Composite insulator multispectral feature extraction method and device considering corona characteristics

Abstract

The invention relates to the technical field of image feature extraction, in particular to a composite insulator multispectral feature extraction method and device considering corona characteristics, and the method comprises the steps: collecting a multispectral image of a composite insulator, and carrying out the registration and fusion of the multispectral image, and obtaining a composite image; extracting an insulator image in the composite image; for any connected domain in the insulator image, whether the connected domain is subjected to corona change or not and whether the connected domain is influenced by sunshine or not are analyzed according to brightness differences, texture features and time sequence change features of the same connected domain in HSV spaces of visible light, infrared and ultraviolet images, and according to the time sequence change features of the connected domain influenced by sunshine; and further analyzing the key feature region of the composite insulator so as to extract the key feature region which is greatly influenced by the corona phenomenon in the insulator image. The invention aims to effectively eliminate the interference of environmental factors such as sunlight irradiation, thereby improving the accuracy and reliability of corona detection.

Description

Technical Field

[0001] This application relates to the technical field of image feature extraction, and specifically to a method and device for multi-spectral feature extraction of composite insulators considering corona characteristics. Background Art

[0002] Composite insulators are widely used in power systems, and their performance directly affects the safe operation of the power grid. During operation, when the surface electric field strength of the composite insulator reaches a certain value, corona discharge will occur. The high-energy charged particles generated by corona discharge will bombard the surface of the insulator, destroying its molecular structure. At the same time, corrosive substances such as nitric acid generated by corona discharge will also cause chemical corrosion to the surface of the insulator, resulting in a decrease in the hydrophobicity of the insulator, and defects such as cracks and holes on the surface, affecting its insulation performance. Therefore, timely detection and evaluation of the corona characteristics of composite insulators are crucial for maintaining the reliability of power systems.

[0003] The existing technology mainly uses ultraviolet imaging and infrared imaging technologies to detect composite insulators, and evaluates whether the composite insulators have phenomena such as aging and insulation deterioration by extracting relevant characteristic parameters and conducting data analysis and comparison.

[0004] However, when the existing technology detects composite insulators through multi-spectral imaging methods, it only analyzes and extracts the characteristics of the composite insulators in the ultraviolet image and infrared image at the current moment, but does not deeply analyze the influence differences of sunlight factors and corona characteristics on the composite insulators, resulting in deviations in the extracted characteristics, and further causing false detection phenomena, affecting the accuracy of the detection of composite insulators. Summary of the Invention

[0005] In order to solve the above technical problems, this application provides a method and device for multi-spectral feature extraction of composite insulators considering corona characteristics, and the specific technical solutions adopted are as follows: In a first aspect, an embodiment of this application provides a method for multi-spectral feature extraction of composite insulators considering corona characteristics, and the method includes the following steps: Collect multi-spectral images of the composite insulator, perform registration and fusion to obtain a composite image; the multi-spectral images include visible light images, infrared images, and ultraviolet images; Extract the composite insulator region in the composite image converted to a grayscale image through a neural network to obtain an insulator image; For any connected domain in the insulator image, preliminarily analyze whether the connected domain has corona changes according to the brightness differences of the same connected domain in the HSV spaces of visible light, infrared, and ultraviolet images; Further analyze whether the connected domain is affected by sunlight according to the texture characteristics of the connected domain in the insulator image; According to the sequential change characteristics of the insulator images obtained for the same connected region at different times and affected by sunlight, the key feature regions of the composite insulator are further analyzed; Combined with the preliminary analysis of whether corona changes occur in the same connected region, the key feature regions in the insulator image that are greatly affected by the corona phenomenon are extracted.

[0006] Preferably, the method for preliminarily analyzing whether corona changes occur in the connected region includes: Calculating the similarity of the brightness values in the HSV space of the infrared and ultraviolet images for the same connected region; Calculating the normalized value of the average brightness of any connected region in the HSV space of any spectral image, and calculating the range of the normalized values of the average brightness of all connected regions in the HSV space of any spectral image; Using the similarity, the normalized value of the average brightness calculated for any connected region in all spectral images, and the range, to determine the corona brightness confidence level of the any connected region, and preliminarily analyzing whether corona changes occur in the connected region based on the magnitude of its value.

[0007] Preferably, the method for determining the corona brightness confidence level of any connected region further includes: Denote the corona brightness confidence level of the v-th connected region as , and the expression is: ; where is the similarity between the infrared HSV image and the ultraviolet HSV image of the v-th connected region; , are respectively the normalized values of the average brightness of the v-th connected region in the infrared HSV image, ultraviolet HSV image, and visible light HSV image; , are respectively the brightness ranges of the v-th connected region in the infrared HSV image, ultraviolet HSV image, and visible light HSV image.

