CN115909028A - High-voltage isolating switch state identification method based on gradient image fusion - Google Patents

High-voltage isolating switch state identification method based on gradient image fusion Download PDF

Info

Publication number
CN115909028A
CN115909028A CN202211432924.XA CN202211432924A CN115909028A CN 115909028 A CN115909028 A CN 115909028A CN 202211432924 A CN202211432924 A CN 202211432924A CN 115909028 A CN115909028 A CN 115909028A
Authority
CN
China
Prior art keywords
image
isolating switch
fusion
voltage isolating
visible light
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
Application number
CN202211432924.XA
Other languages
Chinese (zh)
Inventor
张靖
周丽
周路遥
单长吉
颜悦
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Zhaotong University
Original Assignee
Zhaotong University
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Zhaotong University filed Critical Zhaotong University
Priority to CN202211432924.XA priority Critical patent/CN115909028A/en
Publication of CN115909028A publication Critical patent/CN115909028A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y04—INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
    • Y04S—SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
    • Y04S10/00—Systems supporting electrical power generation, transmission or distribution
    • Y04S10/50—Systems or methods supporting the power network operation or management, involving a certain degree of interaction with the load-side end user applications

Landscapes

  • Image Processing (AREA)

Abstract

The invention discloses a high-voltage isolating switch state identification method based on gradient image fusion, which comprises the steps of collecting an infrared image and a visible light image of a high-voltage isolating switch, carrying out image preprocessing and image registration processing on the infrared image and the visible light image of the high-voltage isolating switch, adopting a non-downsampling shear wave transformation algorithm to decompose the registered infrared image and the registered visible light image of the high-voltage isolating switch into a high-frequency sub-band diagram and a low-frequency sub-band diagram respectively, fusing the high-frequency sub-band diagram and the low-frequency sub-band diagram of the infrared image and the visible light image of the high-voltage isolating switch respectively to realize local image fusion, fusing the infrared low-frequency sub-band diagram and the high-frequency sub-band diagram respectively to realize global fusion, forming a gradient image fusion model through local fusion and global fusion of the infrared image and the visible light image of the high-voltage isolating switch, and carrying out pixel integration projection algorithm processing on the fused image to further realize the identification of the state of the high-voltage isolating switch.

