A lunar surface reflectance remote sensing inversion method based on image enhancement

Abstract

The present application belongs to the technical field of lunar surface reflectivity remote sensing inversion, and discloses a lunar surface reflectivity remote sensing inversion method based on image enhancement. The present application realizes lunar surface remote sensing image enhancement on the lunar surface remote sensing image to be processed by performing enhancement processing on the lunar surface remote sensing image, and the method can enhance the effect of the lunar surface remote sensing image quality enhancement processing. Meanwhile, the optimal end-member mineral spectrum is selected from the spectral library by using a sparse unmixing model to fit the observation spectrum through quantitative analysis of the lunar surface minerals. According to the fitting result, the type, distribution and content of the minerals are inversed, and the quantitative inversion result of the lunar surface minerals is presented in the form of a distribution map of each mineral. The present application is an accurate, efficient and intuitive mineral quantitative inversion method.

Description

Technical Field

[0001] This invention belongs to the field of lunar surface reflectance remote sensing inversion technology, and particularly relates to a method for remote sensing inversion of lunar surface reflectance using image enhancement methods based on remote sensing sampling data from lunar probes. Background Technology

[0002] The Moon is the fifth largest satellite in the solar system, with an average radius of approximately 1737.10 kilometers, equivalent to 0.273 times the radius of Earth; its mass is close to 7.342 × 10⁻⁶. 22 One kilogram, equivalent to 0.0123 times the weight of Earth. The surface of the Moon is covered with impact craters, possibly formed by impacts from small celestial bodies. The average distance between the Moon and Earth is approximately 384,400 kilometers, about 30 times the diameter of Earth. Lunar surface reflectance refers to the reflectivity of the lunar surface to light of various wavelengths. It can reflect the distribution of different materials on the lunar surface, thus revealing the chemical composition and geophysical properties of the lunar surface; however, existing remote sensing inversion methods for lunar surface reflectance produce poor-quality images of the lunar surface; they cannot obtain more detailed information about the lunar surface; and the analysis of lunar surface minerals is inaccurate.

[0003] Based on the above analysis, the problems and shortcomings of the existing technology are as follows:

[0004] (1) Existing lunar surface reflectance remote sensing inversion methods have poor lunar surface remote sensing image quality; they cannot obtain more detailed information about the lunar surface.

[0005] (2) The analysis of minerals on the lunar surface is inaccurate.

[0006] (3) Due to the scarcity of remote sensing data from lunar orbiters, there is a lack of effective remote sensing inversion methods for lunar surface reflectance.

[0007] (4) Inaccurate classification of remote sensing images of the lunar surface affects the inversion results. Summary of the Invention

[0008] To address the problems existing in the prior art, this invention provides a remote sensing inversion method for lunar surface reflectance based on image enhancement.

[0009] This invention is implemented as follows: A remote sensing inversion method for lunar surface reflectance based on image enhancement includes:

[0010] Step 1: Identify the target area on the lunar surface; collect data on the target area on the lunar surface using remote sensing equipment; process and analyze the data; generate remote sensing images of the lunar surface; and enhance the remote sensing images of the lunar surface.

[0011] The method for processing and analyzing data:

[0012] The remote sensing data from the lunar orbiter is broken down into datasets that are easy to identify and classify; the lunar orbiter data is then identified and classified according to predetermined rules to obtain remote sensing metadata.

[0013] The remote sensing metadata is further processed and converted into vector data; the vector data is then edited and processed to obtain effective remote sensing data from the lunar orbiter through analysis;

[0014] Step 2: Classify the remote sensing images; using the DEM and solar azimuth parameters matched with the remote sensing images, including the solar elevation angle, determine the shadows in the remote sensing images, divide the pixels of the lunar surface remote sensing images into two categories: non-shadow pixels and shadow pixels, and use different methods to perform reflectance inversion on each of them.

[0015] The method for classifying remote sensing images:

[0016] High-resolution lunar surface remote sensing images are acquired to obtain high-resolution lunar surface remote sensing images. Land features in these high-resolution lunar surface remote sensing images are labeled to obtain labeled high-resolution lunar surface remote sensing images and a label set. The labeled high-resolution lunar surface remote sensing images are segmented to obtain parent objects, and the parent objects are segmented to obtain child objects, where the label of the child object is the label corresponding to the position of the child object on the parent object. The obtained parent and child objects are standardized, and the standardized parent and child objects are divided into training sample sets, validation sets, and test sample sets. A convolutional neural network model based on the parent and child objects is constructed. The convolutional neural networks of the parent and child objects are merged in fully connected layers. The unstandardized parent and child objects are input into the convolutional neural network models based on the parent and child objects respectively. The features extracted by the convolutional neural networks of the parent and child objects are fused in fully connected layers of the network structure. The label set is used as the output to train the network model, resulting in a trained network model.