[0008] Preferably, the method for further analyzing whether the connected region is affected by sunlight includes: Obtaining the degree of dispersion of the texture feature values of all pixel points in the insulator image of the connected region; Forming a sequence of the gray values of all pixel points between the two pixel points with the largest and smallest gray values in the connected region; Denote the proportion of the number of negative elements after differentiating the sequence as the gray value decreasing index of the connected region; Obtaining the average value of the gradient values of all edge pixel points in the connected region; Using the degree of dispersion, gray-scale decreasing exponent, and the mean value of the calculated gradient values within the connected domain, determine the light influence coefficient of the connected domain, and further analyze whether the connected domain is affected by sunlight based on the magnitude of its value.

[0009] Preferably, the method for determining the light influence coefficient of the connected domain further includes: Denote the light influence coefficient of the v-th connected domain as , and the expression is: ; where is the gray-scale decreasing exponent of the v-th connected domain; is the degree of dispersion of all texture feature values in the v-th connected domain; is the mean value of the gradient values of all edge pixel points in the v-th connected domain.

[0010] Preferably, the method for further analyzing the key feature region of the composite insulator includes: Generate connected domains for all insulator images, and denote all corresponding connected domains after registration of this connected domain in other insulator images as the asynchronous connected domains of this connected domain; Calculate the degree of dispersion of the light influence coefficients of this connected domain and all its asynchronous connected domains; Calculate the Hu-moment values between this connected domain and each of its asynchronous connected domains through the Hu-moment function; Denote the product of the mean value of all calculated Hu-moment values and the degree of dispersion of all Hu-moment values as the contour change index of this connected domain; Using the light influence coefficient of this connected domain, the calculated degree of dispersion of the light influence coefficient, and the contour change coefficient, determine the temporal change coefficient of this connected domain, and further analyze whether this connected domain is the key feature region of the composite insulator based on the magnitude of its value.

[0011] Preferably, the method for determining the temporal change coefficient of the connected domain further includes: Denote the temporal change coefficient of the v-th connected domain as , and the expression is: ; where is the light influence coefficient of the v-th connected domain; is the degree of dispersion of the light influence coefficients of the v-th connected domain and all its asynchronous connected domains; is the contour change index of the v-th connected domain.

[0012] Preferably, the method for extracting the key feature region in the insulator image that is greatly affected by the corona phenomenon is: Take the result of reverse fusion of the corona brightness confidence and the temporal change coefficient of the same connected domain as the corona confidence of this connected domain; Obtain the segmentation threshold of the corona confidence of all connected components in the insulator image, and use the connected components corresponding to the corona confidence greater than the segmentation threshold as the key feature regions in the insulator image that are more affected by the corona phenomenon.

[0013] Preferably, the method for determining the corona confidence of the connected component is further determined by the ratio result of the corona brightness confidence and the time series change coefficient of the same connected component.

[0014] In a second aspect, another embodiment of the present application provides a multi-spectral feature extraction device for composite insulators considering corona characteristics, which implements the method for extracting multi-spectral features of composite insulators considering corona characteristics described in any one of the above. The device includes: An image processing module, configured to perform registration, fusion, and extraction operations on the composite insulator region of the acquired multi-spectral image, and input the obtained insulator image into the feature extraction module; A feature extraction module, configured to analyze the brightness, texture, and time series change characteristics of the insulator image to consider the influence of corona on the composite insulator, the influence of sunlight irradiation, and the time series change characteristics; A corona region extraction module, configured to extract the key feature regions of the composite insulator from the insulator image for the various features analyzed by the feature extraction module.

[0015] The present application has at least the following beneficial effects: The present application first constructs the corona brightness confidence according to the corona characteristics of the composite insulator, thereby initially identifying the regions that may be affected by corona; then, considering that sunlight irradiation may cause false detection, by constructing the light influence coefficient, the error influence generated by sunlight factors can be further filtered; finally, considering the time series change characteristics of corona characteristics and sunlight factors, a time series change coefficient is constructed, so as to accurately locate the key regions affected by corona characteristics.

