CN105445283A - Detection method for filthy conditions of insulator images - Google Patents
Detection method for filthy conditions of insulator images Download PDFInfo
- Publication number
- CN105445283A CN105445283A CN201610071453.2A CN201610071453A CN105445283A CN 105445283 A CN105445283 A CN 105445283A CN 201610071453 A CN201610071453 A CN 201610071453A CN 105445283 A CN105445283 A CN 105445283A
- Authority
- CN
- China
- Prior art keywords
- insulator
- sub
- image
- sample
- classifier
- 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
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/84—Systems specially adapted for particular applications
- G01N21/88—Investigating the presence of flaws or contamination
- G01N21/94—Investigating contamination, e.g. dust
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/84—Systems specially adapted for particular applications
- G01N21/88—Investigating the presence of flaws or contamination
- G01N21/8851—Scan or image signal processing specially adapted therefor, e.g. for scan signal adjustment, for detecting different kinds of defects, for compensating for structures, markings, edges
- G01N2021/8883—Scan or image signal processing specially adapted therefor, e.g. for scan signal adjustment, for detecting different kinds of defects, for compensating for structures, markings, edges involving the calculation of gauges, generating models
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/84—Systems specially adapted for particular applications
- G01N21/88—Investigating the presence of flaws or contamination
- G01N21/8851—Scan or image signal processing specially adapted therefor, e.g. for scan signal adjustment, for detecting different kinds of defects, for compensating for structures, markings, edges
- G01N2021/8887—Scan or image signal processing specially adapted therefor, e.g. for scan signal adjustment, for detecting different kinds of defects, for compensating for structures, markings, edges based on image processing techniques
Landscapes
- Physics & Mathematics (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Chemical & Material Sciences (AREA)
- Analytical Chemistry (AREA)
- Biochemistry (AREA)
- General Health & Medical Sciences (AREA)
- General Physics & Mathematics (AREA)
- Immunology (AREA)
- Pathology (AREA)
- Image Analysis (AREA)
Abstract
The invention relates to a detection method for filthy conditions of insulator images and belongs to the field of electric transmission line equipment operation condition overhaul and computer vision. By the aid of the detection method for the filthy conditions of the insulator images, the problems that detection steps are cumbersome, safety is low, and evaluation criteria lack adaptivity in detection schemes for the filthy conditions of the insulator images in the traditional technology are solved. According to the method, the insulator contamination degree is detected by training condition detection models of three sub classifiers of the insulator images through a robot learning method according to different color characteristics of the insulator images under different contamination conditions. The filthy conditions of insulators can be analyzed accurately, reliably, quickly and effectively, so that a basis is provided for cleaning insulator filth in time, and accidents of pollution flashover power outage of the insulators are reduced. The detection method is suitable for detecting the filthy conditions of the insulator images quickly and accurately.
Description
Technical field
The present invention relates to a kind of based on the filthy condition detection method of insulator image, belong to the maintenance of transmission line equipment running status and computer vision field.
Background technology
Insulator as a kind of special insulation control, its prevent electric current from returning become one of essential ground equipment in ultra-high-tension power transmission line with the vital role of support wire.The quality of insulator work state will directly affect use and the operation life of transmission line of electricity.But under insulator is installed on high-altitude open-air atmosphere usually, it in use often by the filth such as industrial dust, birds droppings attachment, can make insulator dielectric effect reduce, easily pollution flashover occurs, cause heavy economic losses.In addition, the pollution level of insulator, will directly determine what process means relevant departments take to this insulator.Therefore, if can accurately detect pollution severity of insulators situation, by the massive losses effectively reducing Insulators Used fault and bring, advance the development that national grid is built further.
At present, pollution severity of insulators assay method mainly contains equivalent salt density (ESDD) method, integration surface pollution layer conductance method, pulse counting method, Leakage Current method and insulator contamination voltage gradient method etc., wherein applying maximum is equivalent salt density method and Leakage Current method: equivalent salt density method measures insulator contamination degree by insulator surface dirt being converted into every square centimeter of method for expressing containing how many milligrams of Nacl, and the method needs to be disassembled from working line by insulator in advance.Leakage Current method is carried out filth by the time of tide relation flowed through between leakage current that the electric current of surperficial pollution layer arrival point and insulator face sudden strain of a muscle state judge by being determined at insulator under working voltage.More or less there is certain potential safety hazard in these physics class methods, and complicated operation.
