Transformer substation non-uniform exposure image enhancement method and system based on adaptive fusion

Abstract

The invention discloses a transformer substation non-uniform exposure image enhancement method and system based on adaptive fusion, and the method comprises the following steps: obtaining non-uniform exposure video frame images in a transformer substation monitoring video, and carrying out the successive preprocessing of the video frame images; processing the brightness channel graph V by adopting a two-branch parallel step; for the H and S parts of the color channel graph, the saturation of the pre-channel graph is enhanced on the basis of the hue difference; combining the enhanced images of the brightness channel and the color channel with an original channel image, respectively designing a self-adaptive channel weight method for fusion, and converting H, S and V channels after fusion into RGB enhanced images for result output; and establishing an evaluation mechanism to evaluate the quality of the enhanced image, and adjusting parameters of an enhancement algorithm according to an evaluation result. According to the invention, the problem of non-uniform image exposure under the complex illumination condition of the substation is effectively solved, and powerful support is provided for accurately monitoring the state of substation equipment.

Description

Technical Field

[0001] The present invention relates to the field of electric power image processing, and in particular to a method and system for enhancing non-uniform exposure images of a substation based on adaptive fusion. Background Art

[0002] With the rapid development of power systems, substations, as a crucial link in power transmission and distribution, have a direct impact on the safety and stability of the entire power grid. To ensure the proper operation of substation equipment, regular inspection and maintenance are crucial. Using image processing technology to monitor the substation's internal environment and equipment has become a highly effective method. However, in practice, the complex and variable lighting conditions within substations, such as the large difference in light intensity between indoors and outdoors and the inherent illumination of equipment, can lead to severe non-uniform exposure in captured images. This not only affects image quality but also poses significant challenges to subsequent image analysis and fault diagnosis.

[0003] While some image enhancement techniques can improve image quality to a certain extent, they often struggle with the non-uniform exposure issues found in substations. For example, traditional histogram equalization methods can cause images to appear overly bright or dark, while local enhancement algorithms can easily introduce noise, affecting image detail. Furthermore, while some deep learning-based methods excel in certain areas, their high computational complexity makes them inadequate for the real-time demands of substation monitoring. Summary of the Invention

[0004] Purpose of the invention: In order to solve the above-mentioned problems, the present invention provides an image enhancement method and system for non-uniform exposure of substations based on adaptive fusion. The present invention aims to achieve effective enhancement of images under complex lighting conditions of substations by fusing enhanced processing of adaptive images of different channels, improve image clarity and visual effects, and thereby improve the accuracy and reliability of substation environmental monitoring results.

[0005] The technical solution is an image enhancement method for substation non-uniform exposure based on adaptive fusion, which includes the following steps:

[0006] Step S1: Obtain non-uniformly exposed video frame images in the substation monitoring video, pre-process the video frame images one by one, and convert the video frame images from RGB format to HSV format to perform enhancement processing on two different components, namely brightness and color;

[0007] Step S2: Process the luminance channel map V in parallel using two branches: one branch uses an improved adaptive Laplace filtering method to enhance the edges and details of the video frame image; the other branch designs a target-based luminance response estimation function, distinguishes the exposure requirements of the foreground and background, calculates the mean of the image blocks, applies the local luminance response estimation function, and sets a luminance enhancement cutoff threshold to adaptively process the enhanced luminance;

[0008] Step S3: For the H and S parts of the color channel image, the saturation of the pre-channel image is enhanced based on the hue difference to adapt to the color enhancement direction in different environments;

[0009] Step S4: For the enhanced images of the brightness channel and the color channel, combined with the original channel images, a method of designing adaptive channel weights is used to fuse them, and the fused H, S, and V channels are converted into an RGB enhanced image for output;

[0010] Step S5: Establish an evaluation mechanism to evaluate the quality of the enhanced image, and adjust the parameters of the enhancement algorithm according to the evaluation results to continuously optimize the effect of image enhancement and form a closed-loop system.

[0011] According to another aspect of the present application, step S1 is specifically as follows:

[0012] Step S11: extract the video frame image of the non-uniform image from the substation monitoring video and perform pre-processing operation. The video frame image is I N (x,y), where (x,y) represents pixel coordinates and N represents the number of video frames;

[0013] Step S12: Convert the original RGB color format to the HSV color format, separate the color-related information, including hue H and saturation S, brightness information, and value V, and perform enhancement processing according to the component type.

[0014] According to another aspect of the present application, step S2 is specifically:

[0015] Step S21: Use a 3×3 matrix Laplacian filter kernel LK to calculate the second-order derivative of the image, thereby highlighting the sudden changes in the image, that is, the edge part:

[0016]

[0017] Where I′(x,y) represents the filtered image, and κ represents the enhancement coefficient, which is used to control the strength of edge enhancement;

[0018] Step S22: Adaptively adjust the image according to the illumination conditions of different regions in the video frame image by calculating the variance to adapt to the transformation of different regions. The specific method is as follows:

[0019] The size of the local area is n×n. The pixel point (x, y) is taken as the center position. The variance V(x, y) of the local area is calculated. Then, the relationship between the enhancement coefficient and the regional variance is expressed as:

[0020]

[0021] Among them, κ represents the enhancement coefficient, V represents the variance of the local area;

[0022] Step S23: Binarize the grayscale image of the V channel using the Otsu threshold method, determine an optimal threshold, divide the image into two parts, the target and the background, and design a target-based brightness response curve; for the pixels in the target area, calculate the brightness value distribution, and design a brightness response curve to adjust the brightness information of the target / background objects based on the brightness characteristics of the target object:

[0023] f(t)=exp[-(t-μ) 2 / 2σ 2 ]

[0024]

[0025] Among them, f(t) is the brightness response curve, (μ,σ 2 ) is the standard deviation of the response pixel area, f Otsu is the average brightness response of the partition block, m is the number of pixels, I Otsu is the brightness of the processed sub-image block;

[0026] Step S24: Exposure compensation is performed on the foreground object, and overexposure and underexposure information of the background area are appropriately suppressed to avoid the problem of over-enhancement of the entire image. Based on the relevant parameters of the foreground object and background area obtained, an enhancement formula is designed:

[0027]

[0028] Where I″ is the pixel enhancement output, z is the brightness value of the pixel, and z max is the maximum pixel brightness value, τ c is the adaptive enhancement factor;

[0029] To avoid over-enhancement, a truncation threshold is introduced as follows:

[0030]

[0031] According to another aspect of the present application, step S3 is specifically:

[0032] For the color channel map, the color saturation is enhanced according to the overall hue segmentation to meet the color requirements in different environments:

[0033]

[0034] Among them, I″′(x,y) is the saturation S channel output after color enhancement, f(H,S) is the designed saturation segment enhancement function, H h min,H h max is the minimum and maximum value of the hue range, s h is the saturation enhancement coefficient.

[0035] According to another aspect of the present application, step S4 is specifically:

[0036] Step S41: After processing the brightness and color enhanced channels separately, they are fused with the original channel images using a weighted combination method. Furthermore, a weight map is constructed using a quality metric method, and the weights are optimized using an improved Gaussian filter to obtain an exposure-enhanced fused RGB image.

[0037] Step S42: For the channel maps enhanced by Laplace brightness and brightness response curve, a weight map is constructed based on the quality metric with the original brightness channel. The influencing factors of image clarity, exposure, and contrast are considered, and the fused weight map is comprehensively calculated:

[0038] I lout (x,y)=w a ×I(x,y)+w b ×I′(x,y)+w c ×I″(x,y);

[0039]

[0040] Among them, I lout (x, y) is the V channel fusion brightness image, I (x, y) is the original V channel brightness image, k is the image to be processed, w a 、w b 、w c are the clarity, exposure and contrast weights, and α is a fixed constant;

[0041] Step S43: For the color channel image that uses the color matrix for brightness enhancement, fuse it with the original color channel according to the weight of the enhancement intensity:

[0042] I cout =d×I(x,y)+(1-d)×I″′(x,y)

[0043] Among them, I coutis the color enhancement fusion output image, c is the weight coefficient in the [0,1] interval;

[0044] Step S44: Convert the enhanced brightness V channel and color H and S channels into RGB format images for output.

