sample is obtained in its corresponding sample list for training.
9. a kind of image classification system characterized by comprising
Balance processing module obtains every for carrying out sample equilibrium treatment to classified multiple original solar radio radiation spectrograms
A classification participates in multiple samples of training opportunity equalization；
Image training module carries out the multiple sample using convolutional neural networks for being input with the multiple sample
Training, obtains solar radio radiation spectrogram disaggregated model；
Categorization module, for be based on the solar radio radiation spectrogram disaggregated model, to non-classified solar radio radiation spectrogram into
Row classification, obtains classification results.
10. a kind of computer readable storage medium characterized by comprising
At least one processor；And
The memory being connect at least one described processor communication；Wherein,
The memory is stored with the instruction that can be executed by least one described processor, and described instruction is by described at least one
It manages device to execute, so that at least one described processor is able to carry out the image classification side as described in claim 1-8 any one
Method.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910419415.5A CN110197212A (en) | 2019-05-20 | 2019-05-20 | Image classification method, system and computer readable storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910419415.5A CN110197212A (en) | 2019-05-20 | 2019-05-20 | Image classification method, system and computer readable storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110197212A true CN110197212A (en) | 2019-09-03 |
Family
ID=67752850
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910419415.5A Pending CN110197212A (en) | 2019-05-20 | 2019-05-20 | Image classification method, system and computer readable storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110197212A (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110648318A (en) * | 2019-09-19 | 2020-01-03 | 泰康保险集团股份有限公司 | Auxiliary analysis method and device for skin diseases, electronic equipment and storage medium |
| CN111175239A (en) * | 2020-01-19 | 2020-05-19 | 北京科技大学 | High-spectrum nondestructive testing and identifying system for imaging of colored drawing cultural relics under deep learning |
| CN111241969A (en) * | 2020-01-06 | 2020-06-05 | 北京三快在线科技有限公司 | Target detection method and device and corresponding model training method and device |
| CN112784677A (en) * | 2020-12-04 | 2021-05-11 | 上海芯翌智能科技有限公司 | Model training method and device, storage medium and computing equipment |
| EP3910822A1 (en) * | 2020-05-15 | 2021-11-17 | Airbus Defence and Space GmbH | Method of and device for recognizing a radio frequency source |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108376235A (en) * | 2018-01-15 | 2018-08-07 | 深圳市易成自动驾驶技术有限公司 | Image detecting method, device and computer readable storage medium |
| CN109614979A (en) * | 2018-10-11 | 2019-04-12 | 北京大学 | A kind of data augmentation method and image classification method based on selection with generation |
-
2019
- 2019-05-20 CN CN201910419415.5A patent/CN110197212A/en active Pending
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108376235A (en) * | 2018-01-15 | 2018-08-07 | 深圳市易成自动驾驶技术有限公司 | Image detecting method, device and computer readable storage medium |
| CN109614979A (en) * | 2018-10-11 | 2019-04-12 | 北京大学 | A kind of data augmentation method and image classification method based on selection with generation |
Non-Patent Citations (1)
| Title |
|---|
| 潘辉等: "基于深度学习的太阳射电分类算法", 《中国科技论文在线》 * |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110648318A (en) * | 2019-09-19 | 2020-01-03 | 泰康保险集团股份有限公司 | Auxiliary analysis method and device for skin diseases, electronic equipment and storage medium |
| CN111241969A (en) * | 2020-01-06 | 2020-06-05 | 北京三快在线科技有限公司 | Target detection method and device and corresponding model training method and device |
| CN111175239A (en) * | 2020-01-19 | 2020-05-19 | 北京科技大学 | High-spectrum nondestructive testing and identifying system for imaging of colored drawing cultural relics under deep learning |
| CN111175239B (en) * | 2020-01-19 | 2021-01-15 | 北京科技大学 | High-spectrum nondestructive testing and identifying system for imaging of colored drawing cultural relics under deep learning |
| EP3910822A1 (en) * | 2020-05-15 | 2021-11-17 | Airbus Defence and Space GmbH | Method of and device for recognizing a radio frequency source |
| CN112784677A (en) * | 2020-12-04 | 2021-05-11 | 上海芯翌智能科技有限公司 | Model training method and device, storage medium and computing equipment |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110197212A (en) | Image classification method, system and computer readable storage medium | |
| US11379711B2 (en) | Video action detection method based on convolutional neural network | |
| US11016214B2 (en) | Dolomite reservoir prediction method and system based on well and seismic combination, and storage medium | |
| CN107393542B (en) | Bird species identification method based on two-channel neural network | |
| CN110334580A (en) | The equipment fault classification method of changeable weight combination based on integrated increment | |
| CN102201236B (en) | Speaker recognition method combining Gaussian mixture model and quantum neural network | |
| CN110334741A (en) | Radar range profile's recognition methods based on Recognition with Recurrent Neural Network | |
| CN108805269A (en) | A method of TRANSFORMATION RATIO is picked up based on LSTM Recognition with Recurrent Neural Network | |
| CN108470190A (en) | The image-recognizing method of impulsive neural networks is customized based on FPGA | |
| CN109256118B (en) | End-to-end Chinese dialect identification system and method based on generative auditory model | |
| AU2021101142A4 (en) | Method and System for Constructing Digital Rock | |
| CN110163213A (en) | Remote sensing image segmentation method based on disparity map and multiple dimensioned depth network model | |
| CN107688201A (en) | Based on RBM earthquake prestack signal clustering methods | |
| CN106295799A (en) | A kind of implementation method of degree of depth study multilayer neural network | |
| CN112529789B (en) | Weak supervision method for removing shadow of urban visible light remote sensing image | |
| CN112966853B (en) | Urban road network short-time traffic flow prediction method based on space-time residual mixed model | |
| CN108447057A (en) | SAR image change detection based on conspicuousness and depth convolutional network | |
| CN108446616A (en) | Method for extracting roads based on full convolutional neural networks integrated study | |
| CN104849650A (en) | Analog circuit fault diagnosis method based on improvement | |
| CN114220458B (en) | Voice recognition method and device based on array hydrophone | |
| CN103325137A (en) | Method for extracting natural gas hydrate reservoir pore skeletal structure | |
| CN103994820B (en) | A kind of moving target recognition methods based on micropore diameter microphone array | |
| CN112818777B (en) | Remote sensing image target detection method based on dense connection and feature enhancement | |
| CN114021458A (en) | Small sample radar radiation source signal identification method based on parallel prototype network | |
| CN101964055B (en) | Visual perception mechansim simulation natural scene type identification method |
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 |
Application publication date: 20190903 |
|
| RJ01 | Rejection of invention patent application after publication |