Uranium ore altered mineral automatic extraction method for satellite-borne hyperspectral data

Abstract

The invention belongs to the technical field of space flight and aviation remote sensing geological exploration, and particularly relates to a uranium ore altered mineral automatic extraction method for spaceborne hyperspectral data, which comprises the following steps of: 1, preprocessing the spaceborne hyperspectral data and optimizing wavebands; 2, constructing a multi-dimensional spectrum geometric feature set; 3, constructing a'net mineralization 'coupling model; step 4, carrying out adaptive threshold segmentation based on histogram curvature extreme values; and 5, morphological filtering post-processing is carried out, and a uranium ore altered mineral distribution diagram is output. According to the method, full-automatic and high-precision extraction of alteration information is achieved by constructing a positive and negative feature coupling model and combining histogram curvature extreme value analysis, and the problems that in existing hyperspectral mineral extraction, manual repeated threshold adjustment is seriously relied on, multi-index combined logic is rigid, and complex background interference is difficult to adapt are effectively solved.

Description

Technical Field

[0001] This invention belongs to the field of aerospace remote sensing geological exploration technology, specifically relating to an automatic extraction method for alteration minerals in uranium deposits from spaceborne hyperspectral data. Background Technology

[0002] In uranium geological exploration, extracting information on wall rock alteration is a crucial method for delineating prospective mineralization areas. The associated alteration minerals (such as illite, kaolinite, and chlorite) exhibit characteristic spectral absorption peaks in the short-wave infrared (SWIR) band. With the widespread availability of hyperspectral data from domestic satellites such as GF-5 and ZY-1 02D, it has become possible to utilize their continuous bands for fine mineral mapping. However, existing technologies face the following major problems in practical engineering applications:

[0003] 1. Subjectivity and Inefficiency of Manual Threshold Setting: Traditional methods such as spectral angle mapping (SAM), matched filtering (MF), or single spectral index methods often produce abundance or index maps with non-standard distributions. Due to terrain cutting, uneven illumination, and the complexity of surface cover, background values ​​vary greatly in different areas of the same image. Technicians typically need to manually set segmentation thresholds repeatedly through visual interpretation. If the threshold is too high, weak alteration information will be missed; if the threshold is too low, not only will a large amount of noise be introduced, but non-mineral background will also be mistakenly identified as anomalies. This "trial and error" method is not only inefficient, but the results also heavily depend on the operator's experience and lack standardization and repeatability.

[0004] 2. Limitations of Simple Multi-Index Joint Constraints: To address the issue of false positives from single indicators, existing technologies often employ a method of "joint constraints from multiple spectral indices" (e.g., index A > threshold 1 and index B < threshold 2). This hard segmentation method based on Boolean logic has two drawbacks: first, it requires simultaneous manual adjustment of multiple thresholds, resulting in an exponential increase in operational complexity; second, this "one-size-fits-all" logic cannot handle mixed pixels in transitional zones, and it is prone to "false positives" for pixels with weak signals but significant potential for mineral exploration, failing to achieve the organic fusion and complementarity of multi-dimensional features.

[0005] 3. Poor data source adaptability: Existing automatic thresholding algorithms typically assume that the image histogram has a bimodal distribution (the background and the target each occupy a certain proportion). However, in uranium exploration, altered minerals are often sparsely distributed in star-shaped or banded patterns, which are typical low-probability targets, resulting in the histogram not having a clear bimodal distribution, making it difficult to directly apply traditional automatic algorithms.

[0006] Therefore, there is an urgent need for an automated extraction method that can automatically fuse spectral geometric features and background interference information based on the characteristics of spaceborne hyperspectral data, and can adaptively determine the optimal segmentation threshold. Summary of the Invention

[0007] The purpose of this invention is to provide an automatic extraction method for alteration minerals in uranium ore from spaceborne hyperspectral data. This method achieves fully automatic and high-precision extraction of alteration information by constructing a "positive and negative feature coupling model" and combining it with "histogram curvature extreme value analysis". It effectively solves the problems of existing hyperspectral mineral extraction, such as heavy reliance on repeated manual threshold adjustment, rigid multi-index joint logic, and difficulty in adapting to complex background interference.