Description

High-voltage isolating switch state identification method based on gradient image fusion
Technical Field
The invention relates to the field of image recognition, in particular to a high-voltage disconnecting switch state recognition method based on gradient image fusion.
Background
The high-voltage isolating switch plays an important role in a power supply system and is important equipment capable of protecting the safety of maintainers and power equipment. However, the mechanical structure is broken down due to the influence of external weather and strong electric field when the power grid works outdoors for a long time, and the condition of opening and closing failure is easily caused, so that the safe operation of the power grid is threatened.
The identification method for the opening and closing state of the isolating switch mainly comprises methods based on temperature detection, image identification, stress strain detection, motor current detection and the like. The method belongs to contact type measurement, and a sensor is easily damaged by high-voltage strong current. An infrared image-based isolation switch state identification method belongs to a non-contact measurement method, and has strong penetration capability and strong anti-interference capability. But the pixel is lower, the detailed information of the image is easy to lose, and the temperature is sensitive. The principle of visible light image recognition is that images of the isolating switch are collected for processing, an intelligent algorithm is used for recognizing the opening and closing state, the visible light image is high in pixel, more image details can be collected, and the visible light image is insensitive to the external temperature. But is susceptible to the influence of the shielding object and is greatly influenced by the external interference. The principle of stress-strain detection is to determine the opening and closing state by detecting the strain condition of the operating lever. The relative position of the movable contact and the static contact can truly reflect the opening and closing state of the isolating switch, so the method has the defect of inaccurate identification. The principle based on motor current detection is that the opening and closing state is judged by analyzing the peak value change condition of the driving motor current in different states, the torque, the current and the acceleration of the driving motor in the opening and closing state of the isolating switch are respectively subjected to simulation calculation, and the opening and closing state is judged through the current commutation time. The method can not directly detect the relative positions of the moving contact and the fixed contact of the isolating switch, and can not ensure the accuracy of opening and closing identification.
Disclosure of Invention
In order to solve the problems, the invention aims to provide a high-voltage isolating switch state identification method based on gradient image fusion, which comprises the steps of collecting an infrared image and a visible light image of a high-voltage isolating switch, carrying out image preprocessing and image registration on the infrared image and the visible light image of the high-voltage isolating switch, adopting a non-downsampling shear wave transformation algorithm to decompose the registered infrared image and visible light image of the high-voltage isolating switch into a high-frequency sub-band diagram and a low-frequency sub-band diagram respectively, adopting a pulse coupling neural network algorithm to fuse the infrared high-frequency sub-band diagram and the visible light high-frequency sub-band diagram of the high-voltage isolating switch together, adopting a visual saliency characteristic segmentation algorithm to fuse the infrared low-frequency sub-band diagram and the visible light low-frequency sub-band diagram together to realize local image fusion, then adopting a non-downsampling shear wave inverse transformation algorithm to fuse the locally fused images together again to realize global fusion, and obtaining an image fused with the infrared image of the visible light image of the high-voltage isolating switch, thereby realizing a gradient image fusion model. And establishing an image fusion quality index evaluation scheme to compare the effect of the gradient image fusion algorithm with the effect of the common image fusion scheme. And then processing the fused image by a pixel integral projection algorithm, thereby realizing the identification of the opening and closing state of the high-voltage isolating switch. And verifying the switching-on and switching-off state identification result of the high-voltage isolating switch through experimental simulation. The advantages of the infrared image and the visible light image of the high-voltage isolating switch can be complemented, the problems that a single infrared image is low in pixel, easy to lose detailed information of the image, sensitive to temperature and the like are solved, and the problems that a single visible light image is easy to be influenced by a shielding object, large in influence of external interference, incapable of working in all weather and the like are solved.
In order to achieve the above purpose, the method for identifying the state of the high-voltage isolating switch based on gradient image fusion provided by the invention is realized as follows:
a high-voltage isolating switch state identification method based on gradient image fusion comprises an image preprocessing scheme, an image registration scheme, a gradient image fusion scheme, a fusion image evaluation scheme, a high-voltage isolating switch state identification scheme and an experimental simulation scheme, wherein the image preprocessing scheme is used for converting an acquired infrared image and a visible light image of a high-voltage isolating switch into gray images and filtering and denoising the two images so as to improve the quality and the processing speed of the images, the image registration scheme is used for aligning the infrared image and the visible light image of the high-voltage isolating switch in space and finding out the mapping relation between the infrared image and the visible light image of the high-voltage isolating switch so that the infrared image and the visible light image of the high-voltage isolating switch can be well fused in space, the gradient image fusion scheme is used for decomposing the registered infrared image and the registered visible light image of the high-voltage isolating switch into a high-frequency sub and a low-frequency sub, the method comprises the steps of fusing a high-frequency sub-band diagram and a low-frequency sub-band diagram of an infrared image and a visible light image of a high-voltage isolating switch respectively to realize local image fusion, fusing the infrared low-frequency sub-band diagram and the high-frequency sub-band diagram which are fused respectively to realize global fusion, and forming a gradient image fusion model through local fusion and global fusion of the infrared image and the visible light image of the high-voltage isolating switch, wherein a fusion image evaluation scheme is used for evaluating the quality of the high-voltage isolating switch after the infrared image and the visible light image are fused by the gradient image fusion model, a high-voltage isolating switch state identification scheme is used for identifying the image of the high-voltage isolating switch after the infrared image and the visible light image are fused to judge whether the high-voltage isolating switch is in a switching-off state or a switching-on state, and an experimental simulation scheme is used for comparing the fused high-voltage isolating switch image and the single visible light image of the high-voltage isolating switch Or the identification accuracy rate of the infrared image of the high-voltage isolating switch.
The image preprocessing scheme of the invention is as follows:
s1, image acquisition:
the high-voltage isolating switch is set to be in an opening state and a closing state respectively, and images of the high-voltage isolating switch in different states are acquired at the same position by a visible light camera and an infrared thermal imager.
S2, graying of image
Carrying out gray processing on the collected infrared light image and the collected visible light image to improve the operation rate of image processing, and realizing the gray processing of the image through the following formula:
Gray(i,j)＝0.299×R(i,j)+0.587×G(i,j)+0.114×B(i,j) (1)
in the formula, gray (i, j) represents the Gray value of a pixel point of the image at the coordinate (i, j), R (i, j) represents the red component of the image, G (i, j) represents the green component of the image, and B (i, j) represents the blue component of the image.
S3, image filtering denoising
Filtering out noise of the image by adopting two-dimensional median filtering, arranging pixel values in a two-dimensional sliding template from large to small to obtain a two-dimensional data sequence, and making the filtered image be g (x, y), then:
g(x,y)＝med{f(x-k),y-l},(k,l∈A) (2)
in the formula, f (x-k, y-l) represents an original image of the isolating switch, A represents a two-dimensional sliding template, and a 3 x 3 template is selected as the template area.
The image registration scheme of the invention is as follows:
carrying out normalization processing on the collected infrared and visible light images by adopting a maximum and minimum value method:
Figure BDA0003945775400000041
where norm is the normalized value, x i The max (x) and min (x) represent the maximum pixel value and the minimum pixel value in the image respectively;
