CN108537780A - A kind of insulator breakdown detection method based on the full convolutional neural networks of second order - Google Patents
A kind of insulator breakdown detection method based on the full convolutional neural networks of second order Download PDFInfo
- Publication number
- CN108537780A CN108537780A CN201810260456.XA CN201810260456A CN108537780A CN 108537780 A CN108537780 A CN 108537780A CN 201810260456 A CN201810260456 A CN 201810260456A CN 108537780 A CN108537780 A CN 108537780A
- Authority
- CN
- China
- Prior art keywords
- image
- fcn
- rebuild
- marker
- order
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0004—Industrial image inspection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20084—Artificial neural networks [ANN]
Landscapes
- Engineering & Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Quality & Reliability (AREA)
- Image Analysis (AREA)
Abstract
A kind of insulator breakdown detection method based on the full convolutional neural networks of second order, is first marked to obtain a large amount of label images insulation subregion, is learnt to characteristics of image using single order FCN, and then is split to complex background insulation subregion；Then operation Optimized Segmentation is rebuild using mathematical morphology as a result, to obtain the accurate positionin of insulation subregion；Based on the segmentation result of region insulation, insulator breakdown region detection is carried out using second-order F CN networks, realizes the accurate positionin of fault zone, characteristics of image need not artificially be extracted, calculation amount is effectively reduced, and can effectively inhibit the interference of complex background, improves the accuracy rate of insulator breakdown identification.
Description
Technical field
The invention belongs to image segmentation and processing technology fields, more particularly to a kind of to be based on the full convolutional neural networks of second order
Insulator breakdown detection method.
Background technology
In order to ensure the safe and reliable operation of whole transmission line of electricity, it is necessary to timely and effectively to electric transmission line isolator into
Row inspection simultaneously finds to debug.With the development of intelligent power grid technology, unmanned plane inspection technology is in polling transmission line
Using increasingly maturation, insulator image recognition of taking photo by plane becomes the important evidence for judging transmission line of electricity operating status.Currently, right
Has related research result in insulator image failure detection of taking photo by plane, traditional method is broadly divided into following two categories, Yi Leishi
Threshold segmentation method based on profile, color, texture and form；Another kind of is the insulator segmentation based on supervised learning.
And there is scholar to propose insulator breakdown recognition methods of the third class based on deep learning.
In the first kind point based in the dividing method of threshold value, pass through the profile of extraction insulation subgraph, color, texture spy
Sign carries out insulator breakdown detection.Li Bing-feng et al. are proposed by original image binaryzation, to insulator position
It is positioned, extracts the insulator contour in bianry image.However, this method can only detect the overall profile of insulator, no
The details profile that insulator can be extracted, is unfavorable for the fault identification in later stage.Therefore, woods, which amasses great fortunes et al., proposes to be based on coloured image
Glass insulator defect diagnostic method colorful insulation subgraph is transformed into using glass insulator bluish-green color characteristic
HSV vision metric spaces identify insulator breakdown by connected region of the search comprising bluish-green pixel, however this method is to ring
Border variation is more sensitive, and is only applicable to glass insulator, and versatility is poor.Therefore, Yang Cuiru et al. proposes to utilize gray scale
Co-occurrence matrix extracts the textural characteristics of insulator to realize that insulator identifies, although this method avoids environmental factor to insulation
The influence of subcharacter can realize the detection of insulator breakdown, but carry out insulator breakdown knowledge only in accordance with single textural characteristics
Not, relatively low for the unconspicuous insulator discrimination of texture.In this regard, Jiang Yun soil et al. proposes insulation based on multi-feature fusion
Sub- fault identification, to detect insulator breakdown, improves event by merging the various features such as profile, color, texture and form
Hinder recognition accuracy, however this method is more sensitive to parameter.
For insulator dividing method of second class based on supervised learning, it is single at et al. propose combining form feature
It is detected with the defects of insulator of BP neural network.Compared to unsupervised learning, there is very strong adaptive learning capacity.Although should
Method learns the feature of insulator from training sample, realizes the Classification and Identification of insulator, but this method convergence rate is slower,
It is easily trapped into local optimum.Therefore, Cheng Haiyan et al. is trained using the not bending moment of the insulation subgraph of positive negative sample
AdaBoost graders.The positioning and identification that insulator is realized using AdaBoost graders, multiple Weak Classifiers are summed into
Strong classifier, to improve nicety of grading.Although supervised learning can effectively promote insulator breakdown Detection accuracy, still need to
Very important person is extraction insulator characteristics of image, and when image background is complicated and changeable, Feature Descriptor is difficult to effectively extract image spy
Sign, causes fault recognition rate low.