[0045] According to another aspect of the present application, step S5 is specifically:

[0046] Step S51: Establish an image enhancement quality evaluation mechanism based on subjective and objective evaluation indicators: The subjective evaluation method involves professional image analysts or substation staff observing the enhanced image and scoring it based on aspects such as image clarity, color naturalness, prominence of edge details, and identifiability of target objects. The average of these scores is taken as the subjective evaluation result (MOS) of the enhanced image.

[0047]

[0048] Among them, goal(pic) is the score of different subjective evaluation factors;

[0049] Step S52: Objective evaluation method mainly uses the structural similarity index (SSIM) and peak signal-to-noise ratio (PSNR) to evaluate the reconstruction quality of the enhanced image in terms of brightness, structure, and corresponding pixel error.

[0050]

[0051] Among them, μ x 、μ y is the pixel average value, is the pixel standard deviation, σ xy is the pixel covariance, β1 and β2 are constant values, M and N are the length and width of the image, To enhance the output image.

[0052] According to another aspect of the present application, a substation non-uniform exposure image enhancement system based on adaptive fusion is also provided, comprising:

[0053] at least one processor; and

[0054] a memory communicatively connected to at least one of the processors; wherein,

[0055] The memory stores instructions that can be executed by the processor, and the instructions are used to be executed by the processor to implement the image enhancement method for substation non-uniform exposure based on adaptive fusion as described in any of the above technical solutions.

[0056] Beneficial effects: 1. By converting the image from RGB format to HSV format, the hue (H), saturation (S) and brightness (V) channels are enhanced respectively. The Laplace filter is used to enhance the image edge of the brightness channel, and a target-based brightness response estimation function is designed to adjust the local brightness. At the same time, a truncation threshold is set to avoid over-enhancement. The hue-based saturation enhancement method is used for the H and S channels to effectively solve the problem of non-uniform image exposure under complex lighting conditions in substations.

[0057] 2. This paper designs an adaptive weighted channel enhancement and a target-based local brightness response method, fully considering the interpretability of different enhancement methods. This adaptive channel fusion approach highlights key information in the image, improves the readability and usability of the image, and provides strong support for accurate monitoring of substation equipment status.

[0058] 3. This invention establishes a comprehensive evaluation mechanism to evaluate the quality of the enhanced image and adjust the parameters of the enhancement algorithm based on the evaluation results. This mechanism ensures that the image enhancement effect can be continuously optimized and continuously adapted to the complex and changing lighting conditions and environment of the substation. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] Figure 1 It is a logic flow chart of the present invention.

[0060] Figure 2 It is the algorithm structure diagram of the present invention. DETAILED DESCRIPTION

[0061] like Figure 1 As shown in FIG, the image enhancement algorithm for substation non-uniform exposure based on adaptive fusion includes the following steps:

[0062] Step S1: Obtain non-uniformly exposed video frame images in the substation monitoring video, pre-process the video frame images one by one, and convert the video frame images from RGB format to HSV format to perform enhancement processing on two different components, namely brightness and color;

[0063] Step S2: Process the luminance channel map V in parallel using two branches: one branch uses an improved adaptive Laplace filtering method to enhance the edges and details of the video frame image; the other branch designs a target-based luminance response estimation function, distinguishes the exposure requirements of the foreground and background, calculates the mean of the image blocks, applies the local luminance response estimation function, and sets a luminance enhancement cutoff threshold to adaptively process the enhanced luminance;

[0064] Step S3: For the H and S parts of the color channel image, the saturation of the pre-channel image is enhanced based on the hue difference to adapt to the color enhancement direction in different environments;

[0065] Step S4: For the enhanced images of the brightness channel and the color channel, combined with the original channel images, a method of designing adaptive channel weights is used to fuse them, and the fused H, S, and V channels are converted into an RGB enhanced image for output;

[0066] Step S5: Establish an evaluation mechanism to evaluate the quality of the enhanced image, and adjust the parameters of the enhancement algorithm according to the evaluation results to continuously optimize the effect of image enhancement and form a closed-loop system.