adjusting the resolution ratio of the infrared image and the visible light image of the high-voltage isolating switch to 2306 × 2658, mapping the visible light image of the high-voltage isolating switch to the infrared image of the high-voltage isolating switch through a transformation model of image characteristics, performing edge detection on the infrared image and the visible light image of the high-voltage isolating switch by using a Canny operator to extract the outline of the image, detecting and extracting the corner points of the image outline through a SURF algorithm, and extracting the characteristic points of the image outline by adopting a black plug matrix:
Figure BDA0003945775400000042
in the formula, L xx (x, σ) represents the convolution of the gaussian second order differential with a point I (x, y) in the image at the scale σ, and is calculated as:
Figure BDA0003945775400000043
g (x, y) represents an image, and L can be calculated in the same manner as above xy (x, σ) and L yy (x,σ)；
And then, matching feature points of the images by using an Euler distance formula, finally fitting the feature points by using a least square method, exchanging parameter estimation, further obtaining an optimal transformation model, and finally obtaining the registered infrared images and visible light images of the high-voltage isolating switch.
The gradient image fusion scheme of the invention is as follows:
the method comprises the following steps of adopting a non-downsampling shear wave transformation algorithm to respectively decompose an infrared image and a visible light image of a registered high-voltage isolating switch into a high-frequency sub-band diagram and a low-frequency sub-band diagram, adopting a pulse coupling neural network algorithm to fuse the infrared high-frequency sub-band diagram and the visible light high-frequency sub-band diagram of the high-voltage isolating switch, adopting a visual saliency characteristic segmentation algorithm to fuse the infrared low-frequency sub-band diagram and the visible light low-frequency sub-band diagram of the high-voltage isolating switch, realizing local image fusion, fusing the images after local fusion again through a non-downsampling shear wave inverse transformation algorithm, realizing global fusion, obtaining an image after fusing the infrared image and the visible light image of the high-voltage isolating switch, and further realizing a gradient image fusion model, wherein the method comprises the specific steps of:
s1, image decomposition
The method comprises the steps that non-downsampling shear wave transformation decomposition comprises a multiscale decomposition part and a multidirectional decomposition part, in a multiscale decomposition level, a non-downsampling pyramid filter is adopted to decompose an infrared image and a visible light image of a high-voltage isolating switch into 1 low-frequency sub-band diagram and a plurality of high-frequency sub-band diagrams respectively, then the non-downsampling shear wave filter is adopted to decompose the decomposed high-frequency sub-band diagrams in multiple directions to construct a Meyer window function, the high-frequency sub-band diagrams and the Meyer window function are subjected to convolution calculation through a convolution algorithm to obtain high-frequency sub-band coefficients in different directions, and the decomposition of the infrared image and the visible light image of the high-voltage isolating switch in different directions can be achieved;
s2. Gradient image fusion design
S21, local image fusion
S211. High-frequency sub-band diagram fusion
The spatial frequency is used as the input of the pulse coupled neural network algorithm, the laplace energy and the connection strength of the pulse coupled neural network. Calculating the ignition frequency of the pulse coupling neural network after n iterations, realizing the fusion of high-frequency sub-band images according to the ignition frequency, and enabling the spatial frequency of the images to be MSF, wherein the method comprises the following steps:
Figure BDA0003945775400000051
in the formula, RF and CF represent a row frequency and a column frequency of an image, respectively;
in the pulse coupling neural network model, the correlation degree between neurons is expressed by connection strength, and the connection strength is expressed as:
Figure BDA0003945775400000052
LP(i,j)＝|2C(i,j)-C(i-step,j)-C(i+step,j)|+|2C(i,j)-C(i,j-step)-C(i,j+step)| (8)
in the equations (7) and (8), step is the distance between pixels, the value is 1,C (i, j) represents the high-frequency subband coefficient, ω (i, j) is the weight of the high-frequency subband coefficient, and MSLR (i, j) is the sum of laplacian energies of the image at the coordinates (i, j);
the firing frequency is expressed as:
Figure BDA0003945775400000061
in the formula, T R (n) and T V (n) the ignition frequencies of the high-frequency sub-band diagram after the infrared image and the visible light image of the high-voltage isolating switch are decomposed through non-subsampled shear wave transformation are respectively used, so that the infrared high-frequency sub-band diagram and the visible light high-frequency sub-band diagram of the high-voltage isolating switch are fused, and a high-frequency sub-band fusion diagram is obtained;
s212. Fusion of low-frequency sub-band diagrams
The invention obtains the significant characteristics (such as outline and background information) in the low-frequency subband image by a hypercomplex number Fourier transform algorithm, divides the low-frequency subband image into two regions of significance and non-significance, and makes a hypercomplex number matrix as:
f(m,n)＝ω 1 f 1 i+ω 2 f 2 j+ω 3 f 3 k (10)
wherein f (m, n) represents a hypercomplex matrix, i, j, k satisfy i 2 ＝j 2 ＝k 2 ＝ijk＝-1，ω 1 ～ω 3 Respectively representing the brightness, direction and weight of the texture information of the image, and respectively taking the values of 0.5,0.25 and f 1 ～f 3 Feature matrices representing these three dimensions, respectively, are supercomplex fourier transformed from f (m, n):
F H [μ,ν]＝||F H [μ,ν]||e μφ(μ,ν) (11)
wherein, | | F H [μ,ν]| | represents the first order norm of the frequency domain of f (m, n);
using Gaussian kernel function to pair | | | F H [μ,ν]The | l is smoothed to obtain | F H [μ,ν]A multi-scale map of,then, obtaining a salient image by utilizing hypercomplex number Fourier inversion transformation, segmenting a salient region and a non-salient region of the image in an image space by taking the minimum entropy of the image as a criterion, dividing image features into the salient region and the non-salient region by adopting an image fusion method based on visual salient characteristic segmentation for fusion, and judging whether the images are similar by comparing the similarity of the two salient images, wherein the expression is as follows:
Figure BDA0003945775400000062
SIM A,B (i, j) denotes the significance similarity, SM A (i, j) and SM B (i, j) respectively represent the significance values of the infrared image and the visible light image of the high-voltage isolating switch, SIM A,B The larger (i, j) is, the more similar the infrared image and the visible light image of the high-voltage isolating switch are, and the fusion output is as follows:
Figure BDA0003945775400000071
in the formula, omega represents a fusion weighted value, T represents a similarity threshold value (the value range is 0.5-1), IR represents an infrared image, and VI represents a visible light image;
the image with larger information entropy is fused by comparing the size of the information entropy as a judgment basis for image fusion, and the judgment basis is as follows:
Figure BDA0003945775400000072
wherein F (i, j) represents the fused image, LR is the information entropy ratio, and LR = LR IR /LR VI When LR is more than 1, the information entropy of the infrared image in the region is higher than that of the visible light image, so the infrared image is taken as a fusion image, conversely, the visible light image is taken as a fusion image, and the infrared low-frequency sub-band diagram and the visible light low-frequency sub-band diagram of the high-voltage isolating switch are fused to obtain a low-frequency sub-band diagramA sub-band fusion graph;
s22, image global fusion
And reconstructing the high-frequency sub-band fusion map and the low-frequency sub-band fusion map of the high-voltage isolating switch by adopting a non-down-sampling shear wave inverse transformation algorithm to obtain a final fusion image of the infrared image and the visible light image of the high-voltage isolating switch, realizing global image fusion, and decomposing the infrared image and the visible light image of the high-voltage isolating switch into a process of local image fusion and global image fusion, namely a gradient image fusion model.
The fusion image evaluation scheme of the invention is as follows:
in order to evaluate the quality of the fused high-voltage isolating switch infrared image and the fused visible light image by the gradient image fusion model algorithm and the common image fusion methods based on the color model algorithm, the space transformation algorithm, the weighted average method, the ratio transformation method, the wavelet transformation method, the principal component analysis method and the like, the retention degree (Q) of the edge information of the fused image is respectively determined AB/F ) The image fused by different image fusion methods is evaluated by six dimensions such as information Entropy (EN), mutual Information (MI), spatial frequency (MSF), standard deviation (STD), structural Similarity (SSIM), and the advantages of a gradient image fusion model algorithm and six common image fusion methods are compared by the six quality evaluation dimensions, wherein the six quality evaluation dimensions are respectively expressed as:
(1)Q AB/F the retention condition of detail information of the image before and after fusion is reflected, and the larger the value is, the more complete the retention of the edge information is indicated. Expressed as:
Figure BDA0003945775400000081
in the formula, Q AF (i, j) is the intensity, Q, at pixel point (i, j) BF (i, j) a direction value, ω, at the pixel point (i, j) A (i,j)、ω B (i, j) representing the weight of the intensity and direction, respectively;
