A method for enhancing topographic survey images for modular substation installation

Abstract

The present invention relates to the field of image enhancement, and in particular to a method for enhancing topographic survey images for modular substation placement. The method comprises: performing convolution processing on a topographic image to obtain first feature maps of different scales; adaptively adjusting the size of a normalization window in an image local contrast normalization algorithm, and using the image local contrast normalization algorithm to normalize the first feature maps of each scale to obtain a second feature map; calculating a cross-scale fusion weight for each subregion in the second feature map; and using the cross-scale fusion weight to fuse subregions at the same location to obtain a fused region, with the multiple fused regions constituting an enhanced topographic image. The present invention can preserve feature details at different scales, avoid oversmoothing of microscopic details or blurring of macroscopic structures, and can meet the needs of modular substation site selection.

Description

Technical Field

[0001] This invention relates to the field of image enhancement, and more particularly to a method for enhancing topographic survey images for modular substation installation. Background Technology

[0002] The site selection of modular substations is highly dependent on topographic conditions, requiring precise analysis of surface morphology, geological stability, and potential risks (such as landslides and faults). However, actual survey images are limited by topographic undulations, weather conditions, and equipment, resulting in local overexposure or underexposure, leading to the loss of details in shadow areas (such as fissures and rock textures) and distortion of features in highlight areas (such as reflection artifacts). Topography has both macroscopic structures (such as ridges and valleys) and microscopic details (such as rock fissures and soil particles).

[0003] Chinese patent document CN111612700B discloses an image enhancement method, which includes: determining the sharpness ratio of pixels in each color channel of an initial RGB image; converting the initial RGB image into at least one reference YUV image; fusing the reference YUV images according to the sharpness ratio of pixels in each color channel of the initial RGB image to obtain a target YUV image; converting the target YUV image into a target RGB image; and determining the target RGB image as the enhanced image of the initial RGB image.

[0004] Existing technologies enhance initial images using traditional single-scale enhancement methods. However, these image enhancement methods cannot preserve feature details at different scales, resulting in overly smoothed micro-details or blurred macro-structures, which cannot meet the needs of modular substation site selection. Therefore, how to preserve feature details at different scales when enhancing images is the problem that this invention addresses. Summary of the Invention

[0005] To address the issue of preserving feature details at different scales when performing image enhancement processing, this invention provides a method for enhancing topographic survey images for modular substation installation.

[0006] This invention provides a method for enhancing topographic survey images for modular substation installation, employing the following technical solution:

[0007] Acquire terrain images and perform convolution processing on the terrain images to obtain first feature maps at different scales;

[0008] The size of the normalization window in the local contrast normalization algorithm is adaptively adjusted, and the local contrast normalization algorithm is used to normalize the first feature map at each scale to obtain the second feature map.

[0009] Each second feature map is divided into multiple sub-regions. For multiple second feature maps, the sub-regions at the same location correspond to each other.

[0010] Calculate the cross-scale fusion weight for each sub-region in the second feature map. The cross-scale fusion weight is positively correlated with the information entropy of the sub-region. Use the cross-scale fusion weight to fuse sub-regions at the same location to obtain a fused region. Multiple fused regions constitute the enhanced terrain image.

[0011] Multiple first feature maps are obtained by convolution processing of the terrain image. The first feature maps are normalized and then the sub-regions are fused using cross-scale fusion weights to obtain an enhanced terrain image. This can preserve feature details at different scales, avoid overly smooth micro-details or blurred macro-structures, and adapt to the needs of modular substation site selection.

[0012] Preferably, the method further includes: calculating the adjustment coefficient for the normalized window size, expressed as:

[0013]

[0014] In the formula, Represents pixels The adjustment factor for the normalization window size of local image contrast. Represents pixels gradient magnitude, Represents pixels The standard deviation of the brightness values ​​of the neighboring pixels, This represents the maximum gradient magnitude of a pixel in the feature map at this scale. This represents the maximum standard deviation of the brightness values ​​of the neighboring pixels of all pixels in the feature map at this scale.

[0015] The above formula is used to calculate the adjustment coefficient, which improves the accuracy of the calculation results and provides a theoretical basis for calculating the size of the normalization window.

[0016] Preferably, the size of the normalization window in the adaptive image local contrast normalization algorithm is:

[0017]

[0018] In the formula, Represents pixels Adjusted normalized window, Indicates the default window size. Represents pixels The adjustment coefficient for the normalization window size of the local contrast of the image, exp represents the exponential function with base e. This indicates the floor function.