In recent years, for above method Problems existing, extend a kind of insulator contamination detection method based on image procossing and pattern-recognition.These methods utilize contaminated insulator imaging surface color information and the pollution-free insulator of standard to compare usually, set up the class of pollution according to difference degree, can realize the gradation for surface pollution evaluation of particular color insulator.But because insulator is of a great variety, and insulator color of image value has notable difference under different light, and difference level evaluation standard does not have adaptivity.
Summary of the invention
Technical matters to be solved by this invention is: propose the filthy condition detection method of a kind of insulator image, solves in conventional art the problem that detecting step is loaded down with trivial details, security is low, judgment criteria does not possess adaptivity that the filthy state-detection scheme of insulator image exists.
The present invention solves the problems of the technologies described above adopted scheme:
The filthy condition detection method of a kind of insulator image, comprises the following steps:
A, the filth of training insulator image detect disaggregated model:
A1, be four kinds by pollution severity of insulators grade classification, respectively: surface cleaning, slight pollution, intermediate pollution and serious pollution;
A2, collect in steps A 1 four kinds divided insulator sample image and carry out gamma correction;
A3, insulator sample image steps A 2 obtained decompose at HIS and RGB color space respectively, obtain H, S, I, R, G, B six passage color components of insulator sample image;
A4, extract all insulator sample images at the average of described six passage color components, variance, gradient and entropy respectively, and these features extracted are formed the data set of a n*24 dimension, wherein n represents number of samples;
The validity feature of A5, extraction data centralization, and form three sub-classifier training sets, positive sample class label, by positive sample and negative sample composition, is set to 1, negative sample class label is set to-1 by each sub-classifier training set;
A6, respectively the data in three sub-classifier training sets to be trained, obtain three corresponding sub-sorter models;
B, when carrying out the filthy state-detection of insulator image, inputting insulator coloured image to be detected, utilizing three sub-classifiers trained to classify, export final classification results:
B1, pre-service is carried out to insulator coloured image to be detected;
B2, step B1 to be decomposed at HSI and RGB color space respectively through pretreated insulator coloured image, obtain H, S, I, R, G, B six passage color components of insulator coloured image;
M the feature on described six passage color components that B3, extraction insulator coloured image obtain according to Fisher criterion, forms the proper vector of a 1*m dimension;
B4, the proper vector that step B3 obtains is sent into first sub-classifier and classified, more selectively classification judgement is carried out in proper vector input second or the 3rd sub-classifier according to classification results;
B5, export insulator contamination grade according to sorter classification results.
As further optimization, in steps A 4, describedly extract all insulator sample images respectively and in the method for the average of described six passage color components, variance, gradient and entropy be:
Average:
Variance:
Gradient:
Entropy:
Wherein, t represent different color channels (value is 1,2 ..., 6), H represents the line number of sample image, and W represents the columns of sample image, and f (i, j) represents the pixel value of sample image at position (i, j) place; N represents the gray level sum of each Color Channel.
As further optimization, in steps A 5, utilize the feature of Fisher criterion to data centralization to select, extract validity feature, concrete grammar is:
Definition data centralization has n sample and belongs to C class w
1, w
2..., w
c, each class comprises n respectively
iindividual sample; Definition
with
represent inter-class variance on sample set of a kth characteristic attribute and variance within clusters respectively, expression formula is:
In formula, x
(k),
m
(k)represent sample x respectively, the average of the i-th class sample, the value of the average of all samples on a kth characteristic attribute;
Fisher criterion function for certain feature is:
Wherein, J
ffor the Fisher criterion of feature, the Fisher criterion function value of certain feature on training sample set is larger, illustrates that this characteristic attribute discrimination is better; K=24;
Calculate the Fisher criterion function value of each insulator sample image 24 features according to above formula, and sort by size, choose front m (m≤k) individual feature as the final proper vector of this insulator sample image.
As further optimization, in steps A 5, in described three sub-classifier training sets, the positive sample packages of first sub-classifier training set is containing surface cleaning and slight pollution two class, and negative sample comprises intermediate pollution and serious pollution two class; Second positive sample of sub-classifier training set only comprises surface cleaning one class, and negative sample only comprises slight pollution one class; The positive sample of the 3rd sub-classifier training set only comprises intermediate pollution one class, and negative sample only comprises serious pollution one class.
