A remote sensing image enhancement method for green exploration

Abstract

The application discloses a kind of green exploration-oriented remote sensing image enhancement methods, comprising: obtaining target area original image and pre-processing generates first image;Through region segmentation algorithm, topography is classified to first image, and second image is formed by extracting wetland and forest area.Light intensity is adjusted luminance to generate third image by analyzing light intensity for wetland boundary illumination change;By adaptive contrast modulation algorithm, wetland water body and vegetation boundary are optimized, and fourth image is obtained.Extract color distribution characteristics, dynamically adjust hue and saturation to generate fifth image, and carry out feature enhancement to forest vegetation area, highlight texture and soil boundary to form sixth image.If ground feature edge sharpness is insufficient, then iteration optimization contrast and color parameters generate final optimized image.The application significantly improves the identification precision and clarity of ecological region under complex terrain, and provides efficient technical support for ecological environment monitoring and protection.

Description

Technical Field

[0001] This invention belongs to the field of image processing technology, and in particular relates to a remote sensing image enhancement method for green exploration. Background Technology

[0002] Remote sensing technology plays a crucial role in green exploration, supporting scientific decision-making for ecological protection and resource development by acquiring surface information. Especially in complex environments such as forests and wetlands, remote sensing imagery provides critical data support for assessing ecological impacts. However, existing remote sensing image enhancement methods often struggle to effectively highlight surface features when dealing with complex terrain, leading to insufficient accuracy in identifying ecologically sensitive areas. For example, traditional methods often fail to clearly distinguish ground features in areas where vegetation cover and soil boundaries are mixed due to insufficient image contrast. Furthermore, existing technologies have poor adaptability to dynamic environments, making it difficult to flexibly adjust image representation according to the needs of different ecological scenarios. The core challenge lies in how to accurately modulate image contrast and color distribution to adapt to the diverse needs of green exploration.

[0003] Image modulation requires dynamic adjustments to brightness, saturation, and hue based on the complexity of surface features. However, current technologies struggle to achieve adaptive optimization when dealing with complex terrain. This results in image data failing to fully reflect ecological characteristics in areas such as wetlands or forests, impacting the accuracy of environmental assessments. For instance, in green exploration, the boundaries between wetland water bodies and surrounding vegetation are often blurred due to changes in lighting. If the modulation algorithm cannot dynamically adapt to differences in lighting and terrain features, the generated images will fail to meet the demands of high-precision ecological analysis. Furthermore, the modulation process must also consider the characteristic differences between different ecological regions. For example, the dense vegetation of a forest differs significantly from the water reflectance characteristics of a wetland, and a single modulation strategy cannot simultaneously meet the enhancement needs of both, leading to the loss of some terrain feature information. Therefore, how to dynamically optimize the contrast and color distribution of remote sensing images through adaptive modulation to clearly reflect ecological characteristics in complex terrain has become a key issue for green exploration. Summary of the Invention

[0004] To address the aforementioned technical problems, this invention proposes a remote sensing image enhancement method for green exploration, thereby resolving the issues present in the prior art.

[0005] To achieve the above objectives, the present invention provides a remote sensing image enhancement method for green exploration, comprising:

[0006] The original multispectral image of the target area is acquired by a multispectral sensor, and the original multispectral image is preprocessed to obtain the first image;

[0007] The first image is segmented into regions, and classification labels are added to the segmented regions to obtain the second image;

[0008] The wetland boundary region of the second image is obtained, illumination analysis is performed on the wetland boundary region, and the wetland boundary region is processed based on the analysis results to obtain the third image;

[0009] An adaptive contrast modulation algorithm is used to process the third image to obtain a fourth image;

[0010] The color distribution features of the fourth image are obtained, and the color of the fourth image is optimized based on the color distribution features to obtain the fifth image; the boundary between soil and vegetation in the forest vegetation area of ​​the fifth image is enhanced to obtain the sixth image.

[0011] The pixel distribution data of the surface feature points of the sixth image is obtained, and the contrast and color of the sixth image are optimized based on the pixel distribution data to achieve remote sensing image enhancement.

[0012] Optionally, the process of obtaining the first image includes:

[0013] A band selection method is used to select band combinations based on the spectral characteristics of the target area to obtain a complete original multispectral image. A digital elevation model is used to perform terrain correction on the original multispectral image to obtain a corrected image. After a second correction is performed on the corrected image using preset radiometric calibration coefficients and an atmospheric correction model, a median filtering algorithm is used to remove noise to obtain a denoised image. Multiband data fusion is performed on the denoised image to obtain a first image.

[0014] Optionally, the process of obtaining the second image includes:

[0015] The spectral features of the first image are obtained. Based on the spectral features, the pixels of the first image are grouped using the K-means clustering algorithm to obtain segmented region images. The texture features of the segmented region images are extracted. Combined with the spectral features, the segmented region images are classified into terrain types using preset classification rules, and classification labels are added to obtain classification label images. If there are blurred boundary regions in the classification label images, the boundary lines are determined using an edge detection algorithm to obtain spatial distribution images. The spatial distribution images are fused with elevation data from multiple sources to obtain a second image with terrain spatial distribution features.

[0016] Optionally, the process of performing illumination analysis on the wetland boundary area and processing the wetland boundary area based on the analysis results to obtain a third image includes:

[0017] Illumination distribution characteristics of the wetland boundary area are obtained through spectral analysis to obtain illumination distribution data. Based on the illumination distribution data, the average illumination intensity of the wetland boundary area is calculated to determine the illumination intensity value. If the illumination intensity value exceeds a preset threshold, the brightness parameter of the wetland boundary area is adjusted using a linear transformation method to generate a temporary image with adjusted brightness. Spatial features are extracted from the temporary image, and the outline of the wetland boundary area is determined using the Canny edge detection algorithm to obtain a boundary enhancement image. The classification label of the wetland boundary area is updated by fusing the terrain classification results of the boundary enhancement image and the second image to generate a label update image. A third image with uniform illumination distribution is generated based on the label update image.