[0016] Aiming at the problem that the prior art does not deeply analyze the influence differences of sunlight factors and corona characteristics on composite insulators, resulting in poor feature extraction effects; the present application can effectively eliminate the interference of environmental factors such as sunlight irradiation by constructing corona confidence, thereby improving the accuracy and reliability of corona detection, and further being able to better extract key features according to the corona characteristics of composite insulators to ensure the safe and stable operation of high-voltage lines. Description of the Drawings

[0017] Figure 1 It is a flowchart of the method for extracting the key feature regions in the image that are more affected by the corona phenomenon provided by the present application; Figure 2Flowchart of the steps of the method for extracting multi - spectral features of composite insulators considering corona characteristics provided by this application. Detailed implementation manner

[0018] This application aims at the lack of in - depth analysis of the differences in the impacts of sunlight factors and corona characteristics on composite insulators, and provides a flowchart of a method for extracting key feature regions in an image that are greatly affected by the corona phenomenon. As shown in the appendix Figure 1 The method includes the following steps: Collect multi - spectral images of composite insulators, perform registration and fusion to obtain a composite image; the multi - spectral images include visible - light images, infrared images, and ultraviolet images. Extract the region of the composite insulator in the grayscale image converted from the composite image through a neural network to obtain an insulator image. For any connected domain in the insulator image, preliminarily analyze whether the connected domain has corona changes according to the brightness differences of the same connected domain in the HSV spaces of visible - light, infrared, and ultraviolet images respectively. According to the texture features of the connected domain in the insulator image, further analyze whether the connected domain is affected by sunlight. According to the temporal - change characteristics of the insulator images obtained at different times for the same connected domain under the influence of sunlight, further analyze the key feature regions of the composite insulator. And combine the preliminary analysis of whether the same connected domain has corona changes to extract the key feature regions in the insulator image that are greatly affected by the corona phenomenon.

[0019] Specifically, this application provides a flowchart of the steps of the method for extracting multi - spectral features of composite insulators considering corona characteristics. As shown in the appendix Figure 2 The method includes the following steps: Step 1: Obtain relevant images of the composite insulator and perform pre - processing.

[0020] This application uses a sensor integration method to integrate a CMOS sensor sensitive to ultraviolet, a near - infrared thermal imaging sensor, and a standard visible - light sensor into the same device, and uses a drone to load the integrated device to take pictures of a 500kV extra - high - voltage line.

[0021] After that, perform registration and fusion on these multi - spectral images of different bands to obtain a fused composite image; image registration algorithms include SIFT, SURF, etc., and image fusion algorithms include cascade feature fusion algorithm, weighted fusion algorithm, etc. This application uses the SIFT algorithm and the cascade feature fusion algorithm for image configuration and fusion. Among them, the multi - spectral images include visible - light images, infrared images, and ultraviolet images.

[0022] This application collects a composite image every t minutes, and a total of N images are collected. Among them, the collection interval t and the collection quantity N can be self-selected by the implementer according to the actual situation. In this embodiment, the value range of t is limited to 20 - 30, the value of t is 25, and the value of N is 10.

[0023] Based on the image processing module in the composite insulator multi-spectral feature extraction device, denoising processing is performed on all the composite images obtained by the integrated device. The denoising algorithms include but are not limited to Gaussian filtering, median filtering, bilateral filtering, etc. In this embodiment, Gaussian filtering is used for denoising processing; after denoising, all the composite images are converted into grayscale images.

[0024] Since the surrounding environmental factors will inevitably be captured during the UAV shooting, this application extracts the composite insulator area in the grayscale image converted from the composite image through a neural network to obtain the insulator image.

[0025] As a preferred implementation manner, this application performs one-hot encoding on the composite image and the label data and then sends them into the neural network for ROI region extraction. The label data is manually marked, and the composite insulator area and the non-composite insulator area are represented by the numbers 0 and 1 respectively. The output of the neural network is the composite insulator area in the composite image, denoted as the insulator image. Among them, the neural network can be ResNet18, CNN, YOLO, etc. In this embodiment, the ResNet18 network is used; the training of the neural network is a well-known technology, and the specific process will not be elaborated here.

[0026] After preprocessing the composite image, the insulator image is further analyzed and feature-extracted through the feature extraction module in the composite insulator multi-spectral feature extraction device, as follows: Step two: Perform a preliminary analysis on the collected images according to the brightness.

[0027] Around the high-voltage wire, due to the action of the strong electric field, air molecules are easily ionized, thus causing the corona phenomenon. When the composite insulator has corona, the heat generated by the corona discharge will cause the local temperature of the insulator to rise. In the infrared image, the higher the temperature, the higher the brightness will be; at the same time, the ultraviolet light generated by the corona discharge will also increase the brightness of the corresponding area in the ultraviolet image, making the brightness change of the pixel points in the infrared image and the ultraviolet image have a certain synchronism. However, in the visible light image, the brightness change in the corona area is not significant and does not have a synchronous change characteristic.

[0028] Therefore, in infrared and ultraviolet images, there will be a large difference in the brightness values of pixel points between the characteristic regions with corona changes and the normal regions, while in visible light images, there will be no such large difference in brightness. Thus, it is possible to preliminarily analyze whether corona changes have occurred in a region based on the significant brightness differences of the same region in infrared, ultraviolet, and visible light images respectively.

