Remote sensing geological prospecting method based on information entropy-partial differential equation combined modeling

Abstract

The application discloses a remote sensing geological prospecting method based on information entropy-partial differential equation joint modeling, and belongs to the technical field of mineral exploration, and comprises the following steps: extracting multiple layers from remote sensing images, wherein the multiple layers comprise a spectral angle mapping layer, an iron staining alteration grading layer, a hydroxyl alteration grading layer and an indicator mineral distribution layer; superimposing the multiple layers to generate a comprehensive map; establishing a target area extraction model, taking the comprehensive map as input and outputting a target area; and vectorizing the target area boundary to delineate the target area position. The remote sensing geological prospecting method based on information entropy-partial differential equation joint modeling realizes mineral spatial distribution visualization innovation, cooperatively applies four mineral identification technologies, quantifies mineral intensity into a transparency parameter, solves the problem of information overload caused by multiple layer superimposition, realizes automatic and rapid identification of a mining area, and has extremely high identification accuracy.

Description

Technical Field

[0001] This invention relates to a remote sensing geological prospecting method using information entropy-partial differential equation joint modeling, belonging to the field of mineral exploration technology. Background Technology

[0002] Traditional mineral exploration methods rely on manual visual interpretation of remote sensing images and ground exploration, which are inefficient and highly subjective. While existing technologies combine remote sensing and machine learning, they still suffer from the following drawbacks:

[0003] 1. The coupling relationship between mineral spectra and spatial features was not considered when overlaying layers;

[0004] 2. Target delineation often relies on a single algorithm, which lacks sensitivity to complex texture areas;

[0005] 3. The output results lack standardized geographic information formats, making it difficult to seamlessly integrate with GIS platforms.

[0006] Therefore, it is necessary to conduct more in-depth research on existing mineral exploration methods in order to solve the above problems. Summary of the Invention

[0007] To overcome the above problems, in-depth research was conducted, and a remote sensing geological prospecting method based on joint modeling of information entropy and partial differential equations was designed, including the following steps:

[0008] S1. Extract multiple layers from remote sensing images, including a spectral angle mapping layer, an iron staining alteration grading layer, a hydroxyl alteration grading layer, and an indicator mineral distribution layer;

[0009] S2. Overlay multiple layers to generate a composite image;

[0010] S3. Establish a target area extraction model, taking the comprehensive map as input and outputting the target area;

[0011] S4. Vectorize the target area boundary to delineate the target area location.

[0012] In a preferred embodiment, the spectral angle filler layer integrates the three-dimensional features of spectral vector angle, mineral abundance, and distribution distance into a single pixel RGB channel.

[0013] In a preferred embodiment, the angle between the spectral vectors Obtained through the following methods:

[0014]

[0015] in, Indicates the USGS spectral library number k Band end-member reflectivity, This indicates the reflectance of ASTER image pixels in the SWIR band. This indicates the total number of bands in the spectral library.

[0016] In a preferred embodiment, the iron staining alteration grading layer divides the satellite image bands into multiple grades, and takes the band ratio of the two middle grades as the iron staining index.

[0017] In the hydroxyl alteration grading layer, the satellite image bands are divided into multiple levels, and the ratio of the bands in the middle two levels is taken as the hydroxyl index.

[0018] The number of levels in the hydroxyl alteration grading layer is the same as that in the iron staining alteration grading layer.

[0019] In a preferred embodiment, the mineral distribution indicator layer selects a combination of bands Band5, Band7, and Band9 as principal components for principal component analysis.

[0020] In a preferred embodiment, in S2, the superposition is represented as:

[0021]

[0022] in, Represents the composite image metadata. Fill the spectral angle map layer values. This is the transparency value. For iron staining and alteration grading layer values, Values ​​for the hydroxyl etching grading layer. This is a layer value indicating the distribution of minerals.

[0023] In a preferred embodiment, in S3, the composite image is preprocessed to remove image noise.

[0024] In a preferred embodiment, the target region extraction model is based on information entropy-partial differential equations, and the model extraction includes the following sub-steps:

[0025] S31. Color space compression reduces data dimensionality;

[0026] S32. Use a sliding window to obtain the entropy value of the window area;

[0027] S33. Compare the region entropy value with the threshold to determine whether the region is a target area.

[0028] In a preferred embodiment, in S31, color space compression is performed using the following encoding method:

[0029]

[0030] in, For compressed colors, The rounding function

[0031] In a preferred embodiment, in S32, the entropy value Represented as:

[0032]

[0033] in, This indicates the window area number 1 k The number of times each color appears This indicates the window area number 1 k The frequency of the occurrence of a color This indicates the size of the sliding window.