As further optimization, in steps A 6, select gaussian radial basis function kernel function to do the kernel function of support vector machine, adopt K retransposing proof method to carry out parameter optimization for three the sub-classifier training sets obtained in steps A 5, training obtains three corresponding sub-sorter models.
As further optimization, in step B1, describedly pre-service is carried out to insulator coloured image to be detected comprise: image denoising and gamma correction process.
As further optimization, in step B4, the described proper vector by step B3 acquisition is sent into first sub-classifier and is classified, more selectively classification judgement is carried out in proper vector input second or the 3rd sub-classifier according to classification results, and concrete grammar is:
First proper vector is inputted first sub-classifier,
If a. the classification results of first sub-classifier is 1, illustrate that the filthy rank of this insulator image is in surface cleaning and slight pollution two classifications, again proper vector is sent into second sub-classifier to classify, if the Output rusults of second sub-classifier is 1, illustrate that this insulator surface cleans, if the Output rusults of second sub-classifier is-1, this insulator slight pollution is described;
If the classification results of b first sub-classifier is-1, illustrate that the filthy rank of this insulator image is clean with in serious pollution two classifications in moderate, again proper vector is sent into the 3rd sub-classifier to classify, if the Output rusults of the 3rd sub-classifier is 1, this insulator intermediate pollution is described, if the Output rusults of the 3rd sub-classifier is-1, this insulator serious pollution is described.
The invention has the beneficial effects as follows: the method is according to the difference of insulator color of image feature under different pollutional condition, utilize machine learning method, by training insulator image state detection model, can fast and effeciently the filthy state (being divided into surface cleaning, slight pollution, intermediate pollution and serious pollution four class) of insulator be made and being analyzed accurately and reliably, basis is provided, to reduce insulator contamination power outage for clearing up insulator contamination in time.
Accompanying drawing explanation
Fig. 1 is training insulator image filth detection disaggregated model process flow diagram;
Fig. 2 is K retransposing checking process flow diagram;
Fig. 3 is to sample to be detected filthy state-detection classification process figure.
Embodiment
The present invention is intended to propose the filthy condition detection method of a kind of insulator image, solves in conventional art the problem that detecting step is loaded down with trivial details, security is low, judgment criteria does not possess adaptivity that the filthy state-detection scheme of insulator image exists.
In specific implementation, the filthy condition detection method of the insulator image in the present invention comprises the following steps:
1, train the filth of insulator image to detect disaggregated model, its concrete steps are as follows:
1) gradation for surface pollution divides: pollution severity of insulators grade classification is four classes by the present invention, respectively: surface cleaning, slight pollution, intermediate pollution and serious pollution.
2) sorter model training: the present invention adopts support vector cassification algorithm to detect the filthy state of insulator image, adopts " one to one " method training insulator contamination state classifier model.This model comprises three sub-classifiers, and these three sub-classifiers are except training set difference, and training step is identical.Sorter training process is as shown in Figure 1:
A, obtain the insulator sample image of clean, slight pollution, intermediate pollution and serious pollution four kind, and gamma correction is carried out to it, eliminate illumination effect; Insulator sample image levels of contamination can be determined by methods such as equivalent salt densities;
B, the insulator image obtained by step a decompose at HSI and RGB color space respectively, obtain H, S, I, R, G, B six passage color components of insulator image.Pollution severity of insulators degree is reflected on visible images, shows as the difference of color.RGB and HSI is two kinds of main color standards.RGB color standard utilizes red (Red), green (Greeb), blue (Blue) three representation in components coloured images; Hsv color standard utilizes color (Hue), saturation degree (Saturation), brightness (Intensity) three representation in components coloured images.Both are described image from different perspectives, and the present invention fully utilizes the information of two kinds of color spaces, can characterize the difference of different gradation for surface pollution insulator image more all sidedly, reduce the uncertainty identified, improve recognition accuracy.
C, extract 4 features such as average, variance, gradient, entropy of all sample images at six passage color components respectively, and form the data set of a n*24 dimension, wherein n represents number of samples.These feature interpretation gray-scale watermark of image at each color component.The computing formula of each eigenwert is as follows:
Average:
Variance:
Gradient:
Entropy:
Wherein, t represent different color channels (value is 1,2 ... 6), H represents the line number of sample image, and W represents the columns of sample image, f (i, j) represent the pixel value of sample image at position (i, j) place, N represents the gray level sum of each Color Channel.