[0029] Accordingly, for any connected region in the insulator image in this application, based on the brightness differences of the same connected region in the HSV spaces of visible light, infrared, and ultraviolet images respectively, it is preliminarily analyzed whether corona changes have occurred in this connected region.

[0030] Taking the i-th insulator image in this application as an example: The connected regions in the i-th insulator image are obtained through a connected region extraction algorithm. The connected region extraction algorithm includes but is not limited to region growing algorithms, seed filling algorithms, etc. In this embodiment, the region growing algorithm is used. The region growing algorithm is a well-known technology and will not be elaborated here.

[0031] Denote the composite image where the i-th insulator image is located before ROI region extraction as the i-th composite image. Obtain the infrared image, ultraviolet image, and visible light image of the i-th composite image before image fusion, and perform HSV image conversion respectively.

[0032] According to the region ranges of each connected region inside the i-th insulator image in the i-th composite image, extract the corresponding connected region areas from the infrared HSV image, ultraviolet HSV image, and visible light HSV image corresponding to the i-th composite image respectively.

[0033] Furthermore, as a preferred implementation manner, this application calculates the similarity of the brightness values of the same connected region in the HSV spaces of infrared and ultraviolet images respectively.

[0034] Taking the v-th connected region in this embodiment as an example: Respectively obtain the brightness value sequences of the v-th connected region in the infrared HSV image and the ultraviolet HSV image. Among them, the arrangement order of the elements in the two brightness value sequences obtained in the two images is the same, and the elements at each corresponding position are the brightness values of the pixel points at the same position. And calculate the absolute value of the correlation coefficient between the two brightness value sequences, which is denoted as the similarity between the v-th connected region in the infrared HSV image and the ultraviolet HSV image. The calculation of the correlation coefficient includes but is not limited to Pearson similarity coefficient, Spearman similarity coefficient, cosine similarity, etc. In this embodiment, the Spearman similarity coefficient is used. The similarity can reflect whether the brightness changes of the v-th connected region in the infrared image and the ultraviolet image have synchrony caused by the corona phenomenon.

[0035] Furthermore, as a preferred embodiment, the present application calculates the normalized value of the brightness mean of any connected domain in the HSV space of any spectral image, and calculates the range of the normalized values of the brightness mean of all connected domains in the HSV space of any spectral image.

[0036] In this embodiment, the infrared HSV image of the infrared image in the HSV space is taken as an example, and the ratio between the brightness mean of the vth connected domain and the maximum brightness mean of all connected domains is calculated, which is recorded as the normalized value of the brightness mean of the vth connected domain in the infrared HSV image; the normalized value of the brightness mean can reflect the brightness prominence of the vth connected domain in the infrared HSV image. The range between the normalized values of the brightness mean of all connected domains in the infrared HSV image is calculated, which is recorded as the brightness range of the infrared HSV image. Among them, the normalized value of the brightness mean of the vth connected domain in the infrared HSV image can also be obtained by normalizing the brightness mean of the vth connected domain in the infrared HSV image using a known normalization method. Specifically, maximum and minimum normalization, linear normalization, etc. can be used.

[0037] It should be noted that the normalized value of the brightness mean and the brightness range of the vth connected domain in the ultraviolet HSV image and the visible light HSV image are calculated using the same method as the normalized value of the brightness mean and the brightness range of the vth connected domain in the infrared HSV image.

[0038] Furthermore, as a preferred embodiment, the present application uses the similarity and the normalized value and range of the brightness mean calculated in all spectral images of any connected domain to determine the corona brightness confidence of any connected domain, and uses its numerical value to preliminarily analyze whether corona changes occur in the connected domain.

[0039] In this embodiment, the corona brightness confidence of the vth connected domain is recorded as , the expression is: Where, is the similarity between the infrared HSV image and the ultraviolet HSV image of the vth connected domain; 、 are the normalized values of the mean brightness of the v-th connected domain in the infrared HSV image, ultraviolet HSV image, and visible light HSV image respectively; 、 are the brightness extremes of the v-th connected domain in the infrared HSV image, ultraviolet HSV image, and visible light HSV image, respectively.

[0040] It should be understood that the corona brightness confidence comprehensively considers the brightness change synchronization and brightness prominence of the connected domain in infrared, ultraviolet and visible light images. The larger it is, the better the brightness synchronization of the v-th connected region in the infrared and ultraviolet images is reflected. and The larger it is, the higher the brightness significance of the v-th connected region in the infrared and ultraviolet images is relative to the visible light image, and the greater the degree of change in the overall brightness is.

