Ultraviolet Discharge Image - Hyperspectral Image - Contaminated Insulator Condition Assessment Method

Abstract

The present invention relates to a method for evaluating the state of contaminated insulators by combining ultraviolet discharge images and hyperspectral images. The ultraviolet discharge images and hyperspectral images of the contaminated insulators are synchronously collected and preprocessed, and the field of view matching of the ultraviolet discharge images and hyperspectral images is carried out. The abnormal discharge areas in the ultraviolet discharge images are extracted and discharge characteristic quantities are calculated. The reflectance curves of the abnormal discharge areas in the hyperspectral images are obtained and spectral curve characteristic quantities are calculated. An electro-optical joint analysis feature matrix Z is constructed based on the discharge characteristic quantities and spectral curve characteristic quantities. The electro-optical joint analysis feature matrix Z is normalized and input into a contaminated level classification model fused with multiple classifiers to evaluate the state of the contaminated insulators. The present invention provides a method for evaluating the insulator contamination state with complementary advantages, which is beneficial to the operation and maintenance and condition-based maintenance work of the insulators of transmission lines by power inspection personnel.

Description

Technical Field

[0001] The present invention relates to a method for evaluating the state of contaminated insulators by combining ultraviolet discharge images and hyperspectral images, and belongs to the on-line detection method for external insulation in the field of power transmission and distribution. Background Art

[0002] According to current research, insulator flashover due to contamination remains a serious problem affecting the safety of power systems, especially under adverse weather conditions. Therefore, detecting the contamination state of transmission line insulators has important practical significance for preventing flashover accidents and maintaining the reliable operation of power systems.

[0003] Existing pollution detection and characterization methods mainly include the equivalent salt deposit density method, surface contamination layer conductivity method, leakage current pulse counting method, large leakage current method, and lightning potential gradient method for insulator contamination. However, traditional detection methods cannot achieve non-contact detection, which not only requires a large amount of manpower, but also the data acquisition process is dangerous, complex, and vulnerable to external environmental influences. Therefore, it is necessary to find an efficient, convenient, accurate, and non-contact method to evaluate the contamination status of insulators.

[0004] Ultraviolet imaging technology is an accurate and effective method for detecting abnormal discharges. As the degree of insulator contamination increases, the probability of abnormal discharges occurring on the insulator surface will increase significantly. Therefore, the degree of contamination of insulators can be effectively characterized by detecting the degree of abnormal discharges occurring on the insulator surface. However, the detection results of ultraviolet imaging technology are greatly affected by conditions such as air humidity and temperature, so the information that can be obtained by ultraviolet imaging technology is limited. Hyperspectral imaging technology is a detection method that uses hundreds of extremely narrow electromagnetic wave cycles to obtain continuous spectra and images from visible light to near-infrared light. For a hyperspectral image, it is a three-dimensional data cube composed of two-dimensional image information and one-dimensional spectral information. Due to the nanoscale spectral resolution of hyperspectral images, they can accurately describe the reflection spectra of objects and have good recognition capabilities. Therefore, hyperspectral imaging technology has been widely applied in fields such as geotelemetry, food safety, and medical diagnosis. However, relying solely on the detection results of hyperspectral imaging technology, its accuracy still needs to be improved. Therefore, the present invention intends to combine electrical characteristics (ultraviolet discharge images) and optical characteristics (hyperspectral images) to achieve the integration of technologies and the joint analysis of data, and to achieve a highly confident joint diagnosis of the contamination state of insulators. Summary of the Invention

[0005] The present application provides a method for evaluating the state of contaminated insulators by combining ultraviolet discharge images and hyperspectral images, aiming to solve the problem of low reliability in traditional insulator contamination state detection.

[0006] To solve the above technical problems, the embodiments of the present application disclose the following technical solutions:

[0007] A method for evaluating the state of contaminated insulators by combining ultraviolet discharge images and hyperspectral images includes the following steps:

[0008] Step S01: Synchronously collect the ultraviolet discharge images and hyperspectral images of the contaminated insulators and perform image preprocessing, and carry out the field-of-view matching of the ultraviolet discharge images and hyperspectral images;

[0009] Step S02: Extract the abnormal discharge areas in the ultraviolet discharge images and calculate the discharge characteristic quantities, obtain the reflectance curves of the abnormal discharge areas in the hyperspectral images and calculate the spectral curve characteristic quantities, and construct an electro-optical joint analysis characteristic matrix Z based on the discharge characteristic quantities and spectral curve characteristic quantities;

[0010] Step S03: Normalize the electro-optical joint analysis characteristic matrix Z and input it into the contaminated level classification model fused by multiple classifiers to evaluate the state of the contaminated insulators.

[0011] In the method for evaluating the state of contaminated insulators by combining ultraviolet discharge images and hyperspectral images, in the step S01, the ultraviolet discharge images of the contaminated insulators are obtained through ultraviolet imaging, and the ultraviolet discharge images are preprocessed by image denoising and image enhancement; the hyperspectral images of the contaminated insulators are obtained through hyperspectral imaging, and the hyperspectral images are preprocessed by multivariate scatter correction, S-G filtering and black and white calibration; extract the feature points and their feature descriptors of the ultraviolet discharge images and hyperspectral images, and match their feature vectors to realize the field-of-view matching of the ultraviolet discharge images and hyperspectral images.

[0012] In the method for evaluating the state of contaminated insulators by combining ultraviolet discharge images and hyperspectral images, in the step S02, based on the ultraviolet discharge images, binaryzation processing is adopted and edge detection is used to extract the position coordinates [x i , y i of the abnormal discharge areas. At the same time, obtain the gray-scale distribution [H 紫 of the ultraviolet discharge images and the discharge area [S]; based on the hyperspectral images, extract the reflectance spectral information [r0,..., r i , y i at the position coordinates [x i of the abnormal discharge areas, and calculate its spectral derivative [di] and integral area [mi]. At the same time, obtain the gray-scale distribution [H i , y i of the hyperspectral images at the position coordinates [x 高 of the abnormal discharge areas; load the gray-scale distribution [H 紫 , discharge area [S], spectral derivative [di], integral area [mi], and gray-scale distribution [H 高 of the hyperspectral images in the ultraviolet discharge images into the electro-optical joint analysis characteristic matrix Z.

[0013] In the described method for evaluating the state of a contaminated insulator by combining ultraviolet discharge images and hyperspectral images, where r i is the spectral data of 176 bands.

[0014] In the described method for evaluating the state of a contaminated insulator by combining ultraviolet discharge images and hyperspectral images, in step S03, the electro-optical joint analysis feature matrix Z is normalized to obtain Z'; a classification model that fuses multiple different classification algorithms is established. In this classification model, each classification algorithm will output a classification result and its confidence level. Subsequently, the classification result is finally output through a weighted model. Z' is input into the classification model to output the recognition result of the contamination level.

[0015] The beneficial effects of the present invention are as follows:

[0016] (1) In the present invention, the abnormal discharge characteristics (electrical characteristics) of the contaminated insulator are obtained by using an ultraviolet imaging system, and the reflectance spectral characteristics (optical characteristics) of the contaminated insulator are obtained by using hyperspectral imaging technology, realizing the evaluation of the insulator contamination state from different angles, so it has a higher confidence level.

[0017] (2) The ultraviolet discharge image is less affected by natural light but more affected by environmental temperature and humidity. The hyperspectral image is less affected by environmental temperature and humidity but depends on environmental light conditions. Therefore, the two technologies can effectively complement each other's advantages.

[0018] (3) The intelligence of this method can provide strong support for the development of standardized power inspection work, avoiding interference with the detection results caused by human factors. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the technical solutions of the present application, the drawings required for use in the embodiments will be briefly introduced below. Obviously, for those of ordinary skill in the art, other drawings can also be obtained based on these drawings without creative efforts.

[0020] Figure 1 is a flowchart of a method for evaluating the state of a contaminated insulator by combining ultraviolet discharge images and hyperspectral images. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0021] To further describe the present invention, the following is a further description thereof with reference to the drawings.