[0008] Technical solution to achieve the purpose of this invention:

[0009] An automated method for extracting alteration minerals from spaceborne hyperspectral data includes:

[0010] Step 1: Preprocessing and band selection of spaceborne hyperspectral data;

[0011] Step 2: Construct a multidimensional spectral geometric feature set;

[0012] Step 3: Construct a coupled model of "net mineralization";

[0013] Step 4: Adaptive threshold segmentation based on histogram curvature extrema;

[0014] Step 5: Morphological filtering post-processing, outputting a distribution map of uranium ore alteration minerals.

[0015] Furthermore, step one includes: performing strip noise repair and atmospheric correction on the spaceborne hyperspectral remote sensing data; extracting an effective subset of bands in the shortwave infrared region related to uranium ore alteration; and using envelope removal technology to convert reflectivity data into absorption depth data to eliminate the background baseline effect caused by terrain illumination.

[0016] Furthermore, in step one, the formula for converting reflectance data into absorption depth data is as follows:

[0017]

[0018] Where CR(λ) is the normalized absorption depth spectrum; R(λ) is the reflectance spectrum; R envelope This represents the outer envelope value of the spectrum.

[0019] Furthermore, step two includes:

[0020] Step 2.1: Extract positive mineralization features: Based on the spectral data after envelope removal, calculate the absorption depth D and absorption symmetry S of the target mineral's characteristic absorption position; use the absorption depth D and absorption symmetry S as continuous variables to quantitatively characterize the spectral geometry of the mineral; the absorption depth D and absorption symmetry S together constitute a positive feature set, which is used to enhance the mineralization signal in the subsequent coupled model.

[0021] Step 2.2: Extract negative disturbance features: Calculate the vegetation index NDVI and soil moisture uptake index H2O for the main disturbance sources. idx The calculated exponential value is directly defined as the background interference intensity factor and input as a suppression term into the subsequent "net mineralization" coupling model.

[0022] Furthermore, in step 2.1, the formula for calculating the absorption depth D is:

[0023] D = 1 - CR(λ) min )

[0024] Where D represents the absorption depth; λ min CR(λ) represents the minimum point within the range of characteristic absorption locations of the target mineral; min ) represents λ min Normalized absorption depth spectrum at location;

[0025] The formula for calculating the absorption symmetry S is:

[0026]

[0027] Where S represents absorption symmetry; W L Indicates the number of bands from the left shoulder of the absorption valley to the bottom of the valley; W R This indicates the number of wave bands from the bottom of the valley to the right shoulder.

[0028] Furthermore, step three includes: generating a "net mineralization probability image" (I) based on a weighted coupling formula for background suppression. net A "positive and negative feature coupling model" is constructed to enhance positive mineralization features through mathematical operations while suppressing negative interference features.

[0029] Furthermore, in step three, the formula for calculating the net mineralization confidence image is as follows:

[0030]

[0031] Among them, I net denoted as Net Mineralization Confidence Image, D represents Absorption Depth; f(S) represents Symmetry Weighting Function, α represents Vegetation Disturbance Inhibition Coefficient, used to adjust the penalty intensity for areas with high vegetation cover, and its value is usually greater than 1; β represents Soil Moisture Inhibition Coefficient, used to adjust the penalty intensity for areas with high humidity.

[0032] Furthermore, step four includes: statistical analysis. net The grayscale histogram of the image is used to calculate the curvature of the histogram curve. The point of maximum curvature at the falling edge of the logarithmic domain histogram is then used as the optimal segmentation threshold.

[0033] Furthermore, in step four, the formula for calculating the curvature of the histogram curve is:

[0034]

[0035] Where K(x) represents curvature; L′ represents the first derivative (slope / rate of change); and L” represents the second derivative (curvature / degree of bending).

[0036] Furthermore, step five includes: generating a binary image based on the step segmentation threshold, performing mathematical morphological opening operations on the binary image to remove isolated noise pixels, retaining mineralized patches with spatial continuity, converting the binary image into a vector polygon, removing fragmented patches, and outputting the final uranium ore alteration mineral distribution map.

[0037] The beneficial technical effects of this invention are as follows:

[0038] 1. The present invention provides an automatic extraction method for altered minerals of uranium ore from spaceborne hyperspectral data, which achieves full automation and requires no manual intervention: the optimal threshold is automatically locked through histogram curvature analysis, which completely solves the problem of repeated trial and error adjustment of thresholds in traditional methods and greatly improves work efficiency.