[0041] Step 3: Further analyze the acquired images according to the texture features.

[0042] Furthermore, since composite insulators are very susceptible to the external environment, there is a large temperature difference between the sunny side and the shady side of the insulator in the sun, and there will be a reflection phenomenon on the surface of the composite insulator, and the brightness difference of visible light may also be large. Therefore, if the insulator features are extracted only through the gray value, there may be deviations, and it is necessary to further combine the corona feature and the sunlight irradiation feature to further screen the feature extraction area.

[0043] Since the brightness of the area generated by sunlight irradiation has texture feature changes, the fluctuation of the gray value within the area shows a smooth gradient. And corona discharge is a local discharge phenomenon caused by a local high electric field, and the area caused by corona usually shows a texture with obvious edges and local sharp changes in the image.

[0044] Accordingly, the present application further analyzes whether the connected region is affected by sunlight irradiation according to the texture features of the connected region in the insulator image.

[0045] As a preferred implementation manner, the present application obtains the degree of dispersion of the texture feature values of all pixel points in the insulator image of the connected region, which is used to reflect the texture difference between each pixel point in the connected region.

[0046] In this embodiment, all pixel points in the v-th connected region of the insulator image are used as inputs, and a texture analysis algorithm is used to obtain the texture feature value of each pixel point in the connected region, and the degree of dispersion between all feature values is calculated. The texture analysis algorithm includes but is not limited to the LBP algorithm, the MLBP algorithm, etc. The present application uses the MLBP algorithm, and the calculated neighborhood size takes the value of a*a. In this embodiment, the value range of a is 3-5, and the value in this embodiment is 3; the degree of dispersion can be calculated by variance, standard deviation, mean square deviation, etc. The present application uses variance calculation. The degree of dispersion between texture feature values can reflect the texture difference between each pixel point in the v-th connected region; the larger the degree of dispersion is, the more likely it is that the v-th connected region is affected by corona phenomenon and thus has a higher brightness.

[0047] Furthermore, since the area irradiated by sunlight has a smooth gradient feature, the change in its grayscale value shows a certain regularity and continuity. Therefore, from the pixel point with the highest brightness to the pixel point with the lowest brightness, its grayscale value must continuously decrease, while the corona area caused by the corona characteristic does not have such a feature.

[0048] As a preferred implementation, in this application, the grayscale values of all pixel points between the two pixel points with the largest and smallest grayscale values in the connected domain are used to form a sequence. The proportion of the number of negative elements after differentiating the sequence is recorded as the grayscale decreasing index of the connected domain, which is used to reflect the smoothness of the grayscale value change within the connected domain.

[0049] In this embodiment, in the v-th connected domain, a straight line with a width of 1 pixel point is constructed between the two pixel points with the largest and smallest grayscale values, and starting from the pixel point with the largest grayscale value until the pixel point with the smallest grayscale value, a pixel value sequence is constructed according to the grayscale values of all pixel points on the straight line.

[0050] It should be noted that: if there are multiple pixel points with the largest or smallest grayscale value in the v-th connected domain, the pixel points with the largest and smallest grayscale values with the closest Euclidean distance to each other are selected to construct the straight line.

[0051] Obtain the first-order difference sequence of the pixel value sequence, and record the ratio of the number of all negative elements in the first-order difference sequence to the data length of the first-order difference sequence as the grayscale decreasing index of the v-th connected domain. The grayscale decreasing index can reflect the smoothness of the grayscale value change within the v-th connected domain, and further indicate whether the region is a grayscale gradient feature caused by sunlight irradiation or affected by other factors.

[0052] Furthermore, as a preferred implementation, this application obtains the mean value of the gradient values of all edge pixel points within the connected domain, which is used to reflect the edge significance degree of the connected domain.

[0053] In this embodiment, all edge pixel points in the v-th connected domain are obtained through an edge detection algorithm, and the mean value of the gradient values of all edge pixel points is obtained through a sobel operator to reflect the edge significance degree of the v-th connected domain. The edge detection algorithm includes but is not limited to Canny, Prewitt, Laplacian, etc. This application uses Canny edge detection.

[0054] Furthermore, as a preferred implementation, this application uses the calculated dispersion degree, grayscale decreasing index, and the mean value of the calculated gradient values within the connected domain to determine the illumination influence coefficient of the connected domain, and further analyzes whether the connected domain is affected by sunlight irradiation based on the value of the illumination influence coefficient.

[0055] In this embodiment, the light influence coefficient of the v-th connected component is denoted as , and the expression is: ; where is the gray-scale decreasing exponent of the v-th connected component; is the dispersion degree of all texture feature values in the v-th connected component; is the mean value of the gradient values of all edge pixel points in the v-th connected component.