For insulator breakdown recognition methods of the third class based on deep learning, CNN is applied to insulator by Chen Qing et al.
In fault identification, compare above two traditional methods, CNN can AUTOMATIC ZONING learnt, wherein the perception of shallower convolutional layer
Domain is smaller, for carrying out the study of local features；Deeper convolutional layer perception domain is big, for learning the height being more abstracted
Layer semantic feature, deep layer convolutional network consider part and the Global Information of image.Deeper convolutional layer learns to arrive simultaneously
Abstract characteristics it is lower to sensibility such as size, the position and direction of object, to contribute to insulator breakdown to identify.Although
In this way, but this method is classified to image pixel, using the image block around the pixel as the input of CNN
It is trained and predicts, obtain segmentation result indirectly.Since there are redundancies in adjacent pixel blocks, lead to image input data amount
It is excessive, processing speed is slower；Neighborhood of pixels size is difficult to determination, can not consider the spatial positional information in image simultaneously, no
It can identify that the specific profile of object is difficult to Accurate Segmentation well.In this regard, Jonathan Long et al. propose base
In the dividing method of full convolutional neural networks (FCN), traditional CNN dividing methods are compared, FCN is a kind of network end to end,
Input picture size is not limited, and dense prediction can be carried out without containing full articulamentum, generates arbitrary size
Divide collection of illustrative plates, improves processing speed.But since pond layer experiences the expansion in the visual field in FCN networks, cause target location etc. thin
The loss for saving information, for the insulation subgraph of complex background, when directly carrying out fault detect, it is difficult to it is exhausted that region be accurately positioned
Edge, it is foreground to cause part background flase drop, to be difficult to effectively promote the fault identification accuracy rate of insulator.
To sum up, low-level image feature extraction and grader choosing of traditional insulator breakdown detection method dependent on image
It selects, effective fault detect, classical deep learning (CNN, FCN) is difficult to realize to the insulation subgraph with complex background
Method, it is larger by complex background interference, discrimination is promoted limited.
Invention content
In order to overcome the disadvantages of the above prior art, the purpose of the present invention is to provide one kind based on the full convolution god of second order
Insulator breakdown detection method through network, need not artificially extract characteristics of image, effectively reduce calculation amount, and can be effective
Inhibit the interference of complex background, improves the accuracy rate of insulator breakdown identification.
In order to achieve the above object, the technical solution that the present invention takes is：
A kind of insulator breakdown detection method based on the full convolutional neural networks of second order, includes the following steps：
Step 1：The insulation subgraph of taking photo by plane of input complex background, is 400 × 600 sizes by image normalization；
Step 2：Full convolution (FCN) network is initialized, wherein convolution kernel size is 3 × 3, learning rate 10-14, iteration time
Number is 100,000 times；
Step 3：Training set, training set label, test set and test set label input single order FCN networks are instructed
Practice, test；
Step 4：Re-establishing filter is carried out to the insulation subgraph of primary segmentation and obtains region insulation subgraph；
Step 5：Image after re-establishing filter is multiplied with artwork to obtain the region insulation subgraph of removal background；
Step 6：Region insulation subgraph and faulty tag are trained and are tested as the input of second-order F CN networks；
Step 7：Export insulator breakdown testing result.
The FCN networks of the step 3, which are trained, to be comprised the following processes：
(a) FCN network trainings propagated forward process：
The propagated forward process of FCN calculates training sample into the reality output after successively transmitting, specific calculating is public
Formula is as follows：
zl+1=wl+1αl+bl+1
αl+1=f (zl+1)
Wherein, l is the network number of plies, zl+1For the weighting input of l+1 layers of neuron, α is each layer of respective image
Input data, w, b are the weight and biasing of each layer of neuron of full convolutional neural networks, and f is linear correction function ReLU.
(b) FCN network trainings back-propagation process：
The back transfer process of the back-propagation process of FCN, as gradient or error, specific formula for calculation are as follows：
Wherein, J (w, b) is object function, and m is number of samples, hw,bxiFor standard output, yiFor predicted value, by random
Gradient descent method finds optimal w, b and so that object function is minimum.
