CN108345898A - A kind of novel line insulator Condition assessment of insulation method - Google Patents
A kind of novel line insulator Condition assessment of insulation method Download PDFInfo
- Publication number
- CN108345898A CN108345898A CN201711496231.6A CN201711496231A CN108345898A CN 108345898 A CN108345898 A CN 108345898A CN 201711496231 A CN201711496231 A CN 201711496231A CN 108345898 A CN108345898 A CN 108345898A
- Authority
- CN
- China
- Prior art keywords
- insulator
- image
- dimension
- value
- gray
- 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
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2415—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
- G01R31/00—Arrangements for testing electric properties; Arrangements for locating electric faults; Arrangements for electrical testing characterised by what is being tested not provided for elsewhere
- G01R31/12—Testing dielectric strength or breakdown voltage ; Testing or monitoring effectiveness or level of insulation, e.g. of a cable or of an apparatus, for example using partial discharge measurements; Electrostatic testing
- G01R31/1218—Testing dielectric strength or breakdown voltage ; Testing or monitoring effectiveness or level of insulation, e.g. of a cable or of an apparatus, for example using partial discharge measurements; Electrostatic testing using optical methods; using charged particle, e.g. electron, beams or X-rays
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
- G01R31/00—Arrangements for testing electric properties; Arrangements for locating electric faults; Arrangements for electrical testing characterised by what is being tested not provided for elsewhere
- G01R31/12—Testing dielectric strength or breakdown voltage ; Testing or monitoring effectiveness or level of insulation, e.g. of a cable or of an apparatus, for example using partial discharge measurements; Electrostatic testing
- G01R31/1227—Testing dielectric strength or breakdown voltage ; Testing or monitoring effectiveness or level of insulation, e.g. of a cable or of an apparatus, for example using partial discharge measurements; Electrostatic testing of components, parts or materials
- G01R31/1245—Testing dielectric strength or breakdown voltage ; Testing or monitoring effectiveness or level of insulation, e.g. of a cable or of an apparatus, for example using partial discharge measurements; Electrostatic testing of components, parts or materials of line insulators or spacers, e.g. ceramic overhead line cap insulators; of insulators in HV bushings
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace 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
- G06V10/30—Noise filtering
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/56—Extraction of image or video features relating to colour
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Data Mining & Analysis (AREA)
- Multimedia (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Computation (AREA)
- Evolutionary Biology (AREA)
- General Engineering & Computer Science (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Probability & Statistics with Applications (AREA)
- Chemical & Material Sciences (AREA)
- Ceramic Engineering (AREA)
- Image Analysis (AREA)
Abstract
The present invention proposes a kind of novel line insulator Condition assessment of insulation method, line insulator visible images acquisition is carried out in conjunction with optical camera image-forming principle, simultaneously denoising has been carried out for line insulation subgraph, the image partition method being combined with morphological method using a kind of maximum variance between clusters, image live part is split, after the dimension-reduction treatment to color characteristic, the intelligent recognition of insulator contamination grade is realized using the artificial intelligence approach based on Radial Basis Probabilistic Neural Networks grader (RBPNN).The present invention method combination image procossing and information merge method insulator surface image is obtained, denoising, dividing processing, the intelligent recognition of insulator contamination grade is realized by artificial intelligence approach.
Description
Technical field
The present invention relates to a kind of novel line insulator Condition assessment of insulation methods, more particularly, to a kind of image procossing
With the line insulator state of insulation intelligent determination method of information fusion.
Background technology
With the development of industrial and mining enterprises of China and power industry, the filth discharged in air is increasing, and haze is increasingly tight
Weight so that the gradation for surface pollution in region residing for power grid is continuously improved, and equipment contamination phenomenon is on the rise.The electricity consumption of another aspect China needs
Continuous growth, installed capacity in power grid and network size is asked to be significantly expanded, voltage class is continuously improved, and grid equipment and circuit are increasingly
Increase, grid pollution flashover accident has increasingly increased trend, studies filthy cumulative process, accurately and in time early warning and pre- antifouling
Sudden strain of a muscle accident is to carry out one of the most effective means of antifouling work, particularly important to power grid security economical operation.
Currently, there are many high-tension insulator filth detection means and method both at home and abroad, including on-line measurement leakage current method, survey
It measures equivalent salt deposit density method, pollution layer conductivity method, measure flashover standard-field strength method, IR thermometry, spectroscopic methodology, microwave irradiation, purple
Outer impulse method, acoustic-emission etc., but these methods both for some filthy parameter or its derive parameter, in conjunction with relevant
Service condition measures insulator contamination situation with special detection device, and field conduct is inconvenient, fail widespread adoption in
Engineering is practical, and the country has scholar to start to inquire into using visible images progress insulator contamination state-detection, these methods are also only
It is that certain research and application are obtained in laboratory or substation's support insulator, there are no applied on line insulator.
