Laplace-usm msr cr transformer oil under image enhancement method

Abstract

The application discloses a Laplace-MSRCR transformer oil under image enhancement method based on Laplace-MSRCR, which comprises the following steps: decomposing an original transformer oil under image into R, G and B three-channel image components; obtaining reflection components of the three-channel images under different scales; sharpening the reflection components of the three-channel images under large and small scales based on a Laplace algorithm; enhancing the reflection components of the three-channel images under medium scales based on a USM algorithm; accumulating the reflection components of the three-channel images under different scales after sharpening; and correcting colors by using a color recovery factor. The application can effectively solve the problem of detail weakening of the MSRCR algorithm, and can effectively enhance the transformer oil under image.

Description

Technical Field

[0001] This invention belongs to the field of image processing technology, and relates to image processing under transformer oil, specifically to an image enhancement method for transformer oil based on Laplace-USM MSRCR. Background Technology

[0002] Oil-immersed transformers in nuclear power plants are crucial substation devices in nuclear power systems, and their operating status directly impacts the stability of the power system. Regular transformer fault inspections can promptly identify damaged internal components, providing guidance for troubleshooting and transformer lifespan assessment, and ensuring the safe operation of the nuclear power plant.

[0003] Due to the unique environment inside transformers and the deterioration and discoloration of transformer oil, the obtained images of the internal structure of transformers obtained under the oil often suffer from low contrast, color distortion, loss of detail, and uneven brightness. Clear, high-contrast images are essential for achieving precise positioning, fault detection, and fault location in transformer internal inspection robots. Therefore, transformer oil image enhancement technology is crucial for the intelligent operation of robots.

[0004] In the field of fluid image enhancement, the only environment similar to the oil-bearing environment is the underwater environment. Underwater images also suffer from problems such as low contrast, color distortion, lack of detail, and short visibility distance. In traditional methods, underwater image enhancement based on physical attenuation models models the attenuation process of underwater images, estimates the model parameters, and finally inverts to obtain a clear underwater image. As early as 2009, He et al. proposed an algorithm suitable for dehazing enhancement processing, based on the dark channel prior algorithm (DCP) of the Jaffe-McGlamey imaging model, which can accurately estimate the transmission map and background light map (He K, Jian S, Fellow, et al. Single Image Haze Removal Using Dark Channel Prior[J].IEEE Transactions on Pattern Analysis & Machine Intelligence, 2011, 33(12): 2341-2353). Later, when applied to underwater environments, Drews et al. proposed the Underwater Dark Channel Prior Algorithm (UDCP) based on the DCP algorithm (Drews JP, Nascimento E, Moraes F, et al. Transmission Estimation in Underwater Single Images [C] / / IEEE International Conference on Computer Vision Workshops.IEEE, 2013). There is also the Underwater Light Attenuation Prior Algorithm (ULAP) proposed by Song et al. (Song W, Wang Y, Huang D, et al. A rapid scene depthestimation model based on underwater light attenuation prior for underwater image restoration [C] / / Advances in Multimedia Information Processing-PCM 2018:19th Pacific-Rim Conference on Multimedia, Hefei, China, September 21-22, 2018, Proceedings, Part I 19. Springer International Publishing, 2018:678-688). These algorithms can accurately estimate TM (transmittance) and BL (background light) to obtain clear underwater images, but they are limited when prior knowledge is unreliable.In addition, underwater optical imaging methods, such as polarization imaging, are limited by the difficulty of obtaining images with the same polarization in water.

[0005] Furthermore, underwater image quality issues are mostly caused by light attenuation and scattering from particles in the water; while underwater transformer oil image problems are more caused by underwater light attenuation and specular or diffuse reflection from metal structures. Therefore, physical attenuation models based on water cannot be directly applied to underwater environments, making physical attenuation modeling methods unsuitable and difficult to implement. Traditional non-physical model-based underwater image enhancement, on the other hand, processes images at the pixel level without requiring modeling or estimation. It directly and subjectively adjusts pixel values ​​to remove blur, enhance features, improve contrast, and correct colors, thereby improving image quality. This approach is more easily adapted for enhancing images in underwater transformer oil environments.