(2) The EN corresponds to the information richness of the image, the quality of the image is in direct proportion to the information entropy value, and the result is expressed as:
Figure BDA0003945775400000082
where L is the gray level of the image, taking the value 255 and P i Representing the probability of occurrence of the corresponding gray level;
(3) MI reflects the amount of information before and after image fusion, and is expressed as;
MI＝MI A,F +MI B,F (17)
in the formula MI A,F And MI B,F Respectively representing the intersection of the information quantity of the infrared image and the visible light image of the high-voltage isolating switch, wherein the larger MI is, the better fusion between the two images is;
(4) The MSF reflects the frequency information of the image in space, and the higher the frequency is, the more the contour and edge information of the image will be, which is expressed as:
Figure BDA0003945775400000083
in the formula, RF and CF represent a row frequency and a column frequency of an image, respectively;
(5) The STD reflects the contrast of the image, and the larger the standard deviation is, the more detail information is kept in the image, and the higher the effect and quality after fusion is, which is expressed as:
Figure BDA0003945775400000091
(6) The SSIM reflects the similarity degree of the original image before and after fusion, and the larger the value is, the higher the similarity degree of the two is, which is expressed as:
Figure BDA0003945775400000092
in the formula, mu A 、μ B 、μ F Respectively represent the mean values of the infrared image, the visible light image and the fusion image,
Figure BDA0003945775400000093
Figure BDA0003945775400000094
respectively represent the variance, sigma, of the three AF 、σ BF Respectively representing the combined variance of the infrared image, the visible light image and the fused image.
The state identification scheme of the high-voltage isolating switch comprises the following steps:
after a high-voltage isolating switch image fused with a global image is segmented through a Qtsu threshold segmentation algorithm, a high-voltage isolating switch is extracted from a background image, horizontal integral projection and vertical integral projection processing are carried out on the high-voltage isolating switch image segmented from a target through a pixel integral projection method, the state of the isolating switch is judged by calculating the ratio between the horizontal integral projection and the vertical integral projection, and when the isolating switch is in a brake-off state, the ratio of the integral in the vertical direction to the integral in the horizontal direction is calculated to be 0.842; when the high-voltage isolating switch is switched on, the ratio of the two is 0.237, according to the requirement that the angle of the switch blade is not lower than 65 degrees when the high-voltage isolating switch is switched on and switched off, when the angle of the switch blade is 65 degrees, the integral projection ratio of the fused image is 0.826, the switch-off state is set when the integral projection ratio of the pixel is greater than 0.826, the switch-on state is set when the integral projection ratio of the pixel is less than 0.237, and the switch-off state or the switch-on failure is indicated when the integral projection ratio of the pixel is between 0.237 and 0.826.
The experimental simulation scheme of the invention is as follows:
the method comprises the steps of respectively collecting 100 infrared images of the high-voltage isolating switch and identifying light images, judging the accuracy of identification of different images, setting different scenes in the image collection process, and collecting an isolating switch image when an obstacle appears near the isolating switch or the visibility is low, wherein the infrared images can be collected, but a camera is shielded and cannot collect a complete image, or the collected image is fuzzy; secondly, the power supply is cut off, no current passes through the isolating switch, no heat is generated, the infrared image cannot acquire a clear isolating switch image, and the camera can compare the identification accuracy of the high-voltage isolating switch image fused by the gradient image fusion model algorithm provided by the invention with the infrared image of the high-voltage isolating switch and the visible image of the high-voltage isolating switch, so that the advantages of the gradient image fusion model algorithm provided by the invention are obtained.
Because the invention adopts the gradient image fusion algorithm to realize the fusion of the infrared image and the visible light image of the high-voltage isolating switch and realizes the identification of the state of the high-voltage isolating switch by the pixel integral projection algorithm, the following beneficial effects can be obtained:
1. the infrared image and the visible light image of the high-voltage isolating switch are aligned in the space through an image registration scheme, and the mapping relation between the infrared image and the visible light image of the high-voltage isolating switch is found out, so that the infrared image and the visible light image of the high-voltage isolating switch can be well fused in the space.
2. Decomposing the registered high-voltage isolating switch infrared image and visible light image into a high-frequency sub-band diagram and a low-frequency sub-band diagram by a gradient image fusion scheme, fusing the high-frequency sub-band diagram and the low-frequency sub-band diagram of the high-voltage isolating switch infrared image and the visible light image respectively to realize local image fusion, fusing the fused infrared low-frequency sub-band diagram and the high-frequency sub-band diagram respectively to realize global fusion, forming a gradient image fusion model by the local fusion and the global fusion of the high-voltage isolating switch infrared image and the visible light image, and identifying the fused image of the high-voltage isolating switch infrared image and the visible light image by a high-voltage isolating switch state identification scheme to judge whether the high-voltage isolating switch is in an opening state or a closing state. The advantages of the infrared image and the visible light image of the high-voltage isolating switch are complementary, and all-weather work is realized.
Drawings
FIG. 1 is a flowchart of a general scheme of a method for identifying the state of a high-voltage isolating switch based on gradient image fusion according to the present invention;
FIG. 2 is a graph of a high-voltage isolator state identification method based on gradient image fusion after image graying and filtering;
FIG. 3 is an image registration flowchart of a high-voltage isolation switch state identification method based on gradient image fusion according to the present invention;
FIG. 4 is a gradient image fusion model diagram of a high-voltage isolation switch state identification method based on gradient image fusion according to the present invention;
FIG. 5 is a pixel integral projection calculation model diagram of the high-voltage isolation switch state identification method based on gradient image fusion according to the present invention;
FIG. 6 is an image of high-voltage isolator infrared image and visible light image fusion of the high-voltage isolator state identification method based on gradient image fusion of the present invention;
FIG. 7 is a comparison graph of the fusion image effect under different image fusion methods of the high-voltage disconnecting switch state identification method based on gradient image fusion according to the present invention;
fig. 8 is a fused image state recognition result diagram of the high-voltage disconnecting switch state recognition method based on gradient image fusion.
Detailed Description
The present invention will be described in further detail with reference to the following examples and drawings.
Fig. 1 to 8 show a high-voltage isolator state identification method based on gradient image fusion according to the present invention, which includes an image preprocessing scheme, an image registration scheme, a gradient image fusion scheme, a fusion image evaluation scheme, a high-voltage isolator state identification scheme, and an experimental simulation scheme.
As shown in fig. 1, the image preprocessing scheme is to convert an infrared image and a visible light image of a high-voltage isolator into gray images, and perform filtering and denoising on the two images to improve the quality and processing speed of the images, the image registration scheme is to align the infrared image and the visible light image of the high-voltage isolator in space and find out the mapping relationship between the infrared image and the visible light image of the high-voltage isolator so that the infrared image and the visible light image of the high-voltage isolator can be fused in space well, the gradient image fusion scheme is to decompose the infrared image and the visible light image of the high-voltage isolator after registration into a high-frequency sub-band diagram and a low-frequency sub-band diagram, fuse the high-frequency sub-band diagram and the low-frequency sub-band diagram of the infrared image and the visible light image of the high-voltage isolator respectively to realize local image fusion, fuse the infrared low-frequency sub-band diagram and the high-frequency sub-band diagram respectively, realize global fusion, the local fusion of the infrared image and the low-frequency sub-band diagram of the high-voltage isolator form a gradient image fusion model through local fusion and the local fusion and global fusion, the high-frequency sub-image fusion model is used for evaluating the high-voltage isolator gradient image fusion model to evaluate whether the high-voltage isolator image fusion or high-voltage isolator image fusion state identification and high-voltage isolation switch identification single high-isolation switch identification scheme for identifying the high-isolation image fusion state.