Invention content
It can not accurately judge gradation for surface pollution to solve line insulator Condition assessment of insulation in the prior art, affect
The problem of timely, accurate early warning of pollution flashover, the present invention provide a kind of novel line insulator Condition assessment of insulation method, pass through
Great number tested data compares analysis and optical camera image-forming principle is combined to propose the basic of line insulator visible images acquisition
Principle, while denoising is carried out for line insulation subgraph, use a kind of maximum variance between clusters and morphological method phase
In conjunction with image partition method, image live part is split, after the dimension-reduction treatment to color characteristic, using based on
The artificial intelligence approach of Radial Basis Probabilistic Neural Networks grader (RBPNN) realizes the intelligent recognition of insulator contamination grade.
In order to achieve the above object, the technical solution adopted in the present invention is：A kind of novel line insulation insulating sublayer shape
State appraisal procedure, which is characterized in that include the following steps：
Step 1：Line insulator visible images acquisition is carried out in conjunction with optical camera image-forming principle, while use is based on
The small wave self-adaption insulator visible images denoising method of Bayes estimations carries out denoising to line insulation subgraph；
Step 2：In order to study insulator disk face part, unrelated image detail is removed, will be insulated first with OTSU methods
Sub-disk face region segmentation comes out, and is then filled up using morphological method to the duck eye formed by noise jamming；
Step 3：The color characteristic of image is selected using Fei Sheer criterion, absolutely by the filth based on Karhunen-Loeve transformation
The information fusion method of edge visible images characteristic parameter carries out dimension-reduction treatment, using based on Radial Basis Probabilistic Neural Networks point
The artificial intelligence approach of class device RBPNN realizes the intelligent recognition of insulator contamination grade.
Further, in step 1, line insulator visible images obtain and denoising method is as follows：
Step 1-1, when shooting, selection and the shaft tower of camera lens exhale high D and insulator structure height H to have relationship, the coke of camera lens
Calculation formula away from f is as follows：
Wherein W is the width of camera CCD sizes；
The image of insulator surface is obtained according to above method；
After obtaining the image of insulator surface, gray processing processing, the ash of processing are carried out to visible images by step 1-2
Degreeization calculation formula is as follows：
Vgray=0.3R+0.59G+0.11B (2)
Wherein R, G, B are the rgb value of visible images, VgrayFor calculated gray value, each pixel of image is adopted
Its gray-scale map is calculated with formula (2)；
Step 1-3 is exhausted to circuit using the small wave self-adaption insulator visible images denoising method estimated based on Bayes
It is as follows that edge subgraph carries out denoising method：
It is the actual value of signal for signal a Z=X+n, X by noise pollution, n is the noise of addition in the signal,
Z is that the observed value of signal is estimated signal X generally according to observed value Z, X's is optimal when signal X can not be measured directly
EstimationFor posterior probability mean value：
Wherein, pn(Z-X) be noise probability density function, p (X) is the prior probability function of signal；
It can pass throughIt acquires, when noise and signal all obey zero-mean, obeys respectivelySide
When the Gaussian Profile of difference, have：
Since noise is mainly distributed on high frequency section, retains low frequency part, high frequency section wavelet coefficient is estimated using Bayes
Then meter carries out wavelet inverse transformation to all wavelet coefficients, obtains image after denoising.
Further, in step 2, using OTSU methods, by insulator disk region segmentation, out steps are as follows：
(1) probability that each gray-scale number of image and each gray-level pixels occur is calculated；
(2) the overall gray level mean value of image, the gray average and background classes A and mesh of calculating background classes A and target class B are obtained
Mark the probability of occurrence of class B；
(3) variance between class Separation Indexes i.e. two classes of background classes A and target class B is calculated；
(4) asking makes the maximum value of variance be optimal threshold T, divides image by segmentation threshold of T, obtains insulator binary map
Picture：
(5) disk part is cut into from insulator bianry image, obtains insulator disk bianry image.
Further, in step 2, progress morphological operation fills up insulator disk, and steps are as follows：
Mathematical morphology is based on set theory, and basic operation is expansion and erosion operation, and with dilation operation
Glycerine enema, a variety of operations of closure operation and operating method are constituted with the various combination mode of erosion operation；
The characteristics of using dilation operation, erosion operation, Glycerine enema, closure operation, remove background, make insulator disk and
Details preserves complete.
Further, in step 3, carrying out selection to the color characteristic of image using Fisher criterion, steps are as follows：
Feature selecting is carried out using Fisher criterion, it is as small as possible that the stronger feature of performance shows as variance within clusters, between class
Variance is as big as possible；
For contaminated insulator characteristics of image, the variance within clusters S of i-th dimension featurew(i)With inter-class variance Sb(i)Calculation formula
It is as follows：
In formula：I is characterized component number；K is gradation for surface pollution, k=1,2 ..., K, and K is the gradation for surface pollution number that is divided；For the i-th dimension feature of the single sample of gradation for surface pollution k；wkThe sample set for being k for gradation for surface pollution；For gradation for surface pollution
For the mean value of the sample i-th dimension characteristic value of k；miFor the mean value of the i-th dimension characteristic value of all gradation for surface pollution samples；For filth
The i-th dimension characteristic value number of the sample of grade k；niFor the sum of the i-th dimension feature of all samples；
Inter-class variance Sb(i)With variance within clusters Sw(i)The ratio between be Fisher discrimination function value, i.e.,：
It is found that J(i)Value is bigger, shows that dimensional feature difference between all kinds of is bigger, classifying quality is better.
