CN116519710A - Method and system for detecting surface pollution state of composite insulator - Google Patents
Method and system for detecting surface pollution state of composite insulator Download PDFInfo
- Publication number
- CN116519710A CN116519710A CN202310478950.4A CN202310478950A CN116519710A CN 116519710 A CN116519710 A CN 116519710A CN 202310478950 A CN202310478950 A CN 202310478950A CN 116519710 A CN116519710 A CN 116519710A
- Authority
- CN
- China
- Prior art keywords
- data
- insulator
- hyperspectral
- database
- pollution
- 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
- 239000012212 insulator Substances 0.000 title claims abstract description 115
- 238000000034 method Methods 0.000 title claims abstract description 58
- 239000002131 composite material Substances 0.000 title claims abstract description 31
- 238000001228 spectrum Methods 0.000 claims abstract description 60
- 230000003595 spectral effect Effects 0.000 claims abstract description 42
- 238000012706 support-vector machine Methods 0.000 claims abstract description 27
- 238000012937 correction Methods 0.000 claims abstract description 24
- 230000009467 reduction Effects 0.000 claims abstract description 24
- 238000012545 processing Methods 0.000 claims abstract description 23
- 230000009466 transformation Effects 0.000 claims abstract description 19
- 238000007781 pre-processing Methods 0.000 claims abstract description 12
- 230000011218 segmentation Effects 0.000 claims abstract description 12
- 238000012847 principal component analysis method Methods 0.000 claims abstract description 6
- 238000012549 training Methods 0.000 claims description 25
- 230000008569 process Effects 0.000 claims description 17
- 150000003839 salts Chemical class 0.000 claims description 16
- 238000001514 detection method Methods 0.000 claims description 14
- 238000013145 classification model Methods 0.000 claims description 11
- 238000009826 distribution Methods 0.000 claims description 11
- 238000010586 diagram Methods 0.000 claims description 9
- 238000012360 testing method Methods 0.000 claims description 8
- 230000008859 change Effects 0.000 claims description 7
- 239000004973 liquid crystal related substance Substances 0.000 claims description 6
- 238000012795 verification Methods 0.000 claims description 5
- 238000005260 corrosion Methods 0.000 claims description 4
- 230000007797 corrosion Effects 0.000 claims description 4
- 238000009499 grossing Methods 0.000 claims description 4
- 230000000877 morphologic effect Effects 0.000 claims description 4
- 239000003086 colorant Substances 0.000 claims description 3
- 238000005286 illumination Methods 0.000 claims description 3
- 238000002310 reflectometry Methods 0.000 claims description 3
- 238000010276 construction Methods 0.000 claims description 2
- 239000003344 environmental pollutant Substances 0.000 claims description 2
- 231100000719 pollutant Toxicity 0.000 claims description 2
- 238000009827 uniform distribution Methods 0.000 claims description 2
- 238000011109 contamination Methods 0.000 claims 1
- 230000006870 function Effects 0.000 description 9
- 238000005516 engineering process Methods 0.000 description 6
- 238000004590 computer program Methods 0.000 description 3
- 230000003287 optical effect Effects 0.000 description 3
- 238000001579 optical reflectometry Methods 0.000 description 3
- 102000008297 Nuclear Matrix-Associated Proteins Human genes 0.000 description 2
- 108010035916 Nuclear Matrix-Associated Proteins Proteins 0.000 description 2
- 238000000701 chemical imaging Methods 0.000 description 2
- 238000013178 mathematical model Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 210000000299 nuclear matrix Anatomy 0.000 description 2
- 238000000513 principal component analysis Methods 0.000 description 2
- 230000000007 visual effect Effects 0.000 description 2
- 238000010521 absorption reaction Methods 0.000 description 1
- 238000009825 accumulation Methods 0.000 description 1
- 230000004075 alteration Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000010292 electrical insulation Methods 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 230000005251 gamma ray Effects 0.000 description 1
- 238000009413 insulation Methods 0.000 description 1
- 239000002245 particle Substances 0.000 description 1
- 230000002265 prevention Effects 0.000 description 1
- 229920002379 silicone rubber Polymers 0.000 description 1
- 239000007787 solid Substances 0.000 description 1
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
- G01N21/01—Arrangements or apparatus for facilitating the optical investigation
-
- 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
- G01N21/01—Arrangements or apparatus for facilitating the optical investigation
- G01N2021/0106—General arrangement of respective parts
- G01N2021/0112—Apparatus in one mechanical, optical or electronic block
Abstract
The invention discloses a method and a system for detecting a pollution state on the surface of a composite insulator, wherein the method comprises the following steps: collecting hyperspectral images of the pollution insulators, and obtaining hyperspectral data of pixel points of the insulator areas and average spectral data of the interested areas; according to the hyperspectral data and the average spectral data, constructing a hyperspectral database, preprocessing the spectral data of the hyperspectral database through black-white correction, standard normal transformation and trend removal correction, and carrying out grouping dimension reduction on the whole wave band by adopting a selective segmentation principal component analysis method; the method comprises the steps of constructing a multi-classification semi-supervised Newton support vector machine model, and using spectrum data after dimension reduction processing as input data of the model to identify the pollution state of a pollution insulator corresponding to the spectrum data; the multi-classification semi-supervised quasi-Newton support vector machine model provided by the invention can be trained by using a small number of marked samples and a large number of unmarked samples, so that the purposes of reducing the marking cost and improving the generalization capability of the model are achieved.