[0019] By adjusting the size of the normalization window, the normalized image can retain as many detailed features as possible.

[0020] The preferred expression for the cross-scale fusion weights is:

[0021]

[0022] In the formula, This represents the cross-scale fusion weight of the sub-region in the p-th second feature map. Let represent the information entropy of the brightness value of the sub-region in the p-th second feature map, where p represents the index of the second feature map. This indicates the number of second feature maps.

[0023] The cross-scale fusion weight of the sub-region is obtained by calculating the information entropy of the sub-region, which can avoid the details being covered during fusion and thus retain more details.

[0024] Preferably, before fusing sub-regions at the same location using cross-scale fusion weights, the method further includes: calculating the thermal conductivity coefficient in the anisotropic diffusion filtering algorithm, and using the anisotropic diffusion filtering algorithm to filter the sub-regions in the second feature map.

[0025] Preferably, the expression for the thermal conductivity coefficient is:

[0026]

[0027] In the formula, Indicated at a scale of The thermal conductivity coefficient of the sub-region in the second feature map is adaptively adjusted. This represents the preset thermal conductivity coefficient. Indicated at a scale of All pixels in the sub-region of the second feature map The average gradient along the axial direction. Indicated at a scale of All pixels in the sub-region of the second feature map The average gradient along the axial direction. This represents the maximum value of the gradient in the subregion.

[0028] By calculating the thermal conductivity coefficient of a sub-region, we can avoid over-smoothing of details and suppress noise.

[0029] Preferably, the expression for fusing sub-regions at the same location using cross-scale fusion weights is:

[0030]

[0031] In the formula, Indicates the location within the merged area The brightness value of cross-scale fusion. Indicates the number of second feature maps. Indicates the position of the p-th second feature map sub-region Cross-scale fusion weights, Indicates the position of the p-th second feature map sub-region The brightness value, where p represents the index of the second feature map.

[0032] Preferably, the method further includes filtering the enhanced terrain image using an anisotropic diffusion filtering algorithm.

[0033] Preferably, the method further includes geometric correction and radiometric correction of the terrain image before performing convolution processing on the terrain image.

[0034] Preferably, before normalizing the first feature maps at each scale using the local contrast normalization algorithm, the method further includes: upsampling the first feature maps at different scales using bilinear interpolation to make the first feature maps the same size as the terrain image.

[0035] The present invention has the following technical effects:

[0036] Multiple first feature maps are obtained by convolution processing of the terrain image. The first feature maps are normalized and then the sub-regions are fused using cross-scale fusion weights to obtain an enhanced terrain image. This can preserve feature details at different scales, avoid overly smooth micro-details or blurred macro-structures, and adapt to the needs of modular substation site selection. Attached Figure Description

[0037] Figure 1 This is a flowchart of a topographic survey image enhancement method for modular substation installation according to the present invention. Detailed Implementation

[0038] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0039] This invention discloses a method for enhancing topographic survey images for modular substation installation, referring to... Figure 1 The process includes the following steps, as detailed below:

[0040] S1: Obtain the terrain image and perform convolution processing on the terrain image to obtain the first feature map at different scales.

[0041] Topographic images are obtained by capturing visible light images of the area to be surveyed using a multispectral camera. Geometric correction is then performed on the topographic images to address distortions caused by factors such as photographic material deformation, lens distortion, atmospheric refraction, Earth curvature, Earth rotation, and topographic relief, where the geometric position, shape, size, and orientation of features on the topographic images do not match the representation requirements in the reference system. Then, radiometric correction is performed on the topographic images to eliminate or correct image brightness distortion caused by radiometric errors.

[0042] Three first feature maps of different scales were obtained by convolving the corrected terrain image with three different sizes of convolution kernels: 3×3, 5×5, and 7×7, respectively, to capture feature details at different levels in the area to be surveyed.

[0043] S2: Adaptively adjust the size of the normalization window in the local contrast normalization algorithm of the image, and use the local contrast normalization algorithm of the image to normalize the first feature map at each scale to obtain the second feature map.

[0044] Because the convolution kernel sizes of the first feature maps at different scales vary, the first feature maps have different numerical ranges, and their statistical distributions show significant differences in mean and variance. For example, the overall brightness of a mountain outline is low due to global shading, resulting in a lower average amplitude, while the brightness of fissures or gentle slopes is high due to local reflections, resulting in a higher average amplitude. Therefore, the local contrast normalization window is adaptively adjusted based on the gradient amplitude of pixels and regional contrast, and the first feature map at each scale is normalized through this adaptive normalization window.

[0045] Bilinear interpolation upsampling is applied to the first feature maps at different scales to make the first feature maps the same size as the terrain image. Then, the adjustment coefficient for the normalization window size in the image local contrast normalization algorithm is calculated, and the expression is:

[0046]

[0047] In the formula, Represents pixels The adjustment factor for the normalization window size of local image contrast. Represents pixels gradient magnitude, Represents pixels The standard deviation of the brightness values ​​of the neighboring pixels, This represents the maximum gradient magnitude of a pixel in the feature map at this scale. This represents the maximum standard deviation of the brightness values ​​of the neighboring pixels of all pixels in the feature map at this scale, where... Its function is to normalize the adjustment coefficients.

[0048] The method for adaptively adjusting the size of the normalization window in the local contrast normalization algorithm is as follows:

[0049]

[0050] In the formula, Represents pixels Adjusted normalized window, This indicates the preset window size. For example, k=3 means the window size is 3×3. Represents pixels The adjustment coefficient for the normalization window size of the local contrast of the image, exp represents the exponential function with base e. This indicates the floor function.

[0051] A higher local contrast and a larger gradient magnitude indicate that the pixel may belong to a region with complex textures such as rock cracks or steep slope edges, containing more details. Its normalization window should be smaller to preserve details. Conversely, a lower local contrast and a smaller gradient magnitude indicate that the pixel may belong to a flat region with simple textures such as sand, water, or homogeneous soil, containing fewer details and stronger noise. Therefore, its normalization window should be larger to suppress noise.

[0052] S3: Divide each second feature map into multiple sub-regions. For multiple second feature maps, the sub-regions at the same position correspond to each other. Calculate the cross-scale fusion weight of each sub-region in the second feature map. The fusion weight is positively correlated with the information entropy of the sub-region.

[0053] Each second feature map is divided into 9 sub-regions. It can be understood that the first sub-region of the three second feature maps corresponds to each other, the second sub-region corresponds to each other, ..., the ninth sub-region corresponds to each other.

[0054] The expression for the cross-scale fusion weights is:

[0055]

[0056] In the formula, This represents the cross-scale fusion weight of the sub-region in the p-th second feature map. Let represent the information entropy of the brightness value of the sub-region in the p-th second feature map, where p represents the index of the second feature map. This indicates the number of second feature maps. It should be noted that each sub-region of the second feature map corresponds to a fusion weight. For example, if the information entropy of the second sub-region in the three second feature maps is 1, 2, and 3 respectively, then the cross-scale fusion weights of the second sub-regions in the three second feature maps are respectively... , , .

[0057] The higher the information entropy, the more complex the key terrain feature details contained in the sub-region of the second feature map at that scale are. Therefore, a higher cross-scale fusion weight is needed to preserve the details and avoid the details being covered during fusion.

[0058] S4: Use cross-scale fusion weights to fuse sub-regions at the same location to obtain fused regions. Multiple fused regions constitute the enhanced terrain image.

[0059] In one embodiment, the thermal conductivity coefficient in the anisotropic diffusion filtering algorithm is calculated, and the anisotropic diffusion filtering algorithm is used to filter the sub-regions in the second feature map.

[0060] The expression for thermal conductivity is:

[0061]

[0062] In the formula, Indicated at a scale of The thermal conductivity coefficient of the sub-region in the second feature map is adaptively adjusted. This represents the preset thermal conductivity coefficient. The value is set manually based on the actual situation. For example, The value is 15. Indicated at a scale of All pixels in the sub-region of the second feature map The average value of the gradient along the axial direction. Indicated at a scale of All pixels in the sub-region of the second feature map The average value of the gradient along the axial direction. This represents the maximum value of the gradient in the sub-region. Its function is to normalize the numerator term.

[0063] A region with a larger gradient magnitude indicates that it contains more texture details, and a smaller thermal conductivity coefficient should be used to avoid over-smoothing of details. A region with a smaller gradient magnitude indicates that it is flatter and more susceptible to noise, and a larger K value should be used to suppress noise.

[0064] The cross-scale fusion weights are used to fuse sub-regions at the same location, as expressed in the following expression:

[0065]

[0066] In the formula, Indicates the location within the merged area The brightness value of cross-scale fusion. Indicates the number of second feature maps. Indicates the position of the p-th second feature map sub-region Cross-scale fusion weights, Indicates the position of the p-th second feature map sub-region The brightness value, where p represents the index of the second feature map.