[0017] High-resolution lunar surface remote sensing images are acquired. These images are then segmented to obtain parent objects, and superpixel segmentation is performed on the parent objects to obtain child objects. The standardized parent and child objects are then input into convolutional neural networks based on the parent and child objects, respectively, to classify the high-resolution lunar surface remote sensing images. After classification, the child objects with their respective categories are combined with the bounding boxes obtained from superpixel segmentation and their categories are fitted to obtain the classification results of the lunar surface remote sensing images.

[0018] Step 3: For non-shaded pixels, reflectivity is inverted by iteratively calculating direct solar irradiance and adjacent slope reflected irradiance; for shaded pixels, reflectivity is inverted using adjacent slope reflected irradiance; and quantitative analysis of lunar surface minerals is performed.

[0019] Furthermore, the method for enhancing remote sensing images of the lunar surface is as follows:

[0020] (1) Adjust the pixel values ​​of the remote sensing images of the lunar surface to be processed to obtain K lunar surface remote sensing images, wherein the pixel values ​​of the K lunar surface remote sensing images are different, and K is a positive integer greater than 1; perform identification and analysis on the lunar surface;

[0021] (2) Extract the local features of the K lunar surface remote sensing images; extract the global features of the lunar surface remote sensing images to be processed; perform lunar surface remote sensing image enhancement processing on the lunar surface remote sensing images to be processed based on the global features and the local features, and obtain the enhanced lunar surface remote sensing image output lunar surface remote sensing image.

[0022] Furthermore, adjusting the pixel values ​​of the lunar surface remote sensing image to obtain K lunar surface remote sensing images includes:

[0023] The K lunar surface remote sensing images are obtained by increasing or decreasing the pixel values ​​of the lunar surface remote sensing image to be processed through nonlinear transformation.

[0024] Semantic segmentation is performed on the remote sensing image of the lunar surface to be processed to obtain a semantic segmentation map;

[0025] Feature extraction is performed on the semantic segmentation map to obtain semantic conditions, which include the semantic information of the remote sensing image of the lunar surface to be processed;

[0026] Wherein, at least one of the local features of the K lunar surface remote sensing images is extracted based on the semantic conditions.

[0027] Furthermore, at least one of the local features of the K lunar surface remote sensing images is extracted based on a first semantic feature and a second semantic feature, which are determined based on the semantic conditions.

[0028] Furthermore, at least one of the local features of the K lunar surface remote sensing images is determined based on a residual estimate, which is determined based on the first semantic feature, the second semantic feature, and the lunar surface remote sensing image features of the lunar surface remote sensing image to be processed.

[0029] Furthermore, the enhancement method also includes:

[0030] The local features of the K lunar surface remote sensing images are fused to obtain local fused features;

[0031] The step of performing lunar surface remote sensing image enhancement processing on the lunar surface remote sensing image to be processed based on the global features and the local features to obtain an enhanced lunar surface remote sensing image includes:

[0032] The lunar surface remote sensing image to be processed is enhanced based on the global features and the local fusion features to obtain the enhanced lunar surface remote sensing image.

[0033] Further, the step of performing lunar surface remote sensing image enhancement processing on the lunar surface remote sensing image to be processed based on the global features and the local features to obtain an enhanced lunar surface remote sensing image includes:

[0034] The global features and the local features are fused to obtain fused features;

[0035] Based on the fusion features, the lunar surface remote sensing image to be processed is subjected to lunar surface remote sensing image enhancement processing to obtain an enhanced lunar surface remote sensing image.

[0036] Furthermore, the feature fusion of the global features and the local features to obtain fused features includes:

[0037] Feature extraction is performed on the global features to obtain the first global features and the second global features;

[0038] The first global feature and the local feature are added together to obtain the candidate fusion feature;

[0039] The candidate fusion feature and the second global feature are connected and convolutionally processed to obtain the fusion feature.

[0040] The feature fusion of the global features and the local features to obtain fused features includes:

[0041] Feature extraction is performed on the global features to obtain the first global features and the second global features;

[0042] The first global feature and the local feature are concatenated and convolved to obtain candidate fused features;

[0043] The candidate fusion feature and the second global feature are added together to obtain the fusion feature.

[0044] The lunar surface remote sensing image to be processed is the lunar surface remote sensing image obtained after downsampling processing of the original lunar surface remote sensing image to be processed.

[0045] The step of performing lunar surface remote sensing image enhancement processing on the lunar surface remote sensing image to be processed based on the global features and the local features to obtain an enhanced lunar surface remote sensing image includes:

[0046] The original lunar surface remote sensing image of the lunar surface to be processed is subjected to lunar surface remote sensing image enhancement processing based on the global features and the local features to obtain the enhanced lunar surface remote sensing image.

[0047] When the lunar surface remote sensing image to be processed is a lunar surface remote sensing image obtained after downsampling processing of the original lunar surface remote sensing image, the step of performing lunar surface remote sensing image enhancement processing on the lunar surface remote sensing image to be processed according to the fusion feature to obtain an enhanced lunar surface remote sensing image includes:

[0048] Based on the original lunar surface remote sensing image of the lunar surface to be processed, the fusion features are upsampled to obtain the target fusion features;

[0049] Based on the target fusion features, the original lunar surface remote sensing image of the lunar surface to be processed is subjected to lunar surface remote sensing image enhancement processing to obtain the enhanced lunar surface remote sensing image output.

[0050] Furthermore, the method for quantitative analysis of lunar surface minerals is as follows:

[0051] 1) Construct a lunar image set and store the acquired hyperspectral remote sensing images of the lunar surface into the lunar image set; hyperspectral remote sensing images of the lunar surface; establish an endmember spectral database based on typical lunar return samples; calculate the single scattering albedo of the spectral data of the return samples in the hyperspectral remote sensing images of the lunar surface and the endmember spectral database respectively;

[0052] 2) Based on the calculated single scattering albedo, the sparse unmixing algorithm is used to select the optimal end-member mineral spectrum from the end-member spectral database to fit the observed spectrum. Then, the observed spectrum is fitted for each pixel, and the mineral type, distribution and content are inverted based on the fitting results.

[0053] Furthermore, the hyperspectral data of the moon stored in raster format is cropped, geometrically corrected, and mosaicked to obtain a hyperspectral image, which initially reflects the spectral characteristics and mineral distribution characteristics of the example area.

[0054] The single-scattering albedo ωmix can be expressed as a linear combination of the mineral components ωi:

[0055] ω mix (λ i )=F1*ω1(λ i )+F2*ω2(λ i )+…+F j *ω j (λ i )+…+F k *ω k (λ i (1)

[0056] Where K is the number of mineral species in the mixture, λi is the wavelength of the i-th band, and Fj is the relative geometric cross-section of the j-th mineral.

[0057] Before performing spectral unmixing, the single scattering irradiance needs to be calculated using L1 normalization to correct for the scale difference in reflectance between the hyperspectral remote sensing image data of the lunar surface and the experimental endmember spectral data in the endmember spectral database.

[0058] A sparse demixing algorithm is used to find the best subset of spectra from a large endmember library to perform optimal modeling for each mixed pixel in the scene. Then, the observed spectrum is fitted to each pixel, and the mixed spectrum of each pixel is demixed based on the fitting results to determine the type and content of minerals.

[0059] Preferably, the sparse unmixing algorithm needs to perform two solutions. The first solution is to select the main mineral spectra in the hyperspectral data to match the endmember spectral library and determine the main range of minerals. The second solution is to further optimize the endmember mineral spectra to simulate the observed spectra. If the error is within a certain range, the calculated mineral type and content are the final mineral content results.

[0060] After retrieving the types, distribution, and abundance of minerals, the quantitative inversion results of lunar surface minerals are presented as a distribution map for each mineral.

[0061] During the sparse unmixing algorithm solution process:

[0062] The optimal subset of the pixels in the scene is found by using a relatively large endmember spectral library;

[0063] Add a sparsity constraint to the standard linear decomposition model;

[0064] By minimizing their respective Lagrange quantities, they can be transformed into an unconstrained form;

[0065] The spectra of the selected end-member minerals were fitted with hyperspectral data obtained from the lunar probe;

[0066] The type and content of minerals can be inverted based on the fitting results.

[0067] Based on the above technical solutions and the technical problems solved, the advantages and positive effects of the technical solutions protected by this invention are analyzed from the following aspects:

[0068] First, addressing the technical problems existing in the prior art and the difficulty in solving them, and closely combining the technical solution to be protected by this invention with the results and data from the research and development process, this paper provides a detailed and in-depth analysis of how the technical solution of this invention solves the technical problems, and the inventive technical effects brought about after solving the problems. The specific description is as follows:

[0069] This invention enhances lunar surface remote sensing images by adjusting the pixel values ​​of the image to be processed, resulting in multiple lunar surface remote sensing images with different pixel values. These multiple images with different pixel values ​​are then used to enhance the quality of the original lunar surface remote sensing image, thereby extracting more local detail information and achieving lunar surface remote sensing image enhancement. This method effectively improves the quality of lunar surface remote sensing image enhancement. Simultaneously, lunar hyperspectral data is preprocessed using quantitative analysis methods for lunar surface minerals (including cropping, geometric correction, and mosaicking). Next, a typical lunar return sample laboratory spectral library is established. Then, the Hapke model is used to calculate the single-scattering albedo of the observed spectral data and sample spectra. Finally, a sparse unmixing model is used to select the optimal end-member mineral spectrum from the spectral library to fit the observed spectrum. Based on the fitting results, the type, distribution, and content of minerals are inverted. The results are presented as a distribution map of each mineral, demonstrating the quantitative inversion of lunar surface minerals. This invention is an accurate, efficient, and intuitive method for quantitative mineral inversion.

