Discharge detection method and system for ultraviolet and visible light image adaptive registration

Abstract

The application discloses a discharge detection method and system of ultraviolet and visible light image adaptive registration, and belongs to the technical field of image processing. The imaging module is used for capturing ultraviolet images and visible light images of a discharge point area of a power equipment respectively. The registration module is used for calculating motion vectors of the visible light image sequence according to optical flow motion estimation and mapping to the ultraviolet image for displacement compensation in a motion scene, and determining translation compensation by histogram matching in a static scene. According to the compensated images, the ultraviolet image is scaled by a scaling ratio calculated by focal length and a center point of the ultraviolet image, and then the scaled ultraviolet image is scaled to a target size. The discharge detection module detects the discharge of the discharge point area according to the registered images. The application improves robustness, avoids boundary offset and nonlinear distortion, realizes edge linear fitting while ensuring accurate alignment of the center point, and significantly improves the calculation efficiency by cooperating with integer operation.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of image processing, and particularly relates to a discharge detection method and system for adaptive registration of ultraviolet and visible light images. BACKGROUND

[0002] In the fields of power equipment detection and high-voltage discharge monitoring, ultraviolet imaging technology is widely used for detecting defects such as corona discharge and partial discharge. Ultraviolet instruments capture the light signals of the discharge area in the ultraviolet band to generate corresponding ultraviolet images, while visible light cameras simultaneously capture visible light images of the same area. However, in actual applications, there is a spatial position deviation between the ultraviolet image and the visible light image, mainly due to the following reasons:

[0003] 1) Optical system difference: There are differences in focal length, distortion, installation position, etc. between the ultraviolet lens and the visible light lens, resulting in that the positions of the same target in the two images do not completely correspond.

[0004] 2) Dynamic tracking problem: When the detection target (such as a discharge point on a high-voltage device) moves, the visible light image can be tracked in real time, but the ultraviolet image may not be able to move synchronously due to different imaging mechanisms, resulting in that the ultraviolet-visible light images cannot be accurately matched.

[0005] Currently, common image registration methods include feature point matching (such as SIFT, ORB) and affine transformation (such as translation, rotation, scaling), but these methods have the following problems in aligning ultraviolet-visible light images:

[0006] 1) Feature point matching failure: The ultraviolet image usually has low signal-to-noise ratio and less texture information, making it difficult to extract feature points and reducing matching accuracy.

[0007] 2) Affine transformation calculation complexity: Traditional methods need to calculate a global transformation matrix (such as a homography matrix), which has large calculation amount and is difficult to adapt to dynamic adjustment under different focal lengths.

[0008] 3) Boundary alignment error: In the zooming process, the scaling ratios of the ultraviolet and visible light images may not be consistent, resulting in that the center points are aligned but the boundaries are shifted, affecting detection accuracy. SUMMARY

[0009] To solve the problems of the prior art and achieve the purpose of accurate alignment of ultraviolet images and visible light images, the present application adopts the following technical solution:

[0010] An adaptive registration method for ultraviolet and visible light images, comprising the following steps:

[0011] Step S1: Initial registration of ultraviolet images and visible light images;

[0012] For the moving scene, the motion vectors of the visible light image sequence are calculated by optical flow motion estimation and mapped to the ultraviolet image for displacement compensation; for the static scene, histogram matching is used to determine the translation compensation;

[0013] Step S2: focal length adaptive zooming;

[0014] The zooming ratio is calculated by the focal length, and the ultraviolet image is zoomed according to the zooming ratio and the center point of the ultraviolet image.

[0015] Step S3: image zooming;

[0016] The cropped ultraviolet image effective registration area is zoomed to the target size.

[0017] Further, the optical flow motion estimation in step S1 obtains the optimal displacement amount by minimizing the pixel intensity difference in the local window between adjacent frames of the visible light, for the alignment compensation of the ultraviolet image.

[0018] Further, by setting a window region, the pixel intensity values of the visible light image at the current time and spatial position in the window region are calculated, and the pixel intensity values of the corresponding spatial position after displacement compensation of the next time ultraviolet image are calculated, and the square of the pixel intensity difference between the current time and the next time after displacement compensation is calculated. The sum of the squares of all pixel intensity differences in the window region is summed to obtain the ultraviolet image displacement compensation when the sum of the squares of the pixel intensity differences is minimized.

[0019] The formula of optical flow motion estimation is as follows:

[0020]

[0021] wherein, represents the two-dimensional displacement vector to be solved, and respectively represent the displacement amount of the ultraviolet image in the horizontal coordinate x and the vertical coordinate y direction that needs to be compensated, arg min represents the optimal displacement parameter that minimizes the objective function, and the best displacement can be solved by minimizing the error function, represents the summation of all pixels in the specified window region W, represents the optical flow calculation window, represents the visible light image, represents the pixel intensity value of the visible light image at time t, spatial position (x, y), represents the pixel intensity value of the corresponding position after displacement of the next time .

[0022] Further, the histogram matching in the step S1 is to calculate the normalized cross-correlation value of the visible light image and the corresponding ultraviolet light image, take the offset corresponding to the maximum normalized cross-correlation value as the alignment parameter, and perform translation on the ultraviolet light image to align the visible light image.

[0023] Further, the normalized cross-correlation value is to multiply the pixel value of the visible light image and the pixel value of the corresponding position of the ultraviolet light image after the offset, sum all the products in the overlapping region of the visible light image and the ultraviolet light image as the numerator, square sum all the pixel values of the visible light image in the overlapping region, square sum all the pixel values of the corresponding position of the ultraviolet light image after the offset, and then multiply the square sum of all the pixel values of the visible light image and the square sum of all the pixel values of the corresponding position of the ultraviolet light image after the offset to obtain the denominator.

[0024] The normalized cross-correlation value formula is as follows:

[0025]

[0026] wherein, represents the normalized cross-correlation value under the displacement , which is used to measure the similarity of two images under a given displacement, respectively represent the displacement to be evaluated, represents the displacement of x in the horizontal direction, represents the displacement of y in the vertical direction, represents the summation of all pixel coordinates (x, y), and the summation range is all the effective pixel points in the image overlapping region, represents the pixel value of the visible light reference image at the coordinate (x, y), represents the square of the pixel value of the visible light image at the coordinate (x, y), is the pixel value of the ultraviolet target image at the corresponding position after the offset (dx, dy), which represents the alignment of the ultraviolet image after the offset (dx, dy) with the visible light image, represents the square of the pixel value of the ultraviolet image at the corresponding position after the offset (dx, dy).

[0027] Further, in the ultraviolet-visible light image registration process, when the optical system is zoomed, the scaling ratio of the ultraviolet and visible light images will change with the focal length, and therefore, in the step S2 of the present application, the linear relationship between the focal length and the horizontal and / or vertical scaling coefficient is calibrated by experiment, the position of the cropping window is dynamically adjusted according to the scaling coefficient and the center point coordinate of the matching region of the ultraviolet image, the effective matching region of the ultraviolet image is obtained, the center point is accurately aligned and the boundary is linearly fitted through the linear relationship calibrated by experiment, the boundary offset and nonlinear distortion problems caused by the traditional fixed scaling method are effectively solved, and the matching demand under different focal lengths can be adaptively met.

[0028] Further, in step S3, using a bilinear interpolation algorithm, the pixel values of the pixel points in the effective matching region of the ultraviolet image are adjusted based on the weight distribution on the nearest integer coordinates of the pixel points in the horizontal and vertical directions, so that the effective matching region of the ultraviolet image is scaled to the target size to obtain the target image.

[0029] Further, the bilinear interpolation algorithm calculates the integer coordinates of the reference pixels nearest to the coordinates of the pixel points in the effective matching region of the ultraviolet image in the horizontal and / or vertical directions, and obtains the difference weight in the horizontal and / or vertical directions based on the distance ratio of the coordinates of the pixel points to the integer coordinates of the reference pixels in the horizontal and / or vertical directions, and obtains the pixel value of the coordinates of the pixel points by summing the products of the integer coordinates of the reference pixels and the corresponding difference weights, to generate the target image after scaling of the effective matching region of the ultraviolet image.

[0030] The pixel value of the target image at coordinates (x, y) is as follows:

[0031]

[0032] wherein dst represents the scaled target image, src represents the effective matching region of the original ultraviolet image, , , , respectively represent the four nearest pixel points in the target image of the pixel points in the effective matching region of the ultraviolet image, u and v are the interpolation weights in the horizontal and vertical directions (0 ≤ u, v ≤ 1) for representing the horizontal distance ratio of the target point to and , and are the integer coordinates of the reference pixels in the source image (calculated by ).

[0033] An ultraviolet and visible light image adaptive registration discharge detection method, which acquires an ultraviolet image and a visible light image of a discharge point region of a power equipment, and performs registration on the ultraviolet and visible light images of the discharge point region of the power equipment according to the ultraviolet and visible light image adaptive registration method, for discharge detection.

[0034] An ultraviolet and visible light image adaptive registration discharge detection system, which comprises an imaging module, a registration module and a discharge detection module.

[0035] The imaging module comprises an ultraviolet imaging module and a visible light imaging module, and is used for capturing ultraviolet images and visible light images of the discharge point area of the power equipment respectively.

[0036] The registration module registers the ultraviolet and visible light images of the discharge point area of the power equipment according to the ultraviolet and visible light image adaptive registration method.

[0037] The discharge detection module detects the discharge of the discharge point area according to the registered images.

[0038] The advantages and beneficial effects of the present application are as follows:

[0039] The discharge detection method and system for ultraviolet and visible light image adaptive registration provided by the present application realize high-precision matching of images by dynamically adjusting the focal length parameter to cut the region and combining the improved interpolation method to accurately align the ultraviolet images and the visible light images through step-by-step optimization, and are especially suitable for the alignment of ultraviolet and visible light images in the discharge detection of power equipment and other application scenarios. BRIEF DESCRIPTION OF DRAWINGS

[0040] Figure 1 is a traditional method of ultraviolet and visible light synchronous comparison effect diagram.

[0041] Figure 2 is a flowchart of the method in the embodiment of the present application.

[0042] Figure 3 is an ultraviolet and visible light synchronous comparison effect diagram after initial registration in the embodiment of the present application.

[0043] Figure 4 is an ultraviolet and visible light synchronous comparison effect diagram after focal length adaptive scaling in the embodiment of the present application.

[0044] Figure 5 is a structural schematic diagram of the system in the embodiment of the present application. DETAILED DESCRIPTION

[0045] The specific embodiments of the present application will be described in detail below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are only used to illustrate and explain the present application, and are not used to limit the present application.

[0046] As shown in Figure 1 , there is a spatial position deviation between the ultraviolet images and the visible light images in actual application, which leads to the failure of registration of the ultraviolet and visible light images. In view of the problem that the traditional feature matching method (such as SIFT / SURF) is complex in calculation and has a high failure rate in registration due to the few features of the ultraviolet images, the present application proposes an ultraviolet and visible light image adaptive registration method, as shown in Figure 2 , which comprises the following steps.

[0047] Step S1: Initial registration of UV and visible images; fast coarse alignment by fusing optical flow motion estimation and histogram cross-correlation matching.

[0048] Firstly, the Lucas-Kanade optical flow method is used to calculate the motion vector of the visible image sequence, which is then mapped to the UV image for displacement compensation. For static scenes, the histogram matching algorithm based on normalized cross-correlation (NCC) is used to determine the optimal translation parameters, thereby achieving fast and robust initial registration and laying the foundation for subsequent fine alignment.

[0049] (1) Displacement estimation based on optical flow method

[0050] If there is motion in the scene (such as camera movement or target displacement), the Lucas-Kanade sparse optical flow method is used to calculate the pixel motion vector of the visible light image, which is then mapped to the UV image:

[0051]

[0052] where, represents the two-dimensional displacement vector to be solved, and represent the displacement amount in the x and y directions of the UV image that needs to be compensated, argmin represents the optimal displacement parameter that minimizes the objective function, which can be solved by minimizing the error function, represents the sum of all pixels in the specified window region W, is the optical flow calculation window (such as 5x5 pixels), represents the visible light image, represents the pixel intensity value of the visible light image at time t, spatial position (x, y), represents the pixel intensity value at the corresponding position after displacement at the next time .

[0053] The core idea of this equation is to minimize the pixel brightness difference within the local window W between adjacent frames to solve the optimal displacement vector ( ). In UV and visible light registration, the displacement vector solved will be used for initial alignment compensation of the UV image.

[0054] (2) Translation compensation based on histogram matching

[0055] For static scenes, the normalized cross-correlation (NCC) of the visible and UV images is calculated to determine the best translation offset (Δx, Δy):

[0056]

[0057] where, denotes the normalized cross-correlation value at displacement , with value range: [-1, 1], 1 means perfect match, used to measure the similarity of two images at a given displacement, denotes the displacement to be evaluated, respectively, where denotes the x-direction (horizontal) displacement (unit: pixel), denotes the y-direction (vertical) displacement (unit: pixel), usually traversed within a certain range (such as ±20 pixels). denotes the summation of all pixel coordinates (x, y), and the summation range is all valid pixel points in the image overlap area. denotes the pixel value of the visible light reference image at coordinates (x, y), usually the gray value (0-255), while denotes the square of the pixel value of the visible light image at (x, y), used as the denominator part of the normalization calculation, denotes the corresponding pixel value of the ultraviolet target image after shifting (dx, dy), which represents the alignment of the ultraviolet image after shifting (dx, dy) with the visible light image, denotes the square of the pixel value of the ultraviolet image at the shifted position, also used for normalization calculation.

[0058] Take the shift amount corresponding to the maximum NCC value as the coarse alignment parameter:

[0059]

[0060] where, denotes the optimal displacement vector finally solved, and and denote the best displacement in x and y directions, respectively. If (3.2, -1.5) is obtained, it means that the optimal registration needs to move right by 3.2 pixels and up by 1.5 pixels. is a mathematical operator used to find the argument that makes the objective function NCC take the maximum value .

[0061] Output: the coarse alignment result of the ultraviolet image after translation transformation:

[0062]

[0063] where denotes the original ultraviolet image, denotes the image after translation transformation, denotes the target image coordinates, denotes the translation displacement.

[0064] Table 1 Performance comparison table of ultraviolet-visible light image initial registration algorithm

[0065] Indicator Traditional SIFT method Invention fusion algorithm Lifting amplitude Feature point extraction success rate 32% No need for feature points ∞ Single frame processing time (ms) 120 18 85% Translation error (pixels) ±3.5 ±0.8 77% Light change robustness Poor Excellent - Dynamic scene adaptability Not support Support -

[0066] Table 1 quantitatively compares the performance difference between the traditional SIFT method and the fusion optical flow-NCC algorithm of the present application. The data shows that the present application is significantly better than the traditional method in terms of feature point dependence, processing speed, registration accuracy, etc.:

[0067] Feature point extraction: the success rate of the traditional method is only 32%, while the present application does not require feature points and directly solves the problem of few features in ultraviolet images;

[0068] Processing speed: the time consumption of a single frame is reduced from 120 ms to 18 ms, with an efficiency improvement of 85%, meeting the real-time requirement;

[0069] Registration accuracy: the translation error is optimized from ±3.5 pixels to ±0.8 pixels, reaching sub-pixel level alignment;

[0070] Adaptability: dynamic scene support is added, and the robustness to light changes is excellent.

[0071] Data verifies the comprehensive advantages of the fusion strategy in speed, accuracy and scene coverage.

[0072] The core idea of the equation is to search for the optimal translation parameter within the preset displacement range by maximizing the normalized cross-correlation value (NCC) between the visible light and ultraviolet images In the ultraviolet-visible light registration of static scenes, the displacement vector solved will realize the pixel-level coarse alignment of the two images, as shown in Figure 3 , laying a foundation for subsequent fine registration. The advantage of this method is its robustness to light changes and its independence on feature point extraction, making it particularly suitable for ultraviolet image matching scenarios with few features.

[0073] Step S2: focal length adaptive scaling;

[0074] In the process of ultraviolet-visible light image registration, when the optical system zooms, Figure 3 the scaling ratio of ultraviolet and visible light images changes with the focal length. The traditional fixed scaling factor method will cause:

[0075] 1) center point alignment but boundary offset: the edge region of the ultraviolet image cannot be matched with the visible light image;

[0076] 2) nonlinear distortion: during zooming, the scaling ratio and focal length are not a simple linear relationship;

[0077] Therefore, based on experimental data, the present application proposes a focal length adaptive dynamic scaling method to achieve accurate matching through the following steps:

[0078] Step S2.1: scaling parameter calculation;

[0079] According to the experimental calibration data, the linear relationship between focal length x and scaling coefficients a and b is established:

[0080]

[0081] where x represents the focal length (or magnification) of the current optical system, a represents the scaling coefficient in the x direction (horizontal scaling ratio), and b represents the scaling coefficient in the y direction (vertical scaling ratio).

