CN109872317A - A kind of defect identification method based on power equipments defect identification learning model - Google Patents
A kind of defect identification method based on power equipments defect identification learning model Download PDFInfo
- Publication number
- CN109872317A CN109872317A CN201910131289.3A CN201910131289A CN109872317A CN 109872317 A CN109872317 A CN 109872317A CN 201910131289 A CN201910131289 A CN 201910131289A CN 109872317 A CN109872317 A CN 109872317A
- Authority
- CN
- China
- Prior art keywords
- learning model
- defect identification
- power
- defect
- point
- 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
Landscapes
- Image Analysis (AREA)
Abstract
The invention discloses a kind of defect identification methods based on power equipments defect identification learning model.The present invention is based on multilayer labels mark and identification, feature extraction, bounding box return and the Novel learning model of classifier technique centralized integration, prior probability is replaced using clustering algorithm in a model, training set is generated to power equipment sample training, it can be with real-time detection power equipment type and defect, the Accuracy and high efficiency that ensure that defect recognition has ensured the safe operation of power equipment.The degree of automation of power equipment state monitoring and defect recognition is improved, power grid big data analysis system for handling is researched and developed, power equipments defect identification technology theory is innovated, the technology of field of power system is promoted to develop.Advanced technology is applied into field of power system, saves human resources and time management cost, promotes field development.
Description
Technical field
The present invention relates to electrical equipment detection technique fields, and in particular to one kind is based on power equipments defect identification learning mould
The defect identification method of type.
Background technique
With the interconnection of bulk power grid and the continuous expansion of power grid scale, the safety and stability problem of power equipment is caused extensively
General concern.Traditional power equipments defect type is that testing staff judges according to experience.On the one hand, as electric power is set
The continuous variation of standby update and running environment, equipment deficiency type increase, and corresponding detection experience is also constantly changing.It is another
Aspect, with the extension of transmission line of electricity and increasing for electrical equipment, equipment deficiency detects the increase of workload and accordingly detects people
The contradiction of member's lazy weight is increasingly sharpened.
In recent years, digital vedio recording and the fast development of computer vision technique and extensive use are that power equipments defect identifies
Provide new direction.The transport inspection department doors at different levels of China's power grid are aided with unmanned plane, helicopter etc. on the basis of manual inspection and take
The technology synergies operations such as camera, infrared thermoviewer are carried, improve operating efficiency to a certain extent.Know in power equipments defect
Other more generally x-ray imaging technology in the prior art, treats detection device using X-ray first and is irradiated acquisition
The X-ray detection picture of equipment various pieces, then pre-processes picture.Subsequent patrol officer penetrates pretreated X
Line detection picture is observed one by one, determines that electric power is set according to experience by gray scale of different zones in X-ray detection picture etc.
Standby failure and the origin cause of formation.But which still relies on manually to various image interpretations to identify equipment deficiency, and workload is huge
While still remain by artificial experience judgement, fail to judge and judge by accident the case where.Therefore, what raising power equipments defect detected can
It is particularly important by property and efficiency.
Using the picture of computer vision technique processing power equipment, part professional patrol officer's lazy weight can be alleviated
The problem of.Image generally comprises method based on template matching, Statistics-Based Method, certainly according to different algorithm image classifications
Plan tree, support vector machines and Pattern Recognition etc..Sample number needed for the method for traditional images processing is generally existing
Measure the problem of huge, precision is not high enough and takes a long time;Current neural network model can only be trained to one layer of label
It practises, hence it is evident that be not suitable for type and the more field of electrical equipment of defect type.
Summary of the invention
It is provided by the invention a kind of based on power equipments defect identification learning model for above-mentioned deficiency in the prior art
Defect identification method solve the problems, such as power equipments defect detection sample size it is huge, precision is not high and takes a long time.
In order to achieve the above object of the invention, the technical solution adopted by the present invention are as follows: one kind is identified based on power equipments defect
The defect identification method of learning model, comprising the following steps:
S1, pass through the image data of X ray image equipment batch capture power equipment;
S2, the convolutional layer that image data is transmitted to defect recognition learning model extract power equipment figure by convolutional layer
As the characteristic pattern of data;
S3, candidate region is generated by region candidate network, and obtains the interested region of image data;
S4, the area-of-interest pond layer that characteristic pattern and candidate region are sent to defect recognition learning model, and export
Target signature;
S5, target signature is sent to by depth convolutional neural networks by full articulamentum, obtains power equipment type;
S6, classification processing is normalized to target signature based on power equipment type, and is handled using frame recurrence
The exact position of detection block is obtained, defects of power equipment identification decision result is obtained.
Further: the image data in the step S2 is transmitted to defect recognition learning model by serial communication mode
Convolutional layer.
Further: the learning model in the step S2 includes convolutional layer, rectifies layer and pond layer, the convolutional layer
Number is 13, and the number of the rectification layer is 13, and the number of the pond layer is 4.
Further: the parameter of the convolutional layer are as follows: convolution kernel size=3, expansion edge=1, convolution kernel step-length=1,
The parameter of the pond layer are as follows: edge=0, pond step-length=2 are expanded in size=2 Chi Huahe.
Further: the depth convolutional neural networks in the step S5 include 1*1 convolution, 3*3 convolution, 5*5 convolution, 3*
3 maximum pondizations and filter cascade.
Further: normalizing classification processing in the step S6 specifically by normalization classification anchor point and obtain mesh to be measured
Mark and background, the frame return processing and obtain precision target specifically by frame recurrence amendment anchor point.
Further: the classification anchor point and amendment anchor point are all made of density peaks clustering algorithm.
Further: the specific steps of the density peaks clustering algorithm are as follows:
A, truncation distance, calculation formula are calculated are as follows:
dc=D (t*m)
In above formula, dcFor distance is truncated, t is truncation distance parameter, and t ∈ (0,1), D are data set, and m is intermediate parameters,
In,I, j is respectively any two points x in data set SiAnd xj, dijFor xiAnd xjThe distance between, N
For the number of point to be clustered, m=0.5N (N-1);
B, point x is calculated by truncation distanceiLocal density values, calculation formula are as follows:
ρi=∑ χ (dij-dc)
In above formula, ρiFor point xiLocal density values, dc> 0, χ (x) are known function, wherein
C, point x is calculatediWith local density values ratio ρiBig data point xjBetween minimum range, calculation formula are as follows:
In above formula, δiFor minimum range;
D, according to local density valuesAnd minimum rangeDecision diagram is drawn, cluster centre point is selected;
E, the data of non-cluster central point are sorted out.
The invention has the benefit that the present invention is returned based on multilayer labels mark and identification, feature extraction, bounding box
With the Novel learning model of classifier technique centralized integration, prior probability is replaced using clustering algorithm in a model, electric power is set
Standby sample training generates training set, with real-time detection power equipment type and defect, can ensure that defect recognition accuracy and
High efficiency has ensured the safe operation of power equipment.The degree of automation for improving power equipment state monitoring and defect recognition, grinds
Power generating network big data analysis system for handling innovates power equipments defect identification technology theory, promotes the technology of field of power system
Development.Advanced technology is applied into field of power system, saves human resources and time management cost, promotes field development.
Detailed description of the invention
Fig. 1 is flow chart of the present invention.
Specific embodiment
A specific embodiment of the invention is described below, in order to facilitate understanding by those skilled in the art this hair
It is bright, it should be apparent that the present invention is not limited to the ranges of specific embodiment, for those skilled in the art,
As long as various change is in the spirit and scope of the present invention that the attached claims limit and determine, these variations are aobvious and easy
See, all are using the innovation and creation of present inventive concept in the column of protection.
As shown in Figure 1, a kind of defect identification method based on power equipments defect identification learning model, including following step
It is rapid:
S1, pass through the image data of X ray image equipment batch capture power equipment;
The power equipment inner structure part or state obtained by the principle of x-ray imaging is more accurate, to be detected
Equipment predetermined region installs high-definition camera, can also use UAV flight's camera, obtain electricity by the mode of data flow
The image information of power equipment.The image data collected is transmitted to processing as original image by way of serial communication
Device, original image data are the input of defect recognition learning model.