Description
Technical Field
The invention relates to the field of overhauling of the running state of power transmission and transformation equipment, in particular to a method and a system for detecting the pollution state of the surface of a composite insulator.
Background
The insulator is the electrical insulation equipment with the largest use amount in the power system, wherein the silicon rubber composite insulator has the advantages of light weight and excellent mechanical property, has good performances in aspects of pollution flashover resistance, corona resistance, hydrophobicity and the like, and is widely applied to a high-voltage insulation guarantee system in China. The pollution state of the composite insulator can be monitored accurately in real time, so that the occurrence of pollution flashover prevention accidents can be effectively prevented, and the stable operation of the power system is ensured. The hyperspectral imaging technology is used as a novel optical detection technology, the filthy state of the surface of the insulator can be represented in the form of the light reflectivity of the surface of the insulator, the light reflectivity is different when the filthy states are different, and the filthy state of the surface of the insulator can be identified by establishing a mathematical model to learn the corresponding relation between the light reflectivity and the filthy state.
In practical engineering, the number of known insulators in the pollution state is very small, a mathematical model is difficult to establish by adopting a full-supervision mode to accurately identify the pollution state on the surface of the composite insulator, secondly, the pollution accumulation conditions of different areas of an insulator string operated on site are different, the pollution flashover voltage of the insulator can be obviously changed by the non-uniformity of the surface pollution, and the distribution condition of the pollution is very important when the pollution state is identified, so that a detection method and a detection system for the surface pollution state of the composite insulator are urgently needed, and the accurate detection of the pollution grade and the dust-salt ratio of the insulator is realized under the condition of lack of a marked sample.
Disclosure of Invention
In order to solve the problems, the invention aims to provide a detection method and a detection system for the surface pollution state of a composite insulator, which are used for accurately detecting the surface pollution state of the composite insulator at a pixel level by utilizing a large number of unlabeled samples under the condition of lack of labeled samples and forming visual visualized images to further represent the surface pollution distribution condition of the insulator.
In order to achieve the technical purpose, the application provides a detection method for a composite insulator surface pollution state, which comprises the following steps:
collecting hyperspectral images of the polluted insulators, and obtaining hyperspectral data of pixel points of the insulator areas and average spectral data of the interested areas through segmentation processing, wherein the interested areas represent areas with uniform distribution of the pollutants of the insulators;
taking the hyperspectral data as unlabeled data to construct a first database;
taking the average spectrum data as marking data, obtaining the pollution state of the region of interest by an equivalent salt deposit method, and constructing a second database based on the average spectrum data;
constructing a hyperspectral database comprising a first database and a second database, preprocessing spectral data of the hyperspectral database through black-white correction, standard normal transformation and trend removal correction, and then adopting a selective segmentation principal component analysis method to perform grouping dimension reduction treatment on all wave bands;
and constructing a multi-classification semi-supervised Newton support vector machine model, taking the spectrum data after the dimension reduction treatment as input data of the model, carrying out two-round recognition, respectively recognizing gray density and salt density, and finally combining the two classification results.
Preferably, in the process of acquiring the insulator region, the process of performing the segmentation process on the hyperspectral image includes:
extracting an RGB three-color chart from a hyperspectral image;
performing Gaussian smoothing on the RGB three-color image;
converting the RGB trichromatic image into an HSV color space, and setting a threshold value according to the self color characteristics of the composite insulator to obtain a mask of the insulator string region;
carrying out corrosion and expansion treatment on the mask by using morphological operation, and attaching the shape of the mask to the shape of the insulator region;
and (3) segmenting the insulator region from the original hyperspectral image by using the mask obtained in the step S204, wherein spectral data of other background regions except the insulator region in the hyperspectral image are set to be 0.
Preferably, in the process of performing black-and-white correction on the hyperspectral database, the step of black-and-white correction includes:
collecting original hyperspectral image I of target insulator under the same illumination condition 0 Full black scaling with 0 reflectivity collected after closing the lens coverImage I D Standard white reference image I of an open lens cover scanning calibration whiteboard w The corrected image I is obtained according to the following formula:
preferably, in the process of performing the standard normal transformation, an SNV transformation is performed on each spectral curve of the data corrected by black-and-white, wherein the SNV transformation is expressed as:
wherein, the liquid crystal display device comprises a liquid crystal display device,m represents the number of wave points, and i represents the ith spectral curve.