Further, in step 3, the information of the contaminated insulator visible images characteristic parameter based on Karhunen-Loeve transformation merges
Method and step is as follows：
(1) the covariance matrix C of initial characteristic data matrix X is calculated；
(2) eigenvalue λ of covariance matrix C is soughtjAnd eigenvalue λjCorresponding orthonomalization feature vector pj；
(3) by eigenvalue λjIt is ranked sequentially by descending, and makes feature vector also by being ranked sequentially accordingly；
(4) selection of feature vector number is with principal component accumulation contribution rate γkFor evaluation index,
Wherein, ρiFor principal component contributor rate, k is the dimension of quasi- selection feature vector, and M is characterized the dimension of parameter；
Work as γkWhen >=85%, the number of selected feature vector contains the main information of raw data matrix, before selection
Several maximum λjCorresponding feature vector constitutes orthogonal matrix P；
(5) pass through formula Y=PTX acquires principal component feature matrix Y after Karhunen-Loeve transformation.
Further, in step 3, the artificial intelligence approach based on Radial Basis Probabilistic Neural Networks grader RBPNN is realized
Steps are as follows for the intelligent recognition of insulator contamination grade：
For the RBPNN of N number of training sample, output is represented by following formula：
I.e.：Y=WH (12)
In formula：
Y1 Y2 … YNTo correspond to the network output valve of N number of training sample respectively, W is the second hidden layer between output layer
Weight matrix, H are the output of the second hidden layer；P is input vector, and Φ () is the Non-linear Kernel function of the first hidden layer, qiFor net
The number of nodes of i-th of classification of the second hidden layer is belonged in the first hidden layer of network；CkFor k-th of hidden centers vectors of the first hidden layer,
Dimension is identical as input vector P, | | | |2For euclidean 2- norms；σkFor the width control of k-th of hidden central Gaussian function
Parameter；
The 0th grade of filth is represented when output is 0, exports that represent I grade when being 1 filthy, exports that represent Section II grade for 2 dirty
It is dirty, insulator contamination grade is worth to according to output.
Compared with prior art, the beneficial effects of the invention are as follows：The invention has the advantages that：
1. the present invention according to line insulator from the ground farther out, ring different in the dimensional orientation and structure type of installation
Border and illumination also different feature, have found a kind of novel line insulator Condition assessment of insulation method.
2. the method that the present invention combines image procossing and information to merge obtains insulator surface image, denoising, divides
Processing is cut, the intelligent recognition of insulator contamination grade is realized by artificial intelligence approach.
Description of the drawings
The present invention will be further described with reference to the accompanying drawings and examples.
The structure flow chart of Fig. 1 present invention.
Fig. 2 shooting points are chosen and camera lens chooses schematic diagram.
The insulator surface figure visible images gray processing comparison of the scenes Fig. 3.
Noise level filter effect of Fig. 4 differences filtering algorithms in σ=15 compares.
Fig. 5 .otsu methods divide the design sketch of disk.
The design sketch of Fig. 6 morphological methods segmentation insulation subgraph.
Fig. 7 .J functional value space three-dimensional distribution maps and film color coloured silk distribution map.
Fig. 8 initial characteristic data three-dimensional distribution maps.
Principal component data three-dimensional distribution map after Fig. 9 .K-L transformation.
Specific implementation mode
For the ease of those of ordinary skill in the art understand and implement the present invention, with reference to embodiment to the present invention make into
The detailed description of one step, it should be understood that implementation example described herein is merely to illustrate and explain the present invention, and is not used to limit
The fixed present invention.
For the special circumstances of line insulator, the present invention proposes a kind of novel line insulator Condition assessment of insulation
Method, insulator surface image is obtained in conjunction with the method that image procossing and information merge, denoising, dividing processing, pass through
It is real using the artificial intelligence approach based on Radial Basis Probabilistic Neural Networks grader after the dimensionality reduction of insulator surface color characteristic
The intelligent recognition of insulator contamination grade is showed.
A kind of novel line insulator Condition assessment of insulation method, first, in conjunction with optical camera image-forming principle into line
Road insulator visible images obtain, while having carried out denoising for line insulation subgraph；Then, using a kind of maximum
The image partition method that Ostu method is combined with morphological method splits image live part；Finally, by right
After the dimension-reduction treatment of color characteristic, insulation is realized using the artificial intelligence approach based on Radial Basis Probabilistic Neural Networks grader
The intelligent recognition of sub- gradation for surface pollution.