S2, the convolutional layer that image data is transmitted to defect recognition learning model extract power equipment figure by convolutional layer
As the characteristic pattern of data;Image data is transmitted to the volume collection layer of defect recognition learning model by serial communication mode.Learn mould
Type includes convolutional layer, rectification layer and pond layer, and the number of the convolutional layer is 13, and the number of the rectification layer is 13, institute
The number for stating pond layer is 4.The parameter of the convolutional layer are as follows: convolution kernel size=3, expansion edge=1, convolution kernel step-length=
1, the parameter of the pond layer are as follows: edge=0, pond step-length=2 are expanded in size=2 Chi Huahe.
The first part of the defect recognition learning model structure is convolutional layer, uses the convolution+rectification+pond on one group of basis
Layer extracts the characteristic pattern of power equipment image data.Characteristic pattern as a shared platform, be supplied to region candidate network layer and
Full articulamentum does subsequent calculating identification, reduces calculating consumption.
S3, candidate region is generated by region candidate network, and obtains the interested region of image data;
Defect recognition learning model using area candidate network generates candidate region, using the mode of multilayer labels, accordingly
The interested region of power equipment image multilayer can be obtained in learning model.When using the model, using target detection mark
Note tool carries out multilayer labels mark to training set, verifying collection and test set image data respectively, can be simultaneously to image data
On device type or part of appliance have zero defect it is equivalent when mark, no longer use the single label of traditional approach.
S4, the area-of-interest pond layer that characteristic pattern and candidate region are sent to defect recognition learning model, and export
Target signature;
The area-of-interest pond layer of the learning model is responsible for collecting target, and calculates target signature, is to image
Information carries out second quantization, the size in fixed interest region.Assuming that input model be M*N size picture, use first
Input data is mapped as (M/16) * (N/16) by spatial_scale parameter, then by each target level and vertical size point
It is not divided into pooled_w and pooled_h parts, finally every a maximum pondization that all carries out is handled, to realize the learning model
Regular length output.
S5, target signature is sent to by depth convolutional neural networks by full articulamentum, obtains power equipment type;It is deep
Degree convolutional neural networks include 1*1 convolution, 3*3 convolution, 5*5 convolution, 3*3 maximum pond and filter cascade.In order to reduce meter
5*5 convolution is reduced to 1*5 convolution sum 5*1 convolution collective effect, the effect of depth convolutional neural networks by calculation amount, the learning model
Similar to classifier.
Full articulamentum ensure that the input size of model subsequent request, can handle any input, depth convolution mind
Classification and Identification is carried out to first layer label through network, obtains some component of power equipment type or large scale equipment.Using more
The mode of layer label avoids the power equipment for picking out single kind in the image data of batch processing from being sent into model and is instructed
Practice identification, eliminates a large amount of data prediction work.
For example, to marks such as power equipment shaft tower, insulator, Bird's Nest, wire clamps as first layer label;Specific to wire clamp
The positions such as image data, then the non-slip groove to wire clamp, steel core, aluminum stranded conductor mark is used as second layer label;It is pin-pointed to equipment
Component to be detected after, then be labeled to whether the component defective as third layer label.
S6, classification processing is normalized to target signature based on power equipment type, and is handled using frame recurrence
The exact position of detection block is obtained, defects of power equipment identification decision result is obtained.
It normalizes classification processing and obtains object to be measured and background, frame recurrence processing specifically by normalization classification anchor point
Amendment anchor point, which is returned, specifically by frame obtains precision target.Classification anchor point and amendment anchor point are all made of density peaks cluster and calculate
Method.
The specific steps of density peaks clustering algorithm are as follows:
A, truncation distance, calculation formula are calculated are as follows:
dc=D (t*m)
In above formula, dcFor distance is truncated, t is truncation distance parameter, and t ∈ (0,1), D are data set, and m is intermediate parameters,
In,I, j is respectively any two points x in data set SiAnd xj, dijFor xiAnd xjThe distance between, N
For the number of point to be clustered, m=0.5N (N-1);
B, point x is calculated by truncation distanceiLocal density values, calculation formula are as follows:
ρi=∑ χ (dij-dc)
In above formula, ρiFor point xiLocal density values, dc> 0, χ (x) are known function, wherein
C, point x is calculatediWith local density values ratio ρiBig data point xjBetween minimum range, calculation formula are as follows:
In above formula, δiFor minimum range;
D, according to local density valuesAnd minimum rangeDecision diagram is drawn, cluster centre point is selected;
E, the data of non-cluster central point are sorted out.
Claims (8)
1. a kind of defect identification method based on power equipments defect identification learning model, which comprises the following steps:
S1, pass through the image data of X ray image equipment batch capture power equipment;
S2, the convolutional layer that image data is transmitted to defect recognition learning model extract power equipment picture number by convolutional layer
According to characteristic pattern;
S3, candidate region is generated by region candidate network, and obtains the interested region of image data;
S4, the area-of-interest pond layer that characteristic pattern and candidate region are sent to defect recognition learning model, and export target
Characteristic pattern;
S5, target signature is sent to by depth convolutional neural networks by full articulamentum, obtains power equipment type;
S6, classification processing is normalized to target signature based on power equipment type, and handles to obtain using frame recurrence
The exact position of detection block obtains defects of power equipment identification decision result.
2. the defect identification method according to claim 1 based on power equipments defect identification learning model, feature exist
In the image data in the step S2 is transmitted to the convolutional layer of defect recognition learning model by serial communication mode.
3. the defect identification method according to claim 1 based on power equipments defect identification learning model, feature exist
In the learning model in the step S2 includes convolutional layer, rectification layer and pond layer, and the number of the convolutional layer is 13, institute
The number for stating rectification layer is 13, and the number of the pond layer is 4.
4. the defect identification method according to claim 3 based on power equipments defect identification learning model, feature exist
In the parameter of the convolutional layer are as follows: edge=1, convolution kernel step-length=1, the ginseng of the pond layer are expanded in convolution kernel size=3
Number are as follows: edge=0, pond step-length=2 are expanded in size=2 Chi Huahe.
5. the defect identification method according to claim 1 based on power equipments defect identification learning model, feature exist
In the depth convolutional neural networks in the step S5 include 1*1 convolution, 3*3 convolution, 5*5 convolution, 3*3 maximum pond and filter
The cascade of wave device.
6. the defect identification method according to claim 1 based on power equipments defect identification learning model, feature exist
In, classification processing is normalized in the step S6 specifically by normalization classification anchor point obtains object to be measured and background, it is described
Frame returns processing and returns amendment anchor point acquisition precision target specifically by frame.
7. the defect identification method according to claim 6 based on power equipments defect identification learning model, feature exist
In the classification anchor point and amendment anchor point are all made of density peaks clustering algorithm.
8. the defect identification method according to claim 7 based on power equipments defect identification learning model, feature exist
In the specific steps of the density peaks clustering algorithm are as follows:
A, truncation distance, calculation formula are calculated are as follows:
dc=D (t*m)
In above formula, dcFor distance is truncated, t is truncation distance parameter, and t ∈ (0,1), D are data set, and m is intermediate parameters, whereinI, j is respectively any two points x in data set SiAnd xj, dijFor xiAnd xjThe distance between, N be to
Cluster the number of point, m=0.5N (N-1);
B, point x is calculated by truncation distanceiLocal density values, calculation formula are as follows:
ρi=∑ χ (dij-dc)
In above formula, ρiFor point xiLocal density values, dc> 0, χ (x) is known function, wherein
C, point x is calculatediWith local density values ratio ρiBig data point xjBetween minimum range, calculation formula are as follows:
In above formula, δiFor minimum range;
D, according to local density valuesAnd minimum rangeDecision diagram is drawn, cluster centre point is selected;
E, the data of non-cluster central point are sorted out.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910131289.3A CN109872317A (en) | 2019-02-22 | 2019-02-22 | A kind of defect identification method based on power equipments defect identification learning model |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910131289.3A CN109872317A (en) | 2019-02-22 | 2019-02-22 | A kind of defect identification method based on power equipments defect identification learning model |