[0034] The beneficial effects of this invention include:

[0035] (1) Four mineral identification technologies are applied in a coordinated manner. By using layered modeling, complementary features of land cover are achieved. The iron staining index and the hydroxyl principal component adopt the same level of classification standard to establish cross-layer intensity correlation.

[0036] (2) Innovative pseudo-color mapping, which breaks through the three-dimensional features of spectral angle, mineral abundance and distribution distance into a single pixel RGB channel, realizes a revolution in the visualization of mineral spatial distribution;

[0037] (3) By setting the transparency channel, the mineral intensity is quantified into a transparency parameter to solve the problem of information overload when multiple layers are overlaid;

[0038] (4) Through the four-stage pipeline of "feature extraction → fusion enhancement → target area identification → vectorization output", the mining area can be automatically and quickly identified with high accuracy. Attached Figure Description

[0039] Figure 1 A flowchart of a remote sensing geological prospecting method based on joint modeling of information entropy and partial differential equations according to a preferred embodiment of the present invention is shown.

[0040] Figure 2 The mineral exploration results in Example 1 are shown. Detailed Implementation

[0041] The present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Through these descriptions, the features and advantages of the present invention will become clearer and more apparent.

[0042] The term “exemplary” as used herein means “serving as an example, embodiment, or illustration.” Any embodiment illustrated herein as “exemplary” is not necessarily to be construed as superior to or better than other embodiments. Although various aspects of embodiments are shown in the accompanying drawings, the drawings are not necessarily drawn to scale unless specifically indicated otherwise.

[0043] The remote sensing geological prospecting method based on joint modeling of information entropy and partial differential equations provided by this invention, such as... Figure 1 As shown, it includes the following steps:

[0044] S1. Extract multiple layers from remote sensing images, including a spectral angle mapping layer, an iron staining alteration grading layer, a hydroxyl alteration grading layer, and an indicator mineral distribution layer;

[0045] S2. Overlay multiple layers to generate a composite image;

[0046] S3. Establish a target area extraction model, taking the comprehensive map as input and outputting the target area;

[0047] S4. Vectorize the target area boundary to delineate the target area location.

[0048] In S1, the spectral angle mapping layer is used to describe the different absorption spectra of different minerals to different light, and to reflect the overall distribution of minerals.

[0049] The iron staining alteration grading layer is used to describe the brownish-red, brownish-yellow and other faded alteration areas on the surface formed by the oxidation of iron-bearing minerals. It is one of the commonly used layers in mineral exploration methods.

[0050] The hydroxyl alteration grading layer is used to describe clay minerals and sulfate minerals containing hydroxyl groups on the surface, and is one of the commonly used layers in mineral exploration methods.

[0051] The indicator mineral distribution layer is used to describe the spatial distribution of key minerals.

[0052] In this invention, multiple mineral identification technologies are used in synergy to achieve complementary features of ground features.

[0053] In a preferred embodiment, the spectral angle fill layer integrates the three-dimensional features of spectral vector angle, mineral abundance, and distribution distance into a single pixel RGB channel, as shown below:

[0054]

[0055] in, R, G, B These are the three channels of the image. Indicates the angle between spectral vectors. This represents the smallest angle between spectral vectors in the layer. Indicates mineral abundance. This indicates the highest mineral abundance in the layer.

[0056] According to the present invention, the included angle of the spectral vectors The angle between the satellite image and the standard mineral spectrum represents the similarity between the two; the smaller the angle, the higher the similarity.

[0057] In a preferred embodiment, the angle between the spectral vectors Obtained through the following methods:

[0058]

[0059] in, Indicates the USGS spectral library number k Band end-member reflectivity, This indicates the reflectance of ASTER image pixels in the SWIR band. This indicates the total number of bands in the spectral library.

[0060] The mineral abundance is calculated from hyperspectral data using a spectral unmixing algorithm. In this invention, the specific acquisition method is not limited. For example, the method described in the literature: Lin Honglei, Zhang Xia, Sun Yanli. 2016. Mineral hyperspectral sparse unmixing based on single scattering albedo. Journal of Remote Sensing, 20(1):53-61.