Preferably, in the process of removing trend correction, a least square method is used for performing binomial fitting to obtain a trend curve:
subtracting the trend curve from the spectral curve to obtain processed spectral data:
wherein b 0 、b 1 、b 2 Is a binomial coefficient, x is a wavelength, data is a spectral curve after SNV processing,data is trend curve * To remove trend corrected spectral curves.
Preferably, in the process of carrying out grouping dimension reduction treatment, when the pollution grade on the surface of the insulator is changed, each grade corresponds to the change condition of each wave band of the spectrum curve, and the whole wave band can be divided into three areas according to the change trend;
by calculating the correlation of each band, three regions are divided into four subgroups T1/T2/T3/T4: 400-429 nm wave band, 452.25-592.7 nm wave band, 604.51-723.53 nm wave band and 747.54-917.42 nm wave band, wherein lower wave band correlation exists among four sub-groups, and higher wave band correlation exists inside a sub-group;
and respectively carrying out main component transformation on each sub group, arranging the characteristic values in a descending order, and respectively selecting characteristic data according to the contribution degree of the components to finish the dimension reduction processing.
Preferably, in the process of constructing the multi-classification semi-supervised Newton support vector machine model, an L-BFGS Newton method framework is adopted on the basis of the semi-supervised support vector machine, so that the convergence of the model is enhanced, and the operation time and the storage space are reduced;
on the basis of a quasi-Newton semi-supervised support vector machine classification method, a one-against-one and a voting strategy are adopted to expand the quasi-Newton semi-supervised support vector machine classification problem, and a multi-classification semi-supervised quasi-Newton support vector machine solving framework is established;
the multi-classification model consists of M classification models, and thus the test sample is to be repeatedly classified M times. At this time, a voting criterion is introduced, taking a classification model between class i and class j as an example, if for x t And if the predicted class i is i, the class i gets the ticket plus 1, otherwise, the class j gets the ticket plus 1. The most recently obtained class is x t Is used to predict the final predicted value of (c). If the flat ticket condition occurs, the flat ticket is selected randomly from several classes.
Preferably, in the process of identifying the pollution state, the first database after pretreatment and dimension reduction treatment is randomly divided into a training set A and a verification set according to the proportion of 5:1;
randomly extracting partial spectrum data from the second database after the preprocessing and dimension reduction processing as a training set B;
setting the data quantity ratio of the training set B to the training set A as 10:1, wherein the training set A and the training set B are jointly used as the training set;
selecting spectral data of each pixel point of a target insulator as spectral data to be tested to form a test set;
training and optimizing all data of the training set through a multi-classification semi-supervised quasi-Newton support vector machine model, and adjusting model parameters by using the verification set;
inputting the spectral data to be tested serving as the test set into a trained model to generate a classification result, and identifying the pollution state of the target insulator;
the pollution state of each pixel point of the target insulator is classified, different colors are used for representing different pollution states, and finally the pollution state classification result of each pixel point of the insulator is represented in a color image form to form a pollution state distribution diagram of the insulator.
The invention discloses a detection system for a composite insulator surface pollution state, which comprises:
the data acquisition module is used for acquiring hyperspectral images of the pollution insulators;
the data processing module is used for acquiring hyperspectral data of pixel points of the insulator region and average spectral data of the region of interest through segmentation processing according to the hyperspectral image, wherein the region of interest represents a region with even distribution of insulator pollution;
the database construction module is used for constructing a first database by taking the hyperspectral data as unlabeled data; the average spectrum data is used as marking data, the pollution state of the region of interest is obtained through an equivalent salt deposit method, a second database is built based on the average spectrum data, and a hyperspectral database comprising a first database and the second database is built;
the preprocessing module is used for preprocessing the spectrum data of the hyperspectral database through black-white correction, standard normal transformation and trend removal correction, and then adopting a selective segmentation principal component analysis method to perform grouping dimension reduction processing on the whole wave band;
the identification module is used for identifying the pollution state of the pollution insulator corresponding to the spectrum data by constructing a multi-classification semi-supervised Newton support vector machine and taking the spectrum data subjected to the dimension reduction treatment as the input data of the multi-classification semi-supervised Newton support vector machine.
The invention discloses the following technical effects:
the multi-classification semi-supervised quasi-Newton support vector machine model provided by the invention can be trained by using a small number of marked samples and a large number of unmarked samples, so that the purposes of reducing the marking cost and improving the generalization capability of the model are achieved;
the invention classifies the pollution state of each pixel point of the target insulator by using the multi-classification semi-supervised Newton support vector machine model, and represents different pollution states by using different colors, and finally represents the pollution state classification result of each pixel point of the insulator in a color image form, thereby intuitively reflecting the pollution state and the pollution distribution condition of the surface of the insulator.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings that are needed in the embodiments will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.