Publications (1)
Publication Number | Publication Date |
---|---|
CN109872317A true CN109872317A (en) | 2019-06-11 |
Family
ID=66919042
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910131289.3A Pending CN109872317A (en) | 2019-02-22 | 2019-02-22 | A kind of defect identification method based on power equipments defect identification learning model |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109872317A (en) |
Cited By (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110378221A (en) * | 2019-06-14 | 2019-10-25 | 安徽南瑞继远电网技术有限公司 | A kind of power grid wire clamp detects and defect identification method and device automatically |
CN110544247A (en) * | 2019-09-03 | 2019-12-06 | 东莞德福得精密五金制品有限公司 | method for inspecting and monitoring operation defects of artificial intelligent cloud computing multi-path equipment system |
CN110868734A (en) * | 2019-12-03 | 2020-03-06 | 长江师范学院 | Unmanned aerial vehicle cluster dynamic topology mining method based on spectrum data analysis |
CN111027631A (en) * | 2019-12-13 | 2020-04-17 | 四川赛康智能科技股份有限公司 | X-ray image classification and identification method for judging crimping defects of high-voltage strain clamp |
CN111523423A (en) * | 2020-04-15 | 2020-08-11 | 四川赛康智能科技股份有限公司 | Power equipment identification method and device |
CN111815567A (en) * | 2020-06-15 | 2020-10-23 | 国网上海市电力公司 | Automatic labeling method and device for high-recognition-rate power equipment |
CN112017173A (en) * | 2020-09-02 | 2020-12-01 | 西南交通大学 | Power equipment defect detection method based on target detection network and structured positioning |
CN112052622A (en) * | 2020-08-11 | 2020-12-08 | 国网河北省电力有限公司 | Defect disposal method for deep multi-view semantic document representation under cloud platform |
CN112434740A (en) * | 2020-11-26 | 2021-03-02 | 西北大学 | Depth learning-based Qin tomb warriors fragment classification method |
CN112991259A (en) * | 2021-01-29 | 2021-06-18 | 合肥晶合集成电路股份有限公司 | Method and system for detecting defects of semiconductor manufacturing process |
CN115277865A (en) * | 2022-07-19 | 2022-11-01 | 浪潮思科网络科技有限公司 | Display method, device and medium for device panel |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20160070950A1 (en) * | 2014-09-10 | 2016-03-10 | Agency For Science, Technology And Research | Method and system for automatically assigning class labels to objects |
CN107203783A (en) * | 2017-05-24 | 2017-09-26 | 西北工业大学 | Remote sensing images Pixel-level sorting technique based on adaptive convolution kernel convolutional neural networks |
CN107992720A (en) * | 2017-12-14 | 2018-05-04 | 浙江工业大学 | Cancer based on coexpression network targets marker mapping method |
CN108257114A (en) * | 2017-12-29 | 2018-07-06 | 天津市万贸科技有限公司 | A kind of transmission facility defect inspection method based on deep learning |
-
2019
- 2019-02-22 CN CN201910131289.3A patent/CN109872317A/en active Pending
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20160070950A1 (en) * | 2014-09-10 | 2016-03-10 | Agency For Science, Technology And Research | Method and system for automatically assigning class labels to objects |
CN107203783A (en) * | 2017-05-24 | 2017-09-26 | 西北工业大学 | Remote sensing images Pixel-level sorting technique based on adaptive convolution kernel convolutional neural networks |
CN107992720A (en) * | 2017-12-14 | 2018-05-04 | 浙江工业大学 | Cancer based on coexpression network targets marker mapping method |
CN108257114A (en) * | 2017-12-29 | 2018-07-06 | 天津市万贸科技有限公司 | A kind of transmission facility defect inspection method based on deep learning |
Cited By (16)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110378221A (en) * | 2019-06-14 | 2019-10-25 | 安徽南瑞继远电网技术有限公司 | A kind of power grid wire clamp detects and defect identification method and device automatically |
CN110544247A (en) * | 2019-09-03 | 2019-12-06 | 东莞德福得精密五金制品有限公司 | method for inspecting and monitoring operation defects of artificial intelligent cloud computing multi-path equipment system |
CN110544247B (en) * | 2019-09-03 | 2023-11-07 | 东莞德福得精密五金制品有限公司 | Method for checking and monitoring operation defects of artificial intelligent cloud computing multi-path equipment system |
CN110868734A (en) * | 2019-12-03 | 2020-03-06 | 长江师范学院 | Unmanned aerial vehicle cluster dynamic topology mining method based on spectrum data analysis |
CN110868734B (en) * | 2019-12-03 | 2022-11-01 | 长江师范学院 | Unmanned aerial vehicle cluster dynamic topology mining method based on spectrum data analysis |