The image preprocessing scheme is as follows:
s1, image acquisition:
the high-voltage isolating switch is set to be in an opening state and a closing state respectively, and a visible light camera and an infrared thermal imager are adopted to collect images of the high-voltage isolating switch in different states at the same position.
S2, graying of image
As shown in fig. 2, since the collected infrared image and visible light image of the high-voltage isolator both belong to color images, the influence of the color information in these images on the identification result is small, and in order to reduce the calculation amount and improve the calculation efficiency, the two images can be grayed. Carrying out gray processing on the collected infrared light image and the collected visible light image to improve the operation rate of image processing, and realizing the gray processing of the image through the following formula:
Gray(i,j)＝0.299×R(i,j)+0.587×G(i,j)+0.114×B(i,j) (1)
in the formula, gray (i, j) represents the Gray value of a pixel point of the image at the coordinate (i, j), R (i, j) represents the red component of the image, G (i, j) represents the green component of the image, and B (i, j) represents the blue component of the image.
S3, image filtering and denoising
As shown in fig. 2, the image is easily interfered by external environment and self instability during the acquisition process, so that the obtained image generates certain noise, and the noise seriously affects the accuracy of identification. Filtering out noise of the image by adopting two-dimensional median filtering, arranging pixel values in a two-dimensional sliding template from large to small to obtain a two-dimensional data sequence, and making the filtered image be g (x, y), then:
g(x,y)＝med{f(x-k),y-l},(k,l∈A) (2)
in the formula, f (x-k, y-l) represents an original image of the isolating switch, A represents a two-dimensional sliding template, and a 3 x 3 template is selected as the area of the template.
As shown in fig. 3, the image registration scheme is:
in the process of image acquisition, target images cannot be aligned in the space due to different shooting angles and object-image distances, so that the mapping relation between the infrared images and the visible light images cannot be found in the process of fusion, better fusion cannot be performed in the space, and therefore registration of the two different images is required.
Carrying out normalization processing on the collected infrared and visible light images by adopting a maximum and minimum value method:
Figure BDA0003945775400000131
where norm is the normalized value, x i The max (x) and min (x) represent the maximum pixel value and the minimum pixel value in the image respectively;
adjusting the resolution ratio of the infrared image and the visible light image of the high-voltage isolating switch to 2306 × 2658, mapping the visible light image of the high-voltage isolating switch to the infrared image of the high-voltage isolating switch through a transformation model of image characteristics, performing edge detection on the infrared image and the visible light image of the high-voltage isolating switch by using a Canny operator to extract the outline of the image, detecting and extracting the corner points of the image outline through a SURF algorithm, and extracting the characteristic points of the image outline by adopting a black plug matrix:
Figure BDA0003945775400000132
in the formula, L xx (x, σ) represents the convolution of the gaussian second order differential with a point I (x, y) in the image at the scale σ, and is calculated as:
Figure BDA0003945775400000133
g (x, y) represents an image, and L can be calculated in the same manner as above xy (x, σ) and L yy (x,σ)；
And matching the characteristic points of the images by using an Euler distance formula, fitting the characteristic points by using a least square method, exchanging parameter estimation, further obtaining an optimal transformation model, and finally obtaining the registered infrared images and visible light images of the high-voltage isolating switch.
As shown in fig. 4, the gradient image fusion scheme is:
the method comprises the following steps of adopting a non-downsampling shear wave transformation algorithm to respectively decompose an infrared image and a visible light image of a registered high-voltage isolating switch into a high-frequency sub-band diagram and a low-frequency sub-band diagram, adopting a pulse coupling neural network algorithm to fuse the infrared high-frequency sub-band diagram and the visible light high-frequency sub-band diagram of the high-voltage isolating switch, adopting a visual saliency characteristic segmentation algorithm to fuse the infrared low-frequency sub-band diagram and the visible light low-frequency sub-band diagram of the high-voltage isolating switch, realizing local image fusion, fusing the images after local fusion again through a non-downsampling shear wave inverse transformation algorithm, realizing global fusion, obtaining an image after fusing the infrared image and the visible light image of the high-voltage isolating switch, and further realizing a gradient image fusion model, wherein the method comprises the specific steps of:
s1, image decomposition
The non-downsampling shear wave transformation decomposition comprises a multiscale decomposition part and a multidirectional decomposition part, in a multiscale decomposition layer, a non-downsampling pyramid filter is adopted to decompose an infrared image and a visible light image of a high-voltage isolating switch into 1 low-frequency sub-band diagram and a plurality of high-frequency sub-band diagrams respectively, then the non-downsampling shear wave filter is adopted to decompose the decomposed high-frequency sub-band diagrams in multiple directions to construct a Meyer window function, the high-frequency sub-band diagrams and the Meyer window function are subjected to convolution calculation through a convolution algorithm to obtain high-frequency sub-band coefficients in different directions, and the decomposition of the infrared image and the visible light image of the high-voltage isolating switch in different directions can be realized;
s2. Gradient image fusion design
S21, local image fusion
S211. High-frequency sub-band diagram fusion
The spatial frequency is used as the input of the pulse coupled neural network algorithm, the laplace energy and the connection strength of the pulse coupled neural network. Calculating the ignition frequency of the pulse coupling neural network after n iterations, realizing the fusion of high-frequency sub-band images according to the ignition frequency, and enabling the spatial frequency of the images to be MSF, wherein the method comprises the following steps:
Figure BDA0003945775400000141
in the formula, RF and CF represent a row frequency and a column frequency of an image, respectively;
in the pulse coupling neural network model, the correlation degree between neurons is expressed by connection strength, and the connection strength is expressed as:
Figure BDA0003945775400000151
LP(i,j)＝|2C(i,j)-C(i-step,j)-C(i+step,j)|+|2C(i,j)-C(i,j-step)-C(i,j+step)| (8)
in the equations (7) and (8), step is the distance between pixels, the value is 1,C (i, j) represents the high-frequency subband coefficient, ω (i, j) is the weight of the high-frequency subband coefficient, and MSLR (i, j) is the sum of laplacian energies of the image at the coordinates (i, j);
the firing frequency is expressed as:
Figure BDA0003945775400000152
in the formula, T R (n) and T V (n) the ignition frequencies of the high-frequency sub-band diagram after the infrared image and the visible light image of the high-voltage isolating switch are decomposed through non-subsampled shear wave transformation are respectively used, so that the infrared high-frequency sub-band diagram and the visible light high-frequency sub-band diagram of the high-voltage isolating switch are fused, and a high-frequency sub-band fusion diagram is obtained;
s212. Fusion of low-frequency sub-band diagrams
The method acquires the significant characteristics (such as outline and background information) in the low-frequency subband image through a hypercomplex number Fourier transform algorithm, divides the low-frequency subband image into two regions of significance and non-significance, and makes a hypercomplex number matrix as follows:
f(m,n)＝ω 1 f 1 i+ω 2 f 2 j+ω 3 f 3 k (10)
wherein f (m, n) represents a hypercomplex matrix, i, j, k satisfy i 2 ＝j 2 ＝k 2 ＝ijk＝-1，ω 1 ～ω 3 Respectively representing the brightness, direction and weight of the texture information of the image, and respectively taking the values of 0.5,0.25 and f 1 ～f 3 Feature matrices representing these three dimensions, respectively, are supercomplex fourier transformed from f (m, n):
F H [μ,ν]＝||F H [μ,ν]||e μφ(μ,ν) (11)
wherein, | | F H [μ,ν]| | represents the first order norm of the frequency domain of f (m, n);
using Gaussian kernel function to pair | | | F H [μ,ν]The | l is smoothed to obtain | F H [μ,ν]Obtaining a salient image by utilizing hypercomplex number inverse Fourier transform, segmenting a salient region and a non-salient region of the image in an image space by taking the minimum entropy of the image as a criterion, dividing image features into the salient region and the non-salient region by adopting an image fusion method based on visual salient characteristic segmentation, and aiming at the salient region, the probability of the occurrence of a target is in direct proportion to the salient value, so that two salient images are comparedWhether the images are similar or not is judged according to the similarity, and the similarity is expressed as:
Figure BDA0003945775400000161