[0067] By traversing each location, three adaptively filtered second feature maps of different scales are fused across scales to obtain an enhanced terrain image. This method better preserves local details and handles residual noise across scales. It should be noted that filtering sub-regions before fusion preserves more detail and is suitable for scenarios with high requirements for enhanced images.

[0068] In one embodiment, sub-regions at the same location are first fused using cross-scale fusion weights to obtain a preliminary enhanced terrain image. Then, an anisotropic diffusion filtering algorithm is used to filter the enhanced terrain image. It should be noted that this filtering and denoising method has low computational cost and is suitable for scenarios with relatively low requirements for image enhancement.

[0069] The above are all preferred embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Therefore, all equivalent changes made in accordance with the structure, shape and principle of the present invention should be covered within the scope of protection of the present invention.

Claims

1. A method for enhancing topographic survey images for modular substation installation, characterized in that, Including the following steps: The terrain image is acquired, and the terrain image is convolved to obtain the first feature map at different scales; the first feature map at each scale is normalized using the local contrast normalization algorithm to obtain the second feature map. The adaptive adjustment of the normalization window size in the local contrast normalization algorithm includes: , Represents pixels Adjusted normalized window, This indicates the default window size, and exp represents an exponential function with base e. Indicates the floor function; Represents pixels The adjustment coefficient for the image local contrast normalization window size is calculated using the following formula: , Represents pixels gradient magnitude, Represents pixels The standard deviation of the brightness values ​​of the neighboring pixels, This represents the maximum gradient magnitude of a pixel in the first feature map at any scale across different scales. The maximum standard deviation of the brightness values ​​of the neighboring pixels of all pixels in the first feature map at any scale in different scales; Each second feature map is divided into multiple sub-regions. For multiple second feature maps, the sub-regions at the same location correspond to each other. Calculate the cross-scale fusion weight for each sub-region in the second feature map. The cross-scale fusion weight is positively correlated with the information entropy of the sub-region. Use the cross-scale fusion weight to fuse sub-regions at the same location to obtain a fused region. Multiple fused regions constitute the enhanced terrain image. Before fusing sub-regions at the same location using cross-scale fusion weights, the following steps are included: calculating the thermal conductivity coefficient in the anisotropic diffusion filtering algorithm, expressed as: , This represents the adaptively adjusted thermal conductivity coefficient of a sub-region in the second feature map at scale j. This represents the preset thermal conductivity coefficient. The value is set manually based on the actual situation. This represents all pixels in the sub-region of the second feature map at scale j. The average value of the gradient along the axial direction. This represents all pixels in the sub-region of the second feature map at scale j. The average value of the gradient along the axial direction. This represents the maximum gradient value in the sub-region; the anisotropic diffusion filtering algorithm is used to filter the sub-region in the second feature map.

2. The topographic survey image enhancement method for modular substation installation according to claim 1, characterized in that, The expression for the cross-scale fusion weights is: In the formula, This represents the cross-scale fusion weight of the sub-region in the p-th second feature map. Let represent the information entropy of the brightness value of the sub-region in the p-th second feature map, where p represents the index of the second feature map. This indicates the number of second feature maps.

3. The topographic survey image enhancement method for modular substation installation according to claim 1, characterized in that, The expression for fusing sub-regions at the same location using cross-scale fusion weights is as follows: In the formula, Indicates the location within the merged area The brightness value of cross-scale fusion, Indicates the number of second feature maps. Indicates the position of the p-th second feature map sub-region Cross-scale fusion weights, Indicates the position of the p-th second feature map sub-region The brightness value, where p represents the index of the second feature map.

4. The topographic survey image enhancement method for modular substation installation according to claim 1, characterized in that, The method also includes using anisotropic diffusion filtering algorithm to filter the enhanced terrain image.

5. The topographic survey image enhancement method for modular substation installation according to claim 1, characterized in that, Before performing convolution processing on the terrain image, geometric and radiometric corrections are also performed on the terrain image.

6. The topographic survey image enhancement method for modular substation installation according to claim 1, characterized in that, Before normalizing the first feature maps at various scales using the local contrast normalization algorithm, the following steps are also taken: bilinear interpolation upsampling is applied to the first feature maps at different scales to make the first feature maps the same size as the terrain image.

Images