FIG. 1 is a flow chart of the present invention;
FIG. 2 is a graph of spectra of composite insulators under different pollution conditions, wherein (a) represents different gray salt ratio spectra of the same pollution level and (b) represents different pollution level spectra of the same gray salt ratio;
FIG. 3 is a graph showing a comparison of the spectral curves before and after pretreatment, wherein (a) represents the spectral curve before pretreatment and (b) represents the spectral curve after pretreatment;
FIG. 4 is a schematic diagram of spectral data band correlation;
Detailed Description
For the purposes of making the objects, technical solutions and advantages of the embodiments of the present application more clear, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application, and it is apparent that the described embodiments are only some embodiments of the present application, but not all embodiments. The components of the embodiments of the present application, which are generally described and illustrated in the figures herein, may be arranged and designed in a wide variety of different configurations. Thus, the following detailed description of the embodiments of the present application, as provided in the accompanying drawings, is not intended to limit the scope of the application, as claimed, but is merely representative of selected embodiments of the application. All other embodiments, which can be made by those skilled in the art based on the embodiments of the present application without making any inventive effort, are intended to be within the scope of the present application.
As shown in fig. 1-4, example 1: a handheld intelligent SPECIM IQ hyperspectral camera is adopted, the spectrum of the camera covers ultraviolet, visible light near infrared, short wave infrared, medium wave infrared, thermal infrared and other wave bands, the wave band range is 400-1000 nm, 204 spectral channels are provided, the spectral resolution is 2.9nm, and the spatial resolution is 512 multiplied by 512 pixels.
As shown in fig. 1, the invention provides a pixel-level composite insulator surface pollution state detection method based on a hyperspectral imaging technology and a multi-classification semi-supervised quasi-newton support vector machine model, which adopts the following technical scheme:
s1, constructing a hyperspectral experiment platform, and acquiring hyperspectral images of the pollution insulators by using a hyperspectral camera;
s2, segmenting the hyperspectral image by utilizing a computer vision technology to obtain an insulator region, extracting hyperspectral data of all pixel points of the region and average spectral data of an interested region, and detecting the pollution state of the interested region, the whole pollution level of the insulator string and the gray salt ratio by adopting an equivalent salt density method;
the step S2 of dividing the hyperspectral image by using the computer vision technology to obtain the insulator region comprises the following specific contents:
s201, extracting an RGB three-color chart from a hyperspectral image;
s202, performing Gaussian smoothing on the RGB three-color image to reduce image noise and reduce detail level;
s203, converting the RGB trichromatic image into an HSV color space, and setting a threshold according to the self color characteristics of the composite insulator to obtain an insulator string region mask;
s204, performing corrosion and expansion treatment on the mask obtained in the S203 by using morphological operation, so that the shape of the mask is more fit with the shape of the insulator region;
s205, segmenting the insulator region from the original hyperspectral image by using the mask obtained in S204, wherein spectral data of other background regions except the insulator region in the hyperspectral image are set to be 0.
In the step S2, the insulator region is segmented from the background of the hyperspectral image, and a foundation is laid for the subsequent extraction of the spectral curve of the insulator region.
In step S2, segmenting a hyperspectral image by using a computer vision technology to obtain an insulator region, wherein the spectrum dimension of the acquired hyperspectral image is 204-dimensional, the acquired hyperspectral image comprises RGB three-primary-color wave bands, a three-primary-color image is obtained from the hyperspectral image, gaussian smoothing processing is adopted to reduce image noise, the level of detail is reduced, the smoothed RGB three-color image is converted into HSV color space, a threshold value is set according to the self color characteristics of a composite insulator to obtain an insulator string region mask, morphological operation is used for carrying out corrosion and expansion treatment on the insulator string region mask, the shape of the mask is more attached to the shape of the insulator region, the insulator region is segmented from the original hyperspectral image by using the mask, and the spectrum data of other background regions except the insulator region in the hyperspectral image are set to be 0;
in step S2, hyperspectral data of all pixel points of the insulator region and average spectral data of the region of interest in the hyperspectral image are extracted. And (3) extracting a region with even distribution of dirt on the insulator as an interested region by adopting ENVI5.3 software, and obtaining an average spectrum curve of the region, wherein the spectrum data of all pixel points of the insulator are written and extracted by program traversal.
In step S2, the pollution grade and the ash-to-salt ratio of the region of interest and the insulator string are obtained by adopting an equivalent salt deposit method. Wherein the equivalent salt deposit method is carried out according to the GB/T4585-2004 standard;
s3, respectively establishing a pollution insulator marked sample spectrum database and an unmarked sample spectrum database, wherein an average spectrum curve of a region of interest is marked data, a pixel spectrum curve is unmarked data, and figure 2 shows pollution insulator spectrum curves under the same gray salt ratio, different pollution grades and different gray salt ratios of the same pollution grade;
step S3, establishing a hyperspectral database comprises the following specific contents:
firstly, extracting all pixel point spectrum curves of an insulator region, establishing an unlabeled sample hyperspectral database, secondly, selecting a region with even insulator pollution distribution as an interested region, extracting all pixel point spectrum curves in the region, calculating an average spectrum curve, obtaining the polluted state of the interested region by an equivalent salt deposit method in S2, and establishing the labeled sample hyperspectral database based on the average spectrum curve.
Step S3 provides a data basis for the subsequent establishment of a corresponding relation between the semi-supervised detection model learning spectrum data and the pollution level.
S4, preprocessing the spectrum data by adopting black-and-white correction, standard normal transformation (SNV) and trend removal correction;
the step of black-and-white correction in step S4 includes:
collecting original hyperspectral image I of target insulator under the same illumination condition 0 Full black calibration image I with reflectivity of 0 acquired after closing lens cover D Standard white reference image I of an open lens cover scanning calibration whiteboard w The corrected image I is obtained according to the following formula:
step S4 is about the design of black-and-white correction, and the influence of hyperspectral image acquisition environment and the dark current in the hyperspectral camera on a spectrum curve is effectively eliminated.
The specific steps of standard normal transformation in the step S4 include:
SNV conversion is carried out on each spectrum curve after black-white correction:
wherein, the liquid crystal display device comprises a liquid crystal display device,m represents the number of wave points, i represents the ith spectral curve, data ik Indicating the spectral reflectance at the kth wavelength of the ith spectral curve.
The standard normal transformation is used for eliminating the influence of solid particle size, surface scattering and optical path change on the optical data.
The specific steps for removing trend correction in step S4 include:
and (3) performing binomial fitting by using a least square method to obtain a trend curve:
subtracting the trend curve from the spectral curve to obtain processed spectral data:
wherein b 0 、b 1 、b 2 Is a binomial coefficient, x is a wavelength, data is a spectral curve after SNV processing,data is trend curve * In order to remove the trend-corrected spectrum curve, fig. 3 shows the spectrum curve after pretreatment, and as can be seen from fig. 3, the absorption valley, reflection peak and extreme point of the spectrum curve after pretreatment are more obvious, which is beneficial to the subsequent classification.
The invention achieves the purpose of eliminating the baseline drift of the diffuse reflection spectrum by matching the removal trend correction with the standard normal transformation.
S5, analyzing spectral data characteristics, calculating band correlation, and carrying out grouping dimension reduction treatment on the whole band by adopting a selective segmentation principal component analysis method;
the specific steps of the spectrum data dimension reduction processing in the step S5 are as follows:
when the surface pollution grade of the insulator in fig. 2 is analyzed, each grade corresponds to the change condition of each wave band of the spectrum curve, and the whole wave band can be divided into three areas according to the change trend: 400-625 nm, 625-740 nm, 740-1000 nm and 740-1000 nm.
The correlation of each wave band is calculated, and the calculation result is shown in fig. 4, wherein the three areas can be divided into four subgroups T1/T2/T3/T4: the four sub-groups have lower band correlation and higher band correlation inside the subgroup, and the band correlation is between 0 and 12 bands, 20 and 68 bands (namely 400 to 429nm and 452.25 to 592.7nm bands, corresponding to other visible light regions), 72 to 112 bands (namely 604.51 to 723.53nm bands, corresponding to red light regions) and 120 to 176 bands (namely 747.54 to 917.42nm bands, corresponding to near infrared regions).
And respectively carrying out main component transformation on each sub group, arranging the characteristic values in a descending order, and respectively selecting characteristic data according to the contribution degree of the components. PC1 and PC2 are selected at T1, PC1 is selected at T2, PC1 to PC3 are selected at T3, and PC1 to PC10 are selected at T4, for a total of 16 features.
The purpose of the spectrum data dimension reduction processing is to reduce the dimension of the hyperspectral data, avoid the redundancy of the spectrum data and the reduction of the accuracy of a classification model caused by the Hughes phenomenon of the spectrum data, and simultaneously, compared with the principal component analysis, the selective segmentation principal component analysis can reduce the calculated amount by 66 percent.
S6, constructing a pixel-level composite insulator surface pollution state detection model based on a multi-classification semi-supervised Newton support vector machine, detecting the pollution level and the gray salt ratio of each pixel point of the insulator, and performing visual treatment on the detection result to form an intuitive insulator surface pollution state distribution diagram.
The specific steps of constructing the semi-supervised detection model of the surface pollution state of the pixel-level composite insulator in the step S6 comprise the following steps:
on the basis of a semi-supervised support vector machine model, an L-BFGS quasi-Newton method framework is adopted to enhance the convergence of the model, and the operation time and the storage space are reduced. Meanwhile, the traditional S3VM is only suitable for the classification problem, and the classification method based on the quasi-Newton semi-supervised support vector machine is used for expanding the classification problem to a plurality of classification problems by adopting one-against-one and a voting strategy, so that a multi-classification semi-supervised quasi-Newton support vector machine solving framework is established and is used for identifying the surface pollution condition of the composite insulator.
