CN110222683A - A kind of quick-fried defect recognition localization method of electric transmission line isolator component based on depth convolutional neural networks - Google Patents
A kind of quick-fried defect recognition localization method of electric transmission line isolator component based on depth convolutional neural networks Download PDFInfo
- Publication number
- CN110222683A CN110222683A CN201910501897.9A CN201910501897A CN110222683A CN 110222683 A CN110222683 A CN 110222683A CN 201910501897 A CN201910501897 A CN 201910501897A CN 110222683 A CN110222683 A CN 110222683A
- Authority
- CN
- China
- Prior art keywords
- insulator
- image
- quick
- transmission line
- fried
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
- G06F18/232—Non-hierarchical techniques
- G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
- G06F18/23213—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions with fixed number of clusters, e.g. K-means clustering
-
- 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
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/22—Image preprocessing by selection of a specific region containing or referencing a pattern; Locating or processing of specific regions to guide the detection or recognition
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/26—Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Computing Systems (AREA)
- Software Systems (AREA)
- Molecular Biology (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Mathematical Physics (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Multimedia (AREA)
- Probability & Statistics with Applications (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Image Analysis (AREA)
Abstract
The invention belongs to information automation technical fields, disclose a kind of quick-fried defect recognition localization method of the electric transmission line isolator based on depth convolutional neural networks, inspection picture is labeled by the means artificially marked and object detection SSD model is trained, then real-time polling transmission line image is identified；The direction that gradient detection calculates insulation subregion is carried out to the insulation subregion identified, the insulation subregion being partitioned into using kmean clustering algorithm in image, correction for direction is carried out to insulation subregion and insulator foreground point is counted in a column direction, can be obtained by this way insulating regions foreground point in the horizontal direction on distribution curve, and then orient quick-fried point on insulator.Effective solution of the present invention in various resolution ratio, various complex scene images orients insulator, and quick-fried defective locations of insulator are recognized accurately, the automatic detecting ability and routing inspection efficiency of transmission line of electricity are largely improved, network system maintenance cost has been greatly reduced.
Description
Technical field
The invention belongs to power industry transmission line of electricity defect recognition technologies, are related to artificial intelligence and field of image processing.
Background technique
Insulator is that the quality of critical component its performance in transmission line of electricity system is directly related to entire network system
Safety and stability.Insulator part long-term work in transmission line of electricity suffers from solarization, drenches with rain, thunder under the natural environment in field
The erosion of the extreme natural environments such as electricity；In addition, also by continued mechanical tension, the influence of electrical flashover and own material aging.
Under the influence of the above factor, the defects of insulator will appear self-destruction, crackle and filth.If detecting and replacing not in time
It serious will endanger the safe operation of transmission line of electricity system.
Current existing quick-fried recognition methods of some insulators can only solve the problems, such as that the insulator under uniform background identifies, but
It can not solve the high-resolution application scenarios of complex background.Wang Yinli etc. proposes the connected region feature of insulator to identify insulation
Quick-fried defect of son, this method obtains preferable detection effect under conditions of front shoots insulator chain, but in oblique bat
This method will just lose the ability of detection insulator in the case where taking the photograph；After the propositions such as Xiong Jie pass through colour space transformation, use
Otsu image segmentation method segmentation insulation subregion, due to the insulation subgraph taken in a natural environment, the background of image can
The histogram that with the COLOR COMPOSITION THROUGH DISTRIBUTION of insulation subregion can be identical image at this time is being not hump as a result, in this case absolutely
The segmentation of edge will be no longer accurate；Zhao Zhenbing etc. proposes the insulation subregion extracted in Aerial Images based on NSCT method, when
Be insulator occurs in image in actual scene size be it is uncertain, i.e., insulator is in image in the image having
In shared area it is larger, the area that insulator occurs in image in some images is smaller, leads to the algorithm of edge extracting
There is no good generalization abilities.
In summary, the quick-fried chip detection method of existing insulator there is a problem of following:
(1) transmission line of electricity is often in complicated natural environment, and high mountain, the woods, meadow are throughout wherein, and illumination becomes
Change complexity, causes the image that actually photographed that there is complicated background.
(2) due to the variation of shooting angle, so that there are great changes for the shape of insulator, it is special that there is no fixed shapes
Sign.