Therefore, a kind of novel line insulator Condition assessment of insulation method of the invention, including：
Step 1：Proposing line insulator in conjunction with optical camera image-forming principle by great number tested data comparison analysis can
The basic principle that light-exposed image obtains, while using the small wave self-adaption insulator visible images denoising estimated based on Bayes
Method carries out denoising to line insulation subgraph.
Step 2：In order to study insulator disk face part, unrelated image detail is removed, will be insulated first with OTSU methods
Sub-disk face region segmentation comes out, and is then filled up using morphological method to the duck eye formed by noise jamming.
Step 3：The color characteristic of image is selected using Fisher criterion, absolutely by the filth based on Karhunen-Loeve transformation
The information fusion method of edge visible images characteristic parameter carries out dimension-reduction treatment, using based on Radial Basis Probabilistic Neural Networks point
The artificial intelligence approach of class device (RBPNN) realizes the intelligent recognition of insulator contamination grade.
Wherein, in step 1, line insulator visible images obtain and denoising method is as follows：
Step 1-1, when shooting, selection and the shaft tower of camera lens exhale high D and insulator structure height H to have relationship, the coke of camera lens
Calculation formula away from f is as follows：
Wherein W is the width of camera CCD sizes.
The image of insulator surface is obtained according to above method.
After obtaining the image of insulator surface, gray processing processing, the ash of processing are carried out to visible images by step 1-2
Degreeization calculation formula is as follows：
Vgray=0.3R+0.59G+0.11B (2)
Wherein R, G, B are the rgb value of visible images, VgrayFor calculated gray value, each pixel of image is adopted
Its gray-scale map is calculated with formula (2).
Step 1-3 uses the small wave self-adaption insulator visible images denoising method estimated based on Bayes, maximum journey
The information of image is remained to degree, method is as follows：
For one by noise pollution signal Z=X+n (X is the actual value of signal, and n be to add noise in the signal,
Z is the observed value of signal), when signal X can not be measured directly, signal X is estimated generally according to observed value Z, X's is optimal
EstimationFor posterior probability (conditional probability) mean value：
Wherein, pn(Z-X) be noise probability density function, p (X) is the prior probability function of signal.
It can pass throughIt acquires, when noise and signal all obey zero-mean, obeys respectivelySide
When the Gaussian Profile of difference, have：
Since noise is mainly distributed on high frequency section, retains low frequency part, high frequency section wavelet coefficient is estimated using Bayes
Then meter carries out wavelet inverse transformation to all wavelet coefficients, obtains image after denoising.
Wherein, in step 2, insulator disk region segmentation is come out using OTSU methods, to being formed because of noise jamming
Duck eye using morphological method complementing method it is as follows：
Step 2-1, using OTSU methods, by insulator disk region segmentation, out steps are as follows：
(1) probability that each gray-scale number of image and each gray-level pixels occur is calculated；
(2) the overall gray level mean value of image, the gray average and background classes A and mesh of calculating background classes A and target class B are obtained
Mark the probability of occurrence of class B；
(3) variance between class Separation Indexes i.e. two classes of background classes A and target class B is calculated；
(4) asking makes the maximum value of variance be optimal threshold T, divides image by segmentation threshold of T, obtains insulator binary map
Picture：
Disk part is cut into from insulator bianry image, obtains insulator disk bianry image.
Step 2-2, progress morphological operation fills up insulator disk, and steps are as follows：
Mathematical morphology is based on set theory, and basic operation is expansion and erosion operation, and with dilation operation
A variety of operations such as Glycerine enema, closure operation and operating method are constituted with the various combination mode of erosion operation.
The characteristics of using dilation operation, erosion operation, Glycerine enema, closure operation, remove background, make insulator disk and
Details preserves complete.
Wherein, in step 3, the color characteristic of image is selected using Fisher criterion, by based on Karhunen-Loeve transformation
The information fusion method of contaminated insulator visible images characteristic parameter carries out dimension-reduction treatment, using based on radial base probabilistic neural
The artificial intelligence approach of network classifier (RBPNN) realizes the intelligent recognition of insulator contamination grade, and method is as follows：
Step 3-1, carrying out selection to the color characteristic of image using Fisher criterion, steps are as follows：
Feature selecting is carried out using Fisher criterion, it is as small as possible that the stronger feature of performance shows as variance within clusters, between class
Variance is as big as possible.
For contaminated insulator characteristics of image, the variance within clusters S of i-th dimension featurew(i)With inter-class variance Sb(i)Calculation formula
It is as follows：
In formula：I is characterized component number；K is gradation for surface pollution, k=1,2 ..., K, and K is the gradation for surface pollution number that is divided；For the i-th dimension feature of the single sample of gradation for surface pollution k；wkThe sample set for being k for gradation for surface pollution；For filth etc.