[0070] This invention improves the processing efficiency of lunar probe data by utilizing remote sensing image recognition and classification, as well as the encoding and conversion of remote sensing metadata, through data processing and analysis methods, and obtains effective lunar probe remote sensing data through analysis.

[0071] This invention provides a method for classifying remote sensing images, which can accurately classify remote sensing images of the lunar surface.

[0072] Second, considering the technical solution as a whole or from a product perspective, the technical effects and advantages of the technical solution to be protected by this invention are specifically described as follows:

[0073] This invention enhances lunar surface remote sensing images by adjusting the pixel values ​​of the image to be processed, resulting in multiple lunar surface remote sensing images with different pixel values. These multiple images with different pixel values ​​are then used to enhance the quality of the original lunar surface remote sensing image, thereby extracting more local detail information and achieving lunar surface remote sensing image enhancement. This method effectively improves the quality of lunar surface remote sensing image enhancement. Simultaneously, lunar hyperspectral data is preprocessed using quantitative analysis methods for lunar surface minerals (including cropping, geometric correction, and mosaicking). Next, a typical lunar return sample laboratory spectral library is established. Then, the Hapke model is used to calculate the single-scattering albedo of the observed spectral data and sample spectra. Finally, a sparse unmixing model is used to select the optimal end-member mineral spectrum from the spectral library to fit the observed spectrum. Based on the fitting results, the type, distribution, and content of minerals are inverted. The results are presented as a distribution map of each mineral, demonstrating the quantitative inversion of lunar surface minerals. This invention is an accurate, efficient, and intuitive method for quantitative mineral inversion. Attached Figure Description

[0074] Figure 1 This is a flowchart of the lunar surface reflectance remote sensing inversion method provided in this embodiment of the invention.

[0075] Figure 2 This is a flowchart of a method for enhancing remote sensing images of the lunar surface provided in an embodiment of the present invention.

[0076] Figure 3 This is a flowchart of a method for quantitative analysis of minerals on the lunar surface provided in an embodiment of the present invention. Detailed Implementation

[0077] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0078] I. Explanatory and Illustrative Embodiments. To enable those skilled in the art to fully understand how the present invention is specifically implemented, this section provides an explanatory and illustrative description of the embodiments described in the claims.

[0079] like Figure 1 As shown, the present invention provides a remote sensing inversion method for lunar surface reflectance based on image enhancement, comprising the following steps:

[0080] S101, Identify the target area on the lunar surface; collect data on the target area on the lunar surface using remote sensing equipment; process and analyze the data; generate remote sensing images of the lunar surface; and enhance the remote sensing images of the lunar surface.

[0081] The method for processing and analyzing data:

[0082] The remote sensing data from the lunar orbiter is broken down into datasets that are easy to identify and classify; the lunar orbiter data is then identified and classified according to predetermined rules to obtain remote sensing metadata.

[0083] The remote sensing metadata is further processed and converted into vector data; the vector data is then edited and processed to obtain effective remote sensing data from the lunar orbiter through analysis;

[0084] S102, classify the remote sensing images; using the DEM and solar azimuth parameters matched with the remote sensing images, including the solar elevation angle, determine the shadows in the remote sensing images, divide the pixels of the lunar surface remote sensing images into two categories: non-shadow pixels and shadow pixels, and use different methods to perform reflectance inversion on each of them.

[0085] The method for classifying remote sensing images:

[0086] High-resolution lunar surface remote sensing images are acquired to obtain high-resolution lunar surface remote sensing images. Land features in these high-resolution lunar surface remote sensing images are labeled to obtain labeled high-resolution lunar surface remote sensing images and a label set. The labeled high-resolution lunar surface remote sensing images are segmented to obtain parent objects, and the parent objects are segmented to obtain child objects, where the label of the child object is the label corresponding to the position of the child object on the parent object. The obtained parent and child objects are standardized, and the standardized parent and child objects are divided into training sample sets, validation sets, and test sample sets. A convolutional neural network model based on the parent and child objects is constructed. The convolutional neural networks of the parent and child objects are merged in fully connected layers. The unstandardized parent and child objects are input into the convolutional neural network models based on the parent and child objects respectively. The features extracted by the convolutional neural networks of the parent and child objects are fused in fully connected layers of the network structure. The label set is used as the output to train the network model, resulting in a trained network model.

[0087] High-resolution lunar surface remote sensing images are acquired. These images are then segmented to obtain parent objects, and superpixel segmentation is performed on the parent objects to obtain child objects. The standardized parent and child objects are then input into convolutional neural networks based on the parent and child objects, respectively, to classify the high-resolution lunar surface remote sensing images. After classification, the child objects with their respective categories are combined with the bounding boxes obtained from superpixel segmentation and their categories are fitted to obtain the classification results of the lunar surface remote sensing images.

[0088] S103: For non-shaded pixels, reflectivity is inverted by iteratively calculating direct solar illuminance and adjacent slope reflected irradiance; for shaded pixels, reflectivity is inverted using adjacent slope reflected irradiance; and quantitative analysis of lunar surface minerals is performed.