[0018] Optionally, the process of processing the third image using an adaptive contrast modulation algorithm to obtain the fourth image includes:

[0019] Texture features of the boundary area between wetland water and vegetation are obtained from the third image using a feature extraction algorithm to generate a texture feature set. Based on the texture feature set, an adaptive contrast modulation algorithm is used to calculate the contrast distribution of the boundary area between wetland water and vegetation to obtain contrast distribution data. If the contrast distribution data exceeds a preset threshold, the contrast parameters are adjusted using a linear transformation method to generate an adjusted temporary image. Spatial features are extracted from the temporary image, and the contour of the boundary between wetland water and vegetation is determined using the Sobel edge detection algorithm. The segmentation label of the boundary area between wetland water and vegetation is updated based on the determined contour to generate a segmentation label updated image. Based on the segmentation label updated image, the contrast distribution is smoothed using a mean filtering method. It is then determined whether the spectral features of the smoothed image meet a preset uniformity standard. If they do, a fourth image is obtained.

[0020] Optionally, the process of obtaining the color distribution features of the fourth image, and optimizing the color of the fourth image based on the color distribution features to obtain the fifth image includes:

[0021] Pixel distribution data of the fourth image is obtained. Based on the pixel distribution data, the color gamut range is calculated using histogram analysis to obtain color gamut distribution characteristics. According to the color gamut distribution characteristics, the brightness characteristics are smoothed using mean filtering to generate a smoothed brightness image. Contrast values ​​are extracted from the smoothed brightness image. If the contrast value exceeds a preset threshold, the pixel distribution is adjusted using linear transformation to obtain an adjusted image. After determining the color boundaries of the adjusted image using the Canny edge detection algorithm, hue data is extracted. If the hue data deviates from the preset color gamut range, the hue is corrected using color mapping to obtain a hue-corrected image. Based on the hue-corrected image, the saturation value is optimized using spatial smoothing to generate an optimized saturation image. Color parameters are extracted from the optimized saturation image, and it is determined whether the color parameters meet a preset uniformity standard. If they do, the fifth image is obtained.

[0022] Optionally, the process of enhancing the boundary between soil and vegetation in the forest vegetation area of ​​the fifth image to obtain the sixth image includes:

[0023] Based on the fifth image, forest vegetation areas and soil areas are determined using a region segmentation algorithm to obtain a segmented image. According to the segmented image, a Gabor filtering algorithm is used to extract texture features from the vegetation areas, generating a texture feature image. Boundary line data is obtained from the texture feature image. If the intersection of the boundary line data and the soil area is below a preset threshold, the boundary lines are enhanced using morphological dilation, and then mean filtering is used to smooth the texture noise outside the vegetation areas, generating a smooth texture image. Texture intensity values ​​of the vegetation areas are extracted from the smooth texture image. If the texture intensity values ​​exceed a preset threshold, the texture distribution is adjusted using linear transformation, generating an adjusted texture image. Based on the adjusted texture image, a spatial clustering method is used to optimize the contrast between the vegetation areas and the soil boundaries, obtaining the sixth image.

[0024] Optionally, pixel distribution data of surface feature points in the sixth image is obtained, and the contrast and color of the sixth image are optimized based on the pixel distribution data. The process includes:

[0025] Pixel distribution data of surface feature points are obtained from the sixth image. An edge detection algorithm is used to determine the edge sharpness value, resulting in an edge sharpness distribution map. If the sharpness value of the surface feature points in the edge sharpness distribution map is lower than a preset threshold, the pixel distribution is adjusted by histogram equalization to generate a contrast-enhanced image. Based on the contrast-enhanced image, the color saturation is adjusted using a color space conversion method to obtain a color-optimized image. Texture intensity data of surface feature points is extracted from the color-optimized image, and a mean filtering method is used to smooth noise in non-feature areas. If the texture intensity of the surface feature points exceeds a preset threshold, the texture distribution is adjusted by linear transformation, and then a spatial clustering method is used to optimize the contrast between the surface feature points and the background area to obtain an image with adjusted contrast. Edge sharpness value and color distribution data are obtained from the image with adjusted contrast, and the sharpness and color parameters are adjusted by iterative optimization methods to obtain the final optimized image.

[0026] Optional, also includes:

[0027] Based on the final optimized image, spatial distribution data of surface feature points are obtained. An initial surface feature distribution map is generated using Kriging interpolation, resulting in a high-precision distribution map. Feature point coordinates of wetlands and forests are extracted from the high-precision distribution map, and density clustering is used to delineate ecologically sensitive areas and determine area boundaries. Feature point density data of wetlands and forests are obtained from the area boundaries. If the density value is lower than a preset threshold, the density distribution is adjusted using a weighted average method to obtain an optimized density distribution map. Spatial correlation data of surface feature points is extracted from the optimized density distribution map, and spatial autocorrelation is calculated using Moran's index analysis to obtain a spatial correlation map of ecologically sensitive areas. Clustering patterns of feature points in wetlands and forests are obtained from the spatial correlation map. If the clustering pattern deviates from a preset threshold, the feature point distribution is adjusted using a spatial smoothing method to obtain a smoothed correlation map. Boundary feature data of ecologically sensitive areas are extracted from the smoothed correlation map, and a vector distribution map is generated using a vector transformation method to obtain a high-precision vector distribution map. Spatial extent data of ecologically sensitive areas of wetlands and forests are obtained from the high-precision vector distribution map, and a final distribution map is generated using a gridding method to obtain a distribution map of ecologically sensitive areas.

[0028] Compared with the prior art, the present invention has the following advantages and technical effects:

[0029] This invention addresses the challenge of accurately identifying ecologically sensitive areas such as wetlands and forests in complex terrain. It acquires raw images using a multispectral sensor and preprocesses them to generate a first image. Then, a region segmentation algorithm is used for terrain classification, extracting wetland and forest areas to form a second image. For wetland boundary illumination variations, the light intensity is analyzed and brightness adjusted to generate a third image. An adaptive contrast modulation algorithm is then used to optimize the boundaries between wetland water bodies and vegetation, resulting in a fourth image. Further, color distribution features are extracted, and hue and saturation are dynamically adjusted to generate a fifth image. Finally, features in forest vegetation areas are enhanced to highlight texture and soil boundaries, forming a sixth image. If the edge sharpness of surface features is insufficient, contrast and color parameters are iteratively optimized to generate a final optimized image, further producing a high-precision surface feature distribution map. This invention, through multi-stage image enhancement and optimization, significantly improves the accuracy and clarity of identifying ecological areas in complex terrain, providing efficient technical support for ecological environment monitoring and protection. Attached Figure Description

[0030] The accompanying drawings, which form part of this application, are used to provide a further understanding of this application. The illustrative embodiments and descriptions of this application are used to explain this application and do not constitute an undue limitation of this application. In the drawings:

[0031] Figure 1 This is a flowchart of a method according to an embodiment of the present invention. Detailed Implementation

[0032] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. This application will now be described in detail with reference to the accompanying drawings and embodiments.

[0033] It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases the steps shown or described may be executed in a different order than that shown here.

[0034] Example 1

[0035] like Figure 1 As shown, this embodiment provides a remote sensing image enhancement method for green exploration, including:

[0036] The original multispectral image of the target area is acquired by a multispectral sensor, and the original multispectral image is preprocessed to obtain the first image;

[0037] As one specific implementation method, the process of obtaining the first image includes:

[0038] A band selection method is used to select band combinations based on the spectral characteristics of the target area to obtain a complete original multispectral image. A digital elevation model is used to perform terrain correction on the original multispectral image to obtain a corrected image. After a second correction is performed on the corrected image using preset radiometric calibration coefficients and an atmospheric correction model, a median filtering algorithm is used to remove noise to obtain a denoised image. Multiband data fusion is performed on the denoised image to obtain a first image.

[0039] For example, to acquire raw images of a target area using a multispectral sensor, a hyperspectral imager can be used to obtain multispectral data of farmland in agricultural scenarios. Assuming the target area is a cornfield, the sensor covers the visible to near-infrared bands, with a wavelength range of 400-1000 nanometers. The band selection is based on the spectral characteristics of corn, such as the chlorophyll absorption peak at 670 nanometers in the red light band and the near-infrared reflection peak at 850 nanometers. A combination of these two bands is chosen to highlight vegetation health information. The generated multispectral image contains complete spectral information, reflecting the spectral characteristics of corn growth, which is helpful for precision agriculture analysis.

[0040] Topographic correction of raw multispectral images can be performed using digital elevation models (DEMs) to address geometric distortions caused by complex terrain. For example, cornfields located in hilly areas experience pixel shifts due to undulating terrain. Using 30-meter resolution SRTMDEM data, an orthorectification algorithm aligns image pixels with actual geographic coordinates, generating a geometrically corrected image. This correction ensures the accuracy of geographic location in subsequent analyses and improves the reliability of crop distribution maps.

[0041] Radiometric calibration and atmospheric correction of geometrically corrected images can eliminate sensor bias and atmospheric effects. For example, digital luminance values ​​are converted to radiance using radiometric calibration coefficients (such as gain and bias values ​​provided in the sensor calibration file), and then the FLAASH atmospheric correction model is applied based on the day's meteorological data (such as water vapor content of 2 g / cm³). 2 By using an aerosol optical thickness of 0.3 mm to remove atmospheric scattering and absorption, an image of the true surface reflectance is generated. This processing accurately reflects the reflectance characteristics of cornfields, providing reliable data for vegetation index calculations.

[0042] The first image is segmented into regions, and classification labels are added to the segmented regions to obtain the second image;

[0043] As one specific implementation method, the process of obtaining the second image includes:

[0044] The spectral features of the first image are obtained. Based on the spectral features, the pixels of the first image are grouped using the K-means clustering algorithm to obtain segmented region images. The texture features of the segmented region images are extracted. Combined with the spectral features, the segmented region images are classified into terrain types using preset classification rules, and classification labels are added to obtain classification label images. If there are blurred boundary regions in the classification label images, the boundary lines are determined using an edge detection algorithm to obtain spatial distribution images. The spatial distribution images are fused with elevation data from multiple sources to obtain a second image with terrain spatial distribution features.

[0045] Based on the spectral features of the first image, when grouping pixels using the K-means clustering algorithm, region segmentation can be achieved based on differences in spectral reflectance. K-means clustering groups pixels with similar spectral features into one class by calculating the Euclidean distance between the pixel spectral value and the initial cluster center.

[0046] Assuming the target area is a wetland ecosystem, four cluster centers can be selected, representing water, vegetation, bare soil, and shade, respectively. The spectral value of each pixel is represented by the reflectance of its multispectral bands; for example, the values ​​for blue, green, red, and near-infrared bands are 0.1, 0.15, 0.2, and 0.35, respectively. After iterative calculation, pixels are assigned to the nearest cluster center, generating an image that includes segmented regions such as water and vegetation. This method effectively distinguishes regions with similar spectral characteristics, facilitating subsequent classification.

[0047] When extracting spectral and texture features from segmented region images, spectral reflectance and spatial texture information can be combined to enhance classification accuracy. For example, the spectral features of wetland vegetation may exhibit high reflectance in the near-infrared band, while texture features can be extracted using the gray-level co-occurrence matrix to analyze the spatial relationships between pixels.

[0048] For a wetland, the normalized difference vegetation index (NDVI) in the red and near-infrared bands can be extracted, yielding a value of approximately 0.4. Simultaneously, texture contrast is calculated, resulting in a value of approximately 2.5. These features are input into predefined classification rules, such as a threshold-based decision tree, to categorize the region as wetland, forest, or grassland, generating a labeled image. This approach utilizes multidimensional features to improve the robustness of classification.

[0049] When processing regions with blurred boundaries in categorized label images, edge detection algorithms can effectively determine the boundaries of ecological areas. For example, the Canny edge detection algorithm can be used to identify strong edge pixels at the boundary between wetlands and forests by setting high and low thresholds (such as 0.1 and 0.3).

[0050] For transitional zones between wetlands and forests, the algorithm detects pixels with abrupt changes in reflectivity, forming continuous boundary lines and generating spatial distribution images. This method can clearly delineate the boundaries of ecological areas, facilitating spatial analysis.

[0051] By fusing spatial distribution images and elevation data from multiple sources, a more comprehensive ecological distribution image can be generated by combining digital elevation models and spectral classification results. Assuming the elevation range of wetlands is 0-5 meters and that of forests is 5-15 meters, elevation data can be overlaid with labeled images to generate a second image that includes the spatial distribution characteristics of both wetlands and forests. This fusion integrates spectral and topographic information, enhancing the spatial representation of ecological regions.

[0052] The wetland boundary region of the second image is obtained, illumination analysis is performed on the wetland boundary region, and the wetland boundary region is processed based on the analysis results to obtain the third image;

[0053] As a specific implementation method, the process of performing illumination analysis on the wetland boundary area, processing the wetland boundary area based on the analysis results, and obtaining a third image includes:

[0054] Illumination distribution characteristics of the wetland boundary area are obtained through spectral analysis to obtain illumination distribution data. Based on the illumination distribution data, the average illumination intensity of the wetland boundary area is calculated to determine the illumination intensity value. If the illumination intensity value exceeds a preset threshold, the brightness parameter of the wetland boundary area is adjusted using a linear transformation method to generate a temporary image with adjusted brightness. Spatial features are extracted from the temporary image, and the outline of the wetland boundary area is determined using the Canny edge detection algorithm to obtain a boundary enhancement image. The classification label of the wetland boundary area is updated by fusing the terrain classification results of the boundary enhancement image and the second image to generate a label update image. A third image with uniform illumination distribution is generated based on the label update image.

[0055] For example, when extracting the illumination distribution characteristics of wetland boundary areas based on spectral analysis, multispectral image data can be used to analyze the spectral reflectance characteristics of wetland boundary areas in the visible and near-infrared bands. The core of spectral analysis lies in distinguishing the illumination distribution characteristics of wetland boundary areas through the differences in reflectance across different bands. Assuming a multispectral image containing four bands—blue, green, red, and near-infrared—wetland boundary areas typically exhibit higher reflectance in the near-infrared band and lower reflectance in the visible band. Illumination distribution data can be generated by calculating the average reflectance of each pixel in these bands.

[0056] For example, the near-infrared reflectance of a pixel at the boundary of a wetland is 0.65, while the average reflectance in the visible light band is 0.25, indicating that this area is significantly affected by illumination. Generating illumination distribution data is helpful for subsequent analysis of the illumination characteristics of wetland boundaries.

[0057] When calculating the average illuminance of a wetland boundary area, a weighted average of the reflectance of pixels in a specific region can be calculated based on illuminance distribution data. For example, assuming a wetland boundary area containing 1000 pixels has a total near-infrared reflectance of 600 (weight 0.7) and a total visible light reflectance of 200 (weight 0.3), the average illuminance can be calculated using this weighted average. If the preset threshold is 0.5, and the calculated result is 0.55, it indicates that the illuminance exceeds the threshold. This method effectively quantifies illuminance, providing a basis for subsequent brightness adjustments.

[0058] When adjusting the brightness parameters of wetland boundary regions using linear transformation, pixel brightness values ​​can be adjusted through a linear mapping function. Assuming the brightness value range of a certain wetland boundary region is 0.4-0.8, the goal is to compress the brightness value to 0.3-0.6 to reduce brightness differences in areas with excessive illumination. Linear transformation generates a temporary image with adjusted brightness by scaling the brightness value of each pixel. This adjustment makes the wetland boundary region visually more uniform, facilitating subsequent feature extraction.

[0059] When extracting spatial features from temporary images, the texture and shape features of wetland boundaries can be analyzed. When using the Canny edge detection algorithm, the temporary image is first smoothed using Gaussian filtering, then the gradient intensity is calculated, and pixels with higher gradient values ​​are retained as boundaries.

[0060] The texture features of wetland boundary areas are characterized by regular water ripple patterns. The Canny algorithm can extract clear boundary contours by setting high and low thresholds (such as 0.1 and 0.3), generating boundary-enhanced images. This method can effectively highlight the spatial structure of wetland boundaries.

[0061] When fusing the terrain classification results of the boundary enhancement image and the second image, the classification labels of wetland boundaries can be updated by overlaying classification labels and boundary contours. For example, if the second image labels a certain area as wetland, but the boundary enhancement image shows that some boundary areas should be classified as grassland, the labels can be updated to generate a label-updated image. This fusion improves the classification accuracy of wetland boundaries.