The multi-classification model consists of M classification models, and thus the test sample is to be repeatedly classified M times. At this time, a voting rule is introduced, taking a classification model between class i and class j as an example, if for x t And if the predicted class i is i, the class i gets the ticket plus 1, otherwise, the class j gets the ticket plus 1. The most recently obtained class is x t Is used to predict the final predicted value of (c). If the flat ticket condition occurs, the flat ticket is selected randomly from several classes.
The model obtains the optimal hyperplane by minimizing the objective function, and the specific case is as follows:
wherein, the superscript indicates the related parameters of the classification model between the ith class and the jth class, l ij Representing the number of samples of the union of the ith class and the j class in the marked samples, f ij Representing hyperplane functions of class i and class j bipartite models,to regenerate the norms of the class i and class j bipartite model functions in the Hilbert space. />Representing marking data, < >>In order to mark the tag of the data,indicating unlabeled data. M is the number of bipartite models, and n is the total category number. t is t 1 Index t for union sample of class i and class j 2 Is an index of unlabeled samples. Gamma ray 1 And gamma 2 For model parameters, u is the number of unlabeled exemplars. V (V) L1 And V is equal to L2 Error functions for marked samples and unmarked samples, respectively, are chosen here to replace the hange Loss function with a continuously differentiable exponential function for the purpose of solving:
(5) The optimal solution for (a) is generally expressed as a linear combination of the i-th and j-th labeled samples and all unlabeled sample kernel functions:
wherein c is a penalty coefficient, κ is a kernel function, letThe problem is thus shifted to find the optimal parameter vector +.>n ij ＝l ij +u, at which point equation (5) converts to:
the gradient is as follows:
wherein the method comprises the steps ofa satisfies the following conditions:
k is a nuclear matrix, wherein K is a nuclear matrix,
randomly dividing the marked sample database after preprocessing and dimension reduction processing into a training set A and a verification set according to the proportion of 5:1, randomly extracting part of spectrum data in the unmarked sample database after the dimension reduction processing after preprocessing to be used as a training set B, wherein the ratio of the data quantity of the training set B to the data quantity of the training set A is 10:1, the A and the B are jointly used as the training set, the spectrum data to be tested is used as a test set, and the spectrum data to be tested is the spectrum data of each pixel point of the target insulator. Substituting the training set data into the model, and inputting the data to be predicted into the model after training is finished, so as to obtain a final classification result. And finally, the classification result of the pollution state of each pixel point of the insulator is expressed in a color image form to form a distribution diagram of the pollution state of the insulator.
The present invention is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems) and computer program products according to embodiments of the invention. It will be understood that each flow and/or block of the flowchart illustrations and/or block diagrams, and combinations of flows and/or blocks in the flowchart illustrations and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
In the description of the present invention, it should be understood that the terms "first," "second," and the like are used for descriptive purposes only and are not to be construed as indicating or implying a relative importance or number of technical features indicated. Thus, a feature defining "a first" or "a second" may explicitly or implicitly include one or more such feature. In the description of the present invention, the meaning of "a plurality" is two or more, unless explicitly defined otherwise.
It will be apparent to those skilled in the art that various modifications and variations can be made to the present invention without departing from the spirit or scope of the invention. Thus, it is intended that the present invention also include such modifications and alterations insofar as they come within the scope of the appended claims or the equivalents thereof.
Claims (10)
1. The method for detecting the surface pollution state of the composite insulator is characterized by comprising the following steps of:
collecting hyperspectral images of the polluted insulators, and obtaining hyperspectral data of pixel points of the insulator areas and average spectral data of the interested areas through segmentation processing, wherein the interested areas represent areas with uniform distribution of the pollutants of the insulators;
taking the hyperspectral data as unlabeled data to construct a first database;
the average spectrum data is used as marking data, the pollution state of the region of interest is obtained through an equivalent salt deposit method, and a second database is constructed based on the average spectrum data;
constructing a hyperspectral database comprising the first database and the second database, preprocessing the spectrum data of the hyperspectral database through black-white correction, standard normal transformation and trend removal correction, and then carrying out grouping dimensionality reduction on the whole wave band by adopting a selective segmentation principal component analysis method;
and constructing a multi-classification semi-supervised Newton support vector machine model, taking the spectrum data after the dimension reduction treatment as input data of the model, carrying out two-round recognition, respectively recognizing gray density and salt density, and finally combining the two classification results.