[0089] like Figure 2 As shown, the method for enhancing remote sensing images of the lunar surface provided by this invention is as follows:

[0090] S201, Adjust the pixel values ​​of the lunar surface remote sensing images to be processed to obtain K lunar surface remote sensing images, wherein the pixel values ​​of the K lunar surface remote sensing images are different, and K is a positive integer greater than 1; Perform identification and analysis on the lunar surface;

[0091] S202, extract local features of the K lunar surface remote sensing images; extract global features of the lunar surface remote sensing images to be processed; perform lunar surface remote sensing image enhancement processing on the lunar surface remote sensing images to be processed based on the global features and the local features, to obtain an enhanced lunar surface remote sensing image.

[0092] The present invention provides a method for adjusting the pixel values ​​of a remote sensing image of the lunar surface to obtain K lunar surface remote sensing images, comprising:

[0093] The K lunar surface remote sensing images are obtained by increasing or decreasing the pixel values ​​of the lunar surface remote sensing image to be processed through nonlinear transformation.

[0094] Semantic segmentation is performed on the remote sensing image of the lunar surface to be processed to obtain a semantic segmentation map;

[0095] Feature extraction is performed on the semantic segmentation map to obtain semantic conditions, which include the semantic information of the remote sensing image of the lunar surface to be processed;

[0096] Wherein, at least one of the local features of the K lunar surface remote sensing images is extracted based on the semantic conditions.

[0097] At least one of the local features of the K lunar surface remote sensing images provided by the present invention is extracted based on a first semantic feature and a second semantic feature, wherein the first semantic feature and the second semantic feature are determined based on the semantic conditions.

[0098] At least one of the local features of the K lunar surface remote sensing images provided by the present invention is determined based on a residual estimate, which is determined based on the first semantic feature, the second semantic feature, and the lunar surface remote sensing image features of the lunar surface remote sensing image to be processed.

[0099] The enhancement method provided by this invention also includes:

[0100] The local features of the K lunar surface remote sensing images are fused to obtain local fused features;

[0101] The step of performing lunar surface remote sensing image enhancement processing on the lunar surface remote sensing image to be processed based on the global features and the local features to obtain an enhanced lunar surface remote sensing image includes:

[0102] The lunar surface remote sensing image to be processed is enhanced based on the global features and the local fusion features to obtain the enhanced lunar surface remote sensing image.

[0103] The present invention provides a method for enhancing a lunar surface remote sensing image based on the global features and the local features, resulting in an enhanced output lunar surface remote sensing image. This method includes:

[0104] The global features and the local features are fused to obtain fused features;

[0105] Based on the fusion features, the lunar surface remote sensing image to be processed is subjected to lunar surface remote sensing image enhancement processing to obtain an enhanced lunar surface remote sensing image.

[0106] The present invention provides a feature fusion method for obtaining fused features by fusing the global features and the local features, including:

[0107] Feature extraction is performed on the global features to obtain the first global features and the second global features;

[0108] The first global feature and the local feature are added together to obtain the candidate fusion feature;

[0109] The candidate fusion feature and the second global feature are connected and convolutionally processed to obtain the fusion feature.

[0110] The feature fusion of the global features and the local features to obtain fused features includes:

[0111] Feature extraction is performed on the global features to obtain the first global features and the second global features;

[0112] The first global feature and the local feature are concatenated and convolved to obtain candidate fused features;

[0113] The candidate fusion feature and the second global feature are added together to obtain the fusion feature.

[0114] The lunar surface remote sensing image to be processed is the lunar surface remote sensing image obtained after downsampling processing of the original lunar surface remote sensing image to be processed.

[0115] The step of performing lunar surface remote sensing image enhancement processing on the lunar surface remote sensing image to be processed based on the global features and the local features to obtain an enhanced lunar surface remote sensing image includes:

[0116] The original lunar surface remote sensing image of the lunar surface to be processed is subjected to lunar surface remote sensing image enhancement processing based on the global features and the local features to obtain the enhanced lunar surface remote sensing image.

[0117] When the lunar surface remote sensing image to be processed is a lunar surface remote sensing image obtained after downsampling processing of the original lunar surface remote sensing image, the step of performing lunar surface remote sensing image enhancement processing on the lunar surface remote sensing image to be processed according to the fusion feature to obtain an enhanced lunar surface remote sensing image includes:

[0118] Based on the original lunar surface remote sensing image of the lunar surface to be processed, the fusion features are upsampled to obtain the target fusion features;

[0119] Based on the target fusion features, the original lunar surface remote sensing image of the lunar surface to be processed is subjected to lunar surface remote sensing image enhancement processing to obtain the enhanced lunar surface remote sensing image output.

[0120] like Figure 3 As shown, the method for quantitative analysis of lunar surface minerals provided by this invention is as follows:

[0121] S301, Construct a lunar image set and store the acquired hyperspectral remote sensing images of the lunar surface into the lunar image set; hyperspectral remote sensing images of the lunar surface; establish an endmember spectral database based on typical lunar return samples; calculate the single scattering albedo of the spectral data of the return samples in the hyperspectral remote sensing images of the lunar surface and the endmember spectral database respectively;

[0122] S302, based on the calculated single scattering albedo, uses a sparse unmixing algorithm to select the optimal end-member mineral spectrum from the end-member spectral database to fit the observed spectrum, and then fits the observed spectrum for each pixel, and inversely derives the mineral type, distribution and content based on the fitting results.

[0123] The present invention provides a method for cropping, geometrically correcting, and mosaicking lunar hyperspectral data stored in raster format to obtain hyperspectral images, which initially reflect the spectral characteristics and mineral distribution characteristics of the example area.

[0124] The single-scattering albedo ωmix can be expressed as a linear combination of the mineral components ωi:

[0125] ω mix (λ i )=F1*ω1(λ i )+F2*ω2(λ i )+…+F j *ω j (λ i )+…+F k *ω k (λ i (1)

[0126] Where K is the number of mineral species in the mixture, λi is the wavelength of the i-th band, and Fj is the relative geometric cross-section of the j-th mineral.

[0127] Before performing spectral unmixing, the single scattering irradiance needs to be calculated using L1 normalization to correct for the scale difference in reflectance between the hyperspectral remote sensing image data of the lunar surface and the experimental endmember spectral data in the endmember spectral database.

[0128] A sparse demixing algorithm is used to find the best subset of spectra from a large endmember library to perform optimal modeling for each mixed pixel in the scene. Then, the observed spectrum is fitted to each pixel, and the mixed spectrum of each pixel is demixed based on the fitting results to determine the type and content of minerals.

[0129] Preferably, the sparse unmixing algorithm needs to perform two solutions. The first solution is to select the main mineral spectra in the hyperspectral data to match the endmember spectral library and determine the main range of minerals. The second solution is to further optimize the endmember mineral spectra to simulate the observed spectra. If the error is within a certain range, the calculated mineral type and content are the final mineral content results.

[0130] After retrieving the types, distribution, and abundance of minerals, the quantitative inversion results of lunar surface minerals are presented as a distribution map for each mineral.

[0131] During the sparse unmixing algorithm solution process:

[0132] The optimal subset of the pixels in the scene is found by using a relatively large endmember spectral library;

[0133] Add a sparsity constraint to the standard linear decomposition model;

[0134] By minimizing their respective Lagrange quantities, they can be transformed into an unconstrained form;

[0135] The spectra of the selected end-member minerals were fitted with hyperspectral data obtained from the lunar probe;

[0136] The type and content of minerals can be inverted based on the fitting results.

[0137] II. Application Examples. To demonstrate the inventiveness and technical value of the technical solution of this invention, this section provides application examples of the technical solution of the claims on specific products or related technologies.

[0138] This invention enhances lunar surface remote sensing images by adjusting the pixel values ​​of the image to be processed, resulting in multiple lunar surface remote sensing images with different pixel values. These multiple images with different pixel values ​​are then used to enhance the quality of the original lunar surface remote sensing image, thereby extracting more local detail information and achieving lunar surface remote sensing image enhancement. This method effectively improves the quality of lunar surface remote sensing image enhancement. Simultaneously, lunar hyperspectral data is preprocessed using quantitative analysis methods for lunar surface minerals (including cropping, geometric correction, and mosaicking). Next, a typical lunar return sample laboratory spectral library is established. Then, the Hapke model is used to calculate the single-scattering albedo of the observed spectral data and sample spectra. Finally, a sparse unmixing model is used to select the optimal end-member mineral spectrum from the spectral library to fit the observed spectrum. Based on the fitting results, the type, distribution, and content of minerals are inverted. The results are presented as a distribution map of each mineral, demonstrating the quantitative inversion of lunar surface minerals. This invention is an accurate, efficient, and intuitive method for quantitative mineral inversion.

[0139] It should be noted that embodiments of the present invention can be implemented using hardware, software, or a combination of both. The hardware portion can be implemented using dedicated logic; the software portion can be stored in memory and executed by a suitable instruction execution system, such as a microprocessor or dedicated-design hardware. Those skilled in the art will understand that the above-described devices and methods can be implemented using computer-executable instructions or included in processor control code, for example, such code provided on a carrier medium such as a disk, CD, or DVD-ROM, a programmable memory such as read-only memory (firmware), or a data carrier such as an optical or electronic signal carrier. The devices and modules of the present invention can be implemented using hardware circuitry such as very large-scale integrated circuits or gate arrays, semiconductors such as logic chips, transistors, or programmable hardware devices such as field-programmable gate arrays, programmable logic devices, etc., or using software executed by various types of processors, or using a combination of the above-described hardware circuitry and software, such as firmware.

[0140] III. Evidence of the Relevant Effects of the Embodiments. The embodiments of the present invention have achieved some positive effects during research and development or use, and indeed possess significant advantages compared to existing technologies. The following description, in conjunction with data, charts, and other materials from the experimental process, illustrates these advantages.

[0141] This invention enhances lunar surface remote sensing images by adjusting the pixel values ​​of the image to be processed, resulting in multiple lunar surface remote sensing images with different pixel values. These multiple images with different pixel values ​​are then used to enhance the quality of the original lunar surface remote sensing image, thereby extracting more local detail information and achieving lunar surface remote sensing image enhancement. This method effectively improves the quality of lunar surface remote sensing image enhancement. Simultaneously, lunar hyperspectral data is preprocessed using quantitative analysis methods for lunar surface minerals (including cropping, geometric correction, and mosaicking). Next, a typical lunar return sample laboratory spectral library is established. Then, the Hapke model is used to calculate the single-scattering albedo of the observed spectral data and sample spectra. Finally, a sparse unmixing model is used to select the optimal end-member mineral spectrum from the spectral library to fit the observed spectrum. Based on the fitting results, the type, distribution, and content of minerals are inverted. The results are presented as a distribution map of each mineral, demonstrating the quantitative inversion of lunar surface minerals. This invention is an accurate, efficient, and intuitive method for quantitative mineral inversion.

[0142] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any modifications, equivalent substitutions, and improvements made by those skilled in the art within the scope of the technology disclosed in the present invention, and within the spirit and principles of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A remote sensing inversion method for lunar surface reflectance based on image enhancement, characterized in that, The lunar surface reflectance remote sensing inversion method includes the following steps: Step 1: Identify the target area on the lunar surface; collect data on the target area on the lunar surface using remote sensing equipment; process and analyze the data; generate remote sensing images of the lunar surface; and enhance the remote sensing images of the lunar surface. The method for processing and analyzing data: The remote sensing data from the lunar orbiter is broken down into datasets that are easy to identify and classify; the lunar orbiter data is then identified and classified according to predetermined rules to obtain remote sensing metadata. The remote sensing metadata is further processed and converted into vector data; the vector data is then edited and processed to obtain effective remote sensing data from the lunar orbiter through analysis; Step 2: Classify the remote sensing images; using the DEM and solar azimuth parameters matched with the remote sensing images, including the solar elevation angle, determine the shadows in the remote sensing images, divide the pixels of the lunar surface remote sensing images into two categories: non-shadow pixels and shadow pixels, and use different methods to perform reflectance inversion on each of them. The method for classifying remote sensing images: High-resolution lunar surface remote sensing images are acquired to obtain high-resolution lunar surface remote sensing images. Land features in these high-resolution lunar surface remote sensing images are labeled to obtain labeled high-resolution lunar surface remote sensing images and a label set. The labeled high-resolution lunar surface remote sensing images are segmented to obtain parent objects, and the parent objects are segmented to obtain child objects, where the label of the child object is the label corresponding to the position of the child object on the parent object. The obtained parent and child objects are standardized, and the standardized parent and child objects are divided into training sample sets, validation sets, and test sample sets. A convolutional neural network model based on the parent and child objects is constructed. The convolutional neural networks of the parent and child objects are merged in fully connected layers. The unstandardized parent and child objects are input into the convolutional neural network models based on the parent and child objects respectively. The features extracted by the convolutional neural networks of the parent and child objects are fused in fully connected layers of the network structure. The label set is used as the output to train the network model, resulting in a trained network model. High-resolution lunar surface remote sensing images are acquired. These images are then segmented to obtain parent objects, and superpixel segmentation is performed on the parent objects to obtain child objects. The standardized parent and child objects are then input into convolutional neural networks based on the parent and child objects, respectively, to classify the high-resolution lunar surface remote sensing images. After classification, the child objects with their respective categories are combined with the bounding boxes obtained from superpixel segmentation and their categories are fitted to obtain the classification results of the lunar surface remote sensing images. Step 3: For non-shaded pixels, reflectivity is inverted by iteratively calculating direct solar irradiance and adjacent slope reflected irradiance; for shaded pixels, reflectivity is inverted using adjacent slope reflected irradiance; and quantitative analysis of lunar surface minerals is performed.

2. The lunar surface reflectance remote sensing inversion method as described in claim 1, characterized in that, The method for enhancing remote sensing images of the lunar surface is as follows: (1) Adjust the pixel values ​​of the remote sensing images of the lunar surface to be processed to obtain K lunar surface remote sensing images, wherein the pixel values ​​of the K lunar surface remote sensing images are different, and K is a positive integer greater than 1; perform identification and analysis on the lunar surface; (2) Extract local features from the K lunar surface remote sensing images; Global features are extracted from the lunar surface remote sensing image to be processed; lunar surface remote sensing image enhancement processing is performed on the lunar surface remote sensing image to be processed based on the global features and the local features to obtain the enhanced lunar surface remote sensing image output.