Grade is the mean value of the sample i-th dimension characteristic value of k；miFor the mean value of the i-th dimension characteristic value of all gradation for surface pollution samples；For dirt
The i-th dimension characteristic value number of the sample of dirty grade k；niFor the sum of the i-th dimension feature of all samples.
Inter-class variance Sb(i)With variance within clusters Sw(i)The ratio between be Fisher discrimination function value, i.e.,：
It is found that J(i)Value is bigger, shows that dimensional feature difference between all kinds of is bigger, classifying quality is better.
Step 3-2, the information fusion method step of the contaminated insulator visible images characteristic parameter based on Karhunen-Loeve transformation is such as
Under：
(1) the covariance matrix C of initial characteristic data matrix X is calculated；
(2) eigenvalue λ of covariance matrix C is soughtjAnd eigenvalue λiCorresponding orthonomalization feature vector pj；
(3) by eigenvalue λjIt is ranked sequentially by descending, and makes feature vector also by being ranked sequentially accordingly；
(4) selection of feature vector number is with principal component accumulation contribution rate γkFor evaluation index.
Wherein, ρjFor principal component contributor rate, k is the dimension of quasi- selection feature vector, and M is characterized the dimension of parameter.
Work as γkWhen >=85%, the number of selected feature vector contains the main information of raw data matrix, before selection
Several maximum λjCorresponding feature vector constitutes orthogonal matrix P；
(6) pass through formula Y=PTX acquires principal component feature matrix Y (score vector matrix) after Karhunen-Loeve transformation.
Step 3-3, steps are as follows for the artificial intelligence approach based on Radial Basis Probabilistic Neural Networks grader (RBPNN)：
For the RBPNN of N number of training sample, output is represented by down：
In formula：
Y1 Y2 … YNTo correspond to the network output valve of N number of training sample respectively, W is the second hidden layer between output layer
Weight matrix, H are the output of the second hidden layer；P is input vector (R dimensions), and Φ () is that the Non-linear Kernel function of the first hidden layer is (high
This function), qiTo belong to the number of nodes (hidden centers vectors number) of i-th of classification of the second hidden layer in the first hidden layer of network；CkIt is
K-th of hidden centers vectors of one hidden layer, dimension is identical as input vector P, | | | |2For euclidean 2- norms；σkFor kth
The width control parameter of a hidden central Gaussian function.
The 0th grade of filth is represented when output is 0, exports that represent I grade when being 1 filthy, exports that represent Section II grade for 2 dirty
It is dirty, insulator contamination grade is worth to according to output.
The present invention can be further understood by following embodiment.
It is chosen by Fig. 2 shooting points and camera lens chooses schematic diagram, the focal length f of camera lens is obtained according to formula (1), according to annex table
Middle relationship obtains suitable shooting angle, is found by testing repeatedly and carrying out analysis to recognition result, camera and insulator disk
Face horizontal sextant angle is that 10 °~25 ° or so the upper disk surface image recognition effects shot are preferable.
After obtaining insulator surface image, gray processing pretreatment is carried out to live insulator surface image according to formula (2),
As shown in Figure 3.
Be filtered using several method as shown in figure 4, from Fig. 4 c, d, e visual contrast in it can also be seen that base
Insulator image clearly after the small echo self-adaptive solution method denoising of Bayes estimations, image detail retain completely, are better than another two
Kind method.
Insulator disk region segmentation is come out using OTSU methods, a is clean insulator original image in Fig. 5, and b is to use
The insulator disk bianry image that OTSU methods are partitioned into.As can be seen that due to the influence of shade and noise so that segmentation is not accurate enough
Really, using dilation operation, erosion operation, Glycerine enema, closure operation the characteristics of, using the structural element B as shown in formula (14)
Image after the segmentation of OTSU methods is operated, the insulator disk bianry image as shown in Fig. 6 .a is obtained, with the disk two
Value image is multiplied with filtered former insulator disk image, and the insulator disk image after being divided is as shown in Fig. 6 .b.
As seen from Figure 6, background is completely removed, and insulator disk and details preserve completely, clear-cut.
To 180 of front, totally 5 class insulator visible images seek the J values of its 36 dimensional feature component.Make 36 Wei Te
The J functional value space three-dimensional distribution maps and film color coloured silk distribution map of sign are shown in Fig. 7.There are two lofty perch in figure, respectively S components are equal
Value, S component intermediate values usually require that J functional values are more than 1, therefore select very poor R maximum values, R, R variances, the very poor 4 J functional values of G big
In 1 characteristic quantity as rgb color space characteristic quantity；Select H intermediate values, S mean values, S intermediate values, S variances, V maximum values, very poor 6 of V
Characteristic quantity of the J functional values more than 1 is as HSV color space characteristic quantities, totally 10 characteristic parameters.