[0006] Currently, there is limited research on image enhancement algorithms for transformer oil submersion. Therefore, there is an urgent need to develop a method for image enhancement of transformer oil submersion to provide high-quality image support for transformer fault diagnosis and transformer life assessment, thereby ensuring the safe operation of nuclear power plants. Summary of the Invention

[0007] The purpose of this invention is to address the problem that existing technologies cannot effectively enhance underwater images by providing a method for enhancing underwater transformer oil images based on the Laplace-USM MSRCR, which processes low-quality underwater transformer oil images into high-quality images.

[0008] To achieve the above objectives, the present invention employs the following technical solutions.

[0009] The present invention provides a transformer oil submersible image enhancement method based on Laplace-USM MSRCR, which includes the following steps:

[0010] S1 acquires the original transformer oil submersible image and decomposes the original transformer oil submersible image into three-channel image components: R, G, and B.

[0011] S2 converts the R, G, and B image components into the product of the incident and reflected components, respectively, and performs a logarithmic transformation.

[0012] S3 estimates the reflection components of the R, G, and B channels of the image at three different given scales (large, medium, and small) based on the R, G, and B channel image components processed in step S2.

[0013] S4 uses the Laplace algorithm to sharpen the reflection components of the R, G, and B three-channel images at both large and small scales.

[0014] S5 uses the USM algorithm to sharpen the reflection components of the R, G, and B three-channel images at a mesoscale.

[0015] S6 accumulates and calculates the reflection components at different scales after sharpening obtained in steps S4 and S5.

[0016] S7 introduces a color restoration factor to correct the colors of each channel, and then merges the color-corrected channels to obtain an enhanced image.

[0017] Optionally, in step S2, the R, G, and B image components are converted into the product of the incident and reflected components, respectively, and then logarithmically transformed, as follows:

[0018] LogI R (x,y)=LogR R (x,y)+LogL R (x,y)

[0019] LogI G (x,y)=LogR G (x,y)+LogL G (x,y)

[0020] LogI B (x,y)=LogR B (x,y)+LogL B (x,y)

[0021] Among them, R R (x,y) represents the reflection component of the R-channel image, and L... R (x,y) represents the incident components of the R-channel image, R G (x,y) represents the reflection component of the G-channel image, L G (x,y) represents the incident components of the G-channel image, R B (x,y) represents the reflection component of the B-channel image, L B (x,y) represents the incident components of the B-channel image.

[0022] Optionally, step S3 specifically includes:

[0023] S31 constructs Gaussian wrapping functions at different scales based on three scale parameters;

[0024] S32 performs convolution calculations on the R, G, and B channel image components processed in step S31 with Gaussian wrapping functions at different scales to obtain estimated values ​​of the incident components of the R, G, and B channel images at different scales.

[0025] S33 calculates the reflection components of the R, G, and B channel images at different scales based on the incident components of the R, G, and B channel images estimated in step S32, specifically as follows:

[0026] LogR Ri (x,y)=LogI R (x,y)-Log[G i (x,y)*I R (x,y)]

[0027] LogR Gi (x,y)=LogI G (x,y)-Log[G i (x,y)*I G (x,y)]

[0028] LogR Bi (x,y)=LogI B (x,y)-Log[G i (x,y)*I B (x,y)]

[0029] Among them, R Ri (x,y) represent the reflectance components of the R-channel image at different scales, I R (x,y) represents the R-channel image components, G i (x,y) represents the Gaussian wrapping function at different scales, where i is the scale index and R0 is the Gaussian wrapping function at different scales. Gi (x,y) represent the reflection components of the G-channel image at different scales, I G (x,y) represents the G-channel image components, R Bi (x,y) represent the reflection components of the B-channel image at different scales, I B (x,y) represents the B-channel image component.