SIM A,B (i, j) represents the significance similarity, SM A (i, j) and SM B (i, j) respectively represent the significance values of the infrared image and the visible light image of the high-voltage isolating switch, SIM A,B The larger (i, j) is, the more similar the infrared image and the visible light image of the high-voltage isolating switch are, and the fusion output is as follows:
Figure BDA0003945775400000162
in the formula, omega represents a fused weighted value, T represents a similarity threshold (the value range is 0.5-1), IR represents an infrared image, and VI represents a visible light image;
the non-salient region is generally a background image, and the larger the information entropy of the image is, the larger the information amount of the image is. Therefore, images with larger information entropy are fused by comparing the size of the information entropy as a judgment basis for image fusion, and the judgment basis is as follows:
Figure BDA0003945775400000163
wherein F (i, j) represents the fused image, LR is the information entropy ratio, and LR = LR IR /LR VI When LR is larger than 1, the information entropy of the infrared image in the area is higher than that of the visible light image, so that the infrared image is used as a fusion image, and conversely, the visible light image is used as a fusion image, so that the infrared low-frequency sub-band diagram and the visible light low-frequency sub-band diagram of the high-voltage isolating switch are fused to obtain a low-frequency sub-band fusion diagram;
s22, image global fusion
As shown in fig. 6, a non-downsampling shear wave inverse transformation algorithm is adopted to reconstruct the high-frequency sub-band fusion image and the low-frequency sub-band fusion image of the high-voltage isolating switch to obtain a final fusion image of the infrared image and the visible light image of the high-voltage isolating switch, so as to realize global image fusion, and a process of decomposing the infrared image and the visible light image of the high-voltage isolating switch into local image fusion and global image fusion is called as a gradient image fusion model.
The fusion image evaluation scheme of the invention is as follows:
as shown in fig. 7, in order to evaluate the quality of the fused high-voltage isolator infrared image and the fused visible light image by the gradient image fusion model algorithm and the common image fusion methods based on the color model algorithm, the spatial transformation algorithm, the weighted average method, the ratio transformation method, the wavelet transformation method, the principal component analysis method and the like, the degree of retention of the edge information (Q) of the fused image is determined according to the quality of the fused high-voltage isolator infrared image and the fused visible light image AB/F ) The image fused by different image fusion methods is evaluated by six dimensions such as information Entropy (EN), mutual Information (MI), spatial frequency (MSF), standard deviation (STD), structural Similarity (SSIM), and the advantages of a gradient image fusion model algorithm and six common image fusion methods are compared by the six quality evaluation dimensions, wherein the six quality evaluation dimensions are respectively expressed as:
(1)Q AB/F the retention of detail information of the image before and after fusion is reflected, and the larger the value is, the more complete the retention of the edge information is indicated. Expressed as:
Figure BDA0003945775400000171
in the formula, Q AF (i, j) is the intensity, Q, at pixel point (i, j) BF (i, j) the direction value, ω, at the pixel point (i, j) A (i,j)、ω B (i, j) represent the weight of intensity and direction, respectively;
(2) The EN corresponds to the information richness of the image, the quality of the image is in direct proportion to the information entropy value, and the result is expressed as:
Figure BDA0003945775400000172
in the formulaL is the gray level of the image, the value is 255, P is taken i Representing the probability of occurrence of the corresponding gray level;
(3) MI reflects the amount of information before and after image fusion, and is expressed as;
MI＝MI A,F +MI B,F (17)
in the formula MI A,F And MI B,F Respectively representing the intersection of the information quantity of the infrared image and the visible light image of the high-voltage isolating switch, wherein the larger MI is, the better fusion between the two images is;
(5) The MSF reflects the frequency information of the image in space, and the higher the frequency is, the more the contour and edge information of the image will be, which is expressed as:
Figure BDA0003945775400000181
in the formula, RF and CF represent a row frequency and a column frequency of an image, respectively;
(5) The STD reflects the contrast of the image, and the larger the standard deviation is, the more detail information is kept in the image, and the higher the effect and quality after fusion is, which is expressed as:
Figure BDA0003945775400000182
(6) The SSIM reflects the degree of similarity of the original image before and after fusion, and the larger the value, the higher the similarity between the two is, which is expressed as:
Figure BDA0003945775400000183
in the formula, mu A 、μ B 、μ F Respectively represent the mean values of the infrared image, the visible light image and the fusion image,
Figure BDA0003945775400000184
Figure BDA0003945775400000185
respectively represent the variance, sigma, of the three AF 、σ BF Respectively representing the combined variance of the infrared image, the visible light image and the fused image.
As shown in fig. 5 and 8, the scheme for identifying the state of the high-voltage isolating switch is as follows:
after a high-voltage isolating switch image fused with a global image is segmented through a Qtsu threshold segmentation algorithm, a high-voltage isolating switch is extracted from a background image, horizontal integral projection and vertical integral projection processing are carried out on the high-voltage isolating switch image segmented from a target through a pixel integral projection method, the state of the isolating switch is judged by calculating the ratio between the horizontal integral projection and the vertical integral projection, and when the isolating switch is in a brake-off state, the ratio of the integral in the vertical direction to the integral in the horizontal direction is calculated to be 0.842; when the high-voltage isolating switch is switched on, the ratio of the two is 0.237, according to the requirement that the angle of the switch blade is not lower than 65 degrees when the high-voltage isolating switch is switched on and switched off, when the angle of the switch blade is 65 degrees, the integral projection ratio of the fused image is 0.826, the switch-off state is set when the integral projection ratio of the pixel is greater than 0.826, the switch-on state is set when the integral projection ratio of the pixel is less than 0.237, and the switch-off state or the switch-on failure is indicated when the integral projection ratio of the pixel is between 0.237 and 0.826.
The experimental simulation scheme of the invention is as follows:
the method comprises the steps of respectively collecting 100 infrared images of the high-voltage isolating switch and identifying light images, judging the accuracy of identification of different images, setting different scenes in the image collection process, and collecting an isolating switch image when an obstacle appears near the isolating switch or the visibility is low, wherein the infrared image can be collected, but a camera is shielded and cannot collect a complete image, or the collected image is fuzzy; secondly, the power supply is cut off, no current passes through the isolating switch, no heat is generated, the infrared image cannot acquire a clear isolating switch image, and the camera can compare the identification accuracy of the high-voltage isolating switch image fused by the gradient image fusion model algorithm provided by the invention with the infrared image of the high-voltage isolating switch and the visible image of the high-voltage isolating switch, so that the advantages of the gradient image fusion model algorithm provided by the invention are obtained.
The working principle and the working process of the invention are as follows:
the method comprises the steps of firstly collecting an infrared image and a visible light image of a high-voltage isolating switch, carrying out image preprocessing and image registration processing on the infrared image and the visible light image of the high-voltage isolating switch, respectively decomposing the registered infrared image and the registered visible light image of the high-voltage isolating switch into a high-frequency sub-band diagram and a low-frequency sub-band diagram by adopting a non-down sampling shear wave transformation algorithm, fusing the infrared high-frequency sub-band diagram and the visible light high-frequency sub-band diagram of the high-voltage isolating switch together by adopting a pulse coupling neural network algorithm, fusing the infrared low-frequency sub-band diagram and the visible light low-frequency sub-band diagram of the high-voltage isolating switch together by adopting a visual saliency characteristic segmentation algorithm to realize local image fusion, fusing the locally fused images again by adopting a non-down sampling shear wave inverse transformation algorithm to realize global fusion, obtaining an image obtained by fusing the infrared image and the visible light image of the high-voltage isolating switch, and further realizing a gradient image fusion model. And establishing an image fusion quality index evaluation scheme to compare the effect of the gradient image fusion algorithm with the effect of the common image fusion scheme. And then processing the fused image by a pixel integral projection algorithm, thereby realizing the identification of the opening and closing state of the high-voltage isolating switch. And verifying the switching-on and switching-off state identification result of the high-voltage isolating switch through experimental simulation.