The characteristic parameter of sample or very much, using Karhunen-Loeve transformation by the original filth of the insulator comprising 10 dimensional feature vectors
Data dimension is reduced to 3 dimensions.Although each filthy level characteristics data of Fig. 8 Cluster space be it is all kinds of divide, class spacing
From very little, it is all kinds of between erroneous judgement probability it is larger, by Fig. 9 data space distribution it can be seen that：Each pattern class after Karhunen-Loeve transformation
Between distance become larger, Various types of data clusters the different zones in three dimensions, and the possibility that data are judged by accident is smaller.
After extracting insulator visible images characteristic value, the insulator contamination grade based on RBPNN neural networks is used
Recognition methods is trained RBPNN using 1000 training samples, and after iteration 2000 times, network error is less than 0.0382%.
In table 1,1~20 is the initial data of part training sample, and 21~25 be test sample, and all samples all give neural network
Output result and the close instrument of salt measurement result, in order to compare, and whether the test result of checking R BPNN accurate, wherein P1
It is the test output of the desired output or RBPNN of training sample, P2 is the corresponding filthy grade of salt dense instrument measurement result.
1 experiment sample data of table and RBPNN test results
It can be obtained from data comparison, designed RBPNN can realize that of the different insulative based on visible images is filthy
The automatic identification of grade (0~IV grades), and have the characteristics that accurately and efficiently.
It should be understood that the part that this specification does not elaborate belongs to the prior art.
It should be understood that the above-mentioned description for preferred embodiment is more detailed, can not therefore be considered to this
The limitation of invention patent protection range, those skilled in the art under the inspiration of the present invention, are not departing from power of the present invention
Profit requires under protected ambit, can also make replacement or deformation, each fall within protection scope of the present invention, this hair
It is bright range is claimed to be determined by the appended claims.
Claims (7)
1. a kind of novel line insulator Condition assessment of insulation method, which is characterized in that include the following steps：
Step 1：Line insulator visible images acquisition is carried out in conjunction with optical camera image-forming principle, while using based on Bayes
The small wave self-adaption insulator visible images denoising method of estimation carries out denoising to line insulation subgraph；
Step 2：In order to study insulator disk face part, unrelated image detail is removed, first with OTSU methods by insulator disk
Face region segmentation comes out, and is then filled up using morphological method to the duck eye formed by noise jamming；
Step 3：The color characteristic of image is selected using Fei Sheer criterion, passes through the contaminated insulator based on Karhunen-Loeve transformation
The information fusion method of visible images characteristic parameter carries out dimension-reduction treatment, using based on Radial Basis Probabilistic Neural Networks grader
The artificial intelligence approach of RBPNN realizes the intelligent recognition of insulator contamination grade.
2. novel line insulator Condition assessment of insulation method according to claim 1, which is characterized in that in step 1,
Line insulator visible images obtain and denoising method is as follows：
Step 1-1, when shooting, selection and the shaft tower of camera lens exhale high D and insulator structure height H to have relationship, the focal length f's of camera lens
Calculation formula is as follows：
Wherein W is the width of camera CCD sizes；
The image of insulator surface is obtained according to above method；
After obtaining the image of insulator surface, gray processing processing, the gray processing of processing are carried out to visible images by step 1-2
Calculation formula is as follows：
Vgray=0.3R+0.59G+0.11B (2)
Wherein R, G, B are the rgb value of visible images, VgrayFor calculated gray value, formula is used to each pixel of image
(2) its gray-scale map is calculated；
Step 1-3, using the small wave self-adaption insulator visible images denoising method estimated based on Bayes to line insulator
It is as follows that image carries out denoising method：
It is the actual value of signal for signal a Z=X+n, X by noise pollution, n is the noise of addition in the signal, and Z is
The observed value of signal is estimated signal X generally according to observed value Z, the optimal of X estimates instruction when signal X can not be measured directlyFor posterior probability mean value：
Wherein, pn(Z-X) be noise probability density function, p (X) is the prior probability function of signal；
It can pass throughIt acquires, when noise and signal all obey zero-mean, obeys respectivelyThe height of variance
When this distribution, have：
Since noise is mainly distributed on high frequency section, retain low frequency part, high frequency section wavelet coefficient estimated using Bayes,
Then wavelet inverse transformation is carried out to all wavelet coefficients, obtains image after denoising.
3. novel line insulator Condition assessment of insulation method according to claim 2, which is characterized in that in step 2,
Using OTSU methods, by insulator disk region segmentation, out steps are as follows：
(1) probability that each gray-scale number of image and each gray-level pixels occur is calculated；
(2) the overall gray level mean value of image, the gray average and background classes A and target class of calculating background classes A and target class B are obtained
The probability of occurrence of B；
(3) variance between class Separation Indexes i.e. two classes of background classes A and target class B is calculated；
(4) asking makes the maximum value of variance be optimal threshold T, divides image by segmentation threshold of T, obtains insulator bianry image：
(5) disk part is cut into from insulator bianry image, obtains insulator disk bianry image.