[0039] 2. The present invention provides an automatic extraction method for alteration minerals of uranium ore from spaceborne hyperspectral data, which has strong anti-interference ability: the proposed coupling model organically integrates "positive spectral features" and "negative environmental interference" at the mathematical level, which is more robust than simple multi-exponential Boolean logic and effectively reduces the false alarm rate caused by vegetation and water.

[0040] 3. The present invention provides an automatic extraction method for alteration minerals of uranium ore from spaceborne hyperspectral data, which has excellent adaptability and universality: the algorithm does not rely on empirical parameters of a specific region, but adaptively calculates the threshold according to the statistical distribution characteristics of each image, and is particularly suitable for batch processing of large-area, cross-regional spaceborne hyperspectral (GF-5 / 02D) data. Detailed Implementation

[0041] The present invention will be further described in detail below with reference to the embodiments.

[0042] This invention provides an automated method for extracting alteration minerals from spaceborne hyperspectral data, specifically comprising the following steps:

[0043] Step 1: Preprocessing and band selection of spaceborne hyperspectral data

[0044] For spaceborne hyperspectral remote sensing data (GF-5 or ZY-1 02D data), strip noise restoration and atmospheric correction are performed; for the short-wave infrared region (2.0μm-2.5μm) related to uranium ore alteration, an effective band subset is extracted; envelope removal technology is used to convert reflectivity data into absorption depth data to eliminate the background baseline influence caused by topographic illumination.

[0045] The formula for converting reflectance data into absorption depth data is as follows:

[0046]

[0047] Where CR(λ) is the normalized absorption depth spectrum; R(λ) is the reflectance spectrum; R envelope This represents the outer envelope value of the spectrum.

[0048] Step 2: Construct a multidimensional spectral geometric feature set

[0049] Instead of directly using band ratios, the geometric features of the spectral curves are extracted based on mineral crystal field theory.

[0050] Step 2.1: Extract positive mineralization features (enhanced signal)

[0051] Based on the spectral data after envelope removal, the absorption depth (D) and absorption symmetry (S) at the characteristic absorption positions of the target mineral (e.g., 2200 nm) are calculated. Instead of directly setting thresholds for binarization extraction, D and S are used as continuous variables to quantitatively characterize the spectral geometry of the mineral.

[0052] Absorption depth D: Characterizes the relative abundance of the target mineral; the larger the value, the stronger the signal. Absorption symmetry S: Characterizes the geometric similarity between the pixel spectrum and the standard mineral endmember spectrum; the closer the value is to 1, the more regular the morphology and the higher the mineralization confidence. Together, they constitute a positive feature set, used to enhance the mineralization signal in subsequent coupled models.

[0053] The formula for calculating the absorption depth (D) is:

[0054] D = 1 - CR(λ) min )

[0055] Where D represents the absorption depth; λ min CR(λ) represents the minimum point within the range of characteristic absorption locations of the target mineral; min ) represents λ min Normalized absorption depth spectrum of position.

[0056] The formula for calculating absorption symmetry (S) is:

[0057]

[0058] Where S represents absorption symmetry; W L Indicates the number of bands from the left shoulder of the absorption valley to the bottom of the valley; W R This indicates the number of wave bands from the bottom of the valley to the right shoulder.

[0059] Step 2.2: Extract negative interference features (suppress background): Calculate the vegetation index (NDVI) and soil moisture uptake index (H2O) for the main interference sources. idx The extraction logic is as follows: the calculated index values ​​are directly defined as background interference intensity factors. These two indices are not used for direct masking removal, but are instead input as suppression terms into the subsequent "net mineralization" coupled model.

[0060] A higher NDVI value indicates stronger vegetation cover disturbance; H2O idx The higher the value, the stronger the soil moisture interference. Subsequent algorithms use weighted suppression (i.e., reduce the probability of pixels with high interference intensity factors being identified as mineralization points) to achieve interference-resistant extraction.

[0061] Step 3: Construct a coupled model of "net mineralization"

[0062] Abandoning the traditional "AND / OR" logic, a weighted coupling formula based on background suppression is used to generate a "net mineralization probability image" (I). net The "positive and negative feature coupling model" is constructed to enhance positive mineralization features (numerator) through mathematical operations, while suppressing negative interference features (denominator).