Claims (4)

1. A high-voltage isolating switch state identification method based on gradient image fusion is characterized in that: comprises an image preprocessing scheme, an image registration scheme, a gradient image fusion scheme, a fusion image evaluation scheme, a high-voltage isolating switch state identification scheme and an experimental simulation scheme, the image preprocessing scheme is to convert the collected infrared image and visible light image of the high-voltage isolating switch into gray level image, and filter and denoise the two images, in order to improve the quality and the processing speed of the image, the image registration scheme is to align the infrared image and the visible light image of the high-voltage isolating switch in space and find out the mapping relation between the infrared image and the visible light image of the high-voltage isolating switch, the high-voltage isolating switch infrared image and the visible light image can be fused in space well, the gradient image fusion scheme is to decompose the high-voltage isolating switch infrared image and the visible light image after registration into a high-frequency sub-band diagram and a low-frequency sub-band diagram, fuse the high-frequency sub-band diagram and the low-frequency sub-band diagram of the high-voltage isolating switch infrared image and the visible light image respectively to realize local image fusion, fuse the infrared low-frequency sub-band diagram and the high-frequency sub-band diagram which are fused respectively to realize global fusion, the gradient image fusion model is formed by local fusion and global fusion of the infrared image and the visible light image of the high-voltage isolating switch, the fusion image evaluation scheme is used for evaluating the quality of the fusion of the infrared image and the visible light image of the high-voltage isolating switch by the gradient image fusion model, the high-voltage isolating switch state identification scheme is used for identifying the image of the fusion of the infrared image and the visible light image of the high-voltage isolating switch, the experimental simulation scheme is used for comparing the fused high-voltage isolating switch image with the identification accuracy of a single high-voltage isolating switch visible light image or a single high-voltage isolating switch infrared image.
2. The method for recognizing the state of the high-voltage isolating switch based on the gradient image fusion as claimed in claim 1, wherein: the image registration scheme is as follows:
carrying out normalization processing on the collected infrared and visible light images by adopting a maximum and minimum value method:
Figure 907303DEST_PATH_IMAGE001
（3）
in the formula
Figure 302381DEST_PATH_IMAGE002
Is a normalized value>
Figure 5894DEST_PATH_IMAGE003
Is the value of the image pixel point, is greater than or equal to>
Figure 629774DEST_PATH_IMAGE004
And &>
Figure 492557DEST_PATH_IMAGE005
Respectively representing the largest pixel value and the smallest pixel value in the image;
adjusting the resolution ratio of the infrared image and the visible light image of the high-voltage isolating switch to 2306 × 2658, mapping the visible light image of the high-voltage isolating switch to the infrared image of the high-voltage isolating switch through a transformation model of image characteristics, performing edge detection on the infrared image and the visible light image of the high-voltage isolating switch by using a Canny operator to extract the outline of the image, detecting and extracting the corner points of the image outline through a SURF algorithm, and extracting the characteristic points of the image outline by adopting a black plug matrix:
Figure 543689DEST_PATH_IMAGE006
（4）
in the formula (I), the compound is shown in the specification,
Figure 78DEST_PATH_IMAGE007
expressed in a scale>
Figure 620459DEST_PATH_IMAGE008
Second-order difference of Gaussian above and a point in the image->
Figure 150797DEST_PATH_IMAGE009
The calculation formula is:
Figure 700727DEST_PATH_IMAGE010
（5）
Figure 565784DEST_PATH_IMAGE011
representing an image, and in a similar way calculating->
Figure 734728DEST_PATH_IMAGE012
And &>
Figure 181890DEST_PATH_IMAGE013
；
And then, matching feature points of the images by using an Euler distance formula, finally fitting the feature points by using a least square method, exchanging parameter estimation, further obtaining an optimal transformation model, and finally obtaining the registered infrared images and visible light images of the high-voltage isolating switch.
3. The method for recognizing the state of the high-voltage isolating switch based on the gradient image fusion as claimed in claim 1, wherein: the gradient image fusion scheme is as follows:
the method comprises the following steps of adopting a non-downsampling shear wave transformation algorithm to respectively decompose an infrared image and a visible light image of a registered high-voltage isolating switch into a high-frequency sub-band diagram and a low-frequency sub-band diagram, adopting a pulse coupling neural network algorithm to fuse the infrared high-frequency sub-band diagram and the visible light high-frequency sub-band diagram of the high-voltage isolating switch, adopting a visual saliency characteristic segmentation algorithm to fuse the infrared low-frequency sub-band diagram and the visible light low-frequency sub-band diagram of the high-voltage isolating switch, realizing local image fusion, fusing the images after local fusion again through a non-downsampling shear wave inverse transformation algorithm, realizing global fusion, obtaining an image after fusing the infrared image and the visible light image of the high-voltage isolating switch, and further realizing a gradient image fusion model, wherein the method comprises the specific steps of:
s1, image decomposition
The method comprises the steps that non-downsampling shear wave transformation decomposition comprises a multiscale decomposition part and a multidirectional decomposition part, in a multiscale decomposition level, a non-downsampling pyramid filter is adopted to decompose an infrared image and a visible light image of a high-voltage isolating switch into 1 low-frequency sub-band diagram and a plurality of high-frequency sub-band diagrams respectively, then the non-downsampling shear wave filter is adopted to decompose the decomposed high-frequency sub-band diagrams in multiple directions to construct a Meyer window function, the high-frequency sub-band diagrams and the Meyer window function are subjected to convolution calculation through a convolution algorithm to obtain high-frequency sub-band coefficients in different directions, and the decomposition of the infrared image and the visible light image of the high-voltage isolating switch in different directions can be achieved;
s2. Gradient image fusion design
S21, local image fusion
S211. High-frequency sub-band diagram fusion
Adopting spatial frequency as input of a pulse coupling neural network algorithm, laplace energy and connection strength of the pulse coupling neural network;
computing a pulse coupled neural network
Figure 89672DEST_PATH_IMAGE014
The ignition frequency after the sub-iteration is used for realizing the fusion of the high-frequency sub-band images according to the ignition frequency, and the spatial frequency of the image is greater or less>
Figure 130440DEST_PATH_IMAGE015
The method comprises the following steps:
Figure 149081DEST_PATH_IMAGE016
（6）
in the formula (I), the compound is shown in the specification,
Figure 716328DEST_PATH_IMAGE017
and &>
Figure 483427DEST_PATH_IMAGE018
Respectively representing the line frequency and the column frequency of the image;
in the pulse coupling neural network model, the correlation degree between neurons is expressed by connection strength, and the connection strength is expressed as:
Figure 401705DEST_PATH_IMAGE019
（7）
Figure 958457DEST_PATH_IMAGE021
（8）
in the formulae (7) and (8),
Figure 255577DEST_PATH_IMAGE022
is the distance between a pixel and a pixel, takes a value of 1, is combined with a pixel>
Figure 505162DEST_PATH_IMAGE023
Represents a high-frequency subband coefficient, < >>
Figure 645156DEST_PATH_IMAGE024
Is the weight of the high-frequency subband coefficient>
Figure 959594DEST_PATH_IMAGE025
For the image in the coordinate->
Figure 422805DEST_PATH_IMAGE026
The sum of laplace energies of (a);
the firing frequency is expressed as:
Figure 656340DEST_PATH_IMAGE027
（9）
in the formula (I), the compound is shown in the specification,
Figure 158997DEST_PATH_IMAGE028
and &>
Figure 57552DEST_PATH_IMAGE029
Respectively the ignition frequency of the high-frequency sub-band diagram after the infrared image and the visible light image of the high-voltage isolating switch are decomposed by non-subsampled shear wave transformation, so that the infrared high-frequency sub-band diagram and the visible light high-frequency sub-band diagram of the high-voltage isolating switch are fused to obtain a high-frequency sub-band fusion diagram;
s212. Fusion of low-frequency sub-band diagrams
The method acquires the significant characteristics (such as outline and background information) in the low-frequency subband image through a hypercomplex number Fourier transform algorithm, divides the low-frequency subband image into two regions of significance and non-significance, and makes a hypercomplex number matrix as follows:
Figure 453898DEST_PATH_IMAGE030
（10）
in the formula (I), the compound is shown in the specification,
Figure 999280DEST_PATH_IMAGE031
represents a hypercomplex number matrix, based on the value of the signal>
Figure 52816DEST_PATH_IMAGE032
Satisfy +>
Figure 833690DEST_PATH_IMAGE033
，
Figure 959909DEST_PATH_IMAGE034
Respectively representing the brightness, the direction and the weight of the texture information of the image, and respectively taking the value as 0.5,0.25,0.25,
Figure 722198DEST_PATH_IMAGE035
a feature matrix representing these three dimensions, respectively, will->
Figure 58501DEST_PATH_IMAGE036
Performing a hypercomplex fourier transform:
Figure 518432DEST_PATH_IMAGE037
（11）
in the formula (I), the compound is shown in the specification,
Figure 810742DEST_PATH_IMAGE038
represents->
Figure 556981DEST_PATH_IMAGE039
A first order norm of the frequency domain of (a);
using pairs of Gaussian kernel functions
Figure 255947DEST_PATH_IMAGE038
Performing smoothing processing to obtain->
Figure 565575DEST_PATH_IMAGE038
The method comprises the following steps of obtaining a salient image by utilizing hypercomplex number inverse Fourier transform, segmenting a salient region and an insignificant region of the image in an image space by taking the minimum entropy value of the image as a criterion, dividing image features into the salient region and the insignificant region by adopting an image fusion method based on visual salient characteristic segmentation, fusing, judging whether the images are similar by comparing the similarity of the two salient images, and expressing as follows: />
Figure 525440DEST_PATH_IMAGE040
（12）
Figure 380264DEST_PATH_IMAGE041
Represents a significant degree of similarity, and>
Figure 815793DEST_PATH_IMAGE042
and &>
Figure 742161DEST_PATH_IMAGE043
Respectively represents the significance values of the infrared image and the visible light image of the high-voltage isolating switch>
Figure 494216DEST_PATH_IMAGE044
The larger the image is, the more similar the infrared image and the visible light image of the high-voltage isolating switch are, the fusion output is as follows:
Figure 238050DEST_PATH_IMAGE045
（13）
in the formula
Figure 770663DEST_PATH_IMAGE046
A weight value representing fusion +>
Figure 907246DEST_PATH_IMAGE047
Represents a similarity threshold (value range->
Figure 966338DEST_PATH_IMAGE048
），/>
Figure 428543DEST_PATH_IMAGE049
Represents an infrared image, <' > based on>
Figure 448452DEST_PATH_IMAGE050
Representing a visible light image;
the image with larger information entropy is fused by comparing the size of the information entropy as a judgment basis of image fusion, and the judgment basis is as follows:
Figure 841256DEST_PATH_IMAGE051
（14）
in the formula
Figure 567903DEST_PATH_IMAGE052
Represents a fused image, based on the image data and the image data>
Figure 263327DEST_PATH_IMAGE053
Is an information entropy ratio, is greater or less than>
Figure 901025DEST_PATH_IMAGE054
，/>
Figure 644990DEST_PATH_IMAGE055
Respectively represent an infrared image and a visible light image when &>
Figure 288461DEST_PATH_IMAGE056
When the information entropy of the infrared image in the area is higher than that of the visible light image, the infrared image is used as a fusion image, otherwise, the visible light image is used as a fusion image, so that the infrared low-frequency sub-band diagram and the visible light low-frequency sub-band diagram of the high-voltage isolating switch are fused to obtain a low-frequency sub-band fusion diagram;
s22, image global fusion
And reconstructing the high-frequency sub-band fusion map and the low-frequency sub-band fusion map of the high-voltage isolating switch by adopting a non-down-sampling shear wave inverse transformation algorithm to obtain a final fusion image of the infrared image and the visible light image of the high-voltage isolating switch, realizing global image fusion, and decomposing the infrared image and the visible light image of the high-voltage isolating switch into a process of local image fusion and global image fusion, namely a gradient image fusion model.
4. The method for recognizing the state of the high-voltage isolating switch based on the gradient image fusion as claimed in claim 1, wherein: the high-voltage isolating switch state identification scheme is as follows:
after a high-voltage isolating switch image fused with a global image is segmented through a Qtsu threshold segmentation algorithm, a high-voltage isolating switch is extracted from a background image, horizontal integral projection and vertical integral projection processing are carried out on the high-voltage isolating switch image segmented from a target through a pixel integral projection method, the state of the isolating switch is judged by calculating the ratio between the horizontal integral projection and the vertical integral projection, and when the isolating switch is in a brake-off state, the ratio of the integral in the vertical direction to the integral in the horizontal direction is calculated to be 0.842; when the high-voltage isolating switch is switched on, the ratio of the two is 0.237, according to the requirement that the angle of the switch blade is not lower than 65 degrees when the high-voltage isolating switch is switched on and switched off, when the angle of the switch blade is 65 degrees, the integral projection ratio of the fused image is 0.826, the switching-off state is set when the integral projection ratio of the pixel is greater than 0.826, the switching-on state is set when the integral projection ratio of the pixel is less than 0.237, and the switching-off state or the switching-on failure is indicated when the integral projection ratio of the pixel is between 0.237 and 0.826.
CN202211432924.XA 2022-11-16 2022-11-16 High-voltage isolating switch state identification method based on gradient image fusion Pending CN115909028A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202211432924.XA CN115909028A (en) 2022-11-16 2022-11-16 High-voltage isolating switch state identification method based on gradient image fusion