[0067] This proposal proposes an adaptive fusion-based image enhancement algorithm for substations with uneven exposure. Through multi-step processing and optimization, it achieves efficient enhancement of unevenly exposed images from substation monitoring videos. By converting images from RGB to HSV format, brightness and color can be processed independently, avoiding mutual interference between brightness and color and improving the accuracy of the enhancement effect. Adaptive Laplacian filtering effectively enhances image edges and details, improving image clarity and detail. A brightness response estimation function adaptively adjusts brightness enhancement based on the exposure requirements of the foreground and background, avoiding over-enhancement or under-enhancement and enhancing the image's visual quality. Hue and saturation enhancement processes adapt to color enhancement requirements in diverse environments, improving image color rendering and visual quality. Adaptive channel weight fusion dynamically adjusts channel weights based on the enhancement effects of different channels, ensuring the optimal visual quality of the fused image. The fused H, S, and V channels are converted into an enhanced RGB image for output, ensuring that the enhanced image can be directly applied in real-world scenarios. By establishing an evaluation mechanism, the quality of the enhanced image can be objectively evaluated to ensure the stability and reliability of the enhancement effect; the parameters of the enhancement algorithm are adjusted according to the evaluation results to form a closed-loop system that can continuously optimize the image enhancement effect and ensure the adaptability and robustness of the enhancement algorithm in different scenarios.

[0068] This solution efficiently enhances non-uniformly exposed images from substation monitoring videos through multi-channel processing, adaptive Laplacian filtering and luminance response estimation, color channel enhancement, adaptive channel weight fusion, an evaluation mechanism, and closed-loop optimization. Its advantages include separate luminance and color processing, edge and detail enhancement, adaptive luminance enhancement, color enhancement, adaptive fusion, RGB enhanced image output, quality assessment, and closed-loop optimization, ensuring optimal visual quality and stability of the enhanced images.

[0069] According to another aspect of the present application, step S1 is specifically as follows:

[0070] During routine monitoring of substations, video is captured using video acquisition devices. Due to the complex environment of substation monitoring and the diverse and changing lighting conditions, non-uniform exposure can easily occur in the video frames.

[0071] Step S11: extract the video frame image of the non-uniform image from the substation monitoring video and perform pre-processing operation. The video frame image is I N (x,y), where (x,y) represents pixel coordinates and N represents the number of video frames;

[0072] Step S12: Changes in image brightness will cause changes in color information. Directly performing exposure enhancement processing on the RGB non-uniform image is likely to lose the original color information. Therefore, the original RGB color format is converted to the HSV color format, and color-related information is separated, including: hue H and saturation S, brightness information, and lightness V, and enhancement processing is performed separately according to the component type.

[0073] According to another aspect of the present application, step S2 is specifically:

[0074] Step S21: The contrast between the edges of objects in the non-uniformly exposed image is usually weak, and direct exposure enhancement processing will result in further blurring. Therefore, to enhance the edges in the image, a 3×3 Laplacian filter kernel LK is used to calculate the second-order derivative of the image, thereby highlighting the sudden changes in the image, that is, the edges:

[0075]

[0076] Where I′(x,y) represents the filtered image, and κ represents the enhancement coefficient, which is used to control the strength of edge enhancement;

[0077] Step S22: Due to the complexity of lighting conditions, uneven lighting often occurs in different areas of the non-uniformly exposed safety monitoring images of substations. For areas with edges or rich details, a larger enhancement coefficient is required to highlight the features; while for smooth areas, a larger enhancement coefficient may cause excessive noise enhancement, affecting image quality. In summary, the traditional method of fixing the enhancement coefficient cannot effectively enhance the edge information of special areas, and may even cause deterioration of image quality. Therefore, according to the lighting conditions of different areas in the video frame image, adaptive adjustment is performed by calculating the variance to adapt to the transformation of different areas; the specific method is as follows:

[0078] The size of the local area is n×n. The pixel point (x, y) is taken as the center position. The variance V(x, y) of the local area is calculated. Then, the relationship between the enhancement coefficient and the regional variance is expressed as:

[0079]