[0082] Step S2.2: Dynamic cropping window calculation;

[0083] According to the scaling coefficients, the effective matching area of the ultraviolet image is determined:

[0084]

[0085] where (x1, y1) represents the top-left corner coordinates of the cropping window, (x2, y2) represents the bottom-right corner coordinates of the cropping window, and (360, 288) represents the center point coordinates of the ultraviolet image (assuming the original image size is 720x576).

[0086] This algorithm realizes accurate center alignment and boundary linear fitting through experimental linear relationship, effectively solving the boundary offset and nonlinear distortion problems caused by traditional fixed scaling methods, and can adapt to matching requirements under different focal lengths.

[0087] Step S3: Image scaling

[0088] Using the improved bilinear interpolation algorithm, the cropped ultraviolet image effective registration area is scaled to the target size (such as 720x576):

[0089]

[0090] where dst represents the scaled target image, represents the pixel value of the target image at coordinates (x, y), src represents the original ultraviolet image effective registration area, 、 、 、 respectively represent the four nearest pixel points in the target image of the pixel points in the ultraviolet image effective registration area, u and v are the interpolation weights in the horizontal and vertical directions (0 ≤ u, v ≤ 1), respectively, which represent the horizontal distance proportion of the target point to and , and and are the integer coordinates of the reference pixels in the source image (calculated by ).

[0091] In the embodiment of the present application, the ultraviolet radiation characteristics of the electric discharge of the power equipment are simulated by the ethanol combustion flame, and a standardized test environment is constructed. Experimental data show that the ultraviolet radiation intensity of the ethanol flame generated by the alcohol lamp in the 300-400 nm band has high consistency with the ultraviolet spectral characteristics of typical corona discharge, and the relative error is controlled within ±5%, which can be used as a high-fidelity physical simulation source. Figure 1 As shown in the image processing algorithm of the present application, the dynamic response characteristics of the simulated flame in the ultraviolet imaging system can realize synchronous displacement tracking with the visible light image, providing a reliable experimental basis for subsequent registration algorithm verification.

[0092] Table 2 Performance comparison of focal length adaptive zooming algorithm

[0093] Indicator Traditional fixed ratio method Invention dynamic scaling algorithm Lifting amplitude Edge registration error (pixels) 4.8±2.1 0.9±0.3 81% Zoom response delay (ms) 45 12 73% Nonlinear distortion coefficient 0.18 0.03 83% Computing resource occupation (MIPS) 320 150 53% Dynamic zoom support Not support Support -

[0094] As shown in Table 2, the dynamic zooming algorithm of the present application significantly improves the traditional method: the edge registration error is reduced by 81% to 0.9 pixels, the response delay is shortened by 73% to 12 ms, the nonlinear distortion is improved by 83% to 0.03, the computing resources are reduced by 53%, and the dynamic zooming support capability is newly added.

[0095] As shown in Figure 4 , the present application dynamically adjusts the scaling coefficient by the focal length parameter x, uses a dynamic clipping window and an optimized bilinear interpolation algorithm, realizes edge linear fitting while ensuring accurate alignment of the center point, significantly improves the computing efficiency with integer operation, effectively solves the nonlinear matching problem of ultraviolet-visible light images in the zooming process, and has the comprehensive advantages of strong adaptability, high precision and good real-time performance.

[0096] A discharge detection method for adaptive registration of ultraviolet and visible light images, acquires ultraviolet images and visible light images of a power equipment discharge point region, and registers the ultraviolet and visible light images of the power equipment discharge point region according to the adaptive registration method of ultraviolet and visible light images, for discharge detection.

[0097] As shown in Figure 5 , a discharge detection system for adaptive registration of ultraviolet and visible light images, comprising an imaging module, a registration module and a discharge detection module; the imaging module comprises an ultraviolet imaging module and a visible light imaging module, which are respectively used to capture ultraviolet images and visible light images of a power equipment discharge point region; the registration module registers the ultraviolet and visible light images of the power equipment discharge point region according to the adaptive registration method of ultraviolet and visible light images; the discharge detection module detects the discharge of the discharge point region according to the registered images.