[0030] In step S31 above, specifically, the Gaussian wrap function used in this invention is: Where δ is the scaling factor, which is set in advance; for example, 2, 102, 202. The original image is filtered by setting three Gaussian filter kernels with different scales.

[0031] This invention utilizes three scale parameters to construct three Gaussian wrapping functions, and then convolves the R, G, and B channel image components with the three Gaussian wrapping functions respectively to obtain the estimated values ​​of the incident components of the R, G, and B channel images at different scales, i.e., G... i (x,y)*I R (x,y), G i (x,y)*I G (x,y), G i (x,y)*I B (x,y), i=1,2,3.

[0032] Specifically, in step S33 above, calculating the reflection components of the R, G, and B channel images at different scales based on the incident components of the R, G, and B channel images estimated in step S32 is as follows:

[0033] LogR Ri (x,y)=LogI R (x,y)-Log[G i (x,y)*I R (x,y)]

[0034] LogR Gi (x,y)=LogI G (x,y)-Log[G i (x,y)*I G (x,y)]

[0035] LogR Bi (x,y)=LogI B (x,y)-Log[G i (x,y)*I B (x,y)]

[0036] Among them, R Ri (x,y) represent the reflectance components of the R-channel image at different scales, I R (x,y) represents the R-channel image components, G i (x,y) represents the Gaussian wrapping function at different scales, where i is the scale index and R0 is the Gaussian wrapping function at different scales. Gi (x,y) represent the reflection components of the G-channel image at different scales, I G (x,y) represents the G-channel image components, R Bi (x,y) represent the reflection components of the B-channel image at different scales, I B (x,y) represents the B-channel image component.

[0037] Thus, the reflection components of the R, G, and B three-channel images at three scales can be obtained as [R R1 ,R R2 ,R R3 ]、[R G1 ,R G2 ,R G3 ]、[R B1 ,R B2 ,R B3 ].

[0038] In step S4 above, the Laplace algorithm, a sharpening mask algorithm, is used to sharpen the reflection components of the R, G, and B channels at both large and small scales to enhance image edge details; the reflection component after Laplace algorithm sharpening is [R′ R1 ,RR ′3]、[R′ G1 ,R′ G3 ]、[R′ B1 ,R′ B3 The Laplace algorithm used in this invention is a conventional method disclosed in the field. For reference, see Chen Siji, Wang Xiaohong, Li Yunchuan. Research on improved Laplace algorithm for edge detection of UAV images [J]. Surveying and Mapping Engineering, 2021, 30(02):36-44. DOI:10.19349 / j.cnki.issn1006-7949.2021.02.007.

[0039] In step S5 above, the reflection components of the R, G, and B channels of the image at a medium scale are sharpened using the Unsharp Mask (USM) algorithm to enhance the high-frequency components of the image; the reflection component after USM sharpening is R′. R2 、R′ G2 、R′ B2 The USM algorithm used in this invention is a conventional method disclosed in the field. For reference, see Fu Qingqing, Jing Chunlei, Pei Yanliang, et al. Research on underwater image detail enhancement algorithm based on unsharpened mask guided filtering [J]. Acta Oceanologica Sinica, 2020, 42(07):130-138.

[0040] In step S6 above, the reflection components at different scales after sharpening obtained in steps S4 and S5 are summed to calculate the reconstructed reflection components of each channel, i.e.:

[0041]

[0042]

[0043]

[0044] In step S7 above, a color restoration factor is introduced to correct the reconstructed reflection components of each channel, specifically as follows:

[0045]

[0046] Among them, C j This represents the color reproduction coefficient of the j-th channel, j = R, G, B; C j It is expressed as follows:

[0047]

[0048] Here, α represents the adjustment coefficient and β represents the gain coefficient, both of which can be set in advance.

[0049] The reconstructed reflection components of each channel are then merged to obtain the final enhanced image.