3. The lunar surface reflectance remote sensing inversion method as described in claim 2, characterized in that, The process of adjusting the pixel values ​​of the lunar surface remote sensing image to obtain K lunar surface remote sensing images includes: The K lunar surface remote sensing images are obtained by increasing or decreasing the pixel values ​​of the lunar surface remote sensing image to be processed through nonlinear transformation. Semantic segmentation is performed on the remote sensing image of the lunar surface to be processed to obtain a semantic segmentation map; Feature extraction is performed on the semantic segmentation map to obtain semantic conditions, which include the semantic information of the remote sensing image of the lunar surface to be processed; Wherein, at least one of the local features of the K lunar surface remote sensing images is extracted based on the semantic conditions.

4. The lunar surface reflectance remote sensing inversion method as described in claim 3, characterized in that, At least one of the local features of the K lunar surface remote sensing images is extracted based on a first semantic feature and a second semantic feature, which are determined based on the semantic conditions.

5. The lunar surface reflectance remote sensing inversion method as described in claim 4, characterized in that, At least one of the local features of the K lunar surface remote sensing images is determined based on a residual estimate, which is determined based on the first semantic feature, the second semantic feature, and the lunar surface remote sensing image features of the lunar surface remote sensing image to be processed.

6. The lunar surface reflectance remote sensing inversion method as described in claim 2, characterized in that, The enhancement process also includes: The local features of the K lunar surface remote sensing images are fused to obtain local fused features; The step of performing lunar surface remote sensing image enhancement processing on the lunar surface remote sensing image to be processed based on the global features and the local features to obtain an enhanced lunar surface remote sensing image includes: The lunar surface remote sensing image to be processed is enhanced based on the global features and the local fusion features to obtain the enhanced lunar surface remote sensing image.

7. The lunar surface reflectance remote sensing inversion method as described in claim 2, characterized in that, The step of performing lunar surface remote sensing image enhancement processing on the lunar surface remote sensing image to be processed based on the global features and the local features to obtain an enhanced lunar surface remote sensing image includes: The global features and the local features are fused to obtain fused features; Based on the fusion features, the lunar surface remote sensing image to be processed is subjected to lunar surface remote sensing image enhancement processing to obtain an enhanced lunar surface remote sensing image.

8. The lunar surface reflectance remote sensing inversion method as described in claim 7, characterized in that, The feature fusion of the global features and the local features to obtain fused features includes: Feature extraction is performed on the global features to obtain the first global features and the second global features; The first global feature and the local feature are added together to obtain the candidate fusion feature; The candidate fusion feature and the second global feature are connected and convolutionally processed to obtain the fusion feature. The feature fusion of the global features and the local features to obtain fused features includes: Feature extraction is performed on the global features to obtain the first global features and the second global features; The first global feature and the local feature are concatenated and convolved to obtain candidate fused features; The candidate fusion feature and the second global feature are added together to obtain the fusion feature. The lunar surface remote sensing image to be processed is the lunar surface remote sensing image obtained after downsampling processing of the original lunar surface remote sensing image to be processed. The step of performing lunar surface remote sensing image enhancement processing on the lunar surface remote sensing image to be processed based on the global features and the local features to obtain an enhanced lunar surface remote sensing image includes: The original lunar surface remote sensing image of the lunar surface to be processed is subjected to lunar surface remote sensing image enhancement processing based on the global features and the local features to obtain the enhanced lunar surface remote sensing image. When the lunar surface remote sensing image to be processed is a lunar surface remote sensing image obtained after downsampling processing of the original lunar surface remote sensing image, the step of performing lunar surface remote sensing image enhancement processing on the lunar surface remote sensing image to be processed according to the fusion feature to obtain an enhanced lunar surface remote sensing image includes: Based on the original lunar surface remote sensing image of the lunar surface to be processed, the fusion features are upsampled to obtain the target fusion features; Based on the target fusion features, the original lunar surface remote sensing image of the lunar surface to be processed is subjected to lunar surface remote sensing image enhancement processing to obtain the enhanced lunar surface remote sensing image output.

9. The lunar surface reflectance remote sensing inversion method as described in claim 1, characterized in that, The method for quantitative analysis of minerals on the lunar surface is as follows: 1) Construct a lunar image set and store the acquired hyperspectral remote sensing images of the lunar surface into the lunar image set; hyperspectral remote sensing images of the lunar surface; establish an endmember spectral database based on typical lunar return samples; calculate the single scattering albedo of the spectral data of the return samples in the hyperspectral remote sensing images of the lunar surface and the endmember spectral database respectively; 2) Based on the calculated single scattering albedo, the sparse unmixing algorithm is used to select the optimal end-member mineral spectrum from the end-member spectral database to fit the observed spectrum. Then, the observed spectrum is fitted for each pixel, and the mineral type, distribution and content are inverted based on the fitting results.

Images