[0062] An adaptive contrast modulation algorithm is used to process the third image to obtain a fourth image;

[0063] As a specific implementation method, the process of processing the third image using an adaptive contrast modulation algorithm to obtain the fourth image includes:

[0064] Texture features of the boundary area between wetland water and vegetation are obtained from the third image using a feature extraction algorithm to generate a texture feature set. Based on the texture feature set, an adaptive contrast modulation algorithm is used to calculate the contrast distribution of the boundary area between wetland water and vegetation to obtain contrast distribution data. If the contrast distribution data exceeds a preset threshold, the contrast parameters are adjusted using a linear transformation method to generate an adjusted temporary image. Spatial features are extracted from the temporary image, and the contour of the boundary between wetland water and vegetation is determined using the Sobel edge detection algorithm. The segmentation label of the boundary area between wetland water and vegetation is updated based on the determined contour to generate a segmentation label updated image. Based on the segmentation label updated image, the contrast distribution is smoothed using a mean filtering method. It is then determined whether the spectral features of the smoothed image meet a preset uniformity standard. If they do, a fourth image is obtained.

[0065] For example, when extracting texture features from the boundary between wetland water and vegetation in a third image, the gray-level co-occurrence matrix (GLCM) method can be used to analyze the spatial relationships between image pixels. The GLCM generates a texture feature set, such as contrast, correlation, and entropy, by statistically analyzing the gray-level distribution of pixel pairs. Assuming the wetland water area has a smoother texture while the vegetation area exhibits a coarser texture, the feature set can quantify these differences.

[0066] When calculating contrast distribution using an adaptive contrast modulation algorithm, the image's contrast parameters can be dynamically adjusted based on the contrast values ​​of the texture feature set. Assuming a preset contrast threshold of 0.5, if the contrast value of the vegetation area (0.8) exceeds the threshold, the contrast can be reduced through a linear transformation, for example, adjusting the value from 0.8 to 0.4 to generate a temporary image. This adjustment smooths out visual differences in boundary areas, facilitating subsequent processing.

[0067] When updating segmentation labels, the regions separating water bodies and vegetation can be relabeled based on the boundary enhancement image. For example, if some water bodies in the original image are mislabeled as vegetation, the contour information from the boundary enhancement image can be used to relabel these regions as water bodies, generating an updated label image. This method ensures classification accuracy.

[0068] The color distribution features of the fourth image are obtained, and the color of the fourth image is optimized based on the color distribution features to obtain the fifth image; the boundary between soil and vegetation in the forest vegetation area of ​​the fifth image is enhanced to obtain the sixth image.

[0069] As a specific implementation method, the process of obtaining the color distribution features of the fourth image, optimizing the color of the fourth image based on the color distribution features, and obtaining the fifth image includes:

[0070] Pixel distribution data of the fourth image is obtained. Based on the pixel distribution data, the color gamut range is calculated using histogram analysis to obtain color gamut distribution characteristics. According to the color gamut distribution characteristics, the brightness characteristics are smoothed using mean filtering to generate a smoothed brightness image. Contrast values ​​are extracted from the smoothed brightness image. If the contrast value exceeds a preset threshold, the pixel distribution is adjusted using linear transformation to obtain an adjusted image. After determining the color boundaries of the adjusted image using the Canny edge detection algorithm, hue data is extracted. If the hue data deviates from the preset color gamut range, the hue is corrected using color mapping to obtain a hue-corrected image. Based on the hue-corrected image, the saturation value is optimized using spatial smoothing to generate an optimized saturation image. Color parameters are extracted from the optimized saturation image, and it is determined whether the color parameters meet a preset uniformity standard. If they do, the fifth image is obtained.

[0071] In image processing of the boundary area between wetland water bodies and vegetation, when obtaining pixel distribution data from the fifth image, the color gamut characteristics can be characterized by analyzing the pixel value distribution in the RGB color space.

[0072] Specifically, pixel distribution data can be obtained by generating histograms based on the brightness value and color components of each pixel to reflect the color distribution of wetland water bodies and vegetation areas. Assuming that the pixel brightness values ​​of wetland water bodies are concentrated in a lower range, such as 20-80, while the brightness values ​​of vegetation areas are concentrated in 100-150, histogram analysis can clearly distinguish the color gamut ranges of the two. This method helps to accurately capture the color differences between wetlands and vegetation, providing a reliable data foundation for subsequent processing.

[0073] When extracting contrast values ​​and comparing them with a preset threshold, the threshold can be set to 1.5 times the standard deviation.

[0074] If the contrast value of the wetland water area exceeds this threshold, it indicates excessive brightness variation, which may affect boundary recognition. In this case, adjusting the pixel distribution through a linear transformation method, such as linearly compressing the brightness value to the range of 50-200, can generate an adjusted image. This adjustment can effectively balance the brightness difference between the wetland and vegetation areas, improving the accuracy of subsequent edge detection.

[0075] When using the Canny edge detection algorithm to identify color boundaries, the adjusted image can first be Gaussian blurred to reduce noise, and then the boundary between wetland water and vegetation can be detected using a dual threshold method.

[0076] Setting a low threshold of 50 and a high threshold of 150 generates sharp, enhanced boundary images. This method highlights the outline of the boundary between wetland water bodies and vegetation, providing an accurate boundary reference for tone correction.

[0077] When extracting and correcting hue data, if the wetland water body is found to have a bluish hue while the vegetation has a greenish hue, the hue can be adjusted to a preset color gamut range using color mapping methods, such as mapping the water body hue to the standard blue range (H value between 200 and 240). This correction ensures that the color representation of the wetland and vegetation areas is consistent, facilitating subsequent analysis.

[0078] As a specific implementation method, the process of enhancing the boundary between soil and vegetation in the forest vegetation area of ​​the fifth image to obtain the sixth image includes:

[0079] Based on the fifth image, forest vegetation areas and soil areas are determined using a region segmentation algorithm to obtain a segmented image. According to the segmented image, a Gabor filtering algorithm is used to extract texture features from the vegetation areas, generating a texture feature image. Boundary line data is obtained from the texture feature image. If the intersection of the boundary line data and the soil area is below a preset threshold, the boundary lines are enhanced using morphological dilation, and then mean filtering is used to smooth the texture noise outside the vegetation areas, generating a smooth texture image. Texture intensity values ​​of the vegetation areas are extracted from the smooth texture image. If the texture intensity values ​​exceed a preset threshold, the texture distribution is adjusted using linear transformation, generating an adjusted texture image. Based on the adjusted texture image, a spatial clustering method is used to optimize the contrast between the vegetation areas and the soil boundaries, obtaining the sixth image.

[0080] For example, when obtaining pixel distribution data from the fifth image, the distribution of forest vegetation and soil areas can be determined by analyzing the pixel values ​​in the RGB color space.

[0081] When using region segmentation algorithms, images can be divided into vegetation regions and soil regions based on color and texture differences. K-means clustering can be used to cluster pixels according to color features, setting K=2 to distinguish between green vegetation and brown soil.

[0082] When extracting texture features from vegetation areas using the Gabor filtering algorithm, texture details of vegetation can be captured by using filters with different directions and frequencies. By setting filter directions to 0°, 45°, and 90°, and frequencies to 0.1 and 0.2, texture feature images are generated, highlighting leaf veins or grass textures. These texture features can be used to analyze the health status or type of vegetation.

[0083] When extracting boundary line data from texture feature images, the boundary between vegetation and soil can be determined by detecting gradient changes in texture intensity. If the intersection is below a preset threshold (e.g., the boundary pixel ratio is less than 10%), the boundary line can be extended using morphological dilation. For example, dilation using a 3x3 structuring element enhances the continuity of the boundary line, generating a boundary-enhanced image that facilitates subsequent region differentiation.

[0084] When extracting texture intensity values ​​from vegetation areas, the gray-level co-occurrence matrix of the texture feature image can be calculated to obtain the texture intensity. If the intensity value exceeds a preset threshold (e.g., greater than 80), the texture distribution can be adjusted through linear transformation, for example, linearly compressing the intensity value to the range of 50-70 to generate an adjusted texture image. This adjustment can balance texture features and prevent overly prominent textures from affecting subsequent analysis.

[0085] When using spatial clustering methods to optimize the contrast between vegetation areas and soil boundaries, the density-based DBSCAN algorithm can be used to cluster the pixels of vegetation and soil boundaries, thereby enhancing boundary contrast.

[0086] The pixel distribution data of the surface feature points of the sixth image is obtained, and the contrast and color of the sixth image are optimized based on the pixel distribution data to achieve remote sensing image enhancement.

[0087] As a specific implementation method, pixel distribution data of surface feature points in the sixth image is obtained, and the contrast and color of the sixth image are optimized based on the pixel distribution data. The process includes:

[0088] Pixel distribution data of surface feature points are obtained from the sixth image. An edge detection algorithm is used to determine the edge sharpness value, resulting in an edge sharpness distribution map. If the sharpness value of the surface feature points in the edge sharpness distribution map is lower than a preset threshold, the pixel distribution is adjusted by histogram equalization to generate a contrast-enhanced image. Based on the contrast-enhanced image, the color saturation is adjusted using a color space conversion method to obtain a color-optimized image. Texture intensity data of surface feature points is extracted from the color-optimized image, and a mean filtering method is used to smooth noise in non-feature areas. If the texture intensity of the surface feature points exceeds a preset threshold, the texture distribution is adjusted by linear transformation, and then a spatial clustering method is used to optimize the contrast between the surface feature points and the background area to obtain an image with adjusted contrast. Edge sharpness value and color distribution data are obtained from the image with adjusted contrast, and the sharpness and color parameters are adjusted by iterative optimization methods to obtain the final optimized image.

[0089] When obtaining pixel distribution data of surface feature points from the sixth image, gradient-based edge detection methods, such as the Canny algorithm, can be used to identify the pixel distribution of surface feature points. Specifically, the image is first converted to grayscale, and the gradient intensity of each pixel is calculated to generate initial pixel distribution data.

[0090] For example, in an image of a forest vegetation area, the pixel distribution might show that the grayscale values ​​of the vegetation area are concentrated in the 100-150 range, while those of the soil area are concentrated in the 50-80 range. By setting a gradient threshold, such as 20, surface feature points can be filtered out to generate an edge sharpness distribution map. This method can effectively distinguish the boundary between vegetation and soil, providing a clear foundation for subsequent processing.

[0091] In edge sharpness distribution maps, if the sharpness value of surface feature points is lower than a preset threshold, such as less than 30, histogram equalization can be used to adjust the pixel distribution. For example, in a forest image, the pixel grayscale values ​​at the boundary between vegetation and soil may be narrowly distributed, resulting in blurred boundaries. Histogram equalization remaps the grayscale values, making the pixel distribution more uniform and generating a contrast-enhanced image. This method can highlight the edge details of surface feature points, facilitating subsequent analysis.

[0092] Color space conversion methods can be used to adjust the saturation of color-optimized images. Converting the image from RGB to HSV space and adjusting the saturation value of the S channel individually—for example, increasing the saturation of vegetation areas by 10% to make the green more vibrant, while maintaining the original saturation of soil areas—enhances the visual distinction between vegetation and soil, generating a color-optimized image that facilitates further feature extraction. When extracting texture intensity data of surface feature points from the color-optimized image, the gray-level co-occurrence matrix method can be used.

[0093] As one specific implementation method, the method further includes:

[0094] Based on the final optimized image, spatial distribution data of surface feature points are obtained. An initial surface feature distribution map is generated using Kriging interpolation, resulting in a high-precision distribution map. Feature point coordinates of wetlands and forests are extracted from the high-precision distribution map, and density clustering is used to delineate ecologically sensitive areas and determine area boundaries. Feature point density data of wetlands and forests are obtained from the area boundaries. If the density value is lower than a preset threshold, the density distribution is adjusted using a weighted average method to obtain an optimized density distribution map. Spatial correlation data of surface feature points is extracted from the optimized density distribution map, and spatial autocorrelation is calculated using Moran's index analysis to obtain a spatial correlation map of ecologically sensitive areas. Clustering patterns of feature points in wetlands and forests are obtained from the spatial correlation map. If the clustering pattern deviates from a preset threshold, the feature point distribution is adjusted using a spatial smoothing method to obtain a smoothed correlation map. Boundary feature data of ecologically sensitive areas are extracted from the smoothed correlation map, and a vector distribution map is generated using a vector transformation method to obtain a high-precision vector distribution map. Spatial extent data of ecologically sensitive areas of wetlands and forests are obtained from the high-precision vector distribution map, and a final distribution map is generated using a gridding method to obtain a distribution map of ecologically sensitive areas.