[0050] Compared with existing technologies, the underwater image enhancement method based on dynamic color channel compensation provided by this invention has the following beneficial effects:

[0051] This invention first obtains the reflection components of three-channel images at different scales, then sharpens the reflection components of the three-channel images at large and small scales based on the Laplace algorithm, and sharpens the reflection components of the three-channel images at medium scale based on the USM algorithm. Then, the sharpened reflection components of the three channels at different scales are accumulated, and finally, a color restoration factor is used to correct the color of each channel. This can effectively solve the detail weakening problem of the MSRCR algorithm and effectively enhance the image under transformer oil. Attached Figure Description

[0052] Figure 1 This is a schematic diagram of the process of the transformer oil under-oil image enhancement method based on Laplace-USM MSRCR provided by the present invention;

[0053] Figure 2 The images shown are transformer sub-oil images obtained by different methods in the embodiments of the present invention; wherein, (a) is the original transformer sub-oil image, (b) is the transformer sub-oil image processed by the traditional MSRCR algorithm, and (c) is the transformer sub-oil image processed by the Laplace-USM MSRCR transformer sub-oil image enhancement method of the present invention.

[0054] Figure 3 These are magnified images of transformer submerged images obtained by different methods in the embodiments of the present invention; wherein, (a) is the original transformer submerged image, (b) is a magnified image of a specified area in (a), (c) is a magnified image of the specified area in (a) after processing by the traditional MSRCR algorithm, and (d) is a magnified image of the specified area in (a) after processing by the transformer submerged image enhancement method based on Laplace-USM MSRCR of the present invention.

[0055] Specific implementation method

[0056] The specific embodiments of the present invention are described below to enable those skilled in the art to understand the present invention. However, it should be understood that the present invention is not limited to the scope of the specific embodiments. For those skilled in the art, various changes are obvious as long as they are within the spirit and scope of the present invention as defined and determined by the appended claims. All inventions utilizing the concept of the present invention are protected.

[0057] This invention proposes an improved MSRCR algorithm based on the traditional Retinex algorithm. Retinex theory states that the image I(x,y) observed by a human is determined by the reflection component R(x,y) and the incident component L(x,y) of the object, as shown in expression (1):

[0058] I(x,y)=R(x,y)×L(x,y) (1)

[0059] Where (x,y) represents the coordinates of the corresponding image pixel position.

[0060] The SSR algorithm first takes the logarithm of both sides of formula (1) to obtain:

[0061] LogI(x,y)=LogR(x,y)+LogL(x,y) (2)

[0062] Then, rearranging the terms of formula (2) yields:

[0063] LogR(x,y)=LogI(x,y)-LogL(x,y) (3)

[0064] Finally, the incident component is estimated using the center surround function and the observed image, and then the reflection component of the image is obtained:

[0065] LogR SSR (x,y)=LogI(x,y)-Log[G(x,y)*I(x,y)] (4)

[0066] The MSR algorithm adds multiple scales to the SSR algorithm:

[0067]

[0068] In the formula, n is the number of selected scales, and A i These are the scale coefficients.

[0069] The MSRCR algorithm, based on the MSR algorithm, introduces a color restoration factor to ensure image color accuracy. The aforementioned MSRCR algorithm uses logarithmic processing to compress the display range of bright areas, leading to weakened details. To address this issue, this invention enhances the image by sharpening the reflection estimation components at different scales, thereby improving image detail and overall image quality.

[0070] Example 1

[0071] like Figure 1 As shown, the transformer oil submersible image enhancement method based on Laplace-USM MSRCR provided in this embodiment includes the following steps:

[0072] S1 acquires the original image under the transformer oil and decomposes it into three-channel image components: R, G, and B.

[0073] The original transformer oil submersion image obtained in this embodiment is as follows: Figure 2 As shown in (a), the acquired raw transformer oil subsurface image is decomposed into three channels: R, G, and B, to obtain a single-channel image of the three channels, i.e., I. R (x,y), I G (x,y), I B (x,y).

[0074] S2 converts the R, G, and B image components into the product of the incident and reflected components, respectively, and then performs a logarithmic transformation.

