CN113361628A - CNN insulator aging spectrum classification method under multi-task learning - Google Patents
CNN insulator aging spectrum classification method under multi-task learning Download PDFInfo
- Publication number
- CN113361628A CN113361628A CN202110704967.8A CN202110704967A CN113361628A CN 113361628 A CN113361628 A CN 113361628A CN 202110704967 A CN202110704967 A CN 202110704967A CN 113361628 A CN113361628 A CN 113361628A
- Authority
- CN
- China
- Prior art keywords
- cnn
- training data
- aging
- network
- discrimination network
- 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.)
- Granted
Links
Images
Classifications
-
- 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
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- 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
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y04—INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
- Y04S—SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
- Y04S10/00—Systems supporting electrical power generation, transmission or distribution
- Y04S10/50—Systems or methods supporting the power network operation or management, involving a certain degree of interaction with the load-side end user applications
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Health & Medical Sciences (AREA)
- Software Systems (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Mathematical Physics (AREA)
- Computational Linguistics (AREA)
- Health & Medical Sciences (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Image Analysis (AREA)
Abstract
The invention provides a CNN insulator aging spectrum classification method under multitask learning, which comprises the following steps: normalizing the insulator aging image data, dividing the insulator aging image data into training data and data to be processed, constructing a CNN (CNN) discrimination network containing a cross entropy loss function, inputting first training data into the CNN discrimination network for training, and obtaining a first parameter value in the CNN discrimination network; constructing a CNN authentication network containing an authentication target loss function, taking the first parameter value as an initial value of the CNN authentication network, inputting first training data into the CNN authentication network for training, and obtaining a second parameter value in the CNN discrimination network; and constructing a CNN attribute discrimination network containing an integral target loss function, taking the second parameter value as an initial value of the CNN attribute discrimination network, and inputting the first training data and the second training data into the CNN attribute discrimination network for training to obtain a classification result.
Description
Technical Field
The invention relates to the technical field of insulator aging detection, in particular to a CNN insulator aging spectrum classification method under multi-task learning.
Background
The insulator is used as an indispensable insulating element in a power transmission line, and the operation condition of the insulator influences the reliability and safety of a power grid. According to statistics, the accident with the highest proportion of the current power system faults is caused by insulator defects. The current traditional manual inspection mode has certain potential safety hazard, low efficiency, poor effect, easy appearance of missed inspection and other problems. In recent years, robots and unmanned aerial vehicles are used in inspection work instead of manual work, but still have some shortages, such as fatigue is brought to staff's vision by a large amount of video data, and the accuracy of judging the state of the insulator is seriously influenced.
Disclosure of Invention
The invention aims to provide a CNN insulator aging spectrum classification method under multitask learning, and aims to solve the problems of complex feature design extraction, poor classification effect and the like of a traditional insulator aging degree classification algorithm.
The invention is realized by the following technical scheme: a CNN insulator aging spectrum classification method under multitask learning comprises the following steps:
acquiring insulator aging image data, normalizing the insulator aging image data, and dividing the insulator aging image data into training data and data to be processed, wherein the training data comprises first training data and second training data;
constructing a CNN (CNN) discrimination network containing a cross entropy loss function, inputting first training data into the CNN discrimination network for training, and obtaining first parameter values in the CNN discrimination network, wherein the first parameters comprise weight values and offset;
constructing a CNN authentication network containing an authentication target loss function, taking the first parameter value as an initial value of the CNN authentication network, inputting first training data into the CNN authentication network for training, and obtaining a second parameter value in the CNN discrimination network, wherein the second parameter value comprises a weight value and an offset;
constructing a CNN attribute discrimination network containing an overall target loss function, taking the second parameter value as an initial value of the CNN attribute discrimination network, and inputting the first training data and the second training data into the CNN attribute discrimination network for training to obtain a classification result;
and inputting the data to be processed into the CNN attribute discrimination network to obtain a classification result of the insulator aging image.
Optionally, the first training data includes spectral image data of a plurality of different colors obtained from the same type of sample; the second training data includes spectral image data for a plurality of different colors obtained for a plurality of classes of samples.
Optionally, the constructed CNN decision network includes 3 convolutional layers, 3 pooling layers, 2 full-connection layers, and an output layer.
Optionally, the authentication target loss function is formed by the following formula:
L1＝Li+λLv
in the formula, LiAs a cross-entropy loss function, LvFor the triplet loss function, λ is an empirical parameter.
Optionally, the overall objective loss function is formed by:
L＝Ip×Li+In×Li
in the formula IpNormal case image of different illumination colors under the same sample, InIs a counterexample image of different illumination colors under another grade attribute sample.
Optionally, the classification result includes no aging, slight aging, moderate aging and severe aging.
Compared with the prior art, the invention has the following beneficial effects:
the invention provides a CNN insulator aging spectrum classification method under multitask learning, which constructs a multitask learning mode consisting of a CNN judgment network, a CNN authentication judgment network and a CNN authentication judgment attribute classification network and realizes accurate classification of insulator aging spectrum images. Compared with the traditional insulator aging identification algorithm, the method provided by the invention has the advantages that the classification precision is obviously improved, the generalization capability and the robustness are good, and the method is an effective way for detecting the aging degree of the insulator.
Drawings
In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings needed to be used in the description of the embodiments will be briefly introduced below, and it is obvious that the drawings in the following description are only preferred embodiments of the present invention, and it is obvious for those skilled in the art that other drawings can be obtained according to these drawings without inventive exercise.
Fig. 1 is a flowchart of a CNN insulator aging spectrum classification method under multitask learning according to the present invention;
fig. 2 is a parameter setting of the CNN insulator aging spectrum classification structure provided by the present invention.
Fig. 3 is a classification model of CNN insulator aging spectrum for multi-task learning according to the present invention.
Detailed Description
In order to better understand the technical content of the invention, specific embodiments are provided below, and the invention is further described with reference to the accompanying drawings.
In the following description, numerous specific details are set forth in order to provide a more thorough understanding of the present invention. It will be apparent, however, to one skilled in the art, that the present invention may be practiced without one or more of these specific details. In other instances, well-known features have not been described in order to avoid obscuring the invention.
It is to be understood that the present invention may be embodied in many different forms and should not be construed as limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the invention to those skilled in the art.
The terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. As used herein, the singular forms "a", "an" and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise. It will be further understood that the terms "comprises" and/or "comprising," when used in this specification, specify the presence of stated features, integers, steps, operations, elements, and/or components, but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and/or groups thereof. As used herein, the term "and/or" includes any and all combinations of the associated listed items.
In order to provide a thorough understanding of the present invention, a detailed structure will be set forth in the following description in order to explain the present invention. Alternative embodiments of the invention are described in detail below, however, the invention may be practiced in other embodiments that depart from these specific details.
Referring to fig. 1 to 3, a CNN insulator aging spectrum classification algorithm under multitask learning includes the following steps:
step S1, collecting insulator aging image data, normalizing the insulator aging image data, and dividing the insulator aging image data into training data and data to be processed, wherein the training data comprises first training data and second training data;
step S2, constructing a CNN discrimination network containing a cross entropy loss function, inputting first training data into the CNN discrimination network for training, and obtaining a first parameter value in the CNN discrimination network, wherein the first parameter comprises a weight value and an offset;
step S3, constructing a CNN authentication network containing an authentication target loss function, taking the first parameter value as an initial value of the CNN authentication network, inputting first training data into the CNN authentication network for training, and obtaining a second parameter value in the CNN discrimination network, wherein the second parameter value comprises a weight value and an offset;
step S4, constructing a CNN attribute discrimination network containing an integral target loss function, taking the second parameter value as an initial value of the CNN attribute discrimination network, inputting the first training data and the second training data into the CNN attribute discrimination network for training, and obtaining a classification result;
and step S5, inputting the data to be processed into the CNN attribute judgment network to obtain the classification result of the insulator aging image.
Note that, the CNN network in this embodiment is a Convolutional Neural Network (CNN).
In some embodiments of step S1, since the input image size is required to be uniform when the CNN network extracts the image features, all the aged insulator spectral images are normalized to a size of 64 × 64 in this embodiment as the input of the CNN network.
Optionally, the first training data includes spectral image data of a plurality of different colors obtained from the same type of sample; the second training data includes spectral image data for a plurality of different colors obtained for a plurality of classes of samples.
In step S2, the constructed CNN decision network is composed of 3 convolutional layers, 3 pooling layers, 2 full-link layers, and an output layer, and for the first training data, the first parameter value corresponding to the same attribute sample is obtained through operations such as 3 convolutions and 3 pooling, where the convolution operation is used as an important feature extraction method in the CNN decision network, and the equation is as follows:
where S is the result obtained by calculating the output size of the previous layer and the convolution kernel matrix, the weight value and the offset, n is the number of the regions to be calculated obtained by the step length, and x iskThe region matrix is calculated for the kth.
The pooling operation is placed after the convolution operation, reducing the number of parameters for neuron computation, and the maximum pooling equation is as follows:
where S is the pooling result of the current region, xi,jIs a parameter of the current pooling operation.
In step S3, the authentication target loss function is constituted by the following equation:
L1＝Li+λLv
in the formula, LiAs a cross-entropy loss function, LvFor the triplet loss function, λ is the empirical parameter, taken to be 0.001.
The CNN authentication network can authenticate the CNN discrimination network in the first stage, different color illumination images of the same type of sample are given to judge whether the same sample is the same, a triple loss function is adopted for supervision, the distance difference of the illumination images of the same sample in different colors in a feature domain is as small as possible, and the distance of the illumination images of different samples in different colors is larger than a certain discrimination threshold. And adding a first parameter value of the trained CNN discrimination network at the first stage as a CNN authentication network into the triple loss as supervision information, so that the identification efficiency is improved.
The triple loss function is more concise and efficient compared with the contrast loss function, the distance of the images with various colors obtained under the same sample and different illumination in the characteristic domain is smaller than the threshold distance of different samples, and the corresponding triple loss function equation is as follows:
where A is the reference for a sample of multiple colors obtained under different illuminations, P is the positive case, N is the negative case, and α is the loss parameter.
In the training process of the CNN authentication network, due to the fact that the cross entropy loss and the size of the triple loss function are different, fine adjustment is conducted according to the experience super-parameter. By judging the networks after two tasks of the network and the authentication network are trained, the probability distribution of the image characteristics of multiple colors obtained by different illumination under the same sample of the characteristic domain is kept, so that the model has good generalization capability and robustness, the distance relationship between the classes of the sample in the characteristic space domain is also considered, and the images of multiple complex colors obtained by different illumination have better distinguishing effect.
In step S4, the second parameter value obtained by the CNN authentication network in the second stage is used as the initial value of the CNN attribute discrimination network in the third stage, the cross entropy loss generated by the CNN discrimination network and the CNN authentication discrimination attribute classification network is used as the total loss, the local features of the lower layer of the network are shared, and the discrimination, authentication and classification of the illumination images of multiple colors under different samples are completed through the network in the third stage.
Wherein the overall target loss function in the CNN attribute discrimination network is formed by:
L＝Ip×Li+In×Li
in the formula IpNormal case image of different illumination colors under the same sample, InIs a counterexample image of different illumination colors under another grade attribute sample.
Because the attribute relation of different samples is added on the basis of the network in the previous stage, the attributes added with different samples are divided into four grades of the aging degree of the insulator: no aging, mild aging, moderate aging, and severe aging. The cross entropy loss generated by the cross entropy loss function is determined by a discrimination task and the opposite is determined by an attribute classification task, wherein the loss function is used for obtaining images of multiple colors from different illumination of the same sample.
To further illustrate the beneficial effects of the embodiments of the present invention, the following simulation experiments verify the effects of the present invention:
the experimental data contained 7452 (264X 190) spectral images. The method is divided into four classes according to the aging degree of the insulator: unaged, slightly aged, moderately aged, and heavily aged illuminated color images. The input image size of each model is 64 x 64, under model iterative training for 100 times, the traditional algorithms LBP + SVM, CNN and multitask CNN learning classification method Mul-CNN are compared, and the time consumed by the 3 methods and the classification precision result are shown in Table 1:
TABLE 1
| Name of classification algorithm | LBP+SVM | CNN | Mul-CNN |
| Time(s) | 681.3 | 270.6 | 170.9 |
| Accuracy of classification | 0.647 | 0.795 | 0.961 |
As can be seen from table 1, in terms of time consumption: the time consumption of the traditional LBP + SVM is higher than that of other 2 deep learning models, and is about 4 times of that of the insulator aging classification model Mul-CNN with different illumination colors, which shows that the invention combines a plurality of tasks together to train, judge, authenticate and identify attribute classification on insulator aging spectrum images with different illumination colors, can effectively shorten the time for extracting features, and has stronger robustness on data. In terms of classification accuracy: the Mul-CNN accuracy is higher than that of the other 2 methods, which shows that the method provided by the invention has better identification effect and good generalization capability on insulator aging spectrum image classification of different illumination colors.
The above description is only for the purpose of illustrating the preferred embodiments of the present invention and is not to be construed as limiting the invention, and any modifications, equivalents, improvements and the like made within the spirit and principle of the present invention should be included in the scope of the present invention.
Claims (6)
1. A CNN insulator aging spectrum classification method under multitask learning is characterized by comprising the following steps:
acquiring insulator aging image data, normalizing the insulator aging image data, and dividing the insulator aging image data into training data and data to be processed, wherein the training data comprises first training data and second training data;
constructing a CNN (CNN) discrimination network containing a cross entropy loss function, inputting first training data into the CNN discrimination network for training, and obtaining first parameter values in the CNN discrimination network, wherein the first parameters comprise weight values and offset;
constructing a CNN authentication network containing an authentication target loss function, taking the first parameter value as an initial value of the CNN authentication network, inputting first training data into the CNN authentication network for training, and obtaining a second parameter value in the CNN discrimination network, wherein the second parameter value comprises a weight value and an offset;
constructing a CNN attribute discrimination network containing an overall target loss function, taking the second parameter value as an initial value of the CNN attribute discrimination network, and inputting the first training data and the second training data into the CNN attribute discrimination network for training to obtain a classification result;
and inputting the data to be processed into the CNN attribute discrimination network to obtain a classification result of the insulator aging image.
2. The CNN insulator aging spectrum classification method under multitask learning according to claim 1, wherein the first training data comprises spectrum image data of a plurality of different colors obtained from the same class sample; the second training data includes spectral image data for a plurality of different colors obtained for a plurality of classes of samples.
3. The CNN insulator aging spectrum classification method under multitask learning according to claim 1, wherein the constructed CNN discrimination network comprises 3 convolutional layers, 3 pooling layers, 2 full-link layers and an output layer.
4. The CNN insulator aging spectrum classification method under multitask learning according to claim 1, wherein the authentication target loss function is formed by the following formula:
L1＝Li+λLv
in the formula, LiAs a cross-entropy loss function, LvFor the triplet loss function, λ is an empirical parameter.
5. The CNN insulator aging spectrum classification method under multitask learning according to claim 4, wherein the overall objective loss function is formed by the following formula:
L＝Ip×Li+In×Li
in the formula IpNormal case image of different illumination colors under the same sample, InIs a counterexample image of different illumination colors under another grade attribute sample.
6. The CNN insulator aging spectrum classification method under multitask learning of claim 1, wherein the classification results comprise no aging, light aging, moderate aging and severe aging.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110704967.8A CN113361628B (en) | 2021-06-24 | 2021-06-24 | CNN insulator aging spectrum classification method under multi-task learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110704967.8A CN113361628B (en) | 2021-06-24 | 2021-06-24 | CNN insulator aging spectrum classification method under multi-task learning |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113361628A true CN113361628A (en) | 2021-09-07 |
| CN113361628B CN113361628B (en) | 2023-04-14 |
Family
ID=77536129
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110704967.8A Active CN113361628B (en) | 2021-06-24 | 2021-06-24 | CNN insulator aging spectrum classification method under multi-task learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113361628B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114255348A (en) * | 2021-09-27 | 2022-03-29 | 海南电网有限责任公司电力科学研究院 | Insulator aging and fouling spectrum classification method for improving B _ CNN |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108446729A (en) * | 2018-03-13 | 2018-08-24 | 天津工业大学 | Egg embryo classification method based on convolutional neural networks |
| CN109685116A (en) * | 2018-11-30 | 2019-04-26 | 腾讯科技（深圳）有限公司 | Description information of image generation method and device and electronic device |
| CN109858557A (en) * | 2019-02-13 | 2019-06-07 | 安徽大学 | A kind of new hyperspectral image data semisupervised classification algorithm |
| KR20190140824A (en) * | 2018-05-31 | 2019-12-20 | 한국과학기술원 | Training method of deep learning models for ordinal classification using triplet-based loss and training apparatus thereof |
| CN111382709A (en) * | 2020-03-11 | 2020-07-07 | 国网浙江省电力有限公司电力科学研究院 | Insulator image detection method based on unmanned aerial vehicle inspection |
-
2021
- 2021-06-24 CN CN202110704967.8A patent/CN113361628B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108446729A (en) * | 2018-03-13 | 2018-08-24 | 天津工业大学 | Egg embryo classification method based on convolutional neural networks |
| KR20190140824A (en) * | 2018-05-31 | 2019-12-20 | 한국과학기술원 | Training method of deep learning models for ordinal classification using triplet-based loss and training apparatus thereof |
| CN109685116A (en) * | 2018-11-30 | 2019-04-26 | 腾讯科技（深圳）有限公司 | Description information of image generation method and device and electronic device |
| CN109858557A (en) * | 2019-02-13 | 2019-06-07 | 安徽大学 | A kind of new hyperspectral image data semisupervised classification algorithm |
| CN111382709A (en) * | 2020-03-11 | 2020-07-07 | 国网浙江省电力有限公司电力科学研究院 | Insulator image detection method based on unmanned aerial vehicle inspection |
Non-Patent Citations (3)
| Title |
|---|
| 刘欣宇 等: "基于轻量级深度卷积神经网络的绝缘子检测", 《福州大学学报(自然科学版)》 * |
| 张血琴 等: "基于高光谱技术的复合绝缘子表面老化程度评估", 《电工技术学报》 * |
| 黄立达: "航拍图像中基于细粒度分类的绝缘子自爆缺陷检测", 《中国优秀硕士学位论文全文数据库（电子期刊）工程科技II辑》 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114255348A (en) * | 2021-09-27 | 2022-03-29 | 海南电网有限责任公司电力科学研究院 | Insulator aging and fouling spectrum classification method for improving B _ CNN |
| CN114255348B (en) * | 2021-09-27 | 2023-01-10 | 海南电网有限责任公司电力科学研究院 | Insulator aging and fouling spectrum classification method for improving B _ CNN |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113361628B (en) | 2023-04-14 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108615010B (en) | Facial expression recognition method based on parallel convolution neural network feature map fusion | |
| CN111414942B (en) | Remote sensing image classification method based on active learning and convolutional neural network | |
| WO2021134871A1 (en) | Forensics method for synthesized face image based on local binary pattern and deep learning | |
| CN113378632B (en) | Pseudo-label optimization-based unsupervised domain adaptive pedestrian re-identification method | |
| CN109215034B (en) | Weak supervision image semantic segmentation method based on spatial pyramid covering pooling | |
| CN108256482B (en) | Face age estimation method for distributed learning based on convolutional neural network | |
| CN106504233A (en) | Image electric power widget recognition methodss and system are patrolled and examined based on the unmanned plane of Faster R CNN | |
| CN109472284A (en) | A kind of battery core defect classification method based on zero sample learning of unbiased insertion | |
| CN116612098B (en) | Insulator RTV spraying quality evaluation method and device based on image processing | |
| CN106599864A (en) | Deep face recognition method based on extreme value theory | |
| CN111461025A (en) | Signal identification method for self-evolving zero-sample learning | |
| CN113642486A (en) | Unmanned aerial vehicle distribution network inspection method with airborne front-end identification model | |
| CN113361628B (en) | CNN insulator aging spectrum classification method under multi-task learning | |
| CN112070720A (en) | Transformer substation equipment defect identification method based on deep learning model | |
| CN107545281B (en) | Single harmful gas infrared image classification and identification method based on deep learning | |
| CN116012653A (en) | Method and system for classifying hyperspectral images of attention residual unit neural network | |
| CN111598854A (en) | Complex texture small defect segmentation method based on rich robust convolution characteristic model | |
| CN115908842A (en) | Transformer partial discharge data enhancement and identification method | |
| Qiu et al. | Recognition of transmission line related bird species based on image feature extraction and support vector machine | |
| CN111310820A (en) | Foundation meteorological cloud chart classification method based on cross validation depth CNN feature integration | |
| CN113221614A (en) | Power grid insulator damage image identification method based on hybrid neural network | |
| CN110569727B (en) | Transfer learning method combining intra-class distance and inter-class distance for motor imagery classification | |
| CN117173147A (en) | Surface treatment equipment and method for steel strip processing | |
| CN110136098B (en) | Cable sequence detection method based on deep learning | |
| CN113361631A (en) | Insulator aging spectrum classification method based on transfer learning |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |