Insulator pollution detection method and system based on multispectral fusion

Abstract

The invention discloses an insulator pollution detection method and system based on multispectral fusion, and belongs to the technical field of spectral detection. The method comprises the following steps: firstly, reconstructing a three-dimensional geometric model of a target insulator through a binocular vision method, and calculating the surface area of the target insulator according to the three-dimensional geometric model; the abnormal area of the target insulator is extracted, the optimal detection line is determined according to the three-dimensional coordinate of the central point of the abnormal area, the optimal detection equipment is selected according to the optimal detection line, the workload of multidirectional observation is reduced, and the working efficiency is improved. Second high-frequency spectral images are obtained according to the optimal detection equipment, discharge light spots of the multiple second spectral images are fused to generate a light spot contour and a light spot pixel variance, and the relative discharge position and the actual discharge area are calculated according to the light spot contour; and predicting pollution data of the insulator according to the surface area, the light spot pixel variance, the relative discharge position and the actual discharge area.

Description

Technical Field

[0001] The present invention relates to the field of spectrum detection technology, and in particular to an insulator contamination detection method and system based on multi-spectrum fusion. Background Art

[0002] With the development of railway transportation, the safety detection of contact network has become an important part of train operation safety. In the contact network system, insulators are important components connecting wires and supporting structures. Insulators are easily attached to dirt when exposed to the air, affecting electrical performance. The Chinese patent application with publication number CN115661117A discloses a visible light image detection method for contact network insulators. The method constructs a contact network insulator data set through a contact network insulator image set, performs model training based on the contact network insulator data set, and completes the detection of insulators through the trained insulator monitoring model. The data obtained by a single visible light signal is limited, and it is difficult to obtain accurate detection results. The Chinese patent application with publication number CN113406448A discloses a method and device for detecting the electrical status of railway insulators. The method combines visible light image data and ultraviolet signals for superposition analysis, and then obtains an accurate quantitative evaluation of the degree of electrical damage of the insulator and the equipment status. The application has fixed ultraviolet detection equipment at each position of each insulator, which is costly and not suitable for the detection of long-distance railway power grids. Therefore, the current technology needs to be further improved. Summary of the invention

[0003] In response to the above problems, the present invention provides an insulator contamination detection method and system based on multi-spectral fusion. The present invention can select the best detection equipment according to the abnormal area of ​​the target insulator, and then accurately predict the insulator contamination data according to the first spectral image and the second spectral image of the detection equipment. The method is suitable for the detection of long-distance transmission lines.

[0004] The invention objectives of this application can be achieved through the following technical means: An insulator contamination detection method based on multi-spectral fusion comprises the following steps: Step 1: multiple groups of detection equipment receive reflected light from multiple insulators on the transmission line, generate a first spectral image and a second spectral image according to the reflected light, and select one of the insulators as a target insulator; Step 2: Select two groups of detection devices according to the position of the target insulator, extract the first spectrum images of the two groups of detection devices, identify the feature points of the two groups of first spectrum images and generate feature point pairs, and calculate the parallax of each feature point pair; Step 3: Generate a disparity map according to multiple disparities, calculate the three-dimensional data of each pixel in the disparity map and construct a three-dimensional geometric model of the target insulator, and calculate the surface area of ​​the target insulator according to the three-dimensional geometric model; Step 4: extract the abnormal area in the first spectral image, calculate the three-dimensional coordinates of the center point of the abnormal area based on the three-dimensional geometric model, calculate the best detection line, select a group of detection equipment based on the best detection line, and generate correction parameters of the detection equipment; Step 5: Correcting the first spectral image and the second spectral image of the detection device based on the correction parameters to generate a first corrected spectral image and a second corrected spectral image respectively, and generating a light spot profile and a light spot pixel variance according to a plurality of second corrected spectral images; Step 6: Match the spot profile with the first corrected spectrum image, calculate the relative discharge position and spot area of ​​the spot profile, calculate the pixel height of the target insulator in the first corrected spectrum image, and calculate the actual discharge area based on the pixel height and the spot area; Step 7: Repeat steps 5 to 6, generate a spot timing matrix according to the spot pixel variance, actual discharge area and relative discharge position of multiple sampling cycles, and predict the contamination data of the target insulator according to the surface area of ​​the target insulator and the spot timing matrix.

[0005] In the present invention, in step 1, each detection device comprises a visible light imaging system and an ultraviolet imaging system, the visible light imaging system generates a first spectral image, and the ultraviolet imaging system generates a second spectral image.

[0006] In the present invention, in step 2, the internal parameters and external parameters of the visible light imaging system are obtained, and the two sets of first spectral images are stereo corrected according to the internal parameters and external parameters to calculate the parallax d=|x 2- x1|, where x1 and x2 are the horizontal coordinates of the feature point pairs in the two sets of first spectral images respectively.

[0007] In the present invention, in step 3, the pixel value of each pixel point in the disparity map is the disparity of the corresponding feature point pair, and the depth value Z=fB / d of the pixel point is calculated, where B is the distance between the optical centers of the two groups of visible light imaging systems, and f is the focal length of the visible light imaging system. The three-dimensional data of each pixel point is calculated based on the plane coordinates, depth value and internal parameters of the pixel point in the disparity map.

[0008] In the present invention, in step 3, the three-dimensional geometric model is a point cloud data set composed of three-dimensional data of all pixel points. The three-dimensional geometric model of the target insulator is gridded according to the three-dimensional data to generate a three-dimensional grid model, the area of ​​each grid is calculated, and the areas of the grids are added up to obtain the surface area of ​​the target insulator.

[0009] In the present invention, in step 4, the central axis of the three-dimensional geometric model is extracted, the first spectral image is segmented into a first binary image, the connected domain in the first binary image is extracted, the connected domain is the abnormal area, the three-dimensional coordinates of the center point of the abnormal area are calculated, and the optimal detection line passing through the three-dimensional coordinates and perpendicular to the central axis is generated.

[0010] In the present invention, in step 5, the correction parameters include focal length, rotation matrix and translation vector, initial three-dimensional coordinates corresponding to the two-dimensional image coordinates in the first spectral image are preset, the actual three-dimensional coordinates are generated after iterating the initial three-dimensional coordinates according to the correction parameters, the first corrected spectral image coordinates are calculated according to the actual three-dimensional coordinates, the first corrected spectral image is obtained, and then the pixel values ​​of the first spectral image are mapped to the first corrected spectral image.

[0011] In the present invention, in step 6, the spot area and center point coordinates of the spot profile are calculated, and then the actual discharge area is calculated, and the high-voltage side coordinates of the target insulator in the first corrected spectrum image are obtained to generate a relative discharge position.

[0012] An insulator contamination detection system for realizing the insulator contamination detection method based on multi-spectral fusion includes: a detection device, a detection circuit, a data processing device, and a data analysis device, wherein: A plurality of groups of detection devices are installed in the detection line at intervals from each other, and the detection devices are used to collect a first spectral image and a second spectral image; The detection line is arranged on one side of the power transmission line, and the detection line is used to periodically transmit the first spectrum image and the second spectrum image; The data processing device is connected to the plurality of detection devices via the detection circuit, and the data processing device is used to generate the surface area, the light spot pixel variance, the actual discharge area and the relative discharge position; The data analysis device is connected to the data processing device, and the data analysis device is used to predict the contamination data of the target insulator.

[0013] In the present invention, the detection device includes a spectrometer, a visible light imaging system and an ultraviolet imaging system. The spectrometer extracts visible light and ultraviolet light from the reflected light. The visible light imaging system generates a first spectral image based on the visible light, and the ultraviolet imaging system generates a second spectral image based on the ultraviolet light.

[0014] The beneficial effects of the insulator contamination detection method and system based on multi-spectral fusion implemented in the present invention are as follows: the present invention first reconstructs the three-dimensional geometric model of the target insulator through a binocular detection device, calculates the surface area of ​​the target insulator based on the three-dimensional geometric model, extracts the abnormal area in the first spectrum image, and determines the best detection device for photographing the abnormal area based on the three-dimensional coordinates of the abnormal area. The best detection device can improve the detection accuracy of contamination. The light spot profile is generated based on the high-frequency multiple second spectrum images of the detection device, and then the light spot pixel variance, actual discharge area and relative discharge position are obtained, thereby accurately predicting the contamination data of the target insulator. Multiple groups of detection devices correspond to multiple groups of insulators, reducing the number of detection devices. The best detection device is selected according to the location of the contamination during the detection process, which is suitable for the detection of long-distance transmission lines. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 It is a flow chart of the insulator contamination detection method based on multi-spectrum fusion of the present invention; Figure 2 It is a schematic diagram of the relative positions of two groups of detection equipment of the present invention; Figure 3 A schematic diagram of projecting a point cloud data set onto an XOY plane according to the present invention; Figure 4 A schematic diagram of selecting a detection device for the present invention; Figure 5 A schematic diagram of generating an optimal detection line for the present invention; Figure 6 A schematic diagram of the stereo correction of the present invention; Figure 7 A schematic diagram of calculating three-dimensional data of pixel points according to the present invention; Figure 8 It is a partial schematic diagram of a three-dimensional grid model of a target insulator of the present invention; Fig. 9 A schematic diagram of a target insulator profile and a light spot profile of the present invention; Fig.10 is a block diagram of the insulator contamination detection system of the present invention; Fig.11 It is a block diagram of a preferred detection device of the present invention. DETAILED DESCRIPTION

[0016] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0017] Since there are a large number of insulators in the transmission line, and long-term observation of each position of each insulator will greatly increase the workload. The present invention predicts abnormal areas with obvious contamination based on the first spectral image generated by visible light, measures surface spark discharge based on the second spectral image generated by ultraviolet light, and then accurately predicts contamination data. The main point of the present invention is to select the best detection equipment through the first spectral image of the insulator surface, adjust the shooting angle of the best detection equipment, and then obtain data such as the light spot profile based on the first spectral image and the second spectral image of the corresponding detection equipment, and then obtain the insulator contamination data. Embodiment 1

[0018] Reference Figure 1 The insulator contamination spectrum detection method based on multi-spectrum fusion of the present invention described in detail in this embodiment includes the following steps: Step 1: Multiple groups of detection equipment receive reflected light from multiple insulators on the transmission line, generate a first spectral image and a second spectral image based on the reflected light, and select one of the insulators as the target insulator. Each detection device includes a visible light imaging system and an ultraviolet imaging system. The visible light imaging system generates a first spectral image, and the ultraviolet imaging system generates a second spectral image. The spectral image of the present invention refers to a reflected light image with a frequency in a specific range. The surface condition of the target insulator can be analyzed based on the reflected light images of the target insulator in different frequency ranges. The frequency of the first spectral image is 3.8×10 14 to 7.0×10 14 Hz, the frequency of the second spectrum image is 9.0×10 14 to 1.5×10 16 Hz.

[0019] Step 2: Select two groups of detection devices according to the position of the target insulator, extract the first spectrum images of the two groups of detection devices, identify the feature points of the two groups of first spectrum images and generate feature point pairs, and calculate the parallax of each feature point pair. Figure 2 , obtain a set of detection equipment on both sides of the target insulator, and perform stereo correction on the two sets of first spectral images according to the internal and external parameters of the visible light imaging system. The imaging planes of the visible light imaging systems of the two sets of detection equipment after stereo correction are completely parallel and coplanar. For specific implementation methods, refer to Example 2. Use, for example, the SURF (Speeded-Up Robust Features) algorithm to extract feature points and generate a description vector symbol for each feature point. Match the feature points of the two sets of first spectral images through a bidirectional matching algorithm and a description vector symbol to generate feature point pairs. Calculate the disparity d=|x for each feature point pair 2- x1|, where x1 and x2 are the horizontal coordinates of the two feature points in the feature point pair.

[0020] Step 3: Generate a disparity map according to multiple disparities, calculate the three-dimensional data of each pixel in the disparity map and construct a three-dimensional geometric model of the target insulator, and calculate the surface area of ​​the target insulator according to the three-dimensional geometric model. The pixel value of each pixel in the disparity map is the disparity of the corresponding feature point pair, and the depth value Z=fB / d of the pixel is calculated, where B is the distance between the optical centers of the two sets of visible light imaging systems, and f is the focal length of the visible light imaging system. In this embodiment, the focal lengths of the visible light imaging systems of all detection devices are f. The three-dimensional data of each pixel is calculated according to the plane coordinates, depth value and internal parameters of the pixel in the disparity map. The three-dimensional geometric model is a point cloud data set composed of the three-dimensional data of all pixels. The present invention does not limit the method for calculating the surface area of ​​the target insulator, and can use analytical geometry, numerical integration, and the grid accumulation method of embodiment 4. Specifically, the three-dimensional geometric model of the target insulator is first gridded according to the point cloud data set to generate a three-dimensional grid model, the area of ​​each grid is calculated, and the area of ​​each grid is accumulated to obtain the surface area of ​​the target insulator.

[0021] Step 4: Extract the abnormal area in the first spectral image, calculate the three-dimensional coordinates of the center point of the abnormal area based on the three-dimensional geometric model, calculate the best detection line, select a set of detection equipment based on the best detection line, and generate the calibration parameters of the detection equipment. First, extract the central axis of the three-dimensional geometric model, refer to Figure 3 , project the point cloud data set of the three-dimensional geometric model to the XOY plane and the XOZ plane respectively to generate two sets of two-dimensional point sets. In the XOY plane, calculate the covariance matrix of the two-dimensional point set, perform eigenvalue decomposition on the covariance matrix, and obtain eigenvalues ​​and eigenvectors. Select the eigenvector with the largest eigenvalue as the main direction on the XOY plane, divide multiple groups of slices along the main direction, and determine the maximum and minimum values ​​of each slice on the Y axis and X axis. According to the maximum and minimum values, the mean point is calculated. The mean point is the axis point on the XOY plane. The linear regression model is used for fitting, and all axis points are fitted as Y=a1X+a2. Similarly, to find the axis point on the XOZ plane, it can be fitted as Z= a3X+a4, then the central axis of the target insulator in the three-dimensional space can be expressed as , where a1 and a3 are slopes, and a2 and a4 are intercepts.

[0022] Then calculate the three-dimensional coordinates of the center point of the abnormal area. The first spectral image is divided into a first binary image based on a preset threshold. The preset threshold can be automatically determined by the Otsu method. The method automatically finds the threshold that maximizes the inter-class variance between the abnormal area and the non-abnormal area based on the grayscale histogram of the first spectral image. The connected domains in the first binary image are extracted, and the connected domains with too small areas and too long shapes are removed to eliminate the influence of noise points. Among the remaining connected domains, the connected domain with the largest area is selected as the abnormal area. The contour of the abnormal area is extracted, and the two-dimensional image coordinates in the first spectral image are selected so that the sum of the distances from the two-dimensional image coordinates to each point in the contour is minimized. The two-dimensional image coordinates are the center point (geometric center) of the abnormal area. The center point is mapped to a three-dimensional geometric model to obtain the three-dimensional coordinates (X1, Y1, Z1) of the center point.

[0023] Reference Figure 4 and Figure 5 , generate the best detection line passing through the center point and perpendicular to the center axis. Calculate the vertical point coordinates (X0, Y0, Z0) of the center point perpendicular to the center axis, where, , Y0= a1X0+a2, Z0= a3X0+ a4, then the expression of the optimal detection line is . Obtain a set of detection devices closest to the best detection line. This detection device is the best detection device for the target insulator and can obtain the best image of the abnormal area. For the visible light imaging system of the detection device, its internal parameters include focal length, and its external parameters include rotation matrix and translation vector. Then the correction parameters of the visible light imaging system are focal length, rotation matrix and translation vector. Similarly, the correction parameters of the ultraviolet imaging system can be obtained.