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202211432924.XA CN115909028A (en) 2022-11-16 2022-11-16 High-voltage isolating switch state identification method based on gradient image fusion

Publications (1)

Publication Number Publication Date
CN115909028A true CN115909028A (en) 2023-04-04

Family

ID=86472363

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202211432924.XA Pending CN115909028A (en) 2022-11-16 2022-11-16 High-voltage isolating switch state identification method based on gradient image fusion

Country Status (1)

Country Link
CN (1) CN115909028A (en)

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117173703A (en) * 2023-11-02 2023-12-05 温州华嘉电器有限公司 Isolating switch state identification method
CN117409376A (en) * 2023-12-15 2024-01-16 南京中鑫智电科技有限公司 Infrared online monitoring method and system for high-voltage sleeve
CN117972451A (en) * 2024-03-28 2024-05-03 国网安徽省电力有限公司电力科学研究院 GIS isolating switch switching position confirmation method

Cited By (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117173703A (en) * 2023-11-02 2023-12-05 温州华嘉电器有限公司 Isolating switch state identification method
CN117173703B (en) * 2023-11-02 2024-01-16 温州华嘉电器有限公司 Isolating switch state identification method
CN117409376A (en) * 2023-12-15 2024-01-16 南京中鑫智电科技有限公司 Infrared online monitoring method and system for high-voltage sleeve
CN117409376B (en) * 2023-12-15 2024-05-10 南京中鑫智电科技有限公司 Infrared online monitoring method and system for high-voltage sleeve
CN117972451A (en) * 2024-03-28 2024-05-03 国网安徽省电力有限公司电力科学研究院 GIS isolating switch switching position confirmation method
CN117972451B (en) * 2024-03-28 2024-06-11 国网安徽省电力有限公司电力科学研究院 GIS isolating switch switching position confirmation method

Similar Documents

Publication Publication Date Title
CN115909028A (en) High-voltage isolating switch state identification method based on gradient image fusion
CN111079556A (en) Multi-temporal unmanned aerial vehicle video image change area detection and classification method
WO2018023916A1 (en) Shadow removing method for color image and application
CN111666834A (en) Forest fire automatic monitoring and recognizing system and method based on image recognition technology
CN110660065B (en) Infrared fault detection and identification algorithm
CN111339948B (en) Automatic identification method for newly-added buildings of high-resolution remote sensing images
CN109919960B (en) Image continuous edge detection method based on multi-scale Gabor filter
CN106157323B (en) A kind of insulator division and extracting method of dynamic division threshold value and block search combination
CN102494675A (en) High-speed visual capturing method of moving target features
CN111476762B (en) Obstacle detection method and device of inspection equipment and inspection equipment
CN112308873B (en) Edge detection method for multi-scale Gabor wavelet PCA fusion image
Huang et al. Recognition and detection technology of ice-covered insulators under complex environment
Li et al. Infrared-visible image fusion method based on sparse and prior joint saliency detection and LatLRR-FPDE
CN112288682A (en) Electric power equipment defect positioning method based on image registration
CN110751667A (en) Method for detecting infrared dim small target under complex background based on human visual system
CN110047041A (en) A kind of empty-frequency-domain combined Traffic Surveillance Video rain removing method
CN109544608B (en) Unmanned aerial vehicle image acquisition characteristic registration method
CN103632373B (en) A kind of flco detection method of three-frame difference high-order statistic combination OTSU algorithms
Wu et al. Research on crack detection algorithm of asphalt pavement
CN115035350B (en) Edge detection enhancement-based method for detecting small objects on air-ground and ground background
CN110874833A (en) SAR image change detection method based on hypergraph matching
CN113205540B (en) Multi-scale automatic anisotropic morphological direction derivative edge detection method
Zhang et al. Insulator segmentation algorithm based on k-means
CN109215122A (en) A kind of streetscape three-dimensional reconstruction system and method, intelligent carriage
WANG et al. Visual Tracking Inspection Method of Transmission Line based on Infrared Thermal Imaging.

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