[0063] The formula for calculating the net mineralization probability image is as follows:

[0064]

[0065] Among them, I net denoted as Net Mineralization Confidence Image, D represents Absorption Depth; f(S) represents Symmetry Weighting Function, α represents Vegetation Disturbance Inhibition Coefficient, used to adjust the penalty intensity for areas with high vegetation cover, and its value is usually greater than 1; β represents Soil Moisture Inhibition Coefficient, used to adjust the penalty intensity for areas with high humidity.

[0066] The model automatically suppresses the values ​​of high-vegetation and high-moisture areas through mathematical calculations, while enhancing the values ​​of mineralized areas with typical absorption characteristics, compressing multi-dimensional features into a single-dimensional probability space.

[0067] Step 4: Adaptive threshold segmentation based on histogram curvature extrema

[0068] Statistics I netThe image's grayscale histogram. Due to the sparse distribution of mineralized points, the histogram typically exhibits a long tail. This invention does not rely on the bimodal assumption, but instead calculates the second derivative (curvature) of the histogram curve to find the point of maximum curvature at the descending edge of the logarithmic domain histogram. This point statistically represents the critical inflection point between random background noise and structured mineral anomaly signals, and is used as the optimal segmentation threshold T to achieve fully automatic binarization segmentation.

[0069] The formula for calculating the curvature of a histogram curve is:

[0070]

[0071] Where K(x) represents curvature; l′ represents the first derivative (slope / rate of change); and L″ represents the second derivative (curvature / degree of bending).

[0072] Step 5: Morphological filtering post-processing, outputting a distribution map of uranium ore alteration minerals.

[0073] A binary image is generated based on the segmentation threshold obtained in step four. Mathematical morphological opening is performed on the binary image to remove isolated noise pixels and retain mineralized patches with spatial continuity. The binary image is then converted into a vector polygon, and small fragmented patches are removed to output the final uranium ore alteration mineral distribution map.

[0074] Example 1

[0075] Taking the hyperspectral satellite data from China's Gaofen-5 (GF-5) as an example, and considering the relatively low signal-to-noise ratio and significant stripe noise in its shortwave infrared (SWIR) band, this invention elaborates on an automatic extraction method for alteration minerals of uranium ore from spaceborne hyperspectral data, specifically including the following steps:

[0076] Step 1: Refined preprocessing and band selection of spaceborne hyperspectral data

[0077] For the GF-5 AHSI data (330 bands in total), invalid bands severely affected by water vapor absorption were first removed. The short-wave infrared region (2000nm-2500nm) containing characteristic absorption peaks of uranium-bearing alteration minerals (such as illite, kaolinite, and chlorite) was selected as the processing subset. The specific procedures are as follows:

[0078] Step 1.1: Strip Repair

[0079] To address the common vertical stripe noise in the GF-5SWIR band, instead of using global smoothing that destroys spectral characteristics, we employ moment matching or noise reconstruction methods based on MNF (Minimum Noise Separation) transform to directionally repair bad lines.

[0080] Step 1.2: Spectral Smoothing

[0081] To preserve the subtle absorption characteristics of minerals and suppress random noise, Savitzky-Golay filtering was used, with a sliding window size of 5-7 bands and a polynomial order of 2.

[0082] Step 1.3: Envelope Removal: Perform continuum removal on the spectral curve of each pixel, transforming the reflectance spectrum R(λ) into the normalized absorption depth spectrum CR(λ), using the following formula:

[0083]

[0084] Where CR(λ) is the normalized absorption depth spectrum, R(λ) is the reflectance spectrum, and R envelope This is the outer envelope value of the spectrum, thereby eliminating the difference in background base values ​​caused by terrain shadows and uneven illumination.

[0085] Step 2: Construct a multidimensional spectral geometric feature set

[0086] This invention differs from the traditional, simple band ratio method. Instead, it is based on mineral crystal field theory, constructing a feature set from a "morphological" perspective. Taking illite extraction (with its main characteristic absorption peaks around 2190nm-2210nm) as an example:

[0087] Step 2.1: Extract positive mineralization features (signal enhancement)

[0088] Absorption depth (D): Finding the minimum point λ in the 2180nm-2220nm range. min Define D = 1 - CR(λ) min ).