[0056] It should be understood that the light influence coefficient can reflect the influence degree of sunlight irradiation on the v-th connected component. If the light influence coefficient is larger, it reflects that the brightness change of the v-th connected component is mainly affected by sunlight irradiation, rather than the possibility caused by the temperature increase due to the corona phenomenon and the electric field discharge phenomenon.

[0057] Step Four: Perform an overall analysis of the collected images according to the temporal variation.

[0058] Furthermore, by analyzing the variation differences of the corona characteristics and sunlight factors in time series, further accurate feature region extraction is carried out, specifically as follows: Although corona discharge will cause the composite insulator to age, and the aging is irreversible, since the aging requires long-term changes, the temperature of the corona region of the composite insulator will not fluctuate greatly in a short time, and at the same time, the area will not change greatly. And due to the movement of clouds and the change of the sun angle, the light intensity will change, so the area and gray value of the sunlight irradiation region both have large fluctuations, which will also cause the light influence coefficient in different insulator images to fluctuate. Thus, the key feature regions of the composite insulator can be further screened according to the temporal variation of the images at different times.

[0059] Accordingly, this application further analyzes the key feature regions of the composite insulator according to the temporal variation characteristics of the insulator images obtained at different times for the same connected component affected by sunlight irradiation.

[0060] As a preferred embodiment, this application generates connected components for all insulator images, and the corresponding connected components of the connected component registered in other insulator images are respectively denoted as the heterochronous connected components of the connected component.

[0061] This embodiment still takes the v-th connected component in the i-th insulator image as an example. According to the relevant parameters of the above-mentioned region growing algorithm, connected components are generated for all insulator images. Then, the SIFT image matching algorithm is used to register all insulator images, so as to obtain the connected components in other insulator images corresponding to the v-th connected component in the i-th insulator image, which is denoted as the asynchronous connected component of the v-th connected component in the i-th insulator image.

[0062] Furthermore, the present application calculates the degree of dispersion of the illumination influence coefficients of this connected component and all its asynchronous connected components.

[0063] In this embodiment, according to the calculation principle of the illumination influence coefficient of the above-mentioned v-th connected component, the illumination influence coefficients of all asynchronous connected components are obtained, and the degree of dispersion between all illumination influence coefficients is calculated. In this embodiment, variance is used to calculate the degree of dispersion. The degree of dispersion between illumination influence coefficients can reflect the texture feature differences of the v-th connected component at different acquisition times.

[0064] After that, the present application calculates the Hu moment values between this connected component and each of its asynchronous connected components through the Hu moment function; the product of the mean value of all calculated Hu moment values and the degree of dispersion of all Hu moment values is denoted as the contour change index of this connected component.

[0065] In this embodiment, the Hu moment values between the v-th connected component in the i-th insulator image and its asynchronous connected components are calculated through the Hu moment function respectively, the mean value and the degree of dispersion between all Hu moment values are calculated respectively, and the product of the mean value and the degree of dispersion is denoted as the contour change index of the v-th connected component. The Hu moment value can reflect the contour difference between two connected components, and the contour change index can reflect the degree of change of the contour of the v-th connected component in time series. The larger the contour change index, the greater the change in the contour shape of the v-th connected component over time, and the more it does not conform to the change characteristics caused by the corona phenomenon.

[0066] Furthermore, as a preferred implementation manner, the present application uses the illumination influence coefficient of this connected component, the calculated degree of dispersion of the illumination influence coefficient, and the contour change coefficient to determine the time series change coefficient of this connected component, and further analyzes whether this connected component is a key feature region of a composite insulator by using the numerical value thereof, which is used to reflect the stability of this connected component and judge whether this connected component conforms to the key feature region of the change caused by corona.

[0067] In this embodiment, the time series change coefficient of the v-th connected component is denoted as , and the expression is: ; in the formula, is the illumination influence coefficient of the v-th connected component; is the degree of dispersion of the illumination influence coefficients of the v-th connected region and all its asynchronous connected regions; is the contour change index of the v-th connected region.

[0068] It should be understood that the time-series change coefficient can reflect the degree of change in the illumination influence and contour shape of the v-th connected region at different times. The smaller the time-series change coefficient, the better the stability of the v-th connected region is reflected, and the more it conforms to the change characteristics caused by corona.

[0069] Step Five: Extract the key feature regions in the image that are greatly affected by the corona phenomenon based on the above analysis results.

[0070] Based on the above analysis, on the basis of Step 4 of this application, further combined with the preliminary analysis of whether corona changes occur in the same connected region, to extract the key feature regions in the insulator image that are greatly affected by the corona phenomenon.