[0080] Among them, κ represents the enhancement coefficient, V represents the variance of the local area;

[0081] Step S23: Binarize the grayscale image of the V channel using the Otsu threshold method, determine an optimal threshold, divide the image into two parts, the target and the background, and design a target-based brightness response curve; for the pixels in the target area, calculate the brightness value distribution, and design a brightness response curve to adjust the brightness information of the target / background objects based on the brightness characteristics of the target object:

[0082] f(t)=exp[-(t-μ) 2 / 2σ 2 ]

[0083]

[0084] Among them, f(t) is the brightness response curve, (μ,σ 2 ) is the standard deviation of the response pixel area, f Otsu is the average brightness response of the partition block, m is the number of pixels, I Otsu is the brightness of the processed sub-image block;

[0085] Step S24: Considering that the foreground target should obtain a more reasonable exposure compensation, and the overexposure and underexposure information of the background area can be appropriately suppressed to avoid the problem of excessive enhancement of the entire image, the enhancement formula is designed based on the relevant parameters of the foreground target and background area obtained:

[0086]

[0087] Where I″ is the pixel enhancement output, z is the brightness value of the pixel, and z max is the maximum pixel brightness value, τ c is the adaptive enhancement factor;

[0088] To avoid over-enhancement, a truncation threshold is introduced as follows:

[0089]

[0090] This solution efficiently enhances non-uniformly exposed images in substation monitoring videos. Its advantages include edge enhancement, controllable enhancement intensity, adaptive adjustment, local region processing, target and background separation, brightness response curve design, adaptive brightness adjustment, exposure compensation, over- and under-exposure suppression, adaptive enhancement factor, and truncation threshold, ensuring optimal visual quality and stability of the enhanced images.

[0091] According to another aspect of the present application, step S3 is specifically:

[0092] Color enhancement can highlight device details and outlines, enhance color contrast between different areas of the device for easier differentiation, and help detect abnormal color changes in the device to detect potential faults, thereby improving image visual effects and providing a key reference for fault detection and diagnosis. Therefore, for the color channel map, the color saturation is enhanced according to the overall hue segmentation to meet the color requirements of different environments:

[0093]

[0094] Among them, I″′(x,y) is the saturation S channel output after color enhancement, f(H,S) is the designed saturation segment enhancement function, H h min,H h max is the minimum and maximum value of the hue range, s h is the saturation enhancement coefficient.

[0095] According to another aspect of the present application, step S4 is specifically:

[0096] Step S41: After processing the brightness and color enhanced channels separately, they are fused with the original channel images using a weighted combination method. Furthermore, a weight map is constructed using a quality metric method, and the weights are optimized using an improved Gaussian filter to obtain an exposure-enhanced fused RGB image.

[0097] Step S42: For the channel maps enhanced by Laplace brightness and brightness response curve, a weight map is constructed based on the quality metric with the original brightness channel. The influencing factors of image clarity, exposure, and contrast are considered, and the fused weight map is comprehensively calculated:

[0098] I lout (x,y)=w a ×I(x,y)+w b ×I′(x,y)+w c ×I″(x,y);

[0099]

[0100] Among them, I lout (x, y) is the V channel fusion brightness image, I (x, y) is the original V channel brightness image, k is the image to be processed, w a 、w b 、w c are the clarity, exposure and contrast weights, and α is a fixed constant;

[0101] Step S43: For the color channel image that uses the color matrix for brightness enhancement, fuse it with the original color channel according to the weight of the enhancement intensity:

[0102] Icout =d×I(x,y)+(1-d)×I″′(x,y)

[0103] Among them, I cout is the color enhancement fusion output image, c is the weight coefficient in the [0,1] interval;

[0104] Step S44: Convert the enhanced brightness V channel and color H and S channels into RGB format images for output.

[0105] According to another aspect of the present application, step S5 is specifically:

[0106] Step S51: Establish an image enhancement quality evaluation mechanism based on subjective and objective evaluation indicators: The subjective evaluation method involves professional image analysts or substation staff observing the enhanced image and scoring it based on aspects such as image clarity, color naturalness, prominence of edge details, and identifiability of target objects. The average of these scores is taken as the subjective evaluation result (MOS) of the enhanced image.