D, Fisher criterion is utilized to select feature: can other characteristic quantity of Efficient Characterization different gradation for surface pollution insulator image difference in order to find, need to carry out feature selecting to the characteristic quantity in RGB and HSI space, reject and little feature is helped for classification, retain the feature with stronger classification capacity.Fisher criterion main thought is that to differentiate that the stronger feature of performance shows as variance within clusters little as far as possible, and inter-class variance is large as far as possible, specifically comprises:
Definition data centralization has n sample and belongs to C class w
1, w
2..., w
c, each class comprises n respectively
iindividual sample.Definition
with
represent inter-class variance on sample set of a kth characteristic attribute and variance within clusters respectively, expression formula is:
In formula, x
(k),
m
(k)represent sample x respectively, the average of the i-th class sample, the value of the average of all samples on a kth characteristic attribute.The Fisher criterion function of single feature is:
Wherein, J
ffor the Fisher criterion of feature, the Fisher criterion function value of certain feature on training sample set is larger, illustrates that this characteristic attribute discrimination is better.Herein, k=24, we can calculate the Fisher criterion function value of each insulator sample image 24 features according to above formula, and sort by size, choose front m (m≤k) individual feature as the final proper vector of insulator image.
E, according to steps d, m eigenwert of extraction step c the data obtained collection forms three different sub-classifier training sets.Training set is made up of positive sample and negative sample, wherein: the positive sample packages of first sub-classifier training set is containing surface cleaning and slight pollution two class, and negative sample comprises intermediate pollution and serious pollution two class; Second positive sample of sub-classifier training set only comprises surface cleaning one class, and negative sample only comprises slight pollution one class; The positive sample of the 3rd sub-classifier training set only comprises intermediate pollution one class, and negative sample only comprises serious pollution one class.The positive sample class label of each sub-classifier training set and test set is set to 1, and the class label of negative sample is set to-1.
F, select kernel function type, penalty factor involved by support vector machine and kernel function correlation parameter used, three that utilize step e to obtain different training sets training classifier models respectively, obtain three sub-sorter models.
In support vector machine classifier design process, usually adopt gaussian radial basis function kernel function to train, recycling K retransposing proof method carries out optimizing to penalty factor and kernel functional parameter.K retransposing proof method refers to the subset being divided into k part equal training dataset, then in searching parametric procedure, at every turn wherein according to as training data, and will will remain a data as test data by k-1 number.The MSE mean value that program obtains after carrying out k iteration is estimated to expect extensive error, then can select the parameter of one group of optimum.Parameter optimization process flow diagram as shown in Figure 2.
2, the sub-coloured image of isolated input, utilizes three sub-classifiers trained to classify, exports final classification results.Concrete steps are as follows:
1) pre-service is carried out to insulator coloured image to be detected, comprise the process such as image denoising and gamma correction.
2) by step 1) the insulator coloured image that obtains decomposes at HSI and RGB color space respectively, obtains H, S, I, R, G, B six passage color components of insulator coloured image.
3) extract m the characteristic attribute that insulator coloured image obtains according to Fisher criterion, and form the proper vector of a 1*m dimension.
4) proper vector is sent into first sub-classifier to classify, more selectively proper vector input the second (three) individual sub-classifier is carried out classification judgement according to classification results.Assorting process is as shown in Figure 3:
A, first proper vector is inputted first sub-classifier, if classification results is 1, illustrate that the filthy rank of this insulator image is in surface cleaning and slight pollution two classifications; If classification results is-1, illustrate that the filthy rank of this insulator image is clean with in serious pollution two classifications in moderate.
B, then the proper vector being categorized as 1 in a is sent into second sub-classifier and classify, if Output rusults is 1, illustrates that this insulator surface cleans, if Output rusults is-1, this insulator slight pollution is described.
C, finally send into the 3rd sub-classifier classify being categorized as-1 proper vector in a, if Output rusults is 1, this insulator intermediate pollution is described, if Output rusults is-1, this insulator serious pollution is described.
5) insulator contamination grade is exported according to sorter classification results.