4. novel line insulator Condition assessment of insulation method according to claim 3, which is characterized in that in step 2,
Progress morphological operation fills up insulator disk, and steps are as follows：
Mathematical morphology is based on set theory, and basic operation is expansion and erosion operation, and with dilation operation and corruption
The various combination mode of erosion operation constitutes Glycerine enema, a variety of operations of closure operation and operating method；
The characteristics of using dilation operation, erosion operation, Glycerine enema, closure operation, removes background, makes insulator disk and details
It preserves complete.
5. novel line insulator Condition assessment of insulation method according to claim 4, which is characterized in that in step 3,
Carrying out selection to the color characteristic of image using Fisher criterion, steps are as follows：
Feature selecting is carried out using Fisher criterion, it is as small as possible that the stronger feature of performance shows as variance within clusters, inter-class variance
It is as big as possible；
For contaminated insulator characteristics of image, the variance within clusters S of i-th dimension featurew(i)With inter-class variance Sb(i)Calculation formula such as
Under：
In formula：I is characterized component number；K is gradation for surface pollution, k=1,2 ..., K, and K is the gradation for surface pollution number that is divided；For
The i-th dimension feature of the single sample of gradation for surface pollution k；wkThe sample set for being k for gradation for surface pollution；It is k's for gradation for surface pollution
The mean value of sample i-th dimension characteristic value；miFor the mean value of the i-th dimension characteristic value of all gradation for surface pollution samples；For gradation for surface pollution k
Sample i-th dimension characteristic value number；niFor the sum of the i-th dimension feature of all samples；
Inter-class variance Sb(i)With variance within clusters Sw(i)The ratio between be Fisher discrimination function value, i.e.,：
It is found that J(i)Value is bigger, shows that dimensional feature difference between all kinds of is bigger, classifying quality is better.
6. novel line insulator Condition assessment of insulation method according to claim 5, which is characterized in that in step 3,
Steps are as follows for the information fusion method of contaminated insulator visible images characteristic parameter based on Karhunen-Loeve transformation：
(1) the covariance matrix C of initial characteristic data matrix X is calculated；
(2) eigenvalue λ of covariance matrix C is soughtjAnd eigenvalue λjCorresponding orthonomalization feature vector pj；
(3) by eigenvalue λjIt is ranked sequentially by descending, and makes feature vector also by being ranked sequentially accordingly；
(4) selection of feature vector number is with principal component accumulation contribution rate γkFor evaluation index,
Wherein, ρjFor principal component contributor rate, k is the dimension of quasi- selection feature vector, and M is characterized the dimension of parameter；
Work as γkWhen >=85%, the number of selected feature vector contains the main information of raw data matrix, several before selection
Maximum λjCorresponding feature vector constitutes orthogonal matrix P；
(5) pass through formula Y=PTX acquires principal component feature matrix Y after Karhunen-Loeve transformation.
7. novel line insulator Condition assessment of insulation method according to claim 6, which is characterized in that in step 3,
Artificial intelligence approach based on Radial Basis Probabilistic Neural Networks grader RBPNN realizes the intelligent recognition step of insulator contamination grade
It is rapid as follows：
For the RBPNN of N number of training sample, output is represented by following formula：
I.e.：Y=WH (12)
In formula：
Y1 Y2 … YNTo correspond to the network output valve of N number of training sample respectively, W is the second hidden layer to the weights between output layer
Matrix, H are the output of the second hidden layer；P is input vector, and Φ () is the Non-linear Kernel function of the first hidden layer, qiFor network
The number of nodes of i-th of classification of the second hidden layer is belonged in one hidden layer；CkFor k-th of hidden centers vectors of the first hidden layer, dimension
It is identical as input vector P, | | | |2For 2 one norm of euclidean；σkGinseng is controlled for the width of k-th of hidden central Gaussian function
Number；
The 0th grade of filth is represented when output is 0, exports that represent I grade when being 1 filthy, exports filthy, the root that represents Section II grade for 2
It is worth to insulator contamination grade according to output.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711496231.6A CN108345898A (en) | 2017-12-31 | 2017-12-31 | A kind of novel line insulator Condition assessment of insulation method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711496231.6A CN108345898A (en) | 2017-12-31 | 2017-12-31 | A kind of novel line insulator Condition assessment of insulation method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108345898A true CN108345898A (en) | 2018-07-31 |
Family
ID=62963476