(3) due to imaging device disunity used during inspection, cause images to be recognized that there are a variety of resolutions
The focal length of rate, camera lens is multifarious, and the depth of field of image is complicated, causes the characteristics of image difference of insulator huge.
The polling transmission line mode traditional for the inspection of defects of insulator mainly manually steps on tower observation, this method
So that routing inspection efficiency is low, cost of labor is high and larger to the threat of patrol officer.With the development of science and technology, unmanned plane
It has been inexorable trend that automation method for inspecting, which enters power transmission line inspection industry,.Therefore in Aerial Images accurate detection and point
Power components such as insulator is cut, has been a urgent problem to be solved；And existing isolator detecting algorithm cannot overcome
The influence of complex background and image multiresolution, this makes the image segmentation of insulator become problem, and insulator is divided into
Lose the success or failure that often decide defect recognition.
Summary of the invention
In view of the problems of the existing technology, the present invention provides the transmission line insulators based on depth convolutional neural networks
Quick-fried defect recognition localization method of subassembly.Quick-fried defect inspection method of the insulator that the present invention realizes the following steps are included:
The first step obtains high-resolution polling transmission line image data；
Second step is labeled insulation subregion therein by artificial method, that is, records insulation subregion and exist
Location coordinate information in high-definition picture；
Third step, the insulation sub-image data being poured in using step 2 acceptance of the bid carry out deep neural network model offline
Training；
4th step gets the insulation subregion in image by SSD model real-time detection inspection image；
5th step carries out gradient detection to insulation subregion using Sobel gradient operator, obtains the gradient map of insulator
Picture, point biggish to change of gradient in gradient image are detected to obtain the directional information of insulation subregion；
Insulator region segmentation in image is come out using Kmean clustering algorithm, gets insulation subregion by the 6th step
Mask image；
7th step, using insulator direction obtained in step 5, to the insulator mask image progress side in step 6
To correction, i.e., the major axes orientation for the subregion that insulate is rotated to and horizontal direction parallel；
8th step carries out pixel projection to the insulation subregion after progress correction for direction, obtains insulator area pixel
Distribution curve；
9th step, by judging the point on insulator pixel distribution curve, whether Normal Distribution is come to quick-fried, insulator
The position of defect is positioned.
The present invention realizes the detection to insulator in image using the machine learning method of supervised learning, in a manner of manual
It is marked to polling transmission line image is taken during unmanned plane inspection, records the position letter that insulator occurs in image
Breath.Deep neural network model is trained and is tested as monitoring data by the insulation subregion of mark.
The present invention is recorded in taken each during practical polling transmission line using deep neural network model SSD
Insulation subcharacter in kind scene；For theoretically, the neural network model of deep layer can be fitted arbitrarily complicated letter
Number assumes that we can express the insulator in any actual complex scene by deep neural network model based on this
Mode.It is higher can to obtain accuracy of identification for we after the mode using supervised learning is trained deep neural network model
Detection model.
The present invention is by carrying out gradient detection to insulation subregion, getting gradient image and carrying out non-pole to gradient image
Big value inhibits, and this strategy is effectively inhibited since light changes the influence to algorithm detection accuracy, greatly improves calculation
The stability of method.
The present invention puts these and carries out statistical analysis calculating by the maximum point of calculating insulator edge attachments gradient
The first order and second order moments to insulate in subregion out estimate the direction of insulator chain in turn, and this strategy avoids existing algorithms
The middle time space complexity for calculating insulator edge-description algorithm.
Beneficial effect
We effectively raise the detection and identity to insulator using the machine learning method for having supervision first
Can, our model reaches the accuracy of identification of insulator on the true polling transmission line image data set of unmanned plane
91%, and recall rate reached 89%；Secondly we are using gradient image non-maxima suppression and statistical learning method to exhausted
The strategy that edge carries out correction for direction can be effectively by the correction for direction of insulator chain to horizontal direction, in this way to subsequent insulation
The accurate positionin of sub quick-fried point is very helpful.The last experimental result display present invention is to quick-fried, the insulator in actual scene
The verification and measurement ratio of defect reaches 86%.
Detailed description of the invention
Fig. 1 is algorithm flow chart of the invention；
Fig. 2 is SSD modular concept schematic diagram of the present invention；
Fig. 3 is flow chart of the invention.