[0075] In this embodiment, the R, G, and B image components are converted into the product of the incident and reflected components, respectively, and then logarithmically transformed, as follows:

[0076] LogI R (x,y)=LogR R (x,y)+LogL R (x,y)

[0077] LogI G (x,y)=LogR G (x,y)+LogL G (x,y)

[0078] LogI B (x,y)=LogR B (x,y)+LogL B (x,y)

[0079] Among them, R R (x,y) represents the reflection component of the R-channel image, and L... R (x,y) represents the incident components of the R-channel image, R G (x,y) represents the reflection component of the G-channel image, L G (x,y) represents the incident components of the G-channel image, R B (x,y) represents the reflection component of the B-channel image, L B (x,y) represents the incident components of the B-channel image.

[0080] S3 estimates the reflectance components of the R, G, and B channel images at three different given scales (large, medium, and small) based on the R, G, and B channel image components processed in step S2; specifically, this includes:

[0081] S31 constructs Gaussian wrapping functions at different scales based on three scale parameters;

[0082] Specifically, the Gaussian wrap function used in this embodiment is: Where δ is the scaling factor, which is set in advance. In this embodiment, δ = 2, 102, 202.

[0083] S32 performs convolution calculations on the R, G, and B channel image components processed in step S31 with Gaussian wrapping functions at different scales to obtain estimated values ​​of the incident components of the R, G, and B channel images at different scales.

[0084] This invention utilizes three scale parameters to construct three Gaussian wrapping functions, and then convolves the R, G, and B channel image components with the three Gaussian wrapping functions respectively to obtain the estimated values ​​of the incident components of the R, G, and B channel images at different scales, i.e., G... i (x,y)*I R (x,y), G i (x,y)*I G (x,y), G i (x,y)*I B (x,y), i=1,2,3.

[0085] S33 calculates the reflection components of the R, G, and B three-channel images at different scales based on the incident components of the R, G, and B three-channel images estimated in step S32.

[0086] Specifically, the calculation of the reflection components of the R, G, and B channel images at different scales based on the incident components estimated in step S32 is as follows:

[0087] LogR Ri (x,y)=LogI R (x,y)-Log[G i (x,y)*I R (x,y)]

[0088] LogR Gi (x,y)=LogI G (x,y)-Log[G i (x,y)*I G (x,y)]

[0089] LogR Bi (x,y)=LogI B (x,y)-Log[G i (x,y)*I B (x,y)]

[0090] Among them, R Ri (x,y) represent the reflectance components of the R-channel image at different scales, IR (x,y) represents the R-channel image components, G i (x,y) represents the Gaussian wrapping function at different scales, where i is the scale index and R0 is the Gaussian wrapping function at different scales. Gi (x,y) represent the reflection components of the G-channel image at different scales, I G (x,y) represents the G-channel image components, R Bi (x,y) represent the reflection components of the B-channel image at different scales, I B (x,y) represents the B-channel image component.

[0091] Thus, the reflection components of the R, G, and B three-channel images at three scales can be obtained as [R R1 ,R R2 ,R R3 ]、[R G1 ,R G2 ,R G3 ]、[R B1 ,R B2 ,R B3 ].

[0092] S4 uses the Laplace algorithm to sharpen the reflection components of the R, G, and B three-channel images at both large and small scales.

[0093] This embodiment utilizes the Laplace algorithm, a sharpening mask algorithm, to sharpen the reflection components of the R, G, and B channels of the image at both large and small scales, thereby enhancing image edge details; the reflection component after Laplace algorithm sharpening is [R′ R1 ,R R ′3]、[R′ G1 ,R′ G3 ]、[R′ B1 ,R′ B3 ].

[0094] The Laplace algorithm used in this embodiment is a conventional method disclosed in the art. For reference, see Chen Siji, Wang Xiaohong, Li Yunchuan. Research on improved Laplace algorithm for edge detection of UAV images [J]. Surveying and Mapping Engineering, 2021, 30(02):36-44. DOI:10.19349 / j.cnki.issn1006-7949.2021.02.007.

[0095] S5 uses the USM algorithm to sharpen the reflection components of the R, G, and B three-channel images at a medium scale.