[0024] Step 5: Correct the first spectral image and the second spectral image of the detection device based on the correction parameters, generate the first corrected spectral image and the second corrected spectral image respectively, and generate the spot contour and the spot pixel variance according to the multiple second corrected spectral images. Generate a collinearity equation function according to the focal length, rotation matrix and translation vector of the visible light imaging system, preset the initial three-dimensional coordinates corresponding to the two-dimensional image coordinates in the first spectral image, iterate and generate the actual three-dimensional coordinates according to the collinearity equation function, calculate the first corrected spectral image coordinates according to the actual three-dimensional coordinates, obtain the first corrected spectral image, and then map the pixel values ​​of the first spectral image to the first corrected spectral image. Similarly, the second corrected spectral image can be generated according to the correction parameters of the ultraviolet imaging system and the second spectral image.

[0025] In this embodiment, multiple second corrected spectrum images of the detection device are extracted, and the multiple second corrected spectrum images are fused using the average method to generate a mean image, and the average pixel value of the mean image is calculated. The average pixel value is used as a threshold, and then a binary image of the mean image is generated based on the threshold segmentation method. The contour of the mean discharge light spot is extracted from the binary image using an edge detection algorithm, and the contour is fitted using a minimum ellipse fitting method to generate a light spot contour. The light spot pixel variance is the variance of the pixel values ​​in the mean image.

[0026] Step 6: Match the spot profile with the first corrected spectrum image, calculate the relative discharge position and spot area of ​​the spot profile, calculate the pixel height of the target insulator in the first corrected spectrum image, and calculate the actual discharge area based on the pixel height and spot area. Extract the discharge spots of multiple second corrected spectrum images, calculate the similarity between each discharge spot and the spot profile, select the second corrected spectrum image corresponding to the maximum similarity, take the shooting time of the second spectrum image corresponding to the second corrected spectrum image as the ideal time, extract the first spectrum image of the corresponding detection device according to the ideal time, and generate the first corrected spectrum image of the first spectrum image, calculate the spot area and center point coordinates of the spot profile, and then calculate the actual discharge area, and at the same time obtain the high-voltage side coordinates of the target insulator in the first corrected spectrum image to generate the relative discharge position. For the specific calculation method, refer to Example 5.

[0027] Step 7: Repeat steps 5 to 6, generate a spot timing matrix according to the spot pixel variance, actual discharge area and relative discharge position of multiple sampling cycles, and predict the contamination data of the target insulator according to the surface area of ​​the target insulator and the spot timing matrix. The present invention does not limit the training method of the contamination prediction model and the type of contamination data. The contamination data includes the contamination level, the level of influence of contamination on insulation strength and the contamination thickness. The contamination prediction model is trained according to the spot timing matrix and surface area of ​​different insulators, and then the spot timing matrix and surface area of ​​the target insulator are input into the contamination prediction model to predict the contamination data of the target insulator. Embodiment 2

[0028] Reference Figure 6 This embodiment further discloses a method for stereoscopically correcting two sets of first spectral images according to internal parameters and external parameters of visible light imaging systems of two sets of detection devices.

[0029] Step 201: Obtain the internal parameters and external parameters of the visible light imaging systems of the two groups of detection devices. The internal parameters of the visible light imaging systems of the two groups of detection devices are consistent, including focal length, principal point and distortion parameters, and the external parameters include rotation matrix R and translation vector t. The rotation matrix R and translation vector t are respectively the rotation matrix and translation vector of the visible light imaging system of one group of detection devices relative to the visible light imaging system of the other group of detection devices. For example, the internal parameters and external parameters of the first camera unit and the second camera unit are obtained by using a checkerboard calibration plate method, t=[t x ,t y ,t z ] T , t x ,t y ,t z are the translation vectors in the x, y, and z coordinate axes directions respectively.

[0030] Step 202: Decompose the rotation matrix R into matrix R1 and matrix R2, and construct the transformation matrix R3 through the translation vector, where: , , R3=[e1,e2,e3], , , e3=e1×e2.

[0031] Step 203: Multiply the transformation matrix R3 by the translation vector t, the matrices R1 and R2 to obtain new matrices R1', R2' and the translation vector t', that is, R1'=R3×R1, R2'=R3×R2, t'=R3×t.