[0022] See Figure 1, is a flowchart of a method for evaluating the condition of contaminated insulators by combining ultraviolet discharge images and hyperspectral images. It can be seen that the present application provides a method for evaluating the condition of contaminated insulators by combining ultraviolet discharge images and hyperspectral images, including the following steps:

[0023] S01: Synchronously collect the corona ultraviolet discharge images and hyperspectral images of the contaminated insulators and perform image preprocessing, and carry out field matching of the ultraviolet discharge images and hyperspectral images;

[0024] S011: Use an ultraviolet imaging system to obtain the ultraviolet discharge images of the contaminated insulators, and preprocess the ultraviolet discharge images through image denoising and image enhancement;

[0025] Among them, the image denoising method is: where w jk is the wavelet coefficient, λ is the wavelet coefficient threshold, and its determination method is σ is the variance of the noise signal, and N is the signal length.

[0026] Among them, the image enhancement method is V = V'·r γ , r ∈ [0, 1], where V is the transformed gray value, V' is the gray value before transformation, and the γ coefficient is adjusted according to the quality of the ultraviolet discharge image.

[0027] S012: Use a hyperspectral imaging system to obtain the hyperspectral images of the contaminated insulators, and preprocess the hyperspectral images through multivariate scattering correction, S-G filtering, and black and white calibration;

[0028] Among them, the method of multivariate scattering correction is where is the average value of the spectral data of the j-th band after final output correction, R ij is the spectral value of the j-th band of the i-th sample, n is the number of samples, is the average spectral matrix of the full band, m i is the slope (offset) of the linear regression, b i is the intercept (translation) of the linear regression, R i(MSC) is the result after correction.

[0029] Among them, the method of S-G filtering is to set the window width to 2m + 1, that is, the data points in each window are successively x = [-m, -m + 1,..., 0, 1,..., m, m + 1], and a polynomial of degree k - 1 is used to fit the data, then y = a0 + a1x + a2x 2 +... + a k- 1x k-1 , assuming there are n windows in total, then a system of k equations with n equations is formed. To make the equation have a solution, it is necessary to satisfy n ≥ k. Then, the fitting parameters A are determined by least squares fitting: Y(2m+1)×k = X (2m+1)×k ·A k×1 + E (2m+1)×k . The least squares solution of Xiao is as follows: The filtered Y value, i.e., the predicted value of Y is as follows:

[0030] where the black and white calibration method is In the formula, I0 is the original hyperspectral image, B is the all-black calibration image, and W is the all-white calibration image.

[0031] S013: Extract the feature points and their feature descriptors of the preprocessed ultraviolet discharge image and hyperspectral image, match their feature vectors, and achieve the field of view matching of the ultraviolet discharge image and hyperspectral image.

[0032] Among them, the method of field of view matching is to use the SUFR image matching method, extract the feature points and their feature descriptors in the ultraviolet discharge image and hyperspectral image respectively, match their feature vectors, and achieve the field of view matching of the ultraviolet discharge image and hyperspectral image.

[0033] S02: Extract the abnormal discharge points in the ultraviolet discharge image and calculate the discharge characteristic quantities, obtain the reflectance curve of the corresponding discharge area in the hyperspectral image and calculate the spectral curve characteristic quantities, and construct an electro-optical joint analysis feature matrix;

[0034] S021: For the ultraviolet discharge image, perform binarization processing and use the edge detection method to extract the position coordinates [x i , y i of the abnormal discharge area. At the same time, obtain the gray-scale distribution [H 紫 and discharge area [S] in the ultraviolet discharge image;

[0035] S022: For the hyperspectral image, extract the reflectance spectral information [r0,..., r i , y i at the position coordinates [x i , where r i is the spectral data of 176 bands, and further calculate the spectral derivative [di], integral area [mi]. At the same time, obtain the gray-scale distribution [H i , y i of the hyperspectral image at the position coordinates [x 高 ;

[0036] S023: Combine the gray-scale distribution [H 紫 , discharge area [S], spectral derivative [di], integral area [mi], and gray-scale distribution [H 高Load it into the electro-optical joint analysis feature matrix Z.

[0037] S03: After normalizing the electro-optical joint analysis feature matrix, input the matrix into the contamination level classification model that fuses multiple classifiers, so as to realize the state evaluation of contaminated insulators for the combined ultraviolet discharge image - hyperspectral image.

[0038] S031: Perform a normalization operation on the electro-optical joint analysis feature matrix Z to obtain Z';

[0039] S032: Establish a classification model that fuses multiple different classification algorithms. In this classification model, each classification algorithm will output the classification result and its confidence level, and then finally output the classification result through a weighted model. Input Z' into the classification model to output the recognition result of the contamination level.