[0061] In a preferred real-time mode, the objective and constraints of the spectral unmixing algorithm are set as follows:

[0062]

[0063]

[0064] in, This is the USGS endmember spectral matrix. This is a mineral abundance vector. For elements in the vector, The index of the vector element. The total number of elements in the vector. is the pixel observation spectral vector.

[0065] The angle between the spectral vectors determined using the above method ,when When the rad is less than 0.1 rad, the mineral identification accuracy is greater than 90%.

[0066] In this invention, the three-dimensional features of spectral vector angle, mineral abundance, and distribution distance are fused into a single pixel RGB channel, realizing three-dimensional visualization of mineral spatial distribution and significantly improving the intuitiveness of interpretation.

[0067] In a preferred embodiment, the iron staining alteration grading layer divides the satellite image bands into multiple grades, and the ratio of the bands of the two middle grades is taken as the iron staining index.

[0068] More preferably, the satellite image bands are divided into 5 levels, namely: :

[0069]

[0070] Preferably, the iron staining index is obtained in the following way:

[0071]

[0072] in, The mean value of the satellite image band. This represents the standard deviation of the satellite image band.

[0073] In this invention, an adaptive threshold based on the standard deviation multiple is used for grading, replacing the traditional fixed threshold, thereby improving adaptability under different geological backgrounds.

[0074] In a preferred embodiment, the hydroxyl alteration grading layer is similar to the iron staining alteration grading layer in that the satellite image bands are divided into multiple levels, and the ratio of the bands of the two middle levels is taken as the hydroxyl index. The hydroxyl alteration grading layer uses the same level division method as the iron staining alteration grading layer.

[0075] In this invention, the same grading method and index acquisition method are used in the hydroxyl alteration grading layer and the iron staining alteration grading layer to establish cross-layer intensity correlation and enhance the consistency of results.

[0076] In a preferred embodiment, unlike traditional geological prospecting, in this invention, the second principal component (PC2) is forcibly selected as the feature component in the hydroxyl alteration grading layer. Traditional prospecting methods using PCA contain a large amount of information, with different principal components highlighting different information. In this invention, specific principal components are selected as feature components based on experience, removing information with low relevance, retaining the useful parts of mineral exploration, highlighting key mineral features, improving efficiency, and overcoming the deficiency of unclear physical meaning in traditional PCA, establishing a strong correlation between principal components and specific mineral types.

[0077] In a preferred embodiment of the present invention, the indicator mineral distribution layer does not use all bands, but selects a combination of bands Band5, Band7, and Band9 as principal components for principal component analysis (PCA). This allows the indicator mineral distribution layer to focus on the third principal component component, focusing on the sensitive bands for ilmenite, thereby highlighting the information of the region richest in the target mineral. This further overcomes the deficiency of unclear physical meaning in traditional PCA and establishes a strong correlation between principal components and specific mineral types.

[0078] Furthermore, the principal component analysis results are mapped using a heatmap to obtain an indicator mineral distribution layer. Areas with higher brightness in the heatmap are more likely to contain the target mineral and have a higher abundance. The mapping function for this heatmap mapping is:

[0079]

[0080] in, Indicates brightness. This represents the third principal component component of the region. This represents the minimum value of the third principal component across all regions. This represents the maximum value of the third principal component component across all regions.

[0081] According to the present invention, the region H>200 corresponds to a direct indication of ilmenite.

[0082] In S2, multiple independent layers from S1 are overlaid and merged. This overlay is represented as follows:

[0083]

[0084] in, Represents the composite image metadata. Fill the spectral angle map layer values. This is the transparency value. For iron staining and alteration grading layer values, Values ​​for the hydroxyl etching grading layer. This is a layer value indicating the distribution of minerals.

[0085] In this invention, by setting transparency value information, mineral intensity is quantified into a transparency parameter and normalized to between 0 and 1, thereby controlling... The transparency of channel pixels solves the problems of information overload and visual confusion caused by multiple layers being overlaid, making minerals and areas of change visible.

[0086] According to the present invention, Channels: High transparency (α→1) highlights areas of strong alteration; half transparency (α→0) backgrounds areas of weak alteration. It indicates what minerals are present at that location. The channel characterizes the degree of mineral enrichment / alteration intensity at that location.

[0087] In step S3, preferably, the composite image is preprocessed to remove image noise.

[0088] In a preferred embodiment, the preprocessing is achieved using an anisotropic diffusion method based on a modified Perona-Malik model.

[0089] The Perona-Malik model is a classic image denoising algorithm based on partial differential equations. It suppresses noise by constructing a diffusion term. However, due to the smoothing effect of the diffusion process, it often over-smooths the edge information of the image, leading to loss of detail. In this invention, the original Perona-Malik model is improved by setting the diffusion coefficient of the original Perona-Malik model as follows:

[0090]

[0091] in, Indicates the diffusion coefficient. This represents the gradient modulus of the composite graph at that point. This is a configurable parameter.

[0092] This diffusion coefficient controls the image smoothing process, with larger gradient values ​​at edges and smaller gradient values ​​in flat regions. Specifically, in flat regions... Enlarged, strongly diffused, smoothing out noise; in the edge region (mineral boundary), i.e. At this time, the diffusion coefficient decreases rapidly, and diffusion almost stops, thus protecting the mineral edges from being blurred and effectively protecting the target area boundary.

[0093] Preferably, This value is an empirical value, and using this value can achieve the best balance between noise suppression and boundary protection.

[0094] The target region extraction model identifies target regions in the composite map based on information entropy-partial differential equations.

[0095] Furthermore, the target region extraction model extraction includes the following sub-steps:

[0096] S31. Color space compression reduces data dimensionality;

[0097] S32. Use a sliding window to obtain the entropy value of the window area;

[0098] S33. Compare the region entropy value with the threshold to determine whether the region is a target area.

[0099] In S31, the composite image is 24-bit true color (16 million colors). This dimension is too high, which will cause the entropy value calculation dimension explosion problem. By compressing the color space and reducing the color color, the entropy value calculation dimension explosion problem can be solved.

[0100] Preferably, the color space is compressed to 16 bits (4096 colors).

[0101] Preferably, color space compression is performed using the following encoding method:

[0102]

[0103] in, For compressed colors, This is the rounding function.

[0104] In S32, the size of the sliding window can be freely set by those skilled in the art according to actual needs, for example, set to 15×15, and the entropy value... Represented as:

[0105]

[0106] in, This indicates the window area number 1 k The number of times each color appears This indicates the window area number 1 k The frequency of the occurrence of a color This indicates the size of the sliding window; for example, a 15x15 window would have a value of 225.

[0107] In S33, all regions with entropy values ​​greater than the threshold are marked as target regions, resulting in a binary target region mask, where the target region is 1 and the non-target region is 0.

[0108] Preferably, the threshold is 2.3.

[0109] In this invention, information entropy can be used to accurately capture the combination complexity of ilmenite and altered minerals.

[0110] In S4, the target area is divided into polygons, and the GDAL tool is used to project the target area into the WGS84 coordinate system to delineate its location.

[0111] Preferably, the target area is segmented into polygons based on the OpenCV algorithm. Example

[0112] Example 1

[0113] Geological prospecting in a certain area includes the following steps:

[0114] S1. Extract multiple layers from remote sensing images, including a spectral angle mapping layer, an iron staining alteration grading layer, a hydroxyl alteration grading layer, and an indicator mineral distribution layer;

[0115] S2. Overlay multiple layers to generate a composite image;

[0116] S3. Establish a target area extraction model, taking the comprehensive map as input and outputting the target area;

[0117] S4. Vectorize the target area boundary to delineate the target area location.

[0118] In S1, the spectral angle filling layer fuses the three-dimensional features of spectral vector angle, mineral abundance, and distribution distance into a single pixel RGB channel, as shown below:

[0119]

[0120] The angle between the spectral vectors Obtained through the following methods:

[0121]

[0122] The spectral unmixing objective and constraints are set as follows:

[0123]

[0124]

[0125]

[0126] In the iron staining and alteration grading layer, the satellite image bands are divided into 5 levels, namely: :

[0127]

[0128] The iron staining index is obtained through the following methods:

[0129]

[0130] The hydroxyl etching grading layer is the same as the iron staining etching grading layer.

[0131] The mineral distribution layer selects a combination of bands Band5, Band7, and Band9 as principal components. The principal component analysis results are then mapped using a heatmap. The mapping function for this heatmap mapping is:

[0132]

[0133] In S2, the superposition is represented as:

[0134]

[0135]

[0136] In S3, the composite image is preprocessed using an improved Perona-Malik model to remove image noise, with the diffusion coefficient set as follows:

[0137]

[0138] in, .

[0139] The target region extraction model includes the following sub-steps:

[0140] S31. Color space compression reduces data dimensionality;

[0141] S32. Use a sliding window to obtain the entropy value of the window area;

[0142] S33. Compare the region entropy value with the threshold to determine whether the region is a target area.