[0032] Step 204: Construct the internal matrix , where f x and f y are the scaling factors of the focal length in the x-axis and y-axis directions of the first spectral image. In this embodiment, f x =f y , (c x ,c y ) as the main point, and construct projection matrices Q1 and Q2 according to matrices K, R1', R2' and t', Q1=K×[R1'|0], Q2=K×[R2'| t'], where [R1'|0] represents the matrix formed by horizontally splicing matrix R1' and a zero vector, and [R2'| t'] represents the matrix formed by horizontally splicing matrix R2' and translation vector t'. The two sets of first spectral images are reprojected according to projection matrices Q1 and Q2 respectively to complete the stereo correction of the two sets of first spectral images. Embodiment 3

[0033] Reference Figure 7, this embodiment further discloses a method for constructing a three-dimensional geometric model based on two sets of first spectral images after stereo correction.

[0034] Step 301: feature points of the two groups of first spectral images are extracted respectively using a feature point detection algorithm. In this embodiment, a SURF (Speeded-Up Robust Features) algorithm is used to extract feature points, and multiple 4×4 small areas are divided around each feature point, and then a descriptor vector is constructed, which includes the gradient direction and gradient amplitude of each small area.

[0035] Step 302: Taking the feature points of the first spectral image of one group as the reference, for each feature point, calculate the Euclidean distance between the feature point and the descriptor vectors of all feature points in the first spectral image of the other group. Select the feature point with the smallest Euclidean distance for matching, and generate a feature point pair set p1'. Then, taking the feature points of the first spectral image of the other group as the reference, generate a feature point pair set p2' in the same way. Compare the feature point pairs in p1' and p2', and only retain the feature point pairs that exist in both p1' and p2'.

[0036] Step 303: For the feature point pair [p1 (x1, y1), p2 (x2, y2)], calculate its disparity d = |x2-x1| and generate a disparity map. The number of pixels in the disparity map is less than the number of feature points in any set of first spectrum images, and the pixel value of each pixel is the disparity of the corresponding feature point pair.

[0037] Step 304: For each feature point pair [p1 (x1, y1), p2 (x2, y2)], calculate the corresponding three-dimensional data P (X, Y, Z). Calculate the depth value Z = fB / d based on the pixel value and internal parameters, then X = (x1-c x )Z / f,Y=(y1-c y )Z / f, where (c x , c y ) is the projection point of the optical center O1 of the visible light imaging system on the corresponding first spectrum image, that is, (c x , c y ) is the principal point in the internal parameters, B is the distance between the optical center O1 of one visible light imaging system and the optical center O2 of another visible light imaging system, and f is the focal length.

[0038] Step 305: Store each three-dimensional data in an array to generate a point cloud data set, which is a three-dimensional geometric model. Embodiment 4

[0039] This embodiment further discloses a method for constructing a three-dimensional mesh model based on a point cloud data set of a three-dimensional geometric model, and calculating the surface area of ​​a target insulator based on the three-dimensional mesh model. Figure 8 , a three-dimensional mesh model is constructed using the triangular mesh model construction method.

[0040] Step 401: Select an initial seed point P0 from the point cloud data set, select two points P1 and P2 closest to P0, generate an initial triangle G0=(P0, P1, P2), and the initial mesh is the initial triangle.

[0041] Step 402: Select an edge g from all the boundary edges of the current mesh, find the point P closest to the two endpoints of edge g in the point cloud data set, and generate a new triangular mesh with edge g and point P. If the new triangular mesh meets the shape criteria, add the new triangle to the mesh, update the boundary edges of the mesh, and remove the boundary edges from edge g. The shape criteria are that the minimum angle of the new triangular mesh is greater than 15 degrees and the aspect ratio of the new triangular mesh is less than 2.

[0042] Step 403: When no points satisfying the shape criteria can be found in the point cloud data set for all boundary edges, the mesh growth is terminated. The mesh at this time is the final three-dimensional mesh model. The area of ​​each triangular mesh in the three-dimensional mesh model is calculated, and the area of ​​each triangular mesh is accumulated to obtain the surface area of ​​the target insulator. Embodiment 5

[0043] Reference Fig. 9 , this embodiment further discloses a method for calculating the relative discharge position and the actual discharge area according to the first corrected spectrum image and the spot profile.