[0095] Specifically, when acquiring spatial distribution data of surface feature points, the distribution locations of wetlands and forests can be extracted using the pixel coordinates of remote sensing images.

[0096] For example, in a remote sensing image of a wetland ecological area, pixel coordinates show that the wetland is distributed in the lower left corner of the image, and the forest is distributed in the upper right corner, with a spatial resolution of 10 meters. Each pixel records latitude, longitude, and land surface type, generating a dataset containing coordinates and categories. When generating an initial land surface feature distribution map, the Kriging interpolation method can use these coordinate points to interpolate and generate continuous high-precision distribution maps.

[0097] After extracting the coordinates of feature points in wetlands and forests, density clustering method is used to delineate ecologically sensitive areas.

[0098] For example, at the boundary between wetlands and forests, the density of feature points is high. Clustering algorithms divide the central wetland area into a high-sensitivity zone and the forest edge into a medium-sensitivity zone. Boundary determination is achieved by extracting the circumscribed polygons of the clustered regions, resulting in smooth boundary lines that conform to the actual terrain. When the density value is below a preset threshold, it can be adjusted using a weighted average method. For instance, if the density value of a certain area at the wetland edge is 0.3, below the threshold of 0.5, a weighted average is used to incorporate the values ​​of neighboring high-density points, increasing the density to 0.45 and optimizing a more uniform distribution. After extracting spatial correlation data, the Moran's index analysis method is used to calculate spatial autocorrelation. The Moran's index for wetland feature points is 0.7, indicating strong clustering, while the forest index is 0.4, showing weak clustering. A spatial correlation map is generated accordingly, showing a concentrated distribution in wetland areas and a more dispersed distribution in forests. If the clustering pattern deviates from the preset threshold, it can be adjusted using a spatial smoothing method.

[0099] When generating vector distribution maps, boundary feature data is processed using vector transformation methods. For example, wetland boundary point sets are converted into polygon vector format to generate high-precision vector distribution maps with an accuracy of 0.5 meters. When generating the final distribution map using a gridding method, the region can be divided into 1 km x 1 km grids, with each grid labeled with the proportion of wetlands or forests.

[0100] Kriging interpolation ensures the continuity of high-precision distribution maps, density clustering and Moran index analysis jointly reveal the spatial characteristics of ecological regions, and vector transformation and gridding provide intuitive visualization results. These steps are closely linked and progressively advanced, ensuring the accuracy and practicality of ecologically sensitive area delineation.

[0101] The above are merely preferred embodiments of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A remote sensing image enhancement method for green exploration, characterized in that, Includes the following steps: The original multispectral image of the target area is acquired by a multispectral sensor, and the original multispectral image is preprocessed to obtain the first image; The process of obtaining the first image includes: A band selection method is used to select band combinations based on the spectral characteristics of the target area to obtain a complete original multispectral image. A digital elevation model is used to perform terrain correction on the original multispectral image to obtain a corrected image. After a second correction is performed on the corrected image using preset radiometric calibration coefficients and an atmospheric correction model, a median filtering algorithm is used to remove noise to obtain a denoised image. Multiband data fusion is performed on the denoised image to obtain a first image. The first image is segmented into regions, and classification labels are added to the segmented regions to obtain the second image; The wetland boundary region of the second image is obtained, illumination analysis is performed on the wetland boundary region, and the wetland boundary region is processed based on the analysis results to obtain the third image; The process of performing illumination analysis on the wetland boundary area, processing the wetland boundary area based on the analysis results, and obtaining a third image includes: Illumination distribution characteristics of the wetland boundary area are obtained through spectral analysis to generate illumination distribution data. Based on this data, the average illumination intensity of the wetland boundary area is calculated to determine the illumination intensity value. If the illumination intensity value exceeds a preset threshold, a linear transformation method is used to adjust the brightness parameters of the wetland boundary area, generating a temporary image with adjusted brightness. Spatial features are extracted from the temporary image, and the outline of the wetland boundary area is determined using the Canny edge detection algorithm to obtain a boundary enhancement image. The classification label of the wetland boundary area is updated by fusing the boundary enhancement image with the terrain classification results of a second image, generating a label-updated image. A third image with uniform illumination distribution is generated based on the label-updated image. An adaptive contrast modulation algorithm is used to process the third image to obtain a fourth image; The process of processing the third image using an adaptive contrast modulation algorithm to obtain the fourth image includes: Texture features of the boundary area between wetland water and vegetation are obtained from the third image using a feature extraction algorithm to generate a texture feature set. Based on the texture feature set, an adaptive contrast modulation algorithm is used to calculate the contrast distribution of the boundary area between wetland water and vegetation to obtain contrast distribution data. If the contrast distribution data exceeds a preset threshold, the contrast parameters are adjusted using a linear transformation method to generate an adjusted temporary image. Spatial features are extracted from the temporary image, and the contour of the boundary between wetland water and vegetation is determined using the Sobel edge detection algorithm. The segmentation label of the boundary area between wetland water and vegetation is updated based on the determined contour to generate a segmentation label updated image. Based on the segmentation label updated image, the contrast distribution is smoothed using a mean filtering method. It is then determined whether the spectral features of the smoothed image meet a preset uniformity standard. If they do, a fourth image is obtained. The color distribution features of the fourth image are obtained, and the color of the fourth image is optimized based on the color distribution features to obtain the fifth image; the boundary between soil and vegetation in the forest vegetation area of ​​the fifth image is enhanced to obtain the sixth image. The pixel distribution data of the surface feature points of the sixth image is obtained, and the contrast and color of the sixth image are optimized based on the pixel distribution data to achieve remote sensing image enhancement.