[0071] As a preferred implementation manner, the result of the reverse fusion of the corona brightness confidence of the same connected region and the time-series change coefficient is used as the corona confidence of the connected region.

[0072] It can be understood that the reverse fusion is a fusion method such as subtraction and division between data. The specific reverse fusion method is determined by the implementer according to the actual situation to determine a suitable fusion method, and this application does not make special restrictions.

[0073] In this embodiment, the ratio of the corona brightness confidence of the same connected region to the time-series change coefficient is used as the corona confidence of the connected region.

[0074] Taking the corona confidence of the v-th connected region as an example, the expression is: ; in the formula, is the corona brightness confidence of the v-th connected region; is the time-series change coefficient of the v-th connected region; is a preset tuning parameter, and in order to avoid the denominator being 0, it is taken from the range [0.005, 0.01]. The value has no influence on the calculation, and the implementer can take the value by himself.

[0075] The corona confidence can more accurately identify the key regions affected by the corona phenomenon by comprehensively considering factors such as brightness change, sunlight irradiation influence, and time-series change.

[0076] Furthermore, this application obtains the segmentation threshold of the corona confidence of all connected regions in the insulator image, and takes the connected region corresponding to the corona confidence greater than the segmentation threshold as the key feature region in the insulator image that is greatly affected by the corona phenomenon.

[0077] In this embodiment, the corona confidence degrees of all connected components in the i-th insulator image are obtained, and all the corona confidence degrees are used as the input of the Otsu thresholding method to output a segmentation threshold. That is, the connected components are divided into two parts by the Otsu thresholding method. The connected components corresponding to the corona confidence degrees greater than the segmentation threshold are denoted as corona-connected components, which are the key feature regions in the composite insulator that are more affected by the corona phenomenon. By closely monitoring the key feature regions, the safe and stable operation of the high-voltage line can be ensured.

[0078] Based on the same inventive concept as the above method, an embodiment of the present application also provides a multi-spectral feature extraction device for a composite insulator considering corona characteristics. The device implements the multi-spectral feature extraction method for a composite insulator considering corona characteristics described in any one of the above. The device includes: An image processing module, configured to perform registration, fusion, and extraction operations on the obtained multi-spectral image for the composite insulator region, and input the obtained insulator image into the feature extraction module; A feature extraction module, configured to analyze the brightness, texture, and temporal variation characteristics of the insulator image to consider the influence of the corona phenomenon, sunlight irradiation, and temporal variation characteristics on the composite insulator; A corona region extraction module, configured to extract the key feature regions of the composite insulator from the insulator image for various features analyzed by the feature extraction module.

[0079] Those skilled in the art will readily think of other embodiments of the present application after considering the specification and practicing the invention herein. The present application is intended to cover any variations, uses, or adaptations of the present application, which follow the general principles of the present application and include the common general knowledge or conventional technical means in the technical field not invented by the present application.

[0080] It should be understood that the present application is not limited to the exact structures described above and shown in the drawings, and various modifications and changes can be made without departing from its scope.

Claims

1. A method for extracting multi - spectral features of composite insulators considering corona characteristics, characterized in that The method includes: Collecting multi - spectral images of composite insulators, registering and fusing them to obtain a composite image; the multi - spectral images include visible light images, infrared images, and ultraviolet images; Extracting the composite insulator region in the composite image converted to a grayscale image through a neural network to obtain an insulator image; For any connected component in the insulator image, preliminarily analyzing whether corona change occurs in this connected component according to the brightness differences of the same connected component in the HSV spaces of visible light, infrared, and ultraviolet images; Further analyzing whether this connected component is affected by sunlight according to the texture features of this connected component in the insulator image; According to the temporal change characteristics of the insulator images obtained at different times for the same connected component being affected by sunlight, further analyzing the key feature regions of the composite insulator; And combining the preliminary analysis of whether corona change occurs in the same connected component to extract the key feature regions in the insulator image that are greatly affected by the corona phenomenon.

2. The method for extracting multi-spectral features of a composite insulator considering corona characteristics according to claim 1, characterized in that, The method for preliminarily analyzing whether corona change occurs in this connected component includes: Calculating the similarity of the brightness values of the same connected component in the HSV spaces of infrared and ultraviolet images; Calculating the normalized value of the brightness mean of any connected component in the HSV space of any spectral image, and calculating the range of the normalized values of the brightness means of all connected components in the HSV space of any spectral image; Using the similarity, the normalized value of the brightness mean calculated for any connected component in all spectral images, and the range to determine the corona brightness confidence of the any connected component, and preliminarily analyzing whether corona change occurs in this connected component by the magnitude of its value.