[0044] In this embodiment, the spot profile is fitted as an ellipse, the major axis of the ellipse is L1, the minor axis is L2, and the intersection of the major axis and the minor axis is the coordinate of the center point of the spot profile (x0, y0). The present invention presets an image template of the target insulator, extracts the target insulator pattern of the image template from the first corrected spectrum image, and obtains the target insulator profile of the target insulator pattern by an edge detection method. The coordinates (x', y') of the high-voltage side of the target insulator are obtained according to the target insulator profile, and the relative discharge position is .

[0045] The spot area of ​​the spot contour is S1=πL1L2 / 4. The high-voltage side boundary point and the low-voltage side boundary point of the target insulator are obtained according to the target insulator contour. The pixel distance between the two boundary points is the pixel height h of the target insulator, so the actual discharge area S=S1H 2 / h 2 , where H is the calibrated height of the target insulator, and the calibrated height of the target insulator can be calculated according to the three-dimensional geometric model. Embodiment 6

[0046] like Fig.10 and Fig.11 The insulator contamination detection system for realizing the insulator contamination detection method based on multi-spectral fusion of the present invention comprises: a detection device, a detection line, a data processing device, a data analysis device, a database and a display device. Among them, multiple groups of detection devices are installed on the detection line at intervals, and the detection device is used to collect the first spectrum image and the second spectrum image. The detection line is arranged on one side of the transmission line, and the detection line is used to periodically transmit the first spectrum image and the second spectrum image. The data processing device is connected to the multiple groups of detection devices via the detection line, and the data processing device is connected to the multiple groups of detection devices via the detection line, and the data processing device is used to generate the surface area, the light spot pixel variance, the actual discharge area and the relative discharge position. The data analysis device is connected to the data processing device, and the data analysis device is used to predict the contamination data of the target insulator. The database is connected to the data analysis device, and is used to store the contamination prediction model, the insulator contamination data and the detection line coordinates. The display device is used to display the insulator contamination data. In this embodiment, the detection device includes a spectrometer, a visible light imaging system and an ultraviolet imaging system, wherein the visible light imaging system includes a visible light sensor and a first imaging unit, the ultraviolet imaging system includes an ultraviolet light sensor and a second imaging unit, the spectrometer extracts visible light and ultraviolet light from the reflected light, the visible light sensor receives visible light and transmits it to the first imaging unit, the first imaging unit generates a first spectral image, the ultraviolet light sensor receives ultraviolet light and transmits it to the second imaging unit, and the second imaging unit generates a second spectral image.

[0047] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for detecting insulator contamination based on multi-spectral fusion, characterized in that: The following steps are involved: Step 1: multiple groups of detection equipment receive reflected light from multiple insulators on the transmission line, generate a first spectral image and a second spectral image according to the reflected light, and select one of the insulators as a target insulator; Step 2: Select two groups of detection devices according to the position of the target insulator, extract the first spectrum images of the two groups of detection devices, identify the feature points of the two groups of first spectrum images and generate feature point pairs, and calculate the parallax of each feature point pair; Step 3: Generate a disparity map according to multiple disparities, calculate the three-dimensional data of each pixel in the disparity map and construct a three-dimensional geometric model of the target insulator, and calculate the surface area of ​​the target insulator according to the three-dimensional geometric model; Step 4: extract the abnormal area in the first spectral image, calculate the three-dimensional coordinates of the center point of the abnormal area based on the three-dimensional geometric model, calculate the best detection line, select a group of detection equipment based on the best detection line, and generate correction parameters of the detection equipment; Step 5: Correcting the first spectral image and the second spectral image of the detection device based on the correction parameters, generating a first corrected spectral image and a second corrected spectral image respectively, and generating a light spot profile and a light spot pixel variance according to a plurality of second corrected spectral images; Step 6: Match the spot profile with the first corrected spectrum image, calculate the relative discharge position and spot area of ​​the spot profile, calculate the pixel height of the target insulator in the first corrected spectrum image, and calculate the actual discharge area based on the pixel height and the spot area; Step 7: Repeat steps 5 to 6, generate a spot timing matrix according to the spot pixel variance, actual discharge area and relative discharge position of multiple sampling cycles, and predict the contamination data of the target insulator according to the surface area of ​​the target insulator and the spot timing matrix.