Claims (7)
1. the filthy condition detection method of insulator image, is characterized in that, comprise the following steps:
A, the filth of training insulator image detect disaggregated model:
A1, be four kinds by pollution severity of insulators grade classification, respectively: surface cleaning, slight pollution, intermediate pollution and serious pollution;
A2, collect in steps A 1 four kinds divided insulator sample image and carry out gamma correction;
A3, insulator sample image steps A 2 obtained decompose at HIS and RGB color space respectively, obtain H, S, I, R, G, B six passage color components of insulator sample image;
A4, extract all insulator sample images at the average of described six passage color components, variance, gradient and entropy respectively, and these features extracted are formed the data set of a n*24 dimension, wherein n represents number of samples;
The validity feature of A5, extraction data centralization, and form three sub-classifier training sets, positive sample class label, by positive sample and negative sample composition, is set to 1, negative sample class label is set to-1 by each sub-classifier training set;
A6, respectively the data in three sub-classifier training sets to be trained, obtain three corresponding sub-sorter models;
B, when carrying out the filthy state-detection of insulator image, inputting insulator coloured image to be detected, utilizing three sub-classifiers trained to classify, export final classification results:
B1, pre-service is carried out to insulator coloured image to be detected;
B2, step B1 to be decomposed at HSI and RGB color space respectively through pretreated insulator coloured image, obtain H, S, I, R, G, B six passage color components of insulator coloured image;
M the feature on described six passage color components that B3, extraction insulator coloured image obtain according to Fisher criterion, forms the proper vector of a 1*m dimension;
B4, the proper vector that step B3 obtains is sent into first sub-classifier and classified, more selectively classification judgement is carried out in proper vector input second or the 3rd sub-classifier according to classification results;
B5, export insulator contamination grade according to sorter classification results.
2. the filthy condition detection method of a kind of insulator image as claimed in claim 1, is characterized in that, in steps A 4, describedly extracts all insulator sample images respectively and in the method for the average of described six passage color components, variance, gradient and entropy is:
Average:
Variance:
Gradient:
Entropy:
Wherein, t represent different color channels (value is 1,2 ..., 6), H represents the line number of sample image, and W represents the columns of sample image, and f (i, j) represents the pixel value of sample image at position (i, j) place; N represents the gray level sum of each Color Channel.
3. the filthy condition detection method of a kind of insulator image as claimed in claim 1, it is characterized in that, in steps A 5, utilize the feature of Fisher criterion to data centralization to select, extract validity feature, concrete grammar is:
Definition data centralization has n sample and belongs to C class w
1, w
2..., w
c, each class comprises n respectively
iindividual sample; Definition
with
represent inter-class variance on sample set of a kth characteristic attribute and variance within clusters respectively, expression formula is:
In formula, x
(k),
m
(k)represent sample x respectively, the average of the i-th class sample, the value of the average of all samples on a kth characteristic attribute;
Fisher criterion function for certain feature is:
Wherein, J
ffor the Fisher criterion of feature, the Fisher criterion function value of certain feature on training sample set is larger, illustrates that this characteristic attribute discrimination is better; K=24;
Calculate the Fisher criterion function value of each insulator sample image 24 features according to above formula, and sort by size, choose front m (m≤k) individual feature as the final proper vector of this insulator sample image.
4. the filthy condition detection method of a kind of insulator image as claimed in claim 1, it is characterized in that, in steps A 5, in described three sub-classifier training sets, the positive sample packages of first sub-classifier training set is containing surface cleaning and slight pollution two class, and negative sample comprises intermediate pollution and serious pollution two class; Second positive sample of sub-classifier training set only comprises surface cleaning one class, and negative sample only comprises slight pollution one class; The positive sample of the 3rd sub-classifier training set only comprises intermediate pollution one class, and negative sample only comprises serious pollution one class.
5. the filthy condition detection method of a kind of insulator image as claimed in claim 4, it is characterized in that, in steps A 6, gaussian radial basis function kernel function is selected to do the kernel function of support vector machine, adopt K retransposing proof method to carry out parameter optimization for three the sub-classifier training sets obtained in steps A 5, training obtains three corresponding sub-sorter models.
6. the filthy condition detection method of a kind of insulator image as claimed in claim 4, is characterized in that, in step B1, describedly carries out pre-service to insulator coloured image to be detected and comprises: image denoising and gamma correction process.