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201711496231.6A Pending CN108345898A (en) | 2017-12-31 | 2017-12-31 | A kind of novel line insulator Condition assessment of insulation method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108345898A (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112541478A (en) * | 2020-12-25 | 2021-03-23 | 国网吉林省电力有限公司信息通信公司 | Insulator string stain detection method and system based on binocular camera |
| CN113095499A (en) * | 2021-03-26 | 2021-07-09 | 云南电网有限责任公司电力科学研究院 | Insulator equivalent salt deposit density prediction method |
| CN113125908A (en) * | 2021-04-16 | 2021-07-16 | 华北电力大学 | Insulator contamination degree diagnosis device and detection method thereof |
| CN116071368A (en) * | 2023-04-07 | 2023-05-05 | 国网山西省电力公司电力科学研究院 | Insulator pollution multi-angle image detection and fineness analysis method and device |
| CN116679171A (en) * | 2023-05-15 | 2023-09-01 | 江苏云峰科技股份有限公司 | Insulation state judging system of insulation piece of wind power generation switch |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20150227816A1 (en) * | 2014-02-10 | 2015-08-13 | Huawei Technologies Co., Ltd. | Method and apparatus for detecting salient region of image |
| 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 |
-
2017
- 2017-12-31 CN CN201711496231.6A patent/CN108345898A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20150227816A1 (en) * | 2014-02-10 | 2015-08-13 | Huawei Technologies Co., Ltd. | Method and apparatus for detecting salient region of image |
| 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 |
Non-Patent Citations (4)
| Title |
|---|
| 何洪英: "基于红外热像及人工智能的绝缘子污秽等级识别方法研究", 《中国博士学位论文全文数据库 工程科技Ⅱ辑》 * |
| 光学人生: "镜头焦距与视场角计算方法", 《HTTPS://WWW.SOHU.COM/A/169099063_712214》 * |
| 田治仁 等: "基于彩色可见光图像的绝缘子污秽等级判别", 《电工电能新技术》 * |
| 金立军 等: "基于红外与可见光图像信息融合的绝缘子污秽等级识别", 《中国电机工程学报》 * |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112541478A (en) * | 2020-12-25 | 2021-03-23 | 国网吉林省电力有限公司信息通信公司 | Insulator string stain detection method and system based on binocular camera |
| CN113095499A (en) * | 2021-03-26 | 2021-07-09 | 云南电网有限责任公司电力科学研究院 | Insulator equivalent salt deposit density prediction method |
| CN113125908A (en) * | 2021-04-16 | 2021-07-16 | 华北电力大学 | Insulator contamination degree diagnosis device and detection method thereof |
| CN116071368A (en) * | 2023-04-07 | 2023-05-05 | 国网山西省电力公司电力科学研究院 | Insulator pollution multi-angle image detection and fineness analysis method and device |
| CN116679171A (en) * | 2023-05-15 | 2023-09-01 | 江苏云峰科技股份有限公司 | Insulation state judging system of insulation piece of wind power generation switch |
| CN116679171B (en) * | 2023-05-15 | 2023-11-10 | 江苏云峰科技股份有限公司 | Insulation state judging system of insulation piece of wind power generation switch |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108345898A (en) | A kind of novel line insulator Condition assessment of insulation method | |
| CN107818326B (en) | A kind of ship detection method and system based on scene multidimensional characteristic | |
| CN109886217B (en) | Method for detecting wave height from offshore wave video based on convolutional neural network | |
| Topouzelis et al. | Detection and discrimination between oil spills and look-alike phenomena through neural networks | |
| CN101727662B (en) | SAR image nonlocal mean value speckle filtering method | |
| CN111563893B (en) | Grading ring defect detection method, device, medium and equipment based on aerial image | |
| Angelina et al. | Image segmentation based on genetic algorithm for region growth and region merging | |
| CN103236063B (en) | Based on the SAR image oil spilling detection method of multiple dimensioned spectral clustering and decision level fusion | |
| CN106886760B (en) | A kind of EO-1 hyperion Ship Detection combined based on empty spectrum information | |
| CN106557740B (en) | The recognition methods of oil depot target in a kind of remote sensing images | |
| CN105844298A (en) | High spectral oil overflow image classification method based on Fuzzy ARTMAP neural network | |
| CN112396619A (en) | Small particle segmentation method based on semantic segmentation and internally complex composition | |
| CN103700109B (en) | SAR image change detection based on multiple-objection optimization MOEA/D and fuzzy clustering | |
| CN106960433B (en) | It is a kind of that sonar image quality assessment method is referred to based on image entropy and the complete of edge | |
| Yaohua et al. | A SAR oil spill image recognition method based on densenet convolutional neural network | |
| Arif et al. | Adaptive deep learning detection model for multi-foggy images | |
| CN114266947A (en) | Classification method and device based on fusion of laser point cloud and visible light image | |
| CN103336972A (en) | Foundation cloud picture classification method based on completion local three value model | |
| CN117636268A (en) | Unmanned aerial vehicle aerial natural driving data set construction method oriented to ice and snow environment | |
| Ma et al. | Feature selection and classification of oil spills in SAR image based on statistics and artificial neural network | |
| Xin et al. | A fine extraction algorithm for image-based surface cracks in underwater dams | |
| CN112415514A (en) | Target SAR image generation method and device | |
| CN116519710A (en) | Method and system for detecting surface pollution state of composite insulator | |
| CN114140698A (en) | Water system information extraction algorithm based on FasterR-CNN | |
| CN107992863A (en) | Multiresolution grain worm species visual identity 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: 20180731 |