Steps are as follows for the specific implementation of the step 4：
(a) it initializes；B1It is the circular configuration element that size is (2i+1) × (2i+1), i=1, k=1；Wherein i is round
The radius of structural element, k are to rebuild number；
(b) definition mask image is fmask, tag image fmarker；The circular configuration element of reconstructionInput picture is f,
fmask=f
fmarker=f Θ Bi
h1=fmarker
Wherein, fmarkerThe tag image rebuild for first bite is denoted as h1；
(c) corrode restructing operation, specific formula for calculation is as follows：
Wherein, hkAfter being rebuild for kth time corrosion as a result, as+1 tag image rebuild of kth, hk+1For kth+1 time
Result after corrosion is rebuild；
(d) judge, if hk+1=hk, then corrosion reconstructed results f is obtainedε=hk；Otherwise, k=k+1, return to step (c)；
(e) transfer mask image fmaskWith tag image fmarker；I=1, k=1；
f′mask=(fε)c
f′marker=(fε)cΘBi
h′1=f 'marker
Wherein, (fε)cFor fεComplementary operation, f 'markerFor the tag image that first expansion is rebuild, it is denoted as h '1；
(f) operation is rebuild in expansion, and specific formula for calculation is as follows：
Wherein, h 'kAfter being rebuild for kth time expansion as a result, as+1 tag image rebuild of kth, h 'k+1For kth+1
Result after secondary expansion is rebuild；
(g) judge, if h 'k+1=h 'k, then reconstructed results f is obtained outrec；Otherwise, 1 k=k+, return to step (f).
The present invention having the beneficial effect that compared with prior art：
The problem of needing artificial extraction feature, selection sort device for conventional insulators fault detection method, the present invention adopts
Second-order F CN Fault Models are capable of the validity feature of AUTOMATIC ZONING extraction insulation subgraph, need not artificially be extracted
Characteristics of image while reducing calculation amount, improves fault identification accuracy rate；Hold for classical CNN, FCN dividing method
It is easily interfered by complex background, by the defect that flase drop is insulation subregion, the present invention is partitioned into single order FCN exhausted part background
Edge subgraph carries out mathematical morphology reconstruction, can effectively filter out the interference of complex background, obtains the accurate fixed of insulation subregion
Position, to improve fault identification accuracy rate.Present invention could apply in electric power networks transmission line malfunction inspection, be real
Existing electric power networks intelligence based theoretical.
Description of the drawings
Fig. 1 (a) is the insulator test image 1 in present invention experiment.
Fig. 1 (b) is the failure segmentation result to the subgraph 1 that insulate using control methods CNN.
Fig. 1 (c) is to be superimposed the failure segmentation result for the subgraph 1 that insulate under CNN methods with artwork.
Fig. 1 (d) is the failure segmentation result to the subgraph 1 that insulate using control methods FCN.
Fig. 1 (e) is to be superimposed the failure segmentation result for the subgraph 1 that insulate under FCN methods with artwork.
Fig. 1 (f) is the failure segmentation result to the subgraph 1 that insulate using the method for the present invention.
Fig. 1 (g) is to be superimposed the failure segmentation result for the subgraph 1 that insulate under the method for the present invention with artwork.
Fig. 2 (a) is the insulator test image 2 in present invention experiment.
Fig. 2 (b) is the failure segmentation result to the subgraph 2 that insulate using control methods CNN.
Fig. 2 (c) is to be superimposed the failure segmentation result for the subgraph 2 that insulate under CNN methods with artwork.
Fig. 2 (d) is the failure segmentation result to the subgraph 2 that insulate using control methods FCN.
Fig. 2 (e) is to be superimposed the failure segmentation result for the subgraph 2 that insulate under FCN methods with artwork.
Fig. 2 (f) is the failure segmentation result to the subgraph 2 that insulate using the method for the present invention.
Fig. 2 (g) is to be superimposed the failure segmentation result for the subgraph 2 that insulate under the method for the present invention with artwork.
Specific implementation mode
The present invention is described in further details with reference to the accompanying drawings and examples.
Embodiment one：
In order to test validity and superiority of the present invention to color images, experimental situation of the present invention is IW4206-
2Q deep learning work stations, 16.0464 bit manipulation systems of Ubuntu, 62.8GB memories, NVIDIA GeForce GTX1080*
2 video cards, CPU E5-1602V4 finally realize building for second order fcn network models under caffe deep learning frames.
The primary segmentation that insulation subregion is carried out first with single order FCN networks, for the insulation subgraph that is partitioned into
Line number Morphology Remodeling filters, and obtains region insulation；On the basis of region insulation subgraph event is carried out using second-order F CN
Barrier identification.Steps are as follows for specific implementation：
Step 1：The insulation subgraph of taking photo by plane of input complex background, is 400 × 600 sizes by image normalization.
Step 2：Full convolution (FCN) network is initialized, wherein convolution kernel size is 3 × 3, learning rate 10-14, iteration time
Number is 100,000 times.
Step 3：Training set, training set label, test set and test set label input single order FCN networks are instructed
Practice, test, detailed process is as follows：
(a) FCN network trainings propagated forward process：
The propagated forward process of FCN calculates training sample into the reality output after successively transmitting, specific calculating is public
Formula is as follows：
zl+1=wl+1αl+bl+1
αl+1=f (zl+1)
Wherein, l is the network number of plies, zl+1For the weighting input of l+1 layers of neuron, α is each layer of respective image
Input data, w, b are the weight and biasing of each layer of neuron of full convolutional neural networks, and f is linear correction function ReLU.
(b) FCN network trainings back-propagation process：
The back transfer process of the back-propagation process of FCN, as gradient or error, specific formula for calculation are as follows：
Wherein, J (w, b) is object function, and m is number of samples, hw,bxiFor standard output, yiFor predicted value, by random
Gradient descent method finds optimal w, b and so that object function is minimum.
Step 4：Re-establishing filter is carried out to the insulation subgraph of primary segmentation and obtains region insulation subgraph, specific steps
It is as follows：
(a) it initializes；B1It is the circular configuration element that size is (2i+1) × (2i+1), i=1, k=1；Wherein i is round
The radius of structural element, k are to rebuild number；
(b) definition mask image is fmask, tag image fmarker；For the circular configuration element of reconstructionInput picture is f,
fmask=f
fmarker=f Θ Bi
h1=fmarker
Wherein, fmarkerThe tag image rebuild for first bite is denoted as h1；
(c) corrode restructing operation, specific formula for calculation is as follows：
Wherein, hkAfter being rebuild for kth time corrosion as a result, as+1 tag image rebuild of kth, hk+1For kth+1 time
Result after corrosion is rebuild；
(d) judge, if hk+1=hk, then corrosion reconstructed results f is obtainedε=hk；Otherwise, k=k+1, return to step (c)；
(e) transfer mask image fmaskWith tag image fmarker；I=1, k=1；
f′mask=(fε)c
f′marker=(fε)cΘBi
h′1=f 'marker
Wherein, (fε)cFor fεComplementary operation, f 'markerFor the tag image that first expansion is rebuild, it is denoted as h '1；
(f) operation is rebuild in expansion, and specific formula for calculation is as follows：
Wherein, h 'kAfter being rebuild for kth time expansion as a result, as+1 tag image rebuild of kth, h 'k+1For kth+1
Result after secondary expansion is rebuild；
(g) judge, if h 'k+1=h 'k, then reconstructed results frec is obtained out；Otherwise, 1 k=k+, return to step (f).
Step 5：Image after re-establishing filter is multiplied with artwork to obtain the region insulation subgraph of removal background.
Step 6：Region insulation subgraph and faulty tag are trained and are tested as the input of second-order F CN networks.
Step 7：Export insulator breakdown testing result.
Referring to Fig. 1 (a), the insulator test image 1 in present invention experiment.Utilize three kinds of control methods：CNN, FCN and
The method of the present invention is respectively split test image 1.In emulation experiment, for unified parameters, CNN dividing methods, FCN
Dividing method is all made of the convolution kernel of 3 × 3 sizes, learning rate 10-14, iterations are 100,000 times.Experimental result is referring to figure
2 (b) and 2 (c), due to CNN dividing methods, be using prediction pixel around image block as the input of CNN be trained with
Prediction, obtains segmentation result indirectly；Since there are redundancies, neighborhood of pixels size to be difficult to determine in adjacent pixel blocks, cannot consider
To spatial positional information, failure segmentation result is coarse, especially at two fault distance relatively when, be easy by flase drop be
Failure at one.