2. The insulator contamination detection method based on multi-spectral fusion according to claim 1 is characterized in that: In step 1, each detection device includes a visible light imaging system and an ultraviolet imaging system, the visible light imaging system generates a first spectral image, and the ultraviolet imaging system generates a second spectral image.

3. The insulator contamination detection method based on multi-spectral fusion according to claim 2 is characterized in that: In step 2, the internal parameters and external parameters of the visible light imaging system are obtained, and the two sets of first spectrum images are stereo corrected according to the internal parameters and external parameters to calculate the parallax d=|x 2- x1|, where x1 and x2 are the horizontal coordinates of the feature point pairs in the two sets of first spectral images respectively.

4. The insulator contamination detection method based on multi-spectral fusion according to claim 3 is characterized in that: In step 3, the pixel value of each pixel in the disparity map is the disparity of the corresponding feature point pair, and the depth value Z=fB / d of the pixel is calculated, where B is the distance between the optical centers of the two sets of visible light imaging systems, and f is the focal length of the visible light imaging system. The three-dimensional data of each pixel is calculated based on the plane coordinates, depth value and internal parameters of the pixel in the disparity map.

5. The insulator contamination detection method based on multi-spectral fusion according to claim 4 is characterized in that: In step 3, the three-dimensional geometric model is a point cloud data set composed of three-dimensional data of all pixel points. The three-dimensional geometric model of the target insulator is meshed according to the three-dimensional data to generate a three-dimensional mesh model, the area of ​​each mesh is calculated, and the area of ​​the meshes is accumulated to obtain the surface area of ​​the target insulator.

6. The insulator contamination detection method based on multi-spectral fusion according to claim 1 is characterized in that: In step 4, the central axis of the three-dimensional geometric model is extracted, the first spectral image is segmented into a first binary image, the connected domain in the first binary image is extracted, the connected domain is the abnormal area, the three-dimensional coordinates of the center point of the abnormal area are calculated, and the optimal detection line passing through the three-dimensional coordinates and perpendicular to the central axis is generated.

7. The insulator contamination detection method based on multi-spectral fusion according to claim 1 is characterized in that: In step 5, the correction parameters include focal length, rotation matrix and translation vector, initial three-dimensional coordinates corresponding to the two-dimensional image coordinates in the first spectral image are preset, the actual three-dimensional coordinates are generated after iterating the initial three-dimensional coordinates according to the correction parameters, the first corrected spectral image coordinates are calculated according to the actual three-dimensional coordinates, the first corrected spectral image is obtained, and then the pixel values ​​of the first spectral image are mapped to the first corrected spectral image.

8. The insulator contamination detection method based on multi-spectral fusion according to claim 1 is characterized in that: In step 6, the spot area and center point coordinates of the spot profile are calculated, and then the actual discharge area is calculated. At the same time, the high-voltage side coordinates of the target insulator in the first corrected spectrum image are obtained to generate a relative discharge position.

9. An insulator contamination detection system for realizing the insulator contamination detection method based on multi-spectral fusion as claimed in claim 1, characterized in that: include: Testing equipment, testing circuits, data processing equipment, data analysis equipment, among which, A plurality of groups of detection devices are installed in the detection line at intervals from each other, and the detection devices are used to collect a first spectral image and a second spectral image; The detection line is arranged on one side of the power transmission line, and the detection line is used to periodically transmit the first spectrum image and the second spectrum image; The data processing device is connected to the plurality of detection devices via the detection circuit, and the data processing device is used to generate the surface area, the light spot pixel variance, the actual discharge area and the relative discharge position; The data analysis device is connected to the data processing device, and the data analysis device is used to predict the contamination data of the target insulator.

10. The insulator contamination detection system according to claim 9, characterized in that: The detection device includes a spectrometer, a visible light imaging system and an ultraviolet imaging system. The spectrometer extracts visible light and ultraviolet light from reflected light. The visible light imaging system generates a first spectral image based on the visible light, and the ultraviolet imaging system generates a second spectral image based on the ultraviolet light.

Images