[0096] This embodiment utilizes the Unsharp Mask (USM) algorithm to sharpen the reflection components of the R, G, and B channels at a mesoscale to enhance the high-frequency components of the image; the reflection component after USM sharpening is R′. R2 、R′ G2 、R′ B2 .

[0097] The USM algorithm used in this embodiment is a conventional method disclosed in the field. For reference, see Fu Qingqing, Jing Chunlei, Pei Yanliang, et al. Research on underwater image detail enhancement algorithm based on unsharpened mask guided filtering [J]. Acta Oceanologica Sinica, 2020, 42(07):130-138.

[0098] S6 accumulates and calculates the reflection components at different scales after sharpening obtained in steps S4 and S5.

[0099] In this embodiment, the reflection components at different scales after sharpening obtained in steps S4 and S5 are accumulated to calculate the reconstructed reflection components of each channel, namely:

[0100]

[0101]

[0102]

[0103] S7 introduces a color restoration factor to correct the colors of each channel, and then merges the color-corrected channels to obtain an enhanced image.

[0104] This embodiment introduces a color restoration factor to correct the reconstructed reflection components of each channel in order to ensure the color accuracy of the image. Specifically:

[0105]

[0106] Among them, C j This represents the color reproduction coefficient of the j-th channel, j = R, G, B; C j It is expressed as follows:

[0107]

[0108] Where α represents the adjustment coefficient and β represents the gain coefficient; in this embodiment, α = 128 and β = 1.

[0109] The reconstructed reflection components of each channel are then merged to obtain the final enhanced image.

[0110] Figure 2(b) shows the transformer submersible image 2(a) after processing with the traditional MSRCR algorithm (see Wei, Chen, et al. "Deep retinex decomposition for low-light enhancement." arXiv preprint arXiv:1808.04560 (2018)). Figure 3 (c) is a magnified view of the specified area; Figure 2 (c) The original transformer submersible image was enhanced using a Laplace-USM MSRCR-based transformer submersible image enhancement method. Figure 2 (a) Processed image of the transformer oil. Figure 3 (d) is a magnified view of the specified area. The comparison in the figures shows that the image obtained by the Laplace-USM MSRCR transformer oil under-oil image enhancement method provided by this invention can effectively improve the quality of transformer oil under-oil images and, to a certain extent, solve the detail distortion problem of the MSRCR algorithm. This provides a guarantee for subsequent transformer internal fault detection and feature point extraction.

[0111] To further verify the effectiveness of the proposed method, the Underwater Image Quality Evaluation Metric (UIQM) was used for evaluation. A higher UIQM value indicates better image color balance, sharpness, and contrast. Experimental results are shown in Table 1.

[0112] Table 1 UIQM values ​​for different methods

[0113]

[0114] As shown in Table 1, the image processed by the Laplace-USM MSRCR transformer oil image enhancement method provided by this invention has a UIQM value that is 113.17% higher than that of the original image and 18.3% higher than that of the MSRCR algorithm. This demonstrates the superiority of the Laplace-USM MSRCR transformer oil image enhancement method provided by this invention.

[0115] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0116] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0117] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0118] Specific embodiments have been used to illustrate the principles and implementation methods of this invention. The descriptions of the embodiments above are only for the purpose of helping to understand the method and core ideas of this invention. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this invention. Therefore, the content of this specification should not be construed as a limitation of this invention.

[0119] Those skilled in the art will recognize that the embodiments described herein are intended to help the reader understand the principles of the invention, and should be understood that the scope of protection of the invention is not limited to such specific statements and embodiments. Those skilled in the art can make various other specific modifications and combinations based on the technical teachings disclosed in this invention without departing from the spirit of the invention, and these modifications and combinations are still within the scope of protection of this invention.