[0107]

[0108] Among them, goal(pic) is the score of different subjective evaluation factors;

[0109] Step S52: Objective evaluation method mainly uses the structural similarity index (SSIM) and peak signal-to-noise ratio (PSNR) to evaluate the reconstruction quality of the enhanced image in terms of brightness, structure, and corresponding pixel error.

[0110]

[0111] Among them, μ x 、μ y is the pixel average value, is the pixel standard deviation, σ xy is the pixel covariance, β1 and β2 are constant values, M and N are the length and width of the image, To enhance the output image.

[0112] According to another aspect of the present application, a substation non-uniform exposure image enhancement system based on adaptive fusion is also provided, comprising:

[0113] at least one processor; and

[0114] a memory communicatively connected to at least one of the processors; wherein,

[0115] The memory stores instructions that can be executed by the processor, and the instructions are used to be executed by the processor to implement the image enhancement method for substation non-uniform exposure based on adaptive fusion as described in any of the above technical solutions.

[0116] It should be noted that the various specific technical features described in the above specific embodiments can be combined in any appropriate manner without contradiction. To avoid unnecessary repetition, the present invention will not further describe various possible combinations.

Claims

1. An image enhancement method for substation non-uniform exposure based on adaptive fusion, characterized by: The steps include: Step S1: Obtain non-uniformly exposed video frame images in the substation monitoring video, pre-process the video frame images one by one, and convert the video frame images from RGB format to HSV format to perform enhancement processing on two different components, namely brightness and color; Step S2: Process the luminance channel map V in parallel using two branches: one branch uses an improved adaptive Laplace filtering method to enhance the edges and details of the video frame image; the other branch designs a target-based luminance response estimation function, distinguishes the exposure requirements of the foreground and background, calculates the mean of the image blocks, applies the local luminance response estimation function, and sets a luminance enhancement cutoff threshold to adaptively process the enhanced luminance; Step S3: For the H and S parts of the color channel image, the saturation of the pre-channel image is enhanced based on the hue difference to adapt to the color enhancement direction in different environments; Step S4: For the enhanced images of the brightness channel and the color channel, combined with the original channel images, a method of designing adaptive channel weights is used to fuse them, and the fused H, S, and V channels are converted into an RGB enhanced image for output; Step S5: Establish an evaluation mechanism to evaluate the quality of the enhanced image, and adjust the parameters of the enhancement algorithm according to the evaluation results to continuously optimize the effect of image enhancement and form a closed-loop system.

2. The image enhancement method for substation non-uniform exposure based on adaptive fusion according to claim 1, characterized in that: The step S1 is specifically as follows: Step S11: extract the video frame image of the non-uniform image from the substation monitoring video and perform pre-processing operation. The video frame image is I N (x,y), where (x,y) represents pixel coordinates and N represents the number of video frames; Step S12: Convert the original RGB color format to the HSV color format, separate the color-related information, including hue H and saturation S, brightness information, and value V, and perform enhancement processing according to the component type.

3. The image enhancement method for substation non-uniform exposure based on adaptive fusion according to claim 1, characterized in that: The step S2 is specifically as follows: Step S21: Use a 3×3 matrix Laplacian filter kernel LK to calculate the second-order derivative of the image, thereby highlighting the sudden changes in the image, that is, the edge part: Where I′(x,y) represents the filtered image, and κ represents the enhancement coefficient, which is used to control the strength of edge enhancement; Step S22: Adaptively adjust the image according to the illumination conditions of different regions in the video frame image by calculating the variance to adapt to the transformation of different regions. The specific method is as follows: The size of the local area is n×n. The pixel point (x, y) is taken as the center position. The variance V(x, y) of the local area is calculated. Then, the relationship between the enhancement coefficient and the regional variance is expressed as: Among them, κ represents the enhancement coefficient, V represents the variance of the local area; Step S23: Binarize the grayscale image of the V channel using the Otsu threshold method, determine an optimal threshold, divide the image into two parts, the target and the background, and design a target-based brightness response curve; for the pixels in the target area, calculate the brightness value distribution, and design a brightness response curve to adjust the brightness information of the target / background objects based on the brightness characteristics of the target object: f(t)=exp[-(t-μ) 2 / 2σ 2 ] Among them, f(t) is the brightness response curve, (μ,σ 2 ) is the standard deviation of the response pixel area, f Otsu is the average brightness response of the partition block, m is the number of pixels, I Otsu is the brightness of the processed sub-image block; Step S24: Exposure compensation is performed on the foreground object, and overexposure and underexposure information of the background area are appropriately suppressed to avoid the problem of over-enhancement of the entire image. Based on the relevant parameters of the foreground object and background area obtained, an enhancement formula is designed: Where I″ is the pixel enhancement output, z is the brightness value of the pixel, and z max is the maximum pixel brightness value, τ c is the adaptive enhancement factor; To avoid over-enhancement, a truncation threshold is introduced as follows:

4. The image enhancement method for substation non-uniform exposure based on adaptive fusion according to claim 1, characterized in that: The step S3 is specifically as follows: For the color channel map, the color saturation is enhanced according to the overall hue segmentation to meet the color requirements in different environments: Among them, I″′(x,y) is the saturation S channel output after color enhancement, f(H,S) is the designed saturation segment enhancement function, H h min,H h max is the minimum and maximum value of the hue range, s h is the saturation enhancement coefficient.

5. The image enhancement method for substation non-uniform exposure based on adaptive fusion according to claim 1, characterized in that: The step S4 is specifically as follows: Step S41: After processing the brightness and color enhanced channels separately, they are fused with the original channel images using a weighted combination method. Furthermore, a weight map is constructed using a quality metric method, and the weights are optimized using an improved Gaussian filter to obtain an exposure-enhanced fused RGB image. Step S42: For the channel maps enhanced by Laplace brightness and brightness response curve, a weight map is constructed based on the quality metric with the original brightness channel. The influencing factors of image clarity, exposure, and contrast are considered, and the fused weight map is comprehensively calculated: I lout (x,y)=w a ×I(x,y)+w b ×I′(x,y)+w c ×I″(x,y); Among them, I lout (x, y) is the V channel fusion brightness image, I (x, y) is the original V channel brightness image, k is the image to be processed, w a 、w b 、w c are the clarity, exposure and contrast weights, and α is a fixed constant; Step S43: For the color channel image that uses the color matrix for brightness enhancement, fuse it with the original color channel according to the weight of the enhancement intensity: I cout =d×I(x,y)+(1-d)×I″′(x,y) Among them, I cout is the color enhancement fusion output image, c is the weight coefficient in the [0,1] interval; Step S44: Convert the enhanced brightness V channel and color H and S channels into RGB format images for output.

6. The image enhancement method for substation non-uniform exposure based on adaptive fusion according to claim 1, characterized in that: The step S5 is specifically as follows: Step S51: Establish an image enhancement quality evaluation mechanism based on subjective and objective evaluation indicators: The subjective evaluation method involves professional image analysts or substation staff observing the enhanced image and scoring it based on aspects such as image clarity, color naturalness, prominence of edge details, and identifiability of target objects. The average of these scores is taken as the subjective evaluation result (MOS) of the enhanced image. Among them, goal(pic) is the score of different subjective evaluation factors; Step S52: Objective evaluation method mainly uses the structural similarity index (SSIM) and peak signal-to-noise ratio (PSNR) to evaluate the reconstruction quality of the enhanced image in terms of brightness, structure, and corresponding pixel error. Among them, μ x 、μ y is the pixel average value, is the pixel standard deviation, σ xy is the pixel covariance, β1 and β2 are constant values, M and N are the length and width of the image, To enhance the output image.

7. The image enhancement system for substation non-uniform exposure based on adaptive fusion is characterized by: include: at least one processor; as well as a memory communicatively connected to at least one of the processors; wherein, The memory stores instructions that can be executed by the processor, and the instructions are used to be executed by the processor to implement the image enhancement method for substation non-uniform exposure based on adaptive fusion as described in any one of claims 1 to 5.

Images