2. The method for detecting the surface pollution state of the composite insulator according to claim 1, wherein the method comprises the following steps:
the process of segmenting the hyperspectral image in the process of acquiring the insulator region comprises the following steps:
s201, extracting an RGB three-color chart from the hyperspectral image;
s202, performing Gaussian smoothing on the RGB three-color map;
s203, converting the RGB trichromatic image into an HSV color space, and setting a threshold value according to the self color characteristics of the composite insulator to obtain a mask of the insulator string region
S204, performing corrosion and expansion treatment on the mask by using morphological operation, wherein the mask is used for attaching the shape of the mask to the shape of the insulator region;
s205, segmenting the insulator region from the original hyperspectral image by using the mask obtained in S204, wherein spectral data of other background regions except the insulator region in the hyperspectral image are set to be 0.
3. The method for detecting the surface pollution state of the composite insulator according to claim 2, wherein the method comprises the following steps:
in a process of black-and-white correction of a hyperspectral database, the step of black-and-white correction comprises:
collecting original hyperspectral image I of target insulator under the same illumination condition 0 Full black calibration image I with reflectivity of 0 acquired after closing lens cover D Standard white reference image I of an open lens cover scanning calibration whiteboard w The corrected image I is obtained according to the following formula:
4. a method for detecting a fouling condition on a surface of a composite insulator according to claim 3, wherein:
in the process of carrying out standard normal transformation, carrying out SNV transformation on each spectrum curve of data after black-and-white correction, wherein the SNV transformation is expressed as:
wherein, the liquid crystal display device comprises a liquid crystal display device,m represents the number of wave points, and i represents the ith spectral curve.
5. The method for detecting the surface pollution state of the composite insulator according to claim 4, wherein the method comprises the following steps:
in the process of removing trend correction, a least square method is used for performing binomial fitting to obtain a trend curve:
subtracting the trend curve from the spectrum curve to obtain processed spectrum data:
wherein b 0 、b 1 、b 2 Is a binomial coefficient, x is a wavelength, data is a spectral curve after SNV processing,data is trend curve * To remove trend corrected spectral curves.
6. The method for detecting the surface pollution state of the composite insulator according to claim 5, wherein the method comprises the following steps:
in the process of grouping dimension reduction treatment, when the pollution grade on the surface of the insulator is changed, each grade corresponds to the change condition of each wave band of the spectrum curve, and the whole wave band can be divided into three areas according to the change trend;
by calculating the correlation of each band, three regions are divided into four subgroups T1/T2/T3/T4: 400-429 nm wave band, 452.25-592.7 nm wave band, 604.51-723.53 nm wave band and 747.54-917.42 nm wave band, wherein lower wave band correlation exists among four sub-groups, and higher wave band correlation exists inside a sub-group;
and respectively carrying out main component transformation on each sub group, arranging the characteristic values in a descending order, and respectively selecting characteristic data according to the contribution degree of the components to finish the dimension reduction processing.
7. The method for detecting the surface pollution state of the composite insulator according to claim 6, wherein the method comprises the following steps:
in the process of constructing the multi-classification semi-supervised Newton support vector machine model, an L-BFGS Newton method framework is adopted on the basis of the semi-supervised support vector machine, so that the convergence of the model is enhanced, and the operation time and the storage space are reduced;
on the basis of the quasi-Newton semi-supervised support vector machine classification method, a one-against-one and voting strategy are adopted to expand the quasi-Newton semi-supervised support vector machine classification problem, and a multi-classification semi-supervised quasi-Newton support vector machine solving framework is established.
8. The method for detecting the surface pollution state of the composite insulator according to claim 7, wherein the method comprises the following steps:
the multi-classification model consists of M classification models, and thus the test sample is to be repeatedly classified M times. At this time, a voting standard is introduced, taking a classification model between class i and class j as an example, if for x t And if the predicted class i is i, the class i gets the ticket plus 1, otherwise, the class j gets the ticket plus 1. The most recently obtained class is x t Is used to predict the final predicted value of (c). If outThe flat ticket condition is selected randomly from several classes of flat tickets.
9. The method for detecting the surface pollution state of the composite insulator according to claim 8, wherein the method comprises the following steps:
in the process of identifying the pollution state, the first database after pretreatment and dimension reduction treatment is randomly divided into a training set A and a verification set according to the proportion of 5:1;
randomly extracting part of spectrum data from the second database after preprocessing and dimension reduction processing to serve as a training set B;
setting the data quantity ratio of the training set B to the training set A as 10:1, wherein the training set A and the training set B are jointly used as the training set;
selecting spectral data of each pixel point of a target insulator as spectral data to be tested to form a test set;
training and optimizing all data of a training set through a multi-classification semi-supervised quasi-Newton support vector machine model, and adjusting model parameters by using verification set data;
inputting the spectral data to be tested serving as the test set into a trained model to generate a classification result, and identifying the pollution state of the target insulator;
the pollution state of each pixel point of the target insulator is classified, different colors are used for representing different pollution states, and finally the pollution state classification result of each pixel point of the insulator is represented in a color image form to form a pollution state distribution diagram of the insulator.
10. A detection system for a composite insulator surface contamination condition, comprising:
the data acquisition module is used for acquiring hyperspectral images of the pollution insulators;
the data processing module is used for acquiring hyperspectral data of pixel points of the insulator region and average spectral data of the region of interest through segmentation processing according to the hyperspectral image, wherein the region of interest represents a region with even distribution of insulator pollution;
the database construction module is used for constructing a first database by taking the hyperspectral data as unlabeled data; the average spectrum data is used as marking data, the pollution state of the region of interest is obtained through an equivalent salt deposit method, a second database is constructed based on the average spectrum data, and a hyperspectral database comprising the first database and the second database is constructed;
the preprocessing module is used for preprocessing the spectrum data of the hyperspectral database through black-white correction, standard normal transformation and trend removal correction, and then adopting a selective segmentation principal component analysis method to perform grouping dimension reduction processing on the full wave band;
the identification module is used for identifying the pollution state of the pollution insulator corresponding to the spectrum data by constructing a multi-classification semi-supervised Newton support vector machine and taking the spectrum data subjected to the dimension reduction treatment as the input data of the multi-classification semi-supervised Newton support vector machine.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310478950.4A CN116519710A (en) | 2023-04-28 | 2023-04-28 | Method and system for detecting surface pollution state of composite insulator |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310478950.4A CN116519710A (en) | 2023-04-28 | 2023-04-28 | Method and system for detecting surface pollution state of composite insulator |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116519710A true CN116519710A (en) | 2023-08-01 |
Family
ID=87395481
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310478950.4A Pending CN116519710A (en) | 2023-04-28 | 2023-04-28 | Method and system for detecting surface pollution state of composite insulator |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116519710A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117058541A (en) * | 2023-08-03 | 2023-11-14 | 国网吉林省电力有限公司通化供电公司 | Insulator hyperspectral data acquisition system and method thereof |
-
2023
- 2023-04-28 CN CN202310478950.4A patent/CN116519710A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117058541A (en) * | 2023-08-03 | 2023-11-14 | 国网吉林省电力有限公司通化供电公司 | Insulator hyperspectral data acquisition system and method thereof |
| CN117058541B (en) * | 2023-08-03 | 2024-02-13 | 国网吉林省电力有限公司通化供电公司 | Insulator hyperspectral data acquisition system and method thereof |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112734692B (en) | Defect identification method and device for power transformation equipment | |
| CN112380952B (en) | Power equipment infrared image real-time detection and identification method based on artificial intelligence | |
| Chen et al. | A self organizing map optimization based image recognition and processing model for bridge crack inspection | |
| CN111784633B (en) | Insulator defect automatic detection algorithm for electric power inspection video | |
| Jahanshahi et al. | Parametric performance evaluation of wavelet-based corrosion detection algorithms for condition assessment of civil infrastructure systems | |
| CN111223087B (en) | Automatic bridge crack detection method based on generation countermeasure network | |
| CN113344475B (en) | Transformer bushing defect identification method and system based on sequence modal decomposition | |
| CN106951863B (en) | Method for detecting change of infrared image of substation equipment based on random forest | |
| CN111507426A (en) | No-reference image quality grading evaluation method and device based on visual fusion characteristics | |
| CN116612098B (en) | Insulator RTV spraying quality evaluation method and device based on image processing | |
| CN115661649B (en) | BP neural network-based shipborne microwave radar image oil spill detection method and system | |
| CN111105389A (en) | Detection method for pavement crack by fusing Gabor filter and convolutional neural network | |
| CN116519710A (en) | Method and system for detecting surface pollution state of composite insulator | |
| Huynh et al. | Multi-class support vector machines for paint condition assessment on the Sydney Harbour Bridge using hyperspectral imaging | |
| CN111563408B (en) | High-resolution image landslide automatic detection method with multi-level perception characteristics and progressive self-learning | |
| CN111667473A (en) | Insulator hydrophobicity grade judging method based on improved Canny algorithm | |
| CN104951800A (en) | Resource exploitation-type area-oriented remote sensing image fusion method | |
| Kong et al. | Pixel-level assessment model of contamination conditions of composite insulators based on hyperspectral imaging technology and a semi-supervised ladder network | |
| Liu et al. | Enhanced image no‐reference quality assessment based on colour space distribution | |
| CN101630369B (en) | Pedestrian detection method based on wavelet fractal characteristic | |
| CN117197763A (en) | Road crack detection method and system based on cross attention guide feature alignment network | |
| JP2840347B2 (en) | Board mounting inspection equipment | |
| Ji et al. | No-reference image quality assessment for dehazed images | |
| CN114596433A (en) | Insulator identification method | |
| CN113192018B (en) | Water-cooled wall surface defect video identification method based on fast segmentation convolutional neural network |
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 |