CN111027631A (en) * | 2019-12-13 | 2020-04-17 | 四川赛康智能科技股份有限公司 | X-ray image classification and identification method for judging crimping defects of high-voltage strain clamp |
CN111027631B (en) * | 2019-12-13 | 2023-09-01 | 四川赛康智能科技股份有限公司 | X-ray image classification and identification method for judging crimping defects of high-voltage strain clamp |
CN111523423B (en) * | 2020-04-15 | 2023-08-08 | 四川赛康智能科技股份有限公司 | Power equipment identification method and device |
CN111523423A (en) * | 2020-04-15 | 2020-08-11 | 四川赛康智能科技股份有限公司 | Power equipment identification method and device |
CN111815567A (en) * | 2020-06-15 | 2020-10-23 | 国网上海市电力公司 | Automatic labeling method and device for high-recognition-rate power equipment |
CN112052622A (en) * | 2020-08-11 | 2020-12-08 | 国网河北省电力有限公司 | Defect disposal method for deep multi-view semantic document representation under cloud platform |
CN112017173A (en) * | 2020-09-02 | 2020-12-01 | 西南交通大学 | Power equipment defect detection method based on target detection network and structured positioning |
CN112434740A (en) * | 2020-11-26 | 2021-03-02 | 西北大学 | Depth learning-based Qin tomb warriors fragment classification method |
CN112991259B (en) * | 2021-01-29 | 2023-04-18 | 合肥晶合集成电路股份有限公司 | Method and system for detecting defects of semiconductor manufacturing process |
CN112991259A (en) * | 2021-01-29 | 2021-06-18 | 合肥晶合集成电路股份有限公司 | Method and system for detecting defects of semiconductor manufacturing process |
CN115277865A (en) * | 2022-07-19 | 2022-11-01 | 浪潮思科网络科技有限公司 | Display method, device and medium for device panel |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109872317A (en) | A kind of defect identification method based on power equipments defect identification learning model | |
Jiang et al. | Insulator fault detection in aerial images based on ensemble learning with multi-level perception | |
WO2022077605A1 (en) | Wind turbine blade image-based damage detection and localization method | |
CN110018524B (en) | X-ray security inspection contraband identification method based on vision-attribute | |
CN111949809A (en) | Intelligent processing method for infrared inspection data of power transmission line | |
CN104866829A (en) | Cross-age face verify method based on characteristic learning | |
CN111144325A (en) | Fault identification and positioning method, device and equipment for power equipment of transformer substation | |
CN111652835A (en) | Method for detecting insulator loss of power transmission line based on deep learning and clustering | |
CN107358259A (en) | Covering ice for insulator detection method based on GLOH descriptions and GVF Snake models | |
CN111831430A (en) | Electrical equipment defect identification system based on edge calculation | |
CN115719475B (en) | Three-stage trackside equipment fault automatic detection method based on deep learning | |
CN116824517B (en) | Substation operation and maintenance safety control system based on visualization | |
CN110427943A (en) | A kind of intelligent electric meter technique for partitioning based on R-CNN | |
Li et al. | PEDNet: A lightweight detection network of power equipment in infrared image based on YOLOv4-Tiny | |
CN113205039A (en) | Power equipment fault image identification and disaster investigation system and method based on multiple DCNNs | |
CN116681962A (en) | Power equipment thermal image detection method and system based on improved YOLOv5 | |
Lu et al. | Thermal Fault Diagnosis of Electrical Equipment in Substations Based on Image Fusion. | |
Zu et al. | Detection of common foreign objects on power grid lines based on Faster R-CNN algorithm and data augmentation method | |
CN106951924B (en) | Seismic coherence body image fault automatic identification method and system based on AdaBoost algorithm | |
CN110555384A (en) | Beef marbling automatic grading system and method based on image data | |
Wang et al. | High-Voltage Transmission Line Foreign Object and Power Component Defect Detection Based on Improved YOLOv5 | |
CN113762115B (en) | Distribution network operator behavior detection method based on key point detection | |
CN112508946B (en) | Cable tunnel anomaly detection method based on antagonistic neural network | |
CN114220084A (en) | Distribution equipment defect identification method based on infrared image | |
CN114170449A (en) | Artificial intelligence image recognition device based on degree of depth 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 | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190611 |
|
RJ01 | Rejection of invention patent application after publication |