[0089] Absorption symmetry (S): Calculates the number of bands W from the left shoulder to the bottom of the absorption valley. L The number of wave bands W from the trough to the right shoulder. R The ratio. The formula is defined as:

[0090]

[0091] Physical significance: Absorption valleys of standard minerals usually have specific symmetry (S approaches 1), while absorption pits caused by noise or interference are often distorted in shape.

[0092] Step 2.2: Extract negative interference features (background suppression)

[0093] Vegetation disturbance (V): NDVI was calculated using the 850nm and 660nm bands.

[0094] Dark pixel interference (M): Statistically measure the average reflectance of the SWIR band, set a low-value mask, and identify water bodies or deep shadow areas.

[0095] Step 3: Construct a coupled model of "net mineralization"

[0096] To address the boundary effects caused by traditional multi-exponential "hard threshold" cutting, this invention proposes a continuous positive and negative feature coupling formula. The core of this step lies in compressing multi-dimensional features into a single "mineralization confidence" image.

[0097] Constructing a net mineralization probability image (I net )as follows:

[0098]

[0099] Among them: I net The image shows the net mineralization probability. D is the absorption depth, representing the basic signal of mineral abundance; S is the symmetry; γ is the morphological weighting factor (recommended value 1.5-2.0), used to amplify the penalty for asymmetric noise; α is the vegetation suppression coefficient (recommended value 2.0-5.0). The larger the α value, the stronger the suppression of vegetation cover.

[0100] Implementation Results: Using this formula, the IV of high-vegetation areas (high NDVI) and noise-prone areas (low S) can be compared. net The value will be automatically lowered to near 0, while the value of the actual alteration zone is preserved and enhanced.

[0101] Step 4: Adaptive threshold segmentation based on histogram curvature extrema

[0102] This is the core step in achieving "automation" in this invention. (Regarding I) net Instead of manually setting thresholds for images, algorithms automatically find the critical points between background and anomalies.

[0103] Step 4.1: Histogram Statistics and Smoothing

[0104] Statistical Full Chart I net The grayscale histogram H(x) is calculated. Since mineralization points are often low-probability events, the histogram exhibits an "L-shaped" or "long-tailed" distribution. To ensure stable calculation, the logarithm of the histogram frequency is first taken: L(x) = ln(H(x) + ∈).

[0105] Where L(x) is the logarithmic transformation, and x represents the "net mineralization probability image" (I) generated in step 3. net The gray value in the graph (i.e., a specific mineralization index value); H(x) represents the original gray histogram function, i.e., the number of pixels (frequency) with a gray value of x in the whole graph; ln represents the natural logarithm operation; ∈ represents the numerical stability constant, which takes the value of a very small positive number (preferably ∈ = 1 in this embodiment).

[0106] Step 4.2: Finding the inflection point

[0107] Calculate the curvature or second difference of the L(x) curve. Define the curvature K(x) as:

[0108]

[0109] Where K(x) represents curvature; L′ represents the first derivative (slope / rate of change); and L″ represents the second derivative (curvature / degree of bending).

[0110] Find the maximum value of K(x) on the right descending edge of the histogram (representing the high-value outlier area).

[0111] Step 4.3: Determine the threshold

[0112] The x-value corresponding to this maximum point is the optimal segmentation threshold T. opt .

[0113] Physical principle: This point corresponds to the turning point in statistics from "Gaussian background noise" to "long-tailed mineralization anomaly". Using this as a cut-off point can ensure the extraction rate while suppressing the background to the greatest extent.

[0114] Step 5: Morphological filtering post-processing

[0115] Using the threshold T obtained in step 4 opt Generate a binary image BW. At this point, sporadic "salt-and-pepper noise" may still exist in the image. Apply mathematical morphology operations to impose spatial constraints:

[0116] Step 5.1: Opening Operation

[0117] First, erosion is performed using 3×3 structuring elements to remove isolated noise points smaller than 3×3 pixels; then, dilation is performed to restore the boundary of the main ore body.

[0118] Physical significance: Surface anomalies supported by deep structural channels were screened out, while isolated anomalies far from fault zones were eliminated.

[0119] Step 5.2: Vectorize the output

[0120] The processed binary image is converted into a vector polygon, and small patches with an area smaller than a specified threshold (such as the area of ​​5 pixels) are removed. The final output is a distribution map of uranium alteration minerals that conforms to geological mapping standards.