2. The remote sensing image enhancement method for green exploration according to claim 1, characterized in that, The process of obtaining the second image includes: The spectral features of the first image are obtained. Based on the spectral features, the pixels of the first image are grouped using the K-means clustering algorithm to obtain segmented region images. The texture features of the segmented region images are extracted. Combined with the spectral features, the segmented region images are classified into terrain types using preset classification rules, and classification labels are added to obtain classification label images. If there are blurred boundary regions in the classification label images, the boundary lines are determined using an edge detection algorithm to obtain spatial distribution images. The spatial distribution images are fused with elevation data from multiple sources to obtain a second image with terrain spatial distribution features.

3. The remote sensing image enhancement method for green exploration according to claim 1, characterized in that, The process of obtaining the color distribution features of the fourth image, optimizing the color of the fourth image based on the color distribution features, and obtaining the fifth image includes: Pixel distribution data of the fourth image is obtained. Based on the pixel distribution data, the color gamut range is calculated using histogram analysis to obtain color gamut distribution characteristics. According to the color gamut distribution characteristics, the brightness characteristics are smoothed using mean filtering to generate a smoothed brightness image. Contrast values ​​are extracted from the smoothed brightness image. If the contrast value exceeds a preset threshold, the pixel distribution is adjusted using linear transformation to obtain an adjusted image. After determining the color boundaries of the adjusted image using the Canny edge detection algorithm, hue data is extracted. If the hue data deviates from the preset color gamut range, the hue is corrected using color mapping to obtain a hue-corrected image. Based on the hue-corrected image, the saturation value is optimized using spatial smoothing to generate an optimized saturation image. Color parameters are extracted from the optimized saturation image, and it is determined whether the color parameters meet a preset uniformity standard. If they do, the fifth image is obtained.

4. The remote sensing image enhancement method for green exploration according to claim 1, characterized in that, The process of enhancing the boundary between soil and vegetation in the forest vegetation area of ​​the fifth image to obtain the sixth image includes: Based on the fifth image, forest vegetation areas and soil areas are determined using a region segmentation algorithm to obtain a segmented image. According to the segmented image, a Gabor filtering algorithm is used to extract texture features from the vegetation areas, generating a texture feature image. Boundary line data is obtained from the texture feature image. If the intersection of the boundary line data and the soil area is below a preset threshold, the boundary lines are enhanced using morphological dilation, and then mean filtering is used to smooth the texture noise outside the vegetation areas, generating a smooth texture image. Texture intensity values ​​of the vegetation areas are extracted from the smooth texture image. If the texture intensity values ​​exceed a preset threshold, the texture distribution is adjusted using linear transformation, generating an adjusted texture image. Based on the adjusted texture image, a spatial clustering method is used to optimize the contrast between the vegetation areas and the soil boundaries, obtaining the sixth image.

5. The remote sensing image enhancement method for green exploration according to claim 1, characterized in that, The process of acquiring pixel distribution data of surface feature points in the sixth image and optimizing the contrast and color of the sixth image based on the pixel distribution data includes: Pixel distribution data of surface feature points are obtained from the sixth image. An edge detection algorithm is used to determine the edge sharpness value, resulting in an edge sharpness distribution map. If the sharpness value of the surface feature points in the edge sharpness distribution map is lower than a preset threshold, the pixel distribution is adjusted by histogram equalization to generate a contrast-enhanced image. Based on the contrast-enhanced image, the color saturation is adjusted using a color space conversion method to obtain a color-optimized image. Texture intensity data of surface feature points is extracted from the color-optimized image, and a mean filtering method is used to smooth noise in non-feature areas. If the texture intensity of the surface feature points exceeds a preset threshold, the texture distribution is adjusted by linear transformation, and then a spatial clustering method is used to optimize the contrast between the surface feature points and the background area to obtain an image with adjusted contrast. Edge sharpness value and color distribution data are obtained from the image with adjusted contrast, and the sharpness and color parameters are adjusted by iterative optimization methods to obtain the final optimized image.

6. The remote sensing image enhancement method for green exploration according to claim 1, characterized in that, Also includes: Based on the spatial distribution data of surface feature points obtained from the final optimized image, the initial surface feature distribution map is generated by the Kriging interpolation method to obtain a high-precision distribution map. Based on the high-precision distribution map, feature point coordinates of wetlands and forests are extracted. Density clustering is used to delineate ecologically sensitive areas and determine area boundaries. Feature point density data of wetlands and forests are obtained from the area boundaries. If the density value is lower than a preset threshold, the density distribution is adjusted using a weighted average method to obtain an optimized density distribution map. Spatial correlation data of surface feature points are extracted from the optimized density distribution map, and spatial autocorrelation is calculated using the Moran index analysis method to obtain a spatial correlation map of ecologically sensitive areas. Clustering patterns of feature points of wetlands and forests are obtained from the spatial correlation map. If the clustering pattern deviates from a preset threshold, the feature point distribution is adjusted using a spatial smoothing method to obtain a smoothed correlation map. Boundary feature data of ecologically sensitive areas are extracted from the smoothed correlation map, and vector distribution maps are generated using vector transformation methods to obtain a high-precision vector distribution map. Spatial extent data of ecologically sensitive areas of wetlands and forests are obtained from high-precision vector distribution maps, and the final distribution map is generated using a gridding method, resulting in a distribution map of ecologically sensitive areas.

Images