[0098] The above examples are only used to illustrate the technical solutions of the present application, and are not intended to limit the present application; although the present application has been described in detail with reference to the foregoing examples, those skilled in the art should understand that the technical solutions recorded in the foregoing examples can be modified, or some or all of the technical features can be replaced by equivalents; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for adaptive registration of ultraviolet and visible light images, characterized by The method comprises the following steps: Step S1: initial registration of the ultraviolet image and the visible light image; For a moving scene, motion vectors of the visible light image sequence are calculated by optical flow motion estimation and mapped to the ultraviolet image for displacement compensation; for a static scene, histogram matching is used to determine translation compensation; The optical flow motion estimation obtains an optimal displacement amount by minimizing the pixel brightness difference in a local window between adjacent frames of the visible light image, and the optimal displacement amount is used for alignment compensation of the ultraviolet image; specifically, a window region is set, pixel intensity values of the visible light image at the current time and spatial position in the window region are calculated, and pixel intensity values of the ultraviolet image at the corresponding spatial position after displacement compensation at the next time are calculated, and the square of the pixel intensity difference between the current time and the next time after displacement compensation is calculated, and the sum of the squares of all pixel intensity differences in the window region is calculated to obtain the ultraviolet image displacement compensation at which the sum of the squares of the pixel intensity difference is the smallest; Step S2: focal length adaptive scaling; A scaling ratio is calculated according to the focal length, and the ultraviolet image is scaled according to the scaling ratio and the center point of the ultraviolet image; Step S3: image scaling; The cropped effective registration region of the ultraviolet image is scaled to a target size.

2. The method for self-adapting registration of UV and visible images according to claim 1, characterized in that: In the step S1, the histogram matching calculates the normalized cross-correlation value of the visible light image and the corresponding ultraviolet light image, and uses the offset corresponding to the maximum normalized cross-correlation value as the alignment parameter to perform translation and alignment of the ultraviolet image with the visible light image.

3. The method for self-adapting registration of UV and visible images according to claim 2, characterized in that: The normalized cross-correlation value is obtained by multiplying the pixel value of the visible light image with the pixel value of the ultraviolet image at the corresponding position after the offset, and summing all products in the overlapping region of the visible light image and the ultraviolet image as the numerator. In the overlapping region of the images, the sum of the squares of all visible light image pixel values and the sum of the squares of all ultraviolet image pixel values at the corresponding position after the offset are calculated, and then the product of the sum of the squares of all visible light image pixel values and the sum of the squares of all ultraviolet image pixel values at the corresponding position after the offset is taken to obtain the denominator.

4. The method of claim 1, wherein: In the step S2, the linear relationship between the focal length and the horizontal and / or vertical scaling coefficient is calibrated by experiment, and the position of the cropping window is dynamically adjusted according to the scaling coefficient and the center point coordinates of the matching region of the ultraviolet image to obtain the effective matching region of the ultraviolet image.

5. The method of claim 4, wherein: In the step S3, the bilinear interpolation algorithm is used to allocate weights based on the nearest integer coordinates of the pixel points in the effective matching region of the ultraviolet image in the horizontal and vertical directions to adjust the pixel values of the pixel points, so that the effective matching region of the ultraviolet image is scaled to a target size to obtain a target image.

6. The method of claim 5, wherein: The bilinear interpolation algorithm calculates the nearest reference pixel integer coordinates of the pixel points in the effective matching region of the ultraviolet image in the horizontal and / or vertical directions according to the coordinates of the pixel points, and obtains the difference weight in the horizontal and / or vertical direction based on the distance ratio of the coordinates of the pixel points and the reference pixel integer coordinates in the horizontal and / or vertical direction, and obtains the pixel value of the coordinates of the pixel points based on the sum of the products of the reference pixel integer coordinates and the corresponding difference weight to generate the target image after scaling of the effective matching region of the ultraviolet image.

7. A method for detecting electrical discharge by adaptive registration of ultraviolet and visible light images, characterized by: The UV image and the visible light image of the discharge area of the power equipment are acquired, and the UV image and the visible light image of the discharge area of the power equipment are registered according to the adaptive registration method for the UV image and the visible light image.

8. A discharge detection system for adaptive registration of UV and visible light images, comprising an imaging module, a registration module and a discharge detection module, characterized in that: The imaging module comprises a UV imaging module and a visible light imaging module, and is respectively used for capturing the UV image and the visible light image of the discharge area of the power equipment; The registration module is used for registering the UV image and the visible light image of the discharge area of the power equipment according to the adaptive registration method for the UV image and the visible light image; The discharge detection module is used for performing discharge detection on the discharge area according to the registered images.

Images