Claims

1. A method for enhancing images of transformer oil underfill based on Laplace-USM MSRCR, characterized in that, Includes the following steps: S1 acquires the original transformer oil submersible image and decomposes the original transformer oil submersible image into three-channel image components: R, G, and B. S2 converts the R, G, and B image components into the product of the incident and reflected components, respectively, and performs a logarithmic transformation. S3 estimates the reflectance components of the R, G, and B channel images at three different given scales (large, medium, and small) based on the R, G, and B channel image components processed in step S2. Step S3 specifically includes the following sub-steps: S31 Constructs Gaussian wrap functions at different scales based on three scale parameters; S32 The R, G, and B channel image components processed in step S31 are convolved with Gaussian wrapping functions at different scales to obtain the incident component estimates of the R, G, and B channel images at different scales. S33 Calculates the reflection components of the R, G, and B channel images at different scales based on the incident components estimated in step S32, specifically: ； ； ； in, For the reflection components of the R-channel images at different scales, For the R channel image components, Let i represent the Gaussian wrapping function at different scales, where i is the scale index. For the reflection components of the G-channel image at different scales, For the G channel image components, For the reflection components of B-channel images at different scales, For B-channel image components; S4 uses the Laplace algorithm to sharpen the reflection components of the R, G, and B three-channel images at both large and small scales. S5 uses the USM algorithm to sharpen the reflection components of the R, G, and B three-channel images at a mesoscale. S6. The reflection components at different scales after sharpening obtained in steps S4 and S5 are accumulated and calculated to obtain the reconstructed reflection components of each channel. S7 introduces a color restoration factor to correct the colors of each channel, and then merges the color-corrected channels to obtain an enhanced image.

2. The method for enhancing images of transformer oil based on Laplace-USM MSRCR according to claim 1, characterized in that, In step S2, the R, G, and B image components are converted into the product of the incident and reflected components, respectively, and then logarithmically transformed, as follows: ； ； ； in, For the reflection component of the R channel image, For the incident component of the R channel image, For the reflection component of the G channel image, For the incident component of the G channel image, For the reflection component of the B channel image, This represents the incident component of the B-channel image.

3. The method for enhancing transformer oil images based on Laplace-USM MSRCR according to claim 1, characterized in that, The Gaussian wrap function used in step S31 is ,in This is the scaling factor.

4. The method for enhancing transformer oil images based on Laplace-USM MSRCR according to claim 3, characterized in that, Three Gaussian wrapping functions were constructed using three scale parameters. Then, the R, G, and B channel image components were convolved with each of the three Gaussian wrapping functions to obtain the estimated incident components of the R, G, and B channel images at different scales. , , , i=1,2,3.

5. The method for enhancing images of transformer oil based on Laplace-USM MSRCR according to claim 4, characterized in that, In step S33, specifically, the calculation of the reflection components of the R, G, and B channel images at different scales based on the incident components of the R, G, and B channel images estimated in step S32 is as follows: ； ； ； in, For the reflection components of the R-channel images at different scales, For the R channel image components, Let i represent the Gaussian wrapping function at different scales, where i is the scale index. For the reflection components of the G-channel image at different scales, For the G channel image components, For the reflection components of B-channel images at different scales, For B-channel image components; Thus, the reflectance components of the R, G, and B channel images at three scales can be obtained as follows: , , .

6. The method for enhancing images of transformer oil based on Laplace-USM MSRCR according to claim 1, characterized in that, In step S6, the reflection components at different scales after sharpening obtained in steps S4 and S5 are summed to calculate the reconstructed reflection components of each channel, i.e.: ； ； ； , , The reflection component after sharpening using the Laplace algorithm. , , This is the reflection component after sharpening using the USM algorithm.

7. The method for enhancing transformer oil images based on Laplace-USM MSRCR according to claim 6, characterized in that, In step S7, a color restoration factor is introduced to correct the reconstructed reflection components of each channel, specifically as follows: ； Among them, C j This represents the color reproduction coefficient of the j-th channel, where j = R, G, B; C j It is expressed as follows: ； Where α represents the adjustment coefficient and 𝛽 represents the gain coefficient; The reconstructed reflection components of each channel are then merged to obtain the final enhanced image.

Images