[0143] In S31, color space compression is performed using the following encoding method, compressing the color space to 16 bits:

[0144]

[0145] In S32, the size of the sliding window is set to 15×15, and the entropy value... Represented as:

[0146]

[0147] In S33, all regions with entropy values ​​greater than a threshold are marked as target regions to obtain a binary target region mask, where the target region is 1 and the non-target region is 0, and the threshold is 2.3.

[0148] In S4, the target area is divided into polygons, and the GDAL tool is used to project the polygons into the WGS84 coordinate system to delineate the target area location. The result is as follows: Figure 2 As shown.

[0149] Ground sampling was conducted at the target area location for verification, with an accuracy of approximately 85%. The main target area has been recommended as a priority exploration area. The entire process took approximately 23.5 minutes (based on 500 km², with an Intel Xeon Gold 6230 reference efficiency of 4.7 minutes / 100 km²), and the alteration zone identification error was less than 30 meters.

[0150] In the description of this invention, it should be noted that the terms "upper," "lower," "inner," "outer," "front," and "rear," etc., indicate the orientation or positional relationship based on the orientation or positional relationship in the working state of this invention, and are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of this invention. Furthermore, the terms "first," "second," "third," and "fourth" are used for descriptive purposes only and should not be construed as indicating or implying relative importance.

[0151] In the description of this invention, it should be noted that, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal communication between two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0152] The present invention has been described above with reference to preferred embodiments; however, these embodiments are merely exemplary and illustrative. Various substitutions and modifications can be made to the present invention based on these embodiments, all of which fall within the scope of protection of the present invention.

Claims

1. A remote sensing geological prospecting method based on joint modeling of information entropy and partial differential equations, characterized in that, Includes the following steps: S1. Extract multiple layers from remote sensing images, including a spectral angle mapping layer, an iron staining alteration grading layer, a hydroxyl alteration grading layer, and an indicator mineral distribution layer; S2. Overlay multiple layers to generate a composite image; S3. Establish a target area extraction model, taking the comprehensive map as input and outputting the target area; S4. Vectorize the target area boundary to delineate the target area location; In the spectral angle filling layer, the three-dimensional features of spectral vector angle, mineral abundance, and distribution distance are fused into a single pixel RGB channel; In the iron staining and alteration grading layer, the satellite image bands are divided into multiple levels, and the ratio of the bands of the two middle levels is taken as the iron staining index. In the hydroxyl alteration grading layer, the satellite image bands are divided into multiple levels, and the ratio of the bands in the middle two levels is taken as the hydroxyl index. Among them, the number of levels in the hydroxyl alteration grading layer is the same as that in the iron staining alteration grading layer; In S2, the superposition is represented as: in, Represents the composite image metadata. Fill the spectral angle map layer values. This is the transparency value. For iron staining and alteration grading layer values, Values ​​for the hydroxyl etching grading layer. Values ​​for indicating mineral distribution layers; In S3, the target region extraction model is based on information entropy-partial differential equations, and the extraction process includes the following sub-steps: S31. Color space compression reduces data dimensionality; S32. Use a sliding window to obtain the entropy value of the window area; S33. Compare the region entropy value with the threshold to determine whether the region is a target area.

2. The remote sensing geological prospecting method based on joint modeling of information entropy and partial differential equations according to claim 1, characterized in that, The angle between the spectral vectors Obtained through the following methods: in, Indicates the USGS spectral library number k Band end-member reflectivity, This indicates the reflectance of ASTER image pixels in the SWIR band. This indicates the total number of bands in the spectral library.

3. The remote sensing geological prospecting method based on joint modeling of information entropy and partial differential equations according to claim 1, characterized in that, The mineral distribution layer was analyzed by selecting a combination of bands Band5, Band7, and Band9 as principal components.

4. The remote sensing geological prospecting method based on joint modeling of information entropy and partial differential equations according to claim 1, characterized in that, In S3, the composite image is preprocessed to remove image noise.

5. The remote sensing geological prospecting method based on joint modeling of information entropy and partial differential equations according to claim 1, characterized in that, In S31, color space compression is performed using the following encoding method: in, For compressed colors, This is the rounding function.

6. The remote sensing geological prospecting method based on joint modeling of information entropy and partial differential equations according to claim 1, characterized in that, In S32, the entropy value Represented as: in, This indicates the window area number 1 k The number of times each color appears This indicates the window area number 1 k The frequency of the occurrence of a color This indicates the size of the sliding window.

Images