7. the filthy condition detection method of a kind of insulator image as claimed in claim 4, it is characterized in that, in step B4, the described proper vector by step B3 acquisition is sent into first sub-classifier and is classified, selectively classification is carried out in proper vector input second or the 3rd sub-classifier according to classification results again to judge, concrete grammar is:
First proper vector is inputted first sub-classifier,
If a. the classification results of first sub-classifier is 1, illustrate that the filthy rank of this insulator image is in surface cleaning and slight pollution two classifications, again proper vector is sent into second sub-classifier to classify, if the Output rusults of second sub-classifier is 1, illustrate that this insulator surface cleans, if the Output rusults of second sub-classifier is-1, this insulator slight pollution is described;
If the classification results of b first sub-classifier is-1, illustrate that the filthy rank of this insulator image is clean with in serious pollution two classifications in moderate, again proper vector is sent into the 3rd sub-classifier to classify, if the Output rusults of the 3rd sub-classifier is 1, this insulator intermediate pollution is described, if the Output rusults of the 3rd sub-classifier is-1, this insulator serious pollution is described.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610071453.2A CN105445283A (en) | 2016-02-01 | 2016-02-01 | Detection method for filthy conditions of insulator images |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610071453.2A CN105445283A (en) | 2016-02-01 | 2016-02-01 | Detection method for filthy conditions of insulator images |
Publications (1)
Publication Number | Publication Date |
---|---|
CN105445283A true CN105445283A (en) | 2016-03-30 |
Family
ID=55555733
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201610071453.2A Pending CN105445283A (en) | 2016-02-01 | 2016-02-01 | Detection method for filthy conditions of insulator images |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN105445283A (en) |
Cited By (15)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106546601A (en) * | 2016-10-14 | 2017-03-29 | 南京理工大学 | Based on the photovoltaic panel method for detecting cleaning degree that low-rank is constrained |
CN106780438A (en) * | 2016-11-11 | 2017-05-31 | 广东电网有限责任公司清远供电局 | Defects of insulator detection method and system based on image procossing |
CN107240095A (en) * | 2017-05-25 | 2017-10-10 | 武汉大学 | A kind of DC line pollution severity of insulators state recognition method based on visible images |
CN107292873A (en) * | 2017-06-29 | 2017-10-24 | 西安工程大学 | A kind of close degree detecting method of porcelain insulator ash based on color character |
CN107977959A (en) * | 2017-11-21 | 2018-05-01 | 武汉中元华电科技股份有限公司 | A kind of respirator state identification method suitable for electric operating robot |
CN108108772A (en) * | 2018-01-06 | 2018-06-01 | 天津大学 | A kind of insulator contamination condition detection method based on distribution line Aerial Images |
CN108470140A (en) * | 2018-01-27 | 2018-08-31 | 天津大学 | A kind of transmission line of electricity Bird's Nest recognition methods based on statistical nature and machine learning |
CN108470141A (en) * | 2018-01-27 | 2018-08-31 | 天津大学 | Insulator recognition methods in a kind of distribution line based on statistical nature and machine learning |
CN108596196A (en) * | 2018-05-15 | 2018-09-28 | 同济大学 | A kind of filthy state evaluating method based on insulator characteristics of image dictionary |
CN109470628A (en) * | 2018-09-29 | 2019-03-15 | 江苏新绿能科技有限公司 | Contact net insulator contamination condition detection method |
CN109685038A (en) * | 2019-01-09 | 2019-04-26 | 西安交通大学 | A kind of article clean level monitoring method and its device |
CN111398303A (en) * | 2020-03-24 | 2020-07-10 | 郑州铁路职业技术学院 | Railway power supply insulator contamination online detection system |
CN112529093A (en) * | 2020-12-21 | 2021-03-19 | 上海英十信息科技有限公司 | Method for testing mold cleaning effect based on sample dimension weighting of pre-detection weight |
CN112884720A (en) * | 2021-02-01 | 2021-06-01 | 广东电网有限责任公司广州供电局 | Distribution line pollution flashover insulator detection method and system |
CN112884039A (en) * | 2021-02-05 | 2021-06-01 | 慧目(重庆)科技有限公司 | Water body pollution identification method based on computer vision |
Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US4641967A (en) * | 1985-10-11 | 1987-02-10 | Tencor Instruments | Particle position correlator and correlation method for a surface scanner |
CN103234983A (en) * | 2013-04-27 | 2013-08-07 | 南方电网科学研究院有限责任公司 | Contamination monitoring device for electric transmission line insulators |
CN103411980A (en) * | 2013-07-23 | 2013-11-27 | 同济大学 | External insulation filth status identification method based on visible-light images |
CN104483326A (en) * | 2014-12-19 | 2015-04-01 | 长春工程学院 | High-voltage wire insulator defect detection method and high-voltage wire insulator defect detection system based on deep belief network |
CN204255879U (en) * | 2014-12-10 | 2015-04-08 | 贵州电力试验研究院 | The filthy monitoring device of a kind of passive fiber charged insulating |
CN104698008A (en) * | 2015-03-24 | 2015-06-10 | 浙江中烟工业有限责任公司 | Standard component applied to precision measurement of cigarette dirty spots and manufacturing method thereof, and measurement method |
CN104849287A (en) * | 2015-06-10 | 2015-08-19 | 国家电网公司 | Composite insulator contamination degree non-contact detection method |
-
2016
- 2016-02-01 CN CN201610071453.2A patent/CN105445283A/en active Pending
Patent Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US4641967A (en) * | 1985-10-11 | 1987-02-10 | Tencor Instruments | Particle position correlator and correlation method for a surface scanner |
CN103234983A (en) * | 2013-04-27 | 2013-08-07 | 南方电网科学研究院有限责任公司 | Contamination monitoring device for electric transmission line insulators |
CN103411980A (en) * | 2013-07-23 | 2013-11-27 | 同济大学 | External insulation filth status identification method based on visible-light images |
CN204255879U (en) * | 2014-12-10 | 2015-04-08 | 贵州电力试验研究院 | The filthy monitoring device of a kind of passive fiber charged insulating |
CN104483326A (en) * | 2014-12-19 | 2015-04-01 | 长春工程学院 | High-voltage wire insulator defect detection method and high-voltage wire insulator defect detection system based on deep belief network |
CN104698008A (en) * | 2015-03-24 | 2015-06-10 | 浙江中烟工业有限责任公司 | Standard component applied to precision measurement of cigarette dirty spots and manufacturing method thereof, and measurement method |
CN104849287A (en) * | 2015-06-10 | 2015-08-19 | 国家电网公司 | Composite insulator contamination degree non-contact detection method |
Non-Patent Citations (5)
Title |
---|
丁然: "支持向量机多类分类算法研究", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
张达 等: "基于图像信息融合的绝缘子污秽状态识别", 《系统仿真学报》 * |
李卡麟: "基于二叉树的LS-WSVM模型在早期火灾分类上的研究", 《中国优秀硕士学位论文全文数据库 工程科技II辑》 * |
王红军: "《基于知识的机电系统故障诊断与预测技术》", 31 January 2014 * |
金立军 等: "可见光图像颜色特征与支持向量机相结合的绝缘子污秽状态识别方法", 《高压电器》 * |
Cited By (22)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106546601A (en) * | 2016-10-14 | 2017-03-29 | 南京理工大学 | Based on the photovoltaic panel method for detecting cleaning degree that low-rank is constrained |
CN106780438A (en) * | 2016-11-11 | 2017-05-31 | 广东电网有限责任公司清远供电局 | Defects of insulator detection method and system based on image procossing |
CN106780438B (en) * | 2016-11-11 | 2020-09-25 | 广东电网有限责任公司清远供电局 | Insulator defect detection method and system based on image processing |
CN107240095A (en) * | 2017-05-25 | 2017-10-10 | 武汉大学 | A kind of DC line pollution severity of insulators state recognition method based on visible images |
CN107292873A (en) * | 2017-06-29 | 2017-10-24 | 西安工程大学 | A kind of close degree detecting method of porcelain insulator ash based on color character |
CN107977959A (en) * | 2017-11-21 | 2018-05-01 | 武汉中元华电科技股份有限公司 | A kind of respirator state identification method suitable for electric operating robot |
CN107977959B (en) * | 2017-11-21 | 2021-10-12 | 武汉中元华电科技股份有限公司 | Respirator state identification method suitable for electric power robot |
CN108108772A (en) * | 2018-01-06 | 2018-06-01 | 天津大学 | A kind of insulator contamination condition detection method based on distribution line Aerial Images |
CN108108772B (en) * | 2018-01-06 | 2021-08-10 | 天津大学 | Insulator pollution flashover state detection method based on aerial image of distribution line |
CN108470140A (en) * | 2018-01-27 | 2018-08-31 | 天津大学 | A kind of transmission line of electricity Bird's Nest recognition methods based on statistical nature and machine learning |
CN108470141A (en) * | 2018-01-27 | 2018-08-31 | 天津大学 | Insulator recognition methods in a kind of distribution line based on statistical nature and machine learning |
CN108470140B (en) * | 2018-01-27 | 2021-12-07 | 天津大学 | Power transmission line bird nest identification method based on statistical characteristics and machine learning |