Referring to Fig. 2 (d) and Fig. 2 (e), FCN dividing methods, have merged high dimensional feature and low-level feature information in contrast,
Segmentation precision is improved, but since pond layer experiences the expansion in the visual field, leads to the loss of the detailed information such as target location, part
Background is faulty component by flase drop.
Referring to Fig. 2 (f), in contrast, the method for the present invention rebuilds operation using mathematical morphology and filters out single order FCN segmentations
As a result the complex background in, to obtain the accurate positionin of insulation subregion；Based on the segmentation result of region insulation, utilize
Second-order F CN networks carry out insulator breakdown region detection, realize the accurate positionin of fault zone, divide with classical CNN, FCN
Method is compared, and the interference of complex background can be effectively inhibited, and promotes insulator breakdown recognition accuracy.
Claims (3)
1. a kind of insulator breakdown detection method based on the full convolutional neural networks of second order, which is characterized in that include the following steps：
Step 1：The insulation subgraph of taking photo by plane of input complex background, is 400 × 600 sizes by image normalization；
Step 2：Full convolution (FCN) network is initialized, wherein convolution kernel size is 3 × 3, learning rate 10-14, iterations are
100000 times；
Step 3：Training set, training set label, test set and test set label input single order FCN networks are trained, are surveyed
Examination；
Step 4：Re-establishing filter is carried out to the insulation subgraph of primary segmentation and obtains region insulation subgraph；
Step 5：Image after re-establishing filter is multiplied with artwork to obtain the region insulation subgraph of removal background；
Step 6：Region insulation subgraph and faulty tag are trained and are tested as the input of second-order F CN networks；
Step 7：Export insulator breakdown testing result.
2. a kind of insulator breakdown detection method based on the full convolutional neural networks of second order according to claim 1, special
Sign is that the FCN networks of the step 3, which are trained, to be comprised the following processes：
(a) FCN network trainings propagated forward process：
The propagated forward process of FCN calculates training sample into the reality output after successively transmitting, specific formula for calculation is such as
Under：
zl+1=wl+1αl+bl+1
αl+1=f (zl+1)
Wherein, l is the network number of plies, zl+1For the weighting input of l+1 layers of neuron, α is each layer of input number of respective image
According to w, b are the weight and biasing of each layer of neuron of full convolutional neural networks, and f is linear correction function ReLU.
(b) FCN network trainings back-propagation process：
The back transfer process of the back-propagation process of FCN, as gradient or error, specific formula for calculation are as follows：
Wherein, J (w, b) is object function, and m is number of samples, hw,bxiFor standard output, yiFor predicted value, pass through stochastic gradient
Descent method finds optimal w, b and so that object function is minimum.
3. a kind of insulator breakdown detection method based on the full convolutional neural networks of second order according to claim 1, special
Sign is that steps are as follows for the specific implementation of the step 4：
(a) it initializes；B1It is the circular configuration element that size is (2i+1) × (2i+1), i=1, k=1；Wherein i circular configurations member
The radius of element, k are to rebuild number；
(b) definition mask image is fmask, tag image fmarker；The circular configuration element of reconstructionInput picture is f,
fmask=f
fmarker=f Θ Bi
h1=fmarker
Wherein, fmarkerThe tag image rebuild for first bite is denoted as h1；
(c) corrode restructing operation, specific formula for calculation is as follows：
Wherein, hkAfter being rebuild for kth time corrosion as a result, as+1 tag image rebuild of kth, hk+1For+1 corrosion weight of kth
Build rear result；
(d) judge, if hk+1=hk, then corrosion reconstructed results f is obtainedε=hk；Otherwise, k=k+1, return to step (c)；
(e) transfer mask image fmaskWith tag image fmarker；I=1, k=1；
f′mask=(fε)c
f′marker=(fε)cΘBi
h′1=f 'marker
Wherein, (fε)cFor fεComplementary operation, f 'markerFor the tag image that first expansion is rebuild, it is denoted as h '1；
(f) operation is rebuild in expansion, and specific formula for calculation is as follows：
Wherein, h 'kAfter being rebuild for kth time expansion as a result, as+1 tag image rebuild of kth, h 'k+1For+1 expansion of kth
Result after reconstruction；
(g) judge, if h 'k+1=h 'k, then reconstructed results f is obtained outrec；Otherwise, k=k+1, return to step (f).
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810260456.XA CN108537780A (en) | 2018-03-27 | 2018-03-27 | A kind of insulator breakdown detection method based on the full convolutional neural networks of second order |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810260456.XA CN108537780A (en) | 2018-03-27 | 2018-03-27 | A kind of insulator breakdown detection method based on the full convolutional neural networks of second order |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108537780A true CN108537780A (en) | 2018-09-14 |
Family
ID=63485301
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810260456.XA Pending CN108537780A (en) | 2018-03-27 | 2018-03-27 | A kind of insulator breakdown detection method based on the full convolutional neural networks of second order |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108537780A (en) |
Cited By (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109377483A (en) * | 2018-09-30 | 2019-02-22 | 云南电网有限责任公司普洱供电局 | Porcelain insulator crack detecting method and device |
| CN110008901A (en) * | 2019-04-04 | 2019-07-12 | 天津工业大学 | A kind of insulator breakdown identification and localization method based on Mask R-CNN |
| CN110148136A (en) * | 2019-04-10 | 2019-08-20 | 南方电网科学研究院有限责任公司 | Insulator image segmentation method and device and computer readable storage medium |
| CN111289854A (en) * | 2020-02-26 | 2020-06-16 | 华北电力大学 | Insulator insulation state evaluation method of 3D-CNN and LSTM based on ultraviolet video |
| CN111598778A (en) * | 2020-05-13 | 2020-08-28 | 云南电网有限责任公司电力科学研究院 | Insulator image super-resolution reconstruction method |
| CN112183667A (en) * | 2020-10-31 | 2021-01-05 | 哈尔滨理工大学 | Insulator fault detection method in cooperation with deep learning |
| CN112233092A (en) * | 2020-10-16 | 2021-01-15 | 广东技术师范大学 | Deep learning method for intelligent defect detection of unmanned aerial vehicle power inspection |
| CN112434695A (en) * | 2020-11-20 | 2021-03-02 | 哈尔滨市科佳通用机电股份有限公司 | Upper pull rod fault detection method based on deep learning |
| CN112906620A (en) * | 2021-03-09 | 2021-06-04 | 唐山职业技术学院 | Unmanned aerial vehicle-assisted insulator fault detection method and device and electronic equipment |
| CN114418964A (en) * | 2021-12-28 | 2022-04-29 | 广东电网有限责任公司 | Insulator defect detection method and system based on local rotation feature learning |
| CN117422935A (en) * | 2023-12-13 | 2024-01-19 | 深圳市鑫思诚科技有限公司 | Motorcycle fault non-contact diagnosis method and system |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103336224A (en) * | 2013-07-03 | 2013-10-02 | 同济大学 | Complex information based insulator temperature rise fault comprehensive diagnosis method |
| CN107680090A (en) * | 2017-10-11 | 2018-02-09 | 电子科技大学 | Based on the electric transmission line isolator state identification method for improving full convolutional neural networks |
-
2018
- 2018-03-27 CN CN201810260456.XA patent/CN108537780A/en active Pending
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103336224A (en) * | 2013-07-03 | 2013-10-02 | 同济大学 | Complex information based insulator temperature rise fault comprehensive diagnosis method |
| CN107680090A (en) * | 2017-10-11 | 2018-02-09 | 电子科技大学 | Based on the electric transmission line isolator state identification method for improving full convolutional neural networks |
Non-Patent Citations (1)
| Title |
|---|
| 房友盼: "基于图像识别的实木板材优选系统研究", 《中国优秀博硕士学位论文全文数据库(硕士) 工程科技Ⅰ辑》 * |
Cited By (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109377483A (en) * | 2018-09-30 | 2019-02-22 | 云南电网有限责任公司普洱供电局 | Porcelain insulator crack detecting method and device |
| CN110008901A (en) * | 2019-04-04 | 2019-07-12 | 天津工业大学 | A kind of insulator breakdown identification and localization method based on Mask R-CNN |
| CN110148136A (en) * | 2019-04-10 | 2019-08-20 | 南方电网科学研究院有限责任公司 | Insulator image segmentation method and device and computer readable storage medium |
| CN111289854A (en) * | 2020-02-26 | 2020-06-16 | 华北电力大学 | Insulator insulation state evaluation method of 3D-CNN and LSTM based on ultraviolet video |
| CN111598778A (en) * | 2020-05-13 | 2020-08-28 | 云南电网有限责任公司电力科学研究院 | Insulator image super-resolution reconstruction method |
| CN111598778B (en) * | 2020-05-13 | 2023-11-21 | 云南电网有限责任公司电力科学研究院 | Super-resolution reconstruction method for insulator image |
| CN112233092A (en) * | 2020-10-16 | 2021-01-15 | 广东技术师范大学 | Deep learning method for intelligent defect detection of unmanned aerial vehicle power inspection |
| CN112183667B (en) * | 2020-10-31 | 2022-06-14 | 哈尔滨理工大学 | Insulator fault detection method in cooperation with deep learning |
| CN112183667A (en) * | 2020-10-31 | 2021-01-05 | 哈尔滨理工大学 | Insulator fault detection method in cooperation with deep learning |
| CN112434695A (en) * | 2020-11-20 | 2021-03-02 | 哈尔滨市科佳通用机电股份有限公司 | Upper pull rod fault detection method based on deep learning |
| CN112906620A (en) * | 2021-03-09 | 2021-06-04 | 唐山职业技术学院 | Unmanned aerial vehicle-assisted insulator fault detection method and device and electronic equipment |
| CN114418964A (en) * | 2021-12-28 | 2022-04-29 | 广东电网有限责任公司 | Insulator defect detection method and system based on local rotation feature learning |
| CN117422935A (en) * | 2023-12-13 | 2024-01-19 | 深圳市鑫思诚科技有限公司 | Motorcycle fault non-contact diagnosis method and system |
| CN117422935B (en) * | 2023-12-13 | 2024-03-08 | 深圳市鑫思诚科技有限公司 | Motorcycle fault non-contact diagnosis method and system |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108537780A (en) | A kind of insulator breakdown detection method based on the full convolutional neural networks of second order | |
| CN112199993B (en) | Method for identifying transformer substation insulator infrared image detection model in any direction based on artificial intelligence | |
| CN109118479B (en) | Capsule network-based insulator defect identification and positioning device and method | |
| Zhang et al. | InsuDet: A fault detection method for insulators of overhead transmission lines using convolutional neural networks | |
| CN112183667B (en) | Insulator fault detection method in cooperation with deep learning | |
| CN108734143A (en) | A kind of transmission line of electricity online test method based on binocular vision of crusing robot | |
| CN113205063A (en) | Visual identification and positioning method for defects of power transmission conductor | |
| Savino et al. | Automated classification of civil structure defects based on convolutional neural network | |
| CN107016664A (en) | A kind of bad pin flaw detection method of large circle machine | |
| WO2023137627A1 (en) | Tumor microenvironment spatial relationship modeling system and method based on digital pathology image | |
| Chen et al. | Fault Detection of Insulators Using Second‐order Fully Convolutional Network Model | |
| CN107247952B (en) | Deep supervision-based visual saliency detection method for cyclic convolution neural network | |
| CN113837994B (en) | Photovoltaic panel defect diagnosis method based on edge detection convolutional neural network | |
| Wang et al. | Railway insulator detection based on adaptive cascaded convolutional neural network | |
| CN113205039A (en) | Power equipment fault image identification and disaster investigation system and method based on multiple DCNNs | |
| CN112580575A (en) | Electric power inspection insulator image identification method | |
| CN116486240A (en) | Application of image recognition algorithm in intelligent inspection method of unmanned aerial vehicle of power transmission line | |
| Tao et al. | Electric insulator detection of UAV images based on depth learning | |
| Zhou et al. | [Retracted] A High‐Efficiency Deep‐Learning‐Based Antivibration Hammer Defect Detection Model for Energy‐Efficient Transmission Line Inspection Systems | |
| CN114565919A (en) | Tumor microenvironment spatial relationship modeling system and method based on digital pathological image | |
| Zan et al. | Defect Identification of Power Line Insulators Based on a MobileViT‐Yolo Deep Learning Algorithm | |
| Liu et al. | Channel-Spatial attention convolutional neural networks trained with adaptive learning rates for surface damage detection of wind turbine blades | |
| Jin et al. | PCB bare board defect detection based on improved YOLOv5s | |
| Liang et al. | Multi-perspective feature collaborative perception learning network for non-destructive detection of pavement defects | |
| CN116630989A (en) | Visual fault detection method and system for intelligent ammeter, electronic equipment and storage medium |
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: 20180914 |
|
| RJ01 | Rejection of invention patent application after publication |