Specific embodiment
See Fig. 1, quick-fried defect recognition localization method of electric transmission line isolator component based on depth convolutional neural networks leads to
After crossing insulation subregion of the SSD model inspection into image, is converted by biggish point and is detected for gradient in image and is insulated
The directional information of subregion reuses kmean clustering algorithm and comes out the insulator region segmentation in image, to insulation subregion
Rotation and pixel projection are carried out, the distribution curve of insulator area pixel is obtained and then quick-fried defect of insulator is determined.
Most leading object detecting system is divided into following sections in the world at present: (1) assuming the external of object
Rectangle；(2) to each rectangle frame resampled pixel point or feature；(3) using a high quality separator to rectangle frame into
Row classification.The process of this detection has become the benchmark of object detecting method, because the work of selective search exists
Good detection effect is achieved on PASCAL VOC, MS COCO and ILSVRC data set.Although the identification essence of these methods
It is too big to spend very high but these methods for embedded system calculating intensity, in addition on some high-end hardware this
The recognition performance of a little algorithms also can only achieve the requirement close to real-time.In order to make the algorithm of object identification in the hard of lower end
The requirement that real-time can be reached above part, there has been proposed SSD model, this model carries out one shot to multiple class objects
Detect its detection speed detection technique (YOLO) leading better than before.We are exhausted into image by SSD model inspection
After edge subregion, biggish point is converted to gradient in image and is detected to obtain the directional information of insulation subregion, is reused
Kmean clustering algorithm comes out the insulator region segmentation in image, carries out rotation and pixel projection to insulation subregion, obtains
The distribution curve of insulator area pixel in turn determines quick-fried defect of insulator.
In order to detect quick-fried defect of insulator in the picture, we first have to solve the problems, such as whether are deposited in image
Judged in insulator, if in image there are insulator so we to provide the position where insulator.It is understood that
SSD object detection algorithms all show good detection performance on object detection database common at present, this algorithm
Good balance is reached in the precision of detection and the speed of detection.The working principle of SSD object identification model are as follows: (1) defeated
Each for entering an image to be identified and being identified in the training process identifies the true frame of target；(2) we pass through
The default frame small set of the different vertical-horizontal proportions of each position, example on characteristic pattern of the method for convolution to evaluate different scale
Such as the characteristic pattern of 8x 8, as shown in Fig. 2.(3) we predict each default frame shaped Offset and with it is each to be identified
The similar reliability of target class.
After the location information that there is insulation sub-goal and insulator into image by SSD object identification model inspection,
We begin to judge quick-fried state of insulator using the method for image procossing.
1) insulator foreground area is extracted
In order to further analyze defects of insulator state, it would be desirable to the insulation subregion being accurately partitioned into image.
There are many common Region Segmentation Algorithms in image procossing, as Graph cuts algorithm, algorithm of region growing, kmean clustering algorithm
Deng.Graph algorithm and algorithm of region growing are suitable for the application scenarios of interactive process, in our current application scenarios I
Accurately know there is insulation in present image, and insulator occupied area large percentage in the picture, the back of image
Scape is less complicated, and target class number is appointed as 4 i.e. using kmean algorithm image segmentation by us under these conditions can reach
Preferable segmentation effect.Because we require completely to isolate insulation subregion as far as possible, we are in rgb color space
Under clustered.
2) direction of insulation subregion is extracted
By observe after a large amount of insulator image sample datas we have found that insulation subregion in the picture exist it is big
The corner feature of amount.We calculate the gradient of image using sobel operator, extract change of gradient in image using thresholding
Biggish, these will be evenly distributed on around insulation subregion.The formula that gradient calculates is as follows:
Image can regard a planar object as, and zeroth order square and first moment can be used to calculate the weight of its shape
The heart, and second moment can be used to calculate the direction of shape.According to this theory, we can calculate area in bianry image
This direction of the direction in domain is substantially exactly the external elliptical major axes orientation in region, for the external elliptical main shaft of subgraph that insulate
Direction is the direction of sub-pieces distribution.The calculation formula of each rank square in region is as follows in binary image:
mp,q=∫ ∫ xpyqf(x,y)dxdy
Wherein:
Region direction expression formula are as follows:
Wherein mu indicates that central moment, mu11 indicate mixing first moment about the origin both horizontally and vertically, mu20 table
Show that second geometric moment in the horizontal direction, mu02 indicate second geometric moment in vertical direction.