CN108470141B (en) * | 2018-01-27 | 2021-08-10 | 天津大学 | Statistical feature and machine learning-based insulator identification method in distribution line |
CN108596196A (en) * | 2018-05-15 | 2018-09-28 | 同济大学 | A kind of filthy state evaluating method based on insulator characteristics of image dictionary |
CN108596196B (en) * | 2018-05-15 | 2021-10-08 | 同济大学 | Pollution state evaluation method based on insulator image feature dictionary |
CN109470628A (en) * | 2018-09-29 | 2019-03-15 | 江苏新绿能科技有限公司 | Contact net insulator contamination condition detection method |
CN109685038A (en) * | 2019-01-09 | 2019-04-26 | 西安交通大学 | A kind of article clean level monitoring method and its device |
CN111398303A (en) * | 2020-03-24 | 2020-07-10 | 郑州铁路职业技术学院 | Railway power supply insulator contamination online detection system |
CN112529093A (en) * | 2020-12-21 | 2021-03-19 | 上海英十信息科技有限公司 | Method for testing mold cleaning effect based on sample dimension weighting of pre-detection weight |
CN112884720A (en) * | 2021-02-01 | 2021-06-01 | 广东电网有限责任公司广州供电局 | Distribution line pollution flashover insulator detection method and system |
CN112884039A (en) * | 2021-02-05 | 2021-06-01 | 慧目(重庆)科技有限公司 | Water body pollution identification method based on computer vision |
CN112884039B (en) * | 2021-02-05 | 2022-10-21 | 慧目(重庆)科技有限公司 | Water body pollution identification method based on computer vision |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN105445283A (en) | Detection method for filthy conditions of insulator images | |
CN109389180A (en) | A power equipment image-recognizing method and inspection robot based on deep learning | |
CN105512666A (en) | River garbage identification method based on videos | |
CN108171209A (en) | A kind of face age estimation method that metric learning is carried out based on convolutional neural networks | |
CN105825511A (en) | Image background definition detection method based on deep learning | |
CN114048568B (en) | Rotary machine fault diagnosis method based on multisource migration fusion shrinkage framework | |
CN106408030A (en) | SAR image classification method based on middle lamella semantic attribute and convolution neural network | |
CN111339883A (en) | Method for identifying and detecting abnormal behaviors in transformer substation based on artificial intelligence in complex scene | |
CN106127198A (en) | A kind of image character recognition method based on Multi-classifers integrated | |
CN110705873A (en) | Novel power distribution network operation state portrait analysis method | |
CN106157323A (en) | The insulator division and extracting method that a kind of dynamic division threshold value and block search combine | |
CN103411980A (en) | External insulation filth status identification method based on visible-light images | |
CN101251896B (en) | Object detecting system and method based on multiple classifiers | |
CN106845387A (en) | Pedestrian detection method based on self study | |
CN103824092A (en) | Image classification method for monitoring state of electric transmission and transformation equipment on line | |
CN110176143A (en) | A kind of highway traffic congestion detection method based on deep learning algorithm | |
CN104463242A (en) | Multi-feature motion recognition method based on feature transformation and dictionary study | |
CN105303200A (en) | Human face identification method for handheld device | |
CN103679214A (en) | Vehicle detection method based on online area estimation and multi-feature decision fusion | |
CN115205256A (en) | Power transmission line insulator defect detection method and system based on fusion of transfer learning | |
CN106886757A (en) | A kind of multiclass traffic lights detection method and system based on prior probability image | |
CN104834891A (en) | Method and system for filtering Chinese character image type spam | |
CN116486240A (en) | Application of image recognition algorithm in intelligent inspection method of unmanned aerial vehicle of power transmission line | |
CN103903009A (en) | Industrial product detection method based on machine vision | |
CN111414855B (en) | Telegraph pole sign target detection and identification method based on end-to-end regression model |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
RJ01 | Rejection of invention patent application after publication | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20160330 |