[0040]

[0041] In the formula, f 0、I、II、III、IV are respectively the probabilities that the classification results are the diagnosis results for each contamination level, and f 波谱角1、2、3、4、5 are respectively the probabilities that the spectral angle matching determines the 5 contamination levels, and f BP1、2、3、4、5 are respectively the probabilities that the BP neural network determines the 5 contamination levels, and f 支持向量机1、2、3、4、5 are respectively the probabilities that the support vector machine matching determines the 5 contamination levels, and f 平行六面体1、2、3、4、5 are respectively the probabilities that the parallelepiped determines the 5 contamination levels. a, b, c, d, e are respectively the weight coefficients of various methods. The weight coefficients need to be trained by taking the above formula as the model and substituting a large number of actual measurement results. Finally, compare the values of f 0、I、II、III、IV The one with the largest probability is the classification result.

[0042] Although the embodiments of the present invention have been described above in conjunction with the accompanying drawings, the present invention is not limited to the above specific embodiments and application fields. The above specific embodiments are merely illustrative and guiding, rather than restrictive. Those of ordinary skill in the art can also make many forms under the inspiration of this specification and without departing from the scope protected by the claims of the present invention, and these all belong to the scope of protection of the present invention.

Claims

1. A method for evaluating the condition of contaminated insulators by combining ultraviolet discharge images and hyperspectral images, characterized in that, It includes the following steps: Step S01: Synchronously collect the ultraviolet discharge images and hyperspectral images of the contaminated insulators and perform image preprocessing, and carry out the field-of-view matching of the ultraviolet discharge images and hyperspectral images; Step S02: Extract the abnormal discharge area in the ultraviolet discharge image and calculate the discharge characteristic quantities, obtain the reflectance curve of the abnormal discharge area in the hyperspectral image and calculate the spectral curve characteristic quantities, and construct an electro-optical joint analysis feature matrix Z based on the discharge characteristic quantities and spectral curve characteristic quantities. Among them, based on the ultraviolet discharge image, perform binarization processing and use edge detection to extract the position coordinates [x i , y i , and at the same time obtain the gray-scale distribution [H 紫 and the discharge area [S] of the ultraviolet discharge image; based on the hyperspectral image, extract the position coordinates [x i , y i of the abnormal discharge area and the reflectance spectral information [r0,..., r i , and calculate its spectral derivative [di] and integral area [mi]. At the same time, obtain the gray-scale distribution [H i , y i of the hyperspectral image at the position coordinates of the abnormal discharge area; load the gray-scale distribution [H 高 , discharge area [S], spectral derivative [di], integral area [mi], and gray-scale distribution [H 紫 of the ultraviolet discharge image, and the gray-scale distribution [H 高 of the hyperspectral image into the electro-optical joint analysis feature matrix Z; Step S03: Normalize the electro-optical joint analysis feature matrix Z and input it into the contamination level classification model fused by multiple classifiers to evaluate the state of the contaminated insulators.

2. The method for evaluating the state of a contaminated insulator by combining ultraviolet discharge images and hyperspectral images according to claim 1, characterized in that In the step S01, the ultraviolet discharge images of the contaminated insulators are obtained through ultraviolet imaging, and the ultraviolet discharge images are preprocessed by image denoising and image enhancement; the hyperspectral images of the contaminated insulators are obtained through hyperspectral imaging, and the hyperspectral images are preprocessed by multivariate scatter correction, S-G filtering and black-and-white calibration; the feature points and their feature descriptors of the ultraviolet discharge images and hyperspectral images are extracted, and the feature vectors are matched to achieve the field-of-view matching of the ultraviolet discharge images and hyperspectral images.

3. A method for evaluating the condition of contaminated insulators by combining ultraviolet discharge images and hyperspectral images according to claim 1, characterized in that, where r i is the spectral data of 176 bands.

4. The method for evaluating the state of a contaminated insulator by combining ultraviolet discharge images and hyperspectral images according to claim 1, wherein, In the step S03, the electro-optical joint analysis feature matrix Z is normalized to obtain Z'; a classification model fused by multiple different classification algorithms is established. In this classification model, each classification algorithm will output the classification result and its confidence level, and then the classification result is finally output through the weighted model. Z' is input into the classification model to output the contamination level recognition result.

Images