3) k mean cluster extracts insulation subregion
It is that the image that we carry out defects of insulator identification is all derived from deep learning identification as a result, it is such we can be with
Conclude there is insulation sub-goal certainly in the picture, and occupies biggish area (being greater than 1/3) in analyzed area；Secondly exhausted
There are a large amount of periodic texture features for edge subregion.Based on it is above kmean clustering algorithm can be used according to us will be exhausted
Edge subregion is split, when we use the RGB color value of image to cluster as feature vector, due in glass material
The subregion that insulate under the conditions of the insulator of matter can appear similar to the color characteristic of background area, will appear insulation sub-district at this time
Domain is divided into background area by mistake.In order to reduce the erroneous segmentation of insulation subregion, pixel is added in we in cluster process
Original feature space is become five dimensions (R, G, B, x, y) from three-dimensional (R, G, B) by coordinate information.We obtain after handling in this way
Accurate insulator foreground area.
4) quick-fried defect location of insulator
After being partitioned into insulation subregion by kmean algorithm, we use the insulation subregion calculated in front
Deflection rotates the foreground area of insulator, and the direction of our predetermined angulars is clockwise.By insulator
After foreground area carries out correction for direction, the major axes orientation of our insulation subregion is parallel to horizontal direction or is parallel to vertical
Direction.In this way, we can be scanned the subregion that can must insulate to foreground area along line direction or column direction
Pixel distribution thickness curve, the mean value and variance of our calculated curve data again, if pixel count value of some point deviates
It is considered that the point is quick-fried point when 1 times of variance of mean value.
Claims (7)
1. a kind of quick-fried defect recognition localization method of electric transmission line isolator component based on depth convolutional neural networks, feature
It is, the described quick-fried chip detection method of insulator includes the following steps:
Step 1 obtains high-resolution polling transmission line image data；
Step 2 is labeled insulation subregion therein by artificial method, that is, records insulation subregion in high score
Location coordinate information in resolution image；
Step 3, the insulation sub-image data being poured in using step 2 acceptance of the bid carry out off-line training to deep neural network model；
Step 4 gets the insulation subregion in image by SSD model real-time detection inspection image；
Step 5 carries out gradient detection to insulation subregion using Sobel gradient operator, obtains the gradient image of insulator, right
The biggish point of change of gradient is detected to obtain the directional information of insulation subregion in gradient image；
Step 6 is come out the insulator region segmentation in image using Kmean clustering algorithm, gets covering for insulation subregion
Code image；
Step 7 carries out direction school to the insulator mask image in step 6 using insulator direction obtained in step 5
Just, i.e., the major axes orientation for the subregion that insulate is rotated to and horizontal direction parallel；
Step 8 carries out pixel projection to the insulation subregion after progress correction for direction, obtains the distribution of insulator area pixel
Curve；
Step 9, by judging the point on insulator pixel distribution curve, whether Normal Distribution is come to quick-fried defect of insulator
Position positioned.
2. quick-fried defect of a kind of electric transmission line isolator component based on depth convolutional neural networks according to claim 1
Recognition positioning method, feature exist, and allow the deep neural network model SSD sufficiently to learn using the machine learning mode for having supervision
Practise the mode expression of transmission line part insulator in the picture.
3. quick-fried defect of a kind of electric transmission line isolator component based on depth convolutional neural networks according to claim 1
Recognition positioning method, which is characterized in that online right in real time using the good depth Model of Neural Network (SSD) of off-line learning
Polling transmission line image is detected, and the location information of insulator in the picture is oriented.
4. quick-fried defect of a kind of electric transmission line isolator component based on depth convolutional neural networks according to claim 1
Recognition positioning method, which is characterized in that rapid image gradient operator Sobel is used, the gradient image of insulator local is calculated,
The biggish point of change of gradient in image is extracted using high-pass filter, obedience is uniformly distributed by these points, and big gradient point is uniform
Around insulator edges of regions；The formula that gradient calculates is as follows:
As soon as regarding image as planar object, zeroth order square and first moment can be used to calculate the center of gravity of its shape, and two
Rank square can be used to calculate the direction of shape；Go out this direction of the direction in region in bianry image reality according to this theoretical calculation
It is exactly the external elliptical major axes orientation in region in matter, is sub-pieces point for the insulation external elliptical major axes orientation of subgraph
The direction of cloth；The calculation formula of each rank square in region is as follows in binary image:
mp,q=∫ ∫ xpyqf(x,y)dxdy
Wherein:
Region direction expression formula are as follows:
Wherein mu indicates that central moment, mu11 indicate that mixing first moment about the origin both horizontally and vertically, mu20 indicate
Second geometric moment in horizontal direction, mu02 indicate second geometric moment in vertical direction.