[0121] The present invention has been described in detail above with reference to the embodiments. However, the present invention is not limited to the above embodiments. Within the scope of knowledge possessed by those skilled in the art, various changes can be made without departing from the spirit of the present invention. All contents not described in detail in the present invention can be derived from existing technologies.

Claims

1. An automated method for extracting alteration minerals from spaceborne hyperspectral data, characterized in that, include: Step 1: Preprocessing and band selection of spaceborne hyperspectral data; Step 2: Construct a multidimensional spectral geometric feature set; Step 3: Construct a coupled model for "net mineralization"; Step 4: Adaptive threshold segmentation based on histogram curvature extrema; Step 5: Morphological filtering post-processing, outputting a distribution map of uranium ore alteration minerals.

2. The method for automatic extraction of alteration minerals from spaceborne hyperspectral data according to claim 1, characterized in that, Step one includes: performing strip noise repair and atmospheric correction on the spaceborne hyperspectral remote sensing data; extracting an effective subset of bands in the shortwave infrared region related to uranium ore alteration; and using envelope removal technology to convert reflectivity data into absorption depth data to eliminate the background baseline effect caused by terrain illumination.

3. The method for automatic extraction of alteration minerals from spaceborne hyperspectral data according to claim 2, characterized in that, In step one, the formula for converting reflectance data into absorption depth data is as follows: Where CR(λ) is the normalized absorption depth spectrum; R(λ) is the reflectance spectrum; R envelope This represents the outer envelope value of the spectrum.

4. The method for automatic extraction of alteration minerals from spaceborne hyperspectral data according to claim 3, characterized in that, Step two includes: Step 2.1: Extract positive mineralization features: Based on the spectral data after envelope removal, calculate the absorption depth D and absorption symmetry S of the target mineral's characteristic absorption position; use the absorption depth D and absorption symmetry S as continuous variables to quantitatively characterize the spectral geometry of the mineral; the absorption depth D and absorption symmetry S together constitute a positive feature set, which is used to enhance the mineralization signal in the subsequent coupled model. Step 2.2: Extract negative disturbance features: Calculate the vegetation index NDVI and soil moisture uptake index H2O for the main disturbance sources. idx The calculated exponential value is directly defined as the background interference intensity factor and input as a suppression term into the subsequent "net mineralization" coupling model.

5. The automatic extraction method for alteration minerals of uranium ore from spaceborne hyperspectral data according to claim 4, characterized in that, In step 2.1, the formula for calculating the absorption depth D is: D=1-CR(λ min ) Where D represents the absorption depth; λ min CR(λ) represents the minimum point within the range of characteristic absorption locations of the target mineral; min ) represents λ min Normalized absorption depth spectrum at location; The formula for calculating the absorption symmetry S is: Where S represents absorption symmetry; W L Indicates the number of bands from the left shoulder of the absorption valley to the bottom of the valley; W R This indicates the number of wave bands from the bottom of the valley to the right shoulder.

6. The automatic extraction method for alteration minerals of uranium ore from spaceborne hyperspectral data according to claim 5, characterized in that, Step three includes: generating a "net mineralization probability image" based on a weighted coupling formula for background suppression, constructing a "net mineralization" coupling model, enhancing positive mineralization features through mathematical operations, and suppressing negative interference features at the same time.

7. The method for automatic extraction of alteration minerals from spaceborne hyperspectral data according to claim 6, characterized in that, In step three, the formula for calculating the net mineralization probability image is as follows: Among them, I net denoted as Net Mineralization Confidence Image, D represents Absorption Depth; f(S) represents Symmetry Weighting Function, α represents Vegetation Disturbance Inhibition Coefficient, used to adjust the penalty intensity for areas with high vegetation cover, and its value is usually greater than 1; β represents Soil Moisture Inhibition Coefficient, used to adjust the penalty intensity for areas with high humidity.

8. The automatic extraction method for alteration minerals of uranium ore from spaceborne hyperspectral data according to claim 7, characterized in that, Step four includes: statistical analysis. net The grayscale histogram of the image is used to calculate the curvature of the histogram curve. The point of maximum curvature at the falling edge of the logarithmic domain histogram is then used as the optimal segmