CN114140662A - Insulator lightning stroke image sample amplification method based on cyclic generation countermeasure network - Google Patents
Insulator lightning stroke image sample amplification method based on cyclic generation countermeasure network Download PDFInfo
- Publication number
- CN114140662A CN114140662A CN202111436502.5A CN202111436502A CN114140662A CN 114140662 A CN114140662 A CN 114140662A CN 202111436502 A CN202111436502 A CN 202111436502A CN 114140662 A CN114140662 A CN 114140662A
- Authority
- CN
- China
- Prior art keywords
- image
- umbrella skirt
- insulator
- images
- lightning stroke
- 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
- 239000012212 insulator Substances 0.000 title claims abstract description 100
- 208000025274 Lightning injury Diseases 0.000 title claims abstract description 78
- 238000000034 method Methods 0.000 title claims abstract description 47
- 125000004122 cyclic group Chemical group 0.000 title claims description 7
- 230000003321 amplification Effects 0.000 title abstract description 4
- 238000003199 nucleic acid amplification method Methods 0.000 title abstract description 4
- 238000012360 testing method Methods 0.000 claims abstract description 11
- 239000013598 vector Substances 0.000 claims description 45
- 238000012549 training Methods 0.000 claims description 20
- 238000009826 distribution Methods 0.000 claims description 12
- 230000003416 augmentation Effects 0.000 claims description 11
- 238000013527 convolutional neural network Methods 0.000 claims description 5
- 230000009977 dual effect Effects 0.000 claims description 5
- 230000003190 augmentative effect Effects 0.000 claims description 4
- 238000001914 filtration Methods 0.000 claims description 4
- 238000011156 evaluation Methods 0.000 claims description 3
- 230000008569 process Effects 0.000 abstract description 14
- 239000000463 material Substances 0.000 abstract description 3
- 238000004140 cleaning Methods 0.000 abstract description 2
- 230000006870 function Effects 0.000 description 19
- 208000009119 Giant Axonal Neuropathy Diseases 0.000 description 16
- 201000003382 giant axonal neuropathy 1 Diseases 0.000 description 16
- 238000005516 engineering process Methods 0.000 description 14
- 238000013473 artificial intelligence Methods 0.000 description 6
- 238000003860 storage Methods 0.000 description 6
- 230000005540 biological transmission Effects 0.000 description 5
- 238000012545 processing Methods 0.000 description 5
- 238000004891 communication Methods 0.000 description 4
- 238000010586 diagram Methods 0.000 description 4
- 230000004438 eyesight Effects 0.000 description 4
- 208000037170 Delayed Emergence from Anesthesia Diseases 0.000 description 3
- 241000282414 Homo sapiens Species 0.000 description 3
- 238000004422 calculation algorithm Methods 0.000 description 3
- 230000008859 change Effects 0.000 description 3
- 238000013136 deep learning model Methods 0.000 description 3
- 238000009413 insulation Methods 0.000 description 3
- 238000010801 machine learning Methods 0.000 description 3
- 238000013507 mapping Methods 0.000 description 3
- 238000005070 sampling Methods 0.000 description 3
- 239000000725 suspension Substances 0.000 description 3
- 238000004590 computer program Methods 0.000 description 2
- 238000005520 cutting process Methods 0.000 description 2
- 238000003066 decision tree Methods 0.000 description 2
- 238000013135 deep learning Methods 0.000 description 2
- 230000007547 defect Effects 0.000 description 2
- 230000003993 interaction Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 230000003287 optical effect Effects 0.000 description 2
- APTZNLHMIGJTEW-UHFFFAOYSA-N pyraflufen-ethyl Chemical compound C1=C(Cl)C(OCC(=O)OCC)=CC(C=2C(=C(OC(F)F)N(C)N=2)Cl)=C1F APTZNLHMIGJTEW-UHFFFAOYSA-N 0.000 description 2
- 238000012216 screening Methods 0.000 description 2
- 238000004088 simulation Methods 0.000 description 2
- 230000000007 visual effect Effects 0.000 description 2
- 241001643392 Cyclea Species 0.000 description 1
- 230000003044 adaptive effect Effects 0.000 description 1
- 230000003042 antagnostic effect Effects 0.000 description 1
- 238000003491 array Methods 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000006399 behavior Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000015572 biosynthetic process Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 238000013145 classification model Methods 0.000 description 1
- 239000002131 composite material Substances 0.000 description 1
- 238000013480 data collection Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 230000006866 deterioration Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 239000000284 extract Substances 0.000 description 1
- 239000011521 glass Substances 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000009434 installation Methods 0.000 description 1
- 239000004973 liquid crystal related substance Substances 0.000 description 1
- 238000007477 logistic regression Methods 0.000 description 1
- 230000007774 longterm Effects 0.000 description 1
- 238000007726 management method Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 230000003924 mental process Effects 0.000 description 1
- 238000013508 migration Methods 0.000 description 1
- 230000005012 migration Effects 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 238000003062 neural network model Methods 0.000 description 1
- 239000013307 optical fiber Substances 0.000 description 1
- 238000013439 planning Methods 0.000 description 1
- 238000011176 pooling Methods 0.000 description 1
- 230000002265 prevention Effects 0.000 description 1
- 230000000306 recurrent effect Effects 0.000 description 1
- 238000009877 rendering Methods 0.000 description 1
- 230000000717 retained effect Effects 0.000 description 1
- 239000004065 semiconductor Substances 0.000 description 1
- 230000001953 sensory effect Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 238000012546 transfer Methods 0.000 description 1
- 238000013024 troubleshooting Methods 0.000 description 1
- 230000004382 visual function Effects 0.000 description 1
Images
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
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/217—Validation; Performance evaluation; Active pattern learning 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
- G06N3/044—Recurrent networks, e.g. Hopfield 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
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/047—Probabilistic or stochastic 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
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Molecular Biology (AREA)
- Computational Linguistics (AREA)
- Software Systems (AREA)
- Mathematical Physics (AREA)
- Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computing Systems (AREA)
- General Health & Medical Sciences (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Probability & Statistics with Applications (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses an insulator lightning stroke image sample amplification method based on a circularly generated countermeasure network, which comprises the following steps: extracting an image of the shed part in each insulator image from a plurality of pre-acquired insulator images to obtain a plurality of shed images; dividing the umbrella skirt images into a normal umbrella skirt image group and a lightning stroke umbrella skirt image group according to a lightning stroke classification rule, wherein the umbrella skirt in the lightning stroke umbrella skirt image group has lightning stroke traces; and generating a countermeasure network according to a preset cycle to train and test the normal umbrella skirt image group and the lightning stroke umbrella skirt image group. A batch of insulator lightning stroke data sets which are false and spurious can be effectively generated, and the labor and material cost in the data acquisition and cleaning process is reduced.
Description
Technical Field
The invention relates to the technical field of artificial intelligence such as computer vision, deep learning and the like, in particular to an insulator lightning stroke image sample augmentation method based on a cyclic generation countermeasure network.
Background
The distribution network has the characteristics of wide distribution, more equipment and low insulation level, so that the phenomenon of insulation accidents caused by overlarge voltage can often occur, and the probability of the occurrence of the insulation accidents is very high particularly in thunderstorm weather. Along with the continuous deterioration of the environment, the pollution flashover and brake drop accidents of the power transmission line caused by lightning strike are increased day by day, so that the normal operation of equipment is influenced, and the daily production and life of power consumers are greatly influenced. Transmission line tripping is often due to lightning strikes. Statistically, in recent line tripping, the line tripping caused by lightning strike accounts for about 40% -50%, which accounts for a considerable percentage. Therefore, the problem of insulator lightning stroke damage is very important in the distribution line management process, various schemes are adopted to carry out leakage prevention and emergency accident treatment measures, the lightning protection work of the line is well done, the lightning stroke trip-out rate is reduced, the line accident rate is rapidly reduced, and the influence of the insulator after lightning stroke is reduced to the minimum.
In the past years, insulator lightning strike troubleshooting work has been gradually replaced by artificial intelligence CV image algorithm recognition technology, but the technology particularly relies on a large number of image data sets, so that in the past, data acquisition is usually carried out by simulating data in a laboratory environment by an unmanned aerial vehicle and acquiring a target data set through the operation of artificial PS.
Insulator receives the thunderbolt to damage the back and can change immediately in the real circuit, consequently at unmanned aerial vehicle large-scale data collection in the past, then carries out manual screening, picks out the data set of thunderbolt, and this process need consume a large amount of manpower and materials, perhaps can not acquire the data that want yet. The simulation under the laboratory environment is high in specificity and high in cost, and the data sample under the laboratory simulation environment is single in form, so that the real situation of data under different backgrounds cannot be completely simulated, and the sample background is single.
Therefore, more times, the data can be expanded only by rendering lightning stroke traces on normal insulators in the later period through technologies such as PS and the like. The technical difficulty of the PS is that the software needs a troublesome learning process, professional drawing skills and the like, the process of drawing the lightning traces on the thousands of insulators one by one is complex, the efficiency is low, the similarity of the lightning traces is high and the like, and the effect of the software as a training data set is not obvious.
After the advent of artificial intelligence computer vision technology, more and more tedious manual operations were replaced, freeing up both hands. The computer vision technology is a technology for simulating the visual process of human beings by a computer, and has the capability of feeling the environment and the visual function of human beings. How to generate false and real images is an urgent problem to be solved in the industry.
Disclosure of Invention
The invention provides a method for amplifying insulator lightning stroke samples based on a circularly generated countermeasure network, which can effectively generate a batch of insulator lightning stroke data sets which are false and spurious, and reduce the manpower and material resource costs in the data acquisition and cleaning processes.
The invention provides an insulator lightning stroke image sample amplification method based on a circularly generated countermeasure network, which comprises the following steps:
extracting an image of the shed part in each insulator image from a plurality of pre-acquired insulator images to obtain a plurality of shed images;
dividing the umbrella skirt images into a normal umbrella skirt image group and a lightning stroke umbrella skirt image group according to a lightning stroke classification rule, wherein the umbrella skirt in the lightning stroke umbrella skirt image group has lightning stroke traces;
and generating a countermeasure network according to a preset cycle to train and test the normal umbrella skirt image group and the lightning stroke umbrella skirt image group.
Alternatively to this, the first and second parts may,
the step of extracting the image of the shed part in each insulator image from the plurality of pre-acquired insulator images to obtain a plurality of shed images comprises the following steps:
marking the umbrella skirt part in each insulator image in the plurality of insulator images to obtain a plurality of marked insulator images;
and obtaining a plurality of umbrella skirt images according to the positions and sizes of the marks and the images of the umbrella skirt parts in the insulator images.
Alternatively to this, the first and second parts may,
marking the umbrella skirt part in each insulator image in the plurality of insulator images to obtain a plurality of marked insulator images comprises the following steps:
and using a mark frame to circle the umbrella skirt part, wherein the information of the mark frame comprises a name, Xmin is used as an X coordinate of the upper left corner of the mark frame, Ymin is used as a Y coordinate of the upper left corner of the mark frame, Xmax is used as an X coordinate of the lower right corner of the mark frame, and Ymax is used as a Y coordinate of the lower right corner of the mark frame.
Alternatively to this, the first and second parts may,
the step of dividing the plurality of umbrella skirt images into a normal umbrella skirt image group and a lightning stroke umbrella skirt image group according to the lightning stroke classification rule comprises the following steps:
and (3) enabling the normal umbrella skirt image data and the lightning stroke umbrella skirt image data to be in a proportion of 1: the proportion of 1 is respectively stored in two folders;
each image data is saved as a txt file in a CycleGan data format.
Alternatively to this, the first and second parts may,
the step of extracting the image of the umbrella skirt part in each insulator image from the plurality of insulator images obtained in advance further comprises filtering the images without the insulators before obtaining the plurality of umbrella skirt images.
Alternatively to this, the first and second parts may,
the step of generating a countermeasure network according to a preset cycle to train and test the normal umbrella skirt image group and the lightning stroke umbrella skirt image group comprises the following steps:
building a cyclic generation countermeasure network with a dual structure of a first generator network, a second generator network and a first discriminator network and a second discriminator network;
coding images in the normal umbrella skirt image group into feature vectors with preset sizes by using a convolutional neural network to obtain a first domain containing a plurality of feature vectors, and coding images in the lightning stroke umbrella skirt image group into feature vectors with preset sizes to obtain a second domain containing a plurality of feature vectors;
generating a feature vector of a second domain by the feature vector of the first domain through a first generator, and reconstructing an input image of the normal umbrella skirt image group through a second generator;
the first discriminator is used for discriminating the characteristic vector of the first domain, and the second discriminator is used for discriminating the characteristic vector of the second domain.
Alternatively to this, the first and second parts may,
the reconstruction is to use a deconvolution layer to complete the restoration of an image of the same size as the input image from the feature vectors.
Alternatively to this, the first and second parts may,
the discriminator network is also used for judging whether the patch with the preset size covered by the picture is from the original picture.
Alternatively to this, the first and second parts may,
the step of training the normal umbrella skirt image group and the lightning stroke umbrella skirt image group by generating a countermeasure network according to a preset cycle comprises the following steps:
constructing a reconstruction loss function of a generator, a discrimination loss function of a discriminator and a mean square error loss function;
the reconstruction loss function is recorded as L (G _ AB, G _ BA, a, B) ═ E _ (a-a) [ | | G _ AB (a)) - | | 1 ];
the discrimination loss function is denoted as L _ GAN (G, G _ Y, X, Y) ═ E _ (Y to pdata) (Y) [ log (D _ Y (Y)) ] + E _ (X to pdata (X)) [ log (1-D _ Y (G (X))) ];
the mean square error loss function is denoted as L _ lsgan (G, G _ Y, X, Y) ═ E _ (Y to Pdata (Y) [ (D _ Y (Y) -1)2]+E_(x～Pdata(x))[D_y(G(x))2]。
Alternatively to this, the first and second parts may,
the evaluation index of the normal umbrella skirt image group and the lightning stroke umbrella skirt image group tested by the countermeasure network generated according to the preset circulation IS an Inclusion Score (IS), which IS marked as IS (G) exp (E (x-Pg) D _ KL (p (y | x) | p (y)),
wherein x to Pg represent the generation of pictures from the generator;
p (y | x) represents that the generated picture x is input into the inclusion V3 to obtain a 1000-dimensional vector y;
and p (y) is N generated pictures, each generated picture is input into the inclusion V3 to respectively obtain a probability distribution vector of the picture, and the mean value of the probability distribution vectors is obtained to obtain the edge distribution of the whole pictures generated by the generator on all the categories.
Compared with the prior art, the application has the following beneficial effects:
firstly, extracting an image of a shed part in each insulator image from a plurality of pre-acquired insulator images to obtain a plurality of shed images; dividing the umbrella skirt images into a normal umbrella skirt image group and a lightning stroke umbrella skirt image group according to a lightning stroke classification rule, wherein the umbrella skirt in the lightning stroke umbrella skirt image group has lightning stroke traces; and finally, generating a countermeasure network according to a preset cycle to train and test the normal umbrella skirt image group and the lightning stroke umbrella skirt image group. Because the insulator umbrella skirt images are cut out from the original images to be used as training sets, background interference is reduced, a batch of images which are false and spurious are generated, and the cost problem of manual acquisition and marking is solved. The generalization and the universality of the deep learning model can be ensured. It should be understood that the statements in this section do not necessarily identify key or critical features of the embodiments of the present disclosure, nor do they limit the scope of the present disclosure. Other features of the present disclosure will become apparent from the following description.
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 some embodiments of the present invention, and it is obvious for those skilled in the art to obtain other drawings based on these drawings without creative efforts.
FIG. 1 is a flowchart of a first embodiment of a method for augmenting an insulator lightning strike image sample based on a cycle generation countermeasure network according to the present invention;
FIG. 2 is a flowchart of a second embodiment of the method for augmenting an insulator lightning strike image sample based on a cycle generation countermeasure network according to the present invention;
FIG. 3 is a flowchart of a third embodiment of the method for augmenting an insulator lightning strike image sample based on a cycle generation countermeasure network according to the present invention;
FIG. 4 is a schematic diagram of a schematic structure of the loop generation countermeasure network of the present invention;
FIG. 5 is a schematic diagram of another schematic structure of the loop generation countermeasure network of the present invention;
fig. 6 is another schematic structural diagram of the loop generation countermeasure network of the present invention.
Detailed Description
Hereinafter, embodiments of the present disclosure will be described with reference to the accompanying drawings. It should be understood that the description is illustrative only and is not intended to limit the scope of the present disclosure. Moreover, in the following description, descriptions of well-known structures and techniques are omitted so as to not unnecessarily obscure the concepts of the present disclosure.
The terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the disclosure. The words "a", "an" and "the" and the like as used herein are also intended to include the meanings of "a plurality" and "the" unless the context clearly dictates otherwise. Furthermore, the terms "comprises," "comprising," and the like, as used herein, specify the presence of stated features, steps, operations, and/or components, but do not preclude the presence or addition of one or more other features, steps, operations, or components.
All terms (including technical and scientific terms) used herein have the same meaning as commonly understood by one of ordinary skill in the art unless otherwise defined. It is noted that the terms used herein should be interpreted as having a meaning that is consistent with the context of this specification and should not be interpreted in an idealized or overly formal sense.
The terms appearing hereinafter include the following explanations:
the insulator is an industrial product and is used for insulating and suspending wires in high-voltage and ultrahigh-voltage alternating-current and direct-current transmission lines. The insulator is an important component of the power transmission line, is a unique electric insulating part and an important structural support part, and the reasonability of the performance and the configuration of the insulator directly influences the safe and stable operation of the line. The insulators adopted by the existing power transmission line are mainly divided into two categories: one is a disc-type suspension glass insulator applied to a tension string, and the other is a rod-type suspension composite insulator applied to a suspension string and a jumper string. In long-term operation, the two types of insulators show different operating properties and characteristics.
And generating the confrontation network model, wherein the confrontation network model comprises a Generator and a Discriminator, wherein the Generator is G for short, and the Discriminator is D for short. The two data fields are preset as X and Y respectively. One set of images is stored in data field X and another set of images is stored in data field Y. The generator is used to train a mapping G X → Y such that the output is
X is belonged to X, and the discriminator is used for comparing
And y are distinguished, the purpose of generating the antagonistic network model is to obtain the preference
So that the discriminator cannot distinguish
And y.
The particular generator G is a network that generates an image, and receives a random noise z from which the image is generated, denoted G (z). The discriminator D is a discrimination network for discriminating whether an image is "real". The input parameter is x, x represents an image, and the output D (x) represents the probability that the image x is a real image. If D (x) is 1, it means that 100% is a real image, and D (x) is 0, it means that it is impossible to be a real image. During the training process, the goal of the generator G is to try to generate real images to deceive the discriminator D. The object of the discriminator D is to separate the image generated by the generator G from the real image as much as possible. Thus, the generator G and the discriminator D constitute a dynamic "gaming process". In the most ideal state, the generator G can generate enough images G (z) to be "spurious". It is difficult for the discriminator D to determine whether the image generated by the generator G is real or not, and therefore D (G (z)) is 0.5. In this way a model G of the generator is obtained, which can be used to generate images. The loss function of GAN is as follows: the loss function of GAN is composed of two terms. x represents the real image, z represents the noise input to the generator G, and G (z) represents the image generated by the generator G. D (x) represents the probability that the discriminator D judges whether the real image is real or not. Since x is true, the closer this value is to 1 the better for D. And D (G (z)) is the probability that the discriminator D judges whether the image generated by the generator G is true or not.
Referring to fig. 1, a first embodiment of an insulator lightning strike image sample augmentation method based on a cycle generation countermeasure network according to the present invention includes:
101. extracting an image of the shed part in each insulator image from a plurality of pre-acquired insulator images to obtain a plurality of shed images;
in this embodiment, image data is collected by pre-installation monitoring or unmanned aerial vehicles. And acquiring images of the umbrella skirt part in each insulator image through color level sampling to obtain a plurality of umbrella skirt images.
It should be noted that the image data is stored in the VOC data format as an xml file, the xml file contains category information and location information, and then the VOC data format is converted into a CycleGan training data format as a data set.
102. And dividing the umbrella skirt images into a normal umbrella skirt image group and a lightning stroke umbrella skirt image group according to a lightning stroke classification rule, wherein the umbrella skirt in the lightning stroke umbrella skirt image group has lightning stroke traces.
In this embodiment, a mark frame is adopted for marking, and the marking information includes a name, Xmin is used as an X coordinate of the upper left corner of the mark frame, Ymin is used as a Y coordinate of the upper left corner of the mark frame, Xmax is used as an X coordinate of the lower right corner of the mark frame, and Ymax is used as a Y coordinate of the lower right corner of the mark frame. The specific marking rules are not specifically limited herein.
In this embodiment, the normal shed image group and the lightning-stroke shed image group data are in a ratio of 1: and 1, screening. The lightning stroke and the normal stroke of the insulator umbrella skirt are respectively divided into two folders A and B, wherein 80% of the lightning stroke and the normal stroke are used as a training set, and 20% of the lightning stroke and the normal stroke are used as a testing set.
103. And generating a countermeasure network according to a preset cycle to train and test the normal umbrella skirt image group and the lightning stroke umbrella skirt image group.
Unlike the conventional generative countermeasure network GAN, the input image of the recurrent generative countermeasure network Cycle-GAN in this embodiment may be any two images, and the images of the two domains are converted into each other. The normal umbrella skirt image and the lightning-stroke umbrella skirt image can be regarded as the input of two definition domains, the normal umbrella skirt image can be converted into the domain of the lightning-stroke umbrella skirt image through Cycle-GAN, the definition domain difference between a real image and a generated image is solved, and the style transfer of the image is realized. Generally, Cycle-GAN is a ring structure, which mainly consists of two generators and two discriminators. The Cycle-GAN includes a generator G, a generator F, a discriminator Dx, and a discriminator Dy. The input image X of the X domain generates an image of the Y domain through the generator G, and is reconstructed back to the input image Y of the X domain through the generator F, and the image of the X domain is generated through the generator F and is reconstructed back to the input image of the Y domain through the generator G. The discriminator Dx discriminates an image in the X domain, and the discriminator Dy discriminates an image in the Y domain. Thereby ensuring style migration of the image. And simultaneously ensuring that the geometric shape and the spatial relationship of the objects in the image are not changed while the style is converted. Wherein the loss function of the Cycle-GAN is a Cycle-consistency loss function (Cycle-consistency loss).
In the embodiment, firstly, an image of an umbrella skirt part in each insulator image is extracted from a plurality of pre-acquired insulator images to obtain a plurality of umbrella skirt images; dividing the umbrella skirt images into a normal umbrella skirt image group and a lightning stroke umbrella skirt image group according to a lightning stroke classification rule, wherein the umbrella skirt in the lightning stroke umbrella skirt image group has lightning stroke traces; and finally, generating a countermeasure network according to a preset cycle to train and test the normal umbrella skirt image group and the lightning stroke umbrella skirt image group. Because the insulator umbrella skirt images are cut out from the original images to be used as training sets, background interference is reduced, a batch of images which are false and spurious are generated, and the cost problem of manual acquisition and marking is solved. The generalization and the universality of the deep learning model can be ensured.
The first embodiment of the method for amplifying the insulator lightning strike image sample based on the cycle generation countermeasure network according to the present invention is described above, and a second embodiment is described below, with reference to fig. 2, where the second embodiment of the method for amplifying the insulator lightning strike image sample based on the cycle generation countermeasure network according to the present invention is different from the first embodiment in that:
the step of extracting the image of the umbrella skirt part in each insulator image from the plurality of insulator images obtained in advance further comprises filtering the images without the insulators before obtaining the plurality of umbrella skirt images.
One method that may be implemented in this embodiment includes the following steps:
201. filtering images without insulators from the pre-acquired multiple images;
in this embodiment, first, the data set needs to be processed to clean data not including the target and data with high noise, where the target data is defined as data with edge features.
202. Extracting an image of the shed part in each insulator image from a plurality of pre-acquired insulator images to obtain a plurality of shed images;
203. dividing the umbrella skirt images into a normal umbrella skirt image group and a lightning stroke umbrella skirt image group according to a lightning stroke classification rule, wherein the umbrella skirt in the lightning stroke umbrella skirt image group has lightning stroke traces;
204. and generating a countermeasure network according to a preset cycle to train and test the normal umbrella skirt image group and the lightning stroke umbrella skirt image group.
The steps 202-204 are similar to the steps 101-103 in the first embodiment, and are not described herein again.
In this embodiment, since the data set is cleaned in advance, only the image containing the insulator is retained, thereby further improving the accuracy of the subsequent image data processing. Noise interference is further reduced, a batch of images which are false and spurious are generated, and the cost problem of manual collection and marking is solved. The generalization and the universality of the deep learning model can be ensured.
In the above description of the second embodiment of the method for amplifying the insulator lightning strike image sample based on the cycle generation countermeasure network provided by the present invention, referring to fig. 3, a third embodiment is described, and the third embodiment of the method for amplifying the insulator lightning strike image sample based on the cycle generation countermeasure network provided by the present invention is different from the previous embodiments in that:
the step of training the normal umbrella skirt image group and the lightning stroke umbrella skirt image group by generating a countermeasure network according to a preset cycle comprises the following steps:
building a cyclic generation countermeasure network with a dual structure of a first generator network, a second generator network and a first discriminator network and a second discriminator network;
coding images in the normal umbrella skirt image group into feature vectors with preset sizes by using a convolutional neural network to obtain a first domain containing a plurality of feature vectors, and coding images in the lightning stroke umbrella skirt image group into feature vectors with preset sizes to obtain a second domain containing a plurality of feature vectors;
generating a feature vector of a second domain by the feature vector of the first domain through a first generator, and reconstructing an input image of the normal umbrella skirt image group through a second generator;
and constructing a reconstruction loss function of the generator, a discrimination loss function of the discriminator and a mean square error loss function.
One method that may be implemented in this embodiment includes the following steps:
301. extracting an image of the shed part in each insulator image from a plurality of pre-acquired insulator images to obtain a plurality of shed images;
302. dividing the umbrella skirt images into a normal umbrella skirt image group and a lightning stroke umbrella skirt image group according to a lightning stroke classification rule, wherein the umbrella skirt in the lightning stroke umbrella skirt image group has lightning stroke traces;
the steps 301-302 are similar to the steps 101-102, and will not be described in detail.
303. And building a cyclic generation countermeasure network with a dual structure of a first generator network, a second generator network and a first discriminator network and a second discriminator network.
The generator in this embodiment is composed of an encoder, a converter and a decoder, and first encodes an image, and extracts features from the input image using a convolutional neural network, and both the generator G and the generator F use convolution with a step size greater than 1 to reduce the length and width of a feature map, and compress the image into 256 feature vectors of 64 × 64.
The feature vectors of the image in the DA domain are then converted to feature vectors in the DB domain by combining the dissimilar features of the images. Then using residual concatenation, 6 layers of Reset modules are used, each of which is a neural network layer consisting of two convolutional layers, able to achieve the goal of preserving the original image features at the same time as the conversion. And finally, using the deconvolution layer to finish the work of restoring low-level features from the feature vector, and finally obtaining an image with the same size as the input image.
The network as a whole goes through a process of down-sampling and then up-sampling, with a series of residual blocks in between, the number being determined by the actual situation, with 6 residual blocks being used when the input resolution is 128x128 and 9 residual blocks being used when the input resolution is 256x256 or even higher.
In this embodiment, the machine learning model may be a probability model, a classification model, or other classifier in the prior art or future development technology, for example, the machine learning model may include any one of the following items: a Decision Tree model (XGBoost), a logistic regression model (LR), a deep neural network model (DNN), a Gradient Boosting Decision Tree model (GBDT).
In this embodiment, the arbiter has 5 layers of convolution, the number of channels is reduced to 1, the final pooling is averaged, the size is also reduced to 1 × 1, and finally reshape is performed to (batch size, 1). The discriminator takes an image as input and tries to predict whether it is the original image or the output image of the generator. The discriminator itself belongs to a convolutional network, and it is necessary to extract features from the image and then determine whether the extracted features belong to a particular class by adding a convolutional layer that produces a one-dimensional output.
In order to make the training process of the model more stable, the adaptive loss is improved. The negative log-likelihood target is replaced by a least squares penalty. The losses are made more stable during training, resulting in higher quality results.
304. Coding images in the normal umbrella skirt image group into feature vectors with preset sizes by using a convolutional neural network to obtain a first domain containing a plurality of feature vectors, and coding images in the lightning stroke umbrella skirt image group into feature vectors with preset sizes to obtain a second domain containing a plurality of feature vectors;
305. and generating the feature vector of the first domain into a feature vector of a second domain through a first generator, and reconstructing the feature vector of the second domain back into the input image of the normal umbrella skirt image group through a second generator.
In this embodiment, two generator networks G and F are built, and two discriminator networks Dx and Dy are built. The formation of the loop-forming countermeasure network is a dual structure. This network contains two convolutional layers of step size 2, several residual modules, two transposed convolutional layers of step size 1/2. We used 6 modules to process 128x128 pictures and 9 modules to process 256x256 high resolution training pictures. We used regularization for each example. We used the patch gans of 70x70 as my discriminator network to determine if the picture overlay 70x70 patch is from the original. Such patch-level discriminators have fewer parameters than full-graph discriminators and can process images of arbitrary size in a fully convoluted manner. This mapping is created by training to ensure that there is a meaningful correlation between the input image and the generated image, i.e. that the input and output share some features. An input image is taken from the domain DA, which is passed to a first generator GeneratorA → B, the task of which is to convert a given image from the domain DA into an image in the target domain DB. This newly generated image is then passed to another generator GeneratorB → a, the task of which is to convert back to the image CycleA in the original domain DA. This output image must be similar to the original input image to define a meaningful mapping that does not originally exist in the unpaired dataset.
306. And constructing a reconstruction loss function of the generator, a discrimination loss function of the discriminator and a mean square error loss function.
Referring to fig. 4-6, in this embodiment, the one-way GAN is trained using two loss: the reconstructed Loss of the generator and the discriminated Loss of the discriminator.
1. Reconstruction of Loss: it is desirable that the generated picture Gba (gab (a)) be as similar as possible to the original image a.
L(G_AB，G_BA，A，B)＝E_(a～A)[||G_AB(G_AB(a))-||_1]
2. Judging Loss: both the generated false picture and the original true picture are input to the discriminator. The formula is a loss function L _ GAN (G, G _ Y, X, Y) of class 0, 1 ═ E _ (Y to pdata (Y)) [ log (D _ Y (Y)) ] + E _ (X to pdata (X)) [ log (1-D _ Y (G (X)) ]
Two mirror symmetric GANs are created to form a ring network. Data is converted from two data sources by converting input samples.
Two GANs share two generators and each have one arbiter, i.e. there are two arbiters and two generators in common. One unidirectional GAN has two los, i.e., four los in total. Using the mean square error loss expression:
L_lsgan(G，G_Y，X，Y)＝E_(y～Pdata(y))[(D_Y(y)-1)2]
+E_(x～Pdata(x))[D_y(G(x))2]
defining loss functions of four xx machines, respectively optimizing training G and D, wherein two machines share weight and two discriminators also share weight training, and calculating the loss of each generated image is impossible because a large amount of calculation resources are consumed. An image library is created that stores the previously generated 50 images, rather than just the most recent generator generated images.
Lr is 0.0002. The same learning rate of 0.0002 was maintained for the first 100 cycles, and then linearly decayed to 0 over the next 100 cycles.
The method comprises the steps of dividing lightning stroke and normal state of an insulator shed into two folders A and B respectively, cutting out an insulator shed image from an original image to serve as a training set so as to reduce background interference, carrying out iterative training on a model, observing change of LOSS LOSS of training, and stopping training when the LOSS value continuously drops and tends to be stable and the LOSS value drops below 5.
In the testing stage, the insulator body image is cut out from the original image in the same way, and then the cutting size and position information are recorded. Then, inputting the body data into the algorithm model to generate data of normal-lightning stroke of a batch of insulators, and then restoring the original image according to the size and position information of the insulators in the original image recorded at the beginning, and replacing the newly generated insulators with normal sheds.
In the task of normal- > lightning strike, the complete formula of an evaluation index acceptance Score (IS), acceptance Score IS as follows:
IS(G)＝exp(E_(x～Pg)D_KL(p(y|x)||p(y)))
x to Pg: representing the generation of a picture from a generator.
p (y | x): the generated picture x is input into the inclusion V3, and a 1000-dimensional vector y, that is, the probability distribution of the picture belonging to each category, is obtained. According to the previous assumption, for a clearly generated picture, some dimension value of this vector is exceptionally large, while the remaining dimension values are exceptionally small (i.e., the probability density map is quite sharp).
p (y): each generated picture is input into the inclusion V3, a probability distribution vector of the generated picture is obtained, and the vectors are averaged to obtain the edge distribution of the whole generated picture on all categories.
The GAN network generates a batch of new insulator data with lightning stroke in a one-to-one correspondence mode by using normal insulators, and ensures that the overall shape of the insulator is not changed, but the background environment may be changed in a small degree. And the brightness of the image of the insulator in the dark environment is changed by the new lightning strike insulator generated by the GAN network algorithm. Most of the training data sets are insulator image data with bright light. In other generated data, the brightness of the image background may also be accompanied by slight changes. For later defect identification, the change in brightness does not affect the defect data set as a later one.
Various implementations of the systems and techniques described here above may be implemented in digital electronic circuitry, integrated circuitry, Field Programmable Gate Arrays (FPGAs), Application Specific Integrated Circuits (ASICs), Application Specific Standard Products (ASSPs), system on a chip (SOCs), load programmable logic devices (CPLDs), computer hardware, firmware, software, and/or combinations thereof. These various embodiments may include: implemented in one or more computer programs that are executable and/or interpretable on a programmable system including at least one programmable processor, which may be special or general purpose, receiving data and instructions from, and transmitting data and instructions to, a storage system, at least one input device, and at least one output device.
Program code for implementing the methods of the present application may be written in any combination of one or more programming languages. These program codes may be provided to a processor or controller of a general purpose computer, special purpose computer, or other programmable data processing apparatus, such that the program codes, when executed by the processor or controller, cause the functions/operations specified in the flowchart and/or block diagram to be performed. The program code may execute entirely on the machine, partly on the machine, as a stand-alone software package partly on the machine and partly on a remote machine or entirely on the remote machine or server.
In the context of this application, a machine-readable medium may be a tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device. The machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any suitable combination of the foregoing. More specific examples of a machine-readable storage medium would include an electrical connection based on one or more wires, a portable computer diskette, a hard disk, a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.
To provide for interaction with a user, the systems and techniques described here can be implemented on a computer having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to a user; and a keyboard and a pointing device (e.g., a mouse or a trackball) by which a user can provide input to the computer. Other kinds of devices may also be used to provide for interaction with a user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user may be received in any form, including acoustic, speech, or tactile input.
The systems and techniques described here can be implemented in a computing system that includes a back-end component (e.g., as a data server), or that includes a middleware component (e.g., an application server), or that includes a front-end component (e.g., a user computer having a graphical user interface or a web browser through which a user can interact with an implementation of the systems and techniques described here), or any combination of such back-end, middleware, or front-end components. The components of the system can be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: local Area Networks (LANs), Wide Area Networks (WANs), and the Internet.
The computer system may include clients and servers. A client and server are generally remote from each other and typically interact through a communication network. The relationship of client and server arises by virtue of computer programs running on the respective computers and having a client-server relationship to each other.
Artificial intelligence is the subject of studying computers to simulate some human mental processes and intelligent behaviors (such as learning, reasoning, thinking, planning, etc.), both at the hardware level and at the software level. Artificial intelligence hardware technologies generally include technologies such as sensors, dedicated artificial intelligence chips, cloud computing, distributed storage, big data processing, and the like; the artificial intelligence software technology mainly comprises a computer vision technology, a voice recognition technology, a natural voice processing technology, machine learning/deep learning, a big data processing technology, a knowledge map technology and the like.
It should be understood that various forms of the flows shown above may be used, with steps reordered, added, or deleted. For example, the steps described in the present application may be executed in parallel, sequentially, or in different orders, and the present invention is not limited thereto as long as the desired results of the technical solutions disclosed in the present application can be achieved.
The above-described embodiments should not be construed as limiting the scope of the present application. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions may be made in accordance with design requirements and other factors. Any modification, equivalent replacement, and improvement made within the spirit and principle of the present application shall be included in the protection scope of the present application.
Claims (10)
1. An insulator lightning stroke image sample augmentation method based on a cycle generation countermeasure network is characterized by comprising the following steps:
extracting an image of the shed part in each insulator image from a plurality of pre-acquired insulator images to obtain a plurality of shed images;
dividing the umbrella skirt images into a normal umbrella skirt image group and a lightning stroke umbrella skirt image group according to a lightning stroke classification rule, wherein the umbrella skirt in the lightning stroke umbrella skirt image group has lightning stroke traces;
and generating a countermeasure network according to a preset cycle to train and test the normal umbrella skirt image group and the lightning stroke umbrella skirt image group.
2. The insulator lightning strike image sample augmentation method based on cycle-generated countermeasure network of claim 1, characterized by:
the step of extracting the image of the shed part in each insulator image from the plurality of pre-acquired insulator images to obtain a plurality of shed images comprises the following steps:
marking the umbrella skirt part in each insulator image in the plurality of insulator images to obtain a plurality of marked insulator images;
and obtaining a plurality of umbrella skirt images according to the positions and sizes of the marks and the images of the umbrella skirt parts in the insulator images.
3. The insulator lightning strike image sample augmentation method based on cycle generation countermeasure network of claim 2, characterized by:
marking the umbrella skirt part in each insulator image in the plurality of insulator images to obtain a plurality of marked insulator images comprises the following steps:
and using a mark frame to circle the umbrella skirt part, wherein the information of the mark frame comprises a name, Xmin is used as an X coordinate of the upper left corner of the mark frame, Ymin is used as a Y coordinate of the upper left corner of the mark frame, Xmax is used as an X coordinate of the lower right corner of the mark frame, and Ymax is used as a Y coordinate of the lower right corner of the mark frame.
4. The insulator lightning strike image sample augmentation method based on cycle-generated countermeasure network of claim 1, characterized by:
the step of dividing the plurality of umbrella skirt images into a normal umbrella skirt image group and a lightning stroke umbrella skirt image group according to the lightning stroke classification rule comprises the following steps:
and (3) enabling the normal umbrella skirt image data and the lightning stroke umbrella skirt image data to be in a proportion of 1: the proportion of 1 is respectively stored in two folders;
each image data is saved as a txt file in a CycleGan data format.
5. The insulator lightning strike image sample augmentation method based on cycle-generated countermeasure network of claim 1, characterized by:
the step of extracting the image of the umbrella skirt part in each insulator image from the plurality of insulator images obtained in advance further comprises filtering the images without the insulators before obtaining the plurality of umbrella skirt images.
6. The insulator lightning strike image sample augmentation method based on cycle-generated countermeasure network of claim 1, characterized by:
the step of generating a countermeasure network according to a preset cycle to train and test the normal umbrella skirt image group and the lightning stroke umbrella skirt image group comprises the following steps:
building a cyclic generation countermeasure network with a dual structure of a first generator network, a second generator network and a first discriminator network and a second discriminator network;
coding images in the normal umbrella skirt image group into feature vectors with preset sizes by using a convolutional neural network to obtain a first domain containing a plurality of feature vectors, and coding images in the lightning stroke umbrella skirt image group into feature vectors with preset sizes to obtain a second domain containing a plurality of feature vectors;
generating a feature vector of a second domain by the feature vector of the first domain through a first generator, and reconstructing an input image of the normal umbrella skirt image group through a second generator;
the first discriminator is used for discriminating the characteristic vector of the first domain, and the second discriminator is used for discriminating the characteristic vector of the second domain.
7. The insulator lightning strike image sample augmentation method based on cycle-generated countermeasure network of claim 6, characterized by:
the reconstruction is to use a deconvolution layer to complete the restoration of an image of the same size as the input image from the feature vectors.
8. The insulator lightning strike image sample augmentation method based on cycle-generated countermeasure network of claim 6, characterized by:
the discriminator network is also used for judging whether the patch with the preset size covered by the picture is from the original picture.
9. The method for augmenting the insulator lightning strike image sample based on cycle generation countermeasure network of claim 6, further comprising:
the step of training the normal umbrella skirt image group and the lightning stroke umbrella skirt image group by generating a countermeasure network according to a preset cycle comprises the following steps:
constructing a reconstruction loss function of a generator, a discrimination loss function of a discriminator and a mean square error loss function;
the reconstruction loss function is recorded as L (G _ AB, G _ BA, a, B) ═ E _ (a-a) [ | | G _ AB (a)) - | | 1 ];
the discrimination loss function is denoted as L _ GAN (G, G _ Y, X, Y) ═ E _ (Y to pdata) (Y) [ log (D _ Y (Y)) ] + E _ (X to pdata (X)) [ log (1-D _ Y (G (X))) ];
the mean square error loss function is denoted as L _ lsgan (G, G _ Y, X, Y) ═ E _ (Y to Pdata (Y) [ (D _ Y (Y) -1)2]+E_(x～Pdata(x))[D_y(G(x))2]。
10. The insulator lightning strike image sample augmentation method based on cycle-generated countermeasure network of claim 1, characterized by:
the evaluation index of the normal umbrella skirt image group and the lightning stroke umbrella skirt image group tested by the countermeasure network generated according to the preset circulation IS an Inclusion Score (IS), which IS marked as IS (G) exp (E (x-Pg) D _ KL (p (y | x) | p (y)),
wherein x to Pg represent the generation of pictures from the generator;
p (y | x) represents that the generated picture x is input into the inclusion V3 to obtain a 1000-dimensional vector y;
and p (y) is N generated pictures, each generated picture is input into the inclusion V3 to respectively obtain a probability distribution vector of the picture, and the mean value of the probability distribution vectors is obtained to obtain the edge distribution of the whole pictures generated by the generator on all the categories.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111436502.5A CN114140662A (en) | 2021-11-29 | 2021-11-29 | Insulator lightning stroke image sample amplification method based on cyclic generation countermeasure network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111436502.5A CN114140662A (en) | 2021-11-29 | 2021-11-29 | Insulator lightning stroke image sample amplification method based on cyclic generation countermeasure network |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114140662A true CN114140662A (en) | 2022-03-04 |
Family
ID=80390008
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111436502.5A Pending CN114140662A (en) | 2021-11-29 | 2021-11-29 | Insulator lightning stroke image sample amplification method based on cyclic generation countermeasure network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114140662A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114821681A (en) * | 2022-06-27 | 2022-07-29 | 深圳市魔力信息技术有限公司 | Fingerprint augmentation method |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110222757A (en) * | 2019-05-31 | 2019-09-10 | 华北电力大学（保定） | Based on insulator image pattern extending method, the system for generating confrontation network |
| US20190295302A1 (en) * | 2018-03-22 | 2019-09-26 | Northeastern University | Segmentation Guided Image Generation With Adversarial Networks |
| CN111145116A (en) * | 2019-12-23 | 2020-05-12 | 哈尔滨工程大学 | Sea surface rainy day image sample augmentation method based on generation of countermeasure network |
| CN112541566A (en) * | 2020-09-17 | 2021-03-23 | 中国石油大学(华东) | Image translation method based on reconstruction loss |
| CN113570603A (en) * | 2021-09-26 | 2021-10-29 | 国网浙江省电力有限公司电力科学研究院 | Contact network part defect data generation method and system based on cycleGAN |
| CN113658051A (en) * | 2021-06-25 | 2021-11-16 | 南京邮电大学 | Image defogging method and system based on cyclic generation countermeasure network |
| CN113673568A (en) * | 2021-07-19 | 2021-11-19 | 华南理工大学 | Method, system, computer device and storage medium for detecting tampered image |
-
2021
- 2021-11-29 CN CN202111436502.5A patent/CN114140662A/en active Pending
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20190295302A1 (en) * | 2018-03-22 | 2019-09-26 | Northeastern University | Segmentation Guided Image Generation With Adversarial Networks |
| CN110222757A (en) * | 2019-05-31 | 2019-09-10 | 华北电力大学（保定） | Based on insulator image pattern extending method, the system for generating confrontation network |
| CN111145116A (en) * | 2019-12-23 | 2020-05-12 | 哈尔滨工程大学 | Sea surface rainy day image sample augmentation method based on generation of countermeasure network |
| CN112541566A (en) * | 2020-09-17 | 2021-03-23 | 中国石油大学(华东) | Image translation method based on reconstruction loss |
| CN113658051A (en) * | 2021-06-25 | 2021-11-16 | 南京邮电大学 | Image defogging method and system based on cyclic generation countermeasure network |
| CN113673568A (en) * | 2021-07-19 | 2021-11-19 | 华南理工大学 | Method, system, computer device and storage medium for detecting tampered image |
| CN113570603A (en) * | 2021-09-26 | 2021-10-29 | 国网浙江省电力有限公司电力科学研究院 | Contact network part defect data generation method and system based on cycleGAN |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114821681A (en) * | 2022-06-27 | 2022-07-29 | 深圳市魔力信息技术有限公司 | Fingerprint augmentation method |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107742093B (en) | Real-time detection method, server and system for infrared image power equipment components | |
| Schäfer et al. | Detection of gravitational-wave signals from binary neutron star mergers using machine learning | |
| CN106991666B (en) | A kind of disease geo-radar image recognition methods suitable for more size pictorial informations | |
| CN114022432A (en) | Improved yolov 5-based insulator defect detection method | |
| CN113204903B (en) | Method for predicting thunder and lightning | |
| CN112884758B (en) | Defect insulator sample generation method and system based on style migration method | |
| CN115294473A (en) | Insulator fault identification method and system based on target detection and instance segmentation | |
| EP3920145A2 (en) | Method and apparatus for generating image, device, storage medium and program product | |
| CN113935237A (en) | Power transmission line fault type distinguishing method and system based on capsule network | |
| CN117115715A (en) | Video anomaly detection method based on combination of stream reconstruction and frame prediction | |
| CN114140662A (en) | Insulator lightning stroke image sample amplification method based on cyclic generation countermeasure network | |
| CN112712855B (en) | Joint training-based clustering method for gene microarray containing deletion value | |
| CN117152435A (en) | Remote sensing semantic segmentation method based on U-Net3+ | |
| CN117079005A (en) | Optical cable fault monitoring method, system, device and readable storage medium | |
| NL2033915A (en) | Monthly precipitation forecasting model with the temporal convolutional network | |
| CN116310757A (en) | Multitasking real-time smoke detection method | |
| CN115546629A (en) | Remote sensing image workshop identification method and system based on deep learning | |
| CN112100165B (en) | Traffic data processing method, system, equipment and medium based on quality assessment | |
| CN116071299A (en) | Insulator RTV spraying defect detection method and system | |
| CN115641498A (en) | Medium-term rainfall forecast post-processing correction method based on space multi-scale convolutional neural network | |
| CN114821351A (en) | Railway hazard source identification method and device, electronic equipment and storage medium | |
| CN114548307A (en) | Classification model training method and device, and classification method and device | |
| CN112633142A (en) | Power transmission line violation building identification method and related device | |
| CN113487550B (en) | Target detection method and device based on improved activation function | |
| CN117437230B (en) | Photovoltaic power station power prediction method and system based on image restoration |
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 |