5. quick-fried defect of a kind of electric transmission line isolator component based on depth convolutional neural networks according to claim 1
Recognition positioning method, which is characterized in that the coordinate information of pixel is added using kmean clustering algorithm and in cluster process,
Original feature space is become into five dimensions (R, G, B, x, y) from three-dimensional (R, G, B).
6. quick-fried defect of a kind of electric transmission line isolator component based on depth convolutional neural networks according to claim 1
Recognition positioning method, which is characterized in that using the distribution arrangement of the affine transformation combination insulator chain of image, to insulator chain
Major axes orientation is corrected, it is specified that the direction of angle is clockwise；The major axes orientation of insulator chain is corrected to level side
To.
7. quick-fried defect of a kind of electric transmission line isolator component based on depth convolutional neural networks according to claim 1
Recognition positioning method, which is characterized in that the shadow casting technique in vertical direction is carried out to the insulator foreground pixel after correction for direction,
Get insulator area pixel point count curve；Assuming that the pixel of insulation subregion counts Normal Distribution；So exist
Necessarily there is the feature of substantial deviation curve mean value due to the missing of foreground point in the quick-fried panel region of insulator, meets this feature
Point is exactly quick-fried position of insulator chain.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910501897.9A CN110222683A (en) | 2019-06-11 | 2019-06-11 | A kind of quick-fried defect recognition localization method of electric transmission line isolator component based on depth convolutional neural networks |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910501897.9A CN110222683A (en) | 2019-06-11 | 2019-06-11 | A kind of quick-fried defect recognition localization method of electric transmission line isolator component based on depth convolutional neural networks |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110222683A true CN110222683A (en) | 2019-09-10 |
Family
ID=67816357
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910501897.9A Pending CN110222683A (en) | 2019-06-11 | 2019-06-11 | A kind of quick-fried defect recognition localization method of electric transmission line isolator component based on depth convolutional neural networks |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110222683A (en) |
Cited By (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110632092A (en) * | 2019-10-25 | 2019-12-31 | 云南电网有限责任公司电力科学研究院 | Insulator surface contamination distribution characteristic detection method based on hyperspectral technology |
| CN111398291A (en) * | 2020-03-31 | 2020-07-10 | 南通远景电工器材有限公司 | Flat enameled electromagnetic wire surface flaw detection method based on deep learning |
| CN111784692A (en) * | 2020-08-11 | 2020-10-16 | 国网内蒙古东部电力有限公司 | Method and device for detecting insulator defects in power system and electronic equipment |
| CN112184746A (en) * | 2020-08-27 | 2021-01-05 | 西北工业大学 | Transmission line insulator defect analysis method |
| CN112395972A (en) * | 2020-11-16 | 2021-02-23 | 中国科学院沈阳自动化研究所 | Electric power system insulator string identification method based on unmanned aerial vehicle image processing |
| CN113033556A (en) * | 2021-03-31 | 2021-06-25 | 广东电网有限责任公司 | Insulator rapid distinguishing and positioning method and system based on machine vision |
| CN114266985A (en) * | 2022-01-14 | 2022-04-01 | 南方电网数字电网研究院有限公司 | Glass insulator spontaneous explosion identification method, device, equipment, medium and program product |
| CN114359286A (en) * | 2022-03-21 | 2022-04-15 | 湖南应超智能计算研究院有限责任公司 | Insulator defect identification method, device and medium based on artificial intelligence |
| CN117523324A (en) * | 2024-01-04 | 2024-02-06 | 苏州镁伽科技有限公司 | Image processing method and image sample classification method, device and storage medium |
Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101620676A (en) * | 2009-07-02 | 2010-01-06 | 浙江省电力公司 | Fast image recognition method of insulator contour |
| CN104281857A (en) * | 2013-07-08 | 2015-01-14 | 中国科学院沈阳自动化研究所 | Insulator detecting method based on orthogonal direction projection |
| CN105957081A (en) * | 2016-04-28 | 2016-09-21 | 华北电力大学（保定） | Glass insulator string dropping fault detection method |
| CN105976368A (en) * | 2016-04-28 | 2016-09-28 | 华北电力大学（保定） | Insulator positioning method |
| CN106290388A (en) * | 2016-08-03 | 2017-01-04 | 国网山东省电力公司电力科学研究院 | A kind of insulator breakdown automatic testing method |
| CN106778734A (en) * | 2016-11-10 | 2017-05-31 | 华北电力大学（保定） | A kind of insulator based on rarefaction representation falls to go here and there defect inspection method |
| WO2018086299A1 (en) * | 2016-11-11 | 2018-05-17 | 广东电网有限责任公司清远供电局 | Image processing-based insulator defect detection method and system |
| CN108596886A (en) * | 2018-04-17 | 2018-09-28 | 福州大学 | Aerial Images insulator based on deep learning falls piece fault rapid detecting method |
| CN108776145A (en) * | 2018-04-16 | 2018-11-09 | 华北电力大学（保定） | A kind of insulator falls to go here and there fault detection method and system |
| CN109215020A (en) * | 2018-08-30 | 2019-01-15 | 国网黑龙江省电力有限公司佳木斯供电公司 | Ultra-high-tension power transmission line fault recognition method based on computer vision |
| CN109615611A (en) * | 2018-11-19 | 2019-04-12 | 国家电网有限公司 | A kind of insulator self-destruction defect inspection method based on inspection image |
-
2019
- 2019-06-11 CN CN201910501897.9A patent/CN110222683A/en active Pending
Patent Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101620676A (en) * | 2009-07-02 | 2010-01-06 | 浙江省电力公司 | Fast image recognition method of insulator contour |
| CN104281857A (en) * | 2013-07-08 | 2015-01-14 | 中国科学院沈阳自动化研究所 | Insulator detecting method based on orthogonal direction projection |
| CN105957081A (en) * | 2016-04-28 | 2016-09-21 | 华北电力大学（保定） | Glass insulator string dropping fault detection method |
| CN105976368A (en) * | 2016-04-28 | 2016-09-28 | 华北电力大学（保定） | Insulator positioning method |
| CN106290388A (en) * | 2016-08-03 | 2017-01-04 | 国网山东省电力公司电力科学研究院 | A kind of insulator breakdown automatic testing method |
| CN106778734A (en) * | 2016-11-10 | 2017-05-31 | 华北电力大学（保定） | A kind of insulator based on rarefaction representation falls to go here and there defect inspection method |
| WO2018086299A1 (en) * | 2016-11-11 | 2018-05-17 | 广东电网有限责任公司清远供电局 | Image processing-based insulator defect detection method and system |
| CN108776145A (en) * | 2018-04-16 | 2018-11-09 | 华北电力大学（保定） | A kind of insulator falls to go here and there fault detection method and system |
| CN108596886A (en) * | 2018-04-17 | 2018-09-28 | 福州大学 | Aerial Images insulator based on deep learning falls piece fault rapid detecting method |
| CN109215020A (en) * | 2018-08-30 | 2019-01-15 | 国网黑龙江省电力有限公司佳木斯供电公司 | Ultra-high-tension power transmission line fault recognition method based on computer vision |
| CN109615611A (en) * | 2018-11-19 | 2019-04-12 | 国家电网有限公司 | A kind of insulator self-destruction defect inspection method based on inspection image |
Non-Patent Citations (6)
| Title |
|---|
| 朱邵成等: "基于频率调谐的绝缘子识别与定位", 《电工技术学报》 * |
| 缄默笔记: "图像处理基础知识（二）—— 中心矩求主轴方向", 《HTTPS://BLOG.CSDN.NET/HONG_FANG/ARTICLE/DETAILS/49851569》 * |
| 翟永杰等: "基于空域形态一致性特征的绝缘子串定位方法", 《中国电机工程学报》 * |
| 苏超等: "基于图像处理的绝缘子缺陷检测", 《工业控制计算机》 * |
| 袁利杰: "基于双目立体图像目标区域的三维重建", 《中国优秀硕士学位论文全文数据库信息科技辑》 * |
| 齐南珣: "输电线路巡检图像的智能缺陷检测方法研究", 《中国优秀硕士学位论文全文数据库工程科技Ⅱ辑》 * |
Cited By (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110632092A (en) * | 2019-10-25 | 2019-12-31 | 云南电网有限责任公司电力科学研究院 | Insulator surface contamination distribution characteristic detection method based on hyperspectral technology |
| CN110632092B (en) * | 2019-10-25 | 2022-05-20 | 云南电网有限责任公司电力科学研究院 | Method for detecting distribution characteristics of dirt on surface of insulator based on hyperspectral technology |
| CN111398291A (en) * | 2020-03-31 | 2020-07-10 | 南通远景电工器材有限公司 | Flat enameled electromagnetic wire surface flaw detection method based on deep learning |
| CN111784692A (en) * | 2020-08-11 | 2020-10-16 | 国网内蒙古东部电力有限公司 | Method and device for detecting insulator defects in power system and electronic equipment |
| CN112184746A (en) * | 2020-08-27 | 2021-01-05 | 西北工业大学 | Transmission line insulator defect analysis method |
| CN112395972A (en) * | 2020-11-16 | 2021-02-23 | 中国科学院沈阳自动化研究所 | Electric power system insulator string identification method based on unmanned aerial vehicle image processing |
| CN113033556A (en) * | 2021-03-31 | 2021-06-25 | 广东电网有限责任公司 | Insulator rapid distinguishing and positioning method and system based on machine vision |
| CN114266985A (en) * | 2022-01-14 | 2022-04-01 | 南方电网数字电网研究院有限公司 | Glass insulator spontaneous explosion identification method, device, equipment, medium and program product |
| CN114359286A (en) * | 2022-03-21 | 2022-04-15 | 湖南应超智能计算研究院有限责任公司 | Insulator defect identification method, device and medium based on artificial intelligence |
| CN117523324A (en) * | 2024-01-04 | 2024-02-06 | 苏州镁伽科技有限公司 | Image processing method and image sample classification method, device and storage medium |
| CN117523324B (en) * | 2024-01-04 | 2024-04-16 | 苏州镁伽科技有限公司 | Image processing method and image sample classification method, device and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110222683A (en) | A kind of quick-fried defect recognition localization method of electric transmission line isolator component based on depth convolutional neural networks | |
| CN112199993B (en) | Method for identifying transformer substation insulator infrared image detection model in any direction based on artificial intelligence | |
| CN107742093A (en) | A kind of infrared image power equipment component real-time detection method, server and system | |
| CN110288586A (en) | A kind of multiple dimensioned transmission line of electricity defect inspection method based on visible images data | |
| CN105551032B (en) | The shaft tower image capturing system and its method of a kind of view-based access control model servo | |
| CN109949209B (en) | Rope detection and removal method based on deep learning | |
| CN102831393A (en) | Rapid image recognizing method of power tower pole outline | |
| CN110807765B (en) | Suspension insulator string inclination detection method and system based on image processing | |
| CN117677969A (en) | Defect detection method and device | |
| CN115049663B (en) | Unmanned aerial vehicle-based double-light fusion insulator defect detection method and system | |
| Rao et al. | Wind turbine blade inspection based on unmanned aerial vehicle (UAV) visual systems | |
| CN114724042B (en) | Automatic detection method for zero-value insulator in power transmission line | |
| CN109598772A (en) | Based on the single defect automatic marking image data source extending method of overhead transmission line | |
| CN115147591A (en) | Transformer equipment infrared image voltage heating type defect diagnosis method and system | |
| Fang et al. | A framework of power pylon detection for UAV-based power line inspection | |
| CN114581760A (en) | Equipment fault detection method and system for machine room inspection | |
| CN113723389A (en) | Method and device for positioning strut insulator | |
| CN109104591B (en) | Image acquisition device for power equipment | |
| Di et al. | Research on Real-Time Power Line Damage Detection Method Based on YOLO Algorithm | |
| CN113139955B (en) | Photovoltaic module fault identification method and system based on double-light image | |
| CN112257514B (en) | Infrared vision intelligent detection shooting method for equipment fault inspection | |
| CN114355083A (en) | Inspection robot fault identification method and system based on artificial intelligence algorithm | |
| CN111429411B (en) | X-ray defect image sample generation method for carbon fiber composite core wire | |
| CN113486873A (en) | Transformer substation equipment inspection method and system based on big data and artificial intelligence | |
| CN113313024A (en) | Mask-based power transmission line tower bird nest small sample detection method |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| WD01 | Invention patent application deemed withdrawn after publication | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20190910 |