3. The multi-spectral feature extraction method for composite insulators considering corona characteristics according to claim 2, characterized in that The method for further determining the corona brightness confidence of any connected component further includes: Denote the corona brightness confidence of the v-th connected component as , and the expression is: ; where is the similarity between the v-th connected component in the infrared HSV image and the ultraviolet HSV image; , are the normalized values of the average brightness of the v-th connected component in the infrared HSV image, the ultraviolet HSV image, and the visible light HSV image respectively; , are the brightness ranges of the v-th connected component in the infrared HSV image, the ultraviolet HSV image, and the visible light HSV image respectively.

4. The method for extracting multi-spectral features of a composite insulator considering corona characteristics according to claim 1, characterized in that, The method for further analyzing whether this connected component is affected by sunlight includes: Obtaining the degree of dispersion of the texture feature values of all pixel points of this connected component in the insulator image; Forming a sequence of the gray - scale values of all pixel points between the two pixel points with the largest and smallest gray - scale values in this connected component from the largest to the smallest gray - scale value; recording the proportion of the number of negative elements after differentiating the sequence as the gray - scale decreasing index of this connected component; Obtaining the mean value of the gradient values of all edge pixel points in this connected component; Using the degree of dispersion, the gray - scale decreasing index, and the mean value of the gradient values calculated within this connected component to determine the light - irradiation influence coefficient of this connected component, and further analyzing whether this connected component is affected by sunlight by the magnitude of its value.

5. The method for extracting multi-spectral features of a composite insulator considering corona characteristics according to claim 4, wherein The method for further determining the light - irradiation influence coefficient of this connected component further includes: Denote the illumination influence coefficient of the v-th connected component as , and the expression is as follows: ; where is the gray-scale decreasing exponent of the v-th connected component; is the degree of dispersion of all texture feature values in the v-th connected component; is the mean value of the gradient values of all edge pixel points in the v-th connected component.

6. The method for extracting multi - spectral features of a composite insulator considering corona characteristics according to claim 4, characterized in that, The method for further analyzing the key feature regions of the composite insulator includes: Generating connected components for all insulator images, and respectively denoting all corresponding connected components after registration of this connected component in other insulator images as the different - time connected components of this connected component; Calculating the degree of dispersion of the light - irradiation influence coefficients of this connected component and all its different - time connected components; Calculating the Hu - moment values between this connected component and each of its different - time connected components through the Hu - moment function; denoting the product of the mean value of all calculated Hu - moment values and the degree of dispersion of all Hu - moment values as the contour change index of this connected component. Determine the temporal variation coefficient of the connected component using the illumination influence coefficient of the connected component, the degree of dispersion of the calculated illumination influence coefficient, and the contour change coefficient, and further analyze whether the connected component is a key feature region of the composite insulator based on the value of the temporal variation coefficient.

7. The method for extracting multi - spectral features of a composite insulator considering corona characteristics according to claim 6, characterized in that, The method for determining the temporal variation coefficient of the connected component further includes: Denote the temporal variation coefficient of the v-th connected component as , and the expression is: ; where is the illumination influence coefficient of the v-th connected component; is the degree of dispersion of the illumination influence coefficients of the v-th connected component and all its asynchronous connected components; is the contour change index of the v-th connected component.

8. The method for extracting multi-spectral features of a composite insulator considering corona characteristics according to claim 1, characterized in that, The method for extracting the key feature region in the insulator image that is greatly affected by the corona phenomenon is: Use the result of the inverse fusion of the corona brightness confidence of the same connected component and the temporal variation coefficient as the corona confidence of the connected component; Obtain the segmentation threshold of the corona confidence of all connected components in the insulator image, and use the connected components corresponding to the corona confidence greater than the segmentation threshold as the key feature regions in the insulator image that are greatly affected by the corona phenomenon.

9. The method for extracting multi-spectral features of a composite insulator considering corona characteristics according to claim 8, characterized in that, The determination method of the corona confidence of the connected component is further determined by the ratio result of the corona brightness confidence and the temporal variation coefficient of the same connected component.

10. Composite insulator multi-spectral feature extraction device considering corona characteristics, characterized in that, Implement the multi-spectral feature extraction method for composite insulators considering corona characteristics as described in any one of claims 1-9. The device includes: An image processing module for registering, fusing, and extracting the composite insulator region from the acquired multi-spectral image, and inputting the obtained insulator image into the feature extraction module; A feature extraction module for analyzing the brightness, texture, and temporal variation features of the insulator image to consider the influence of corona, sunlight, and temporal variation features on the composite insulator; A corona region extraction module for extracting the key feature regions of the composite insulator from the insulator image based on the various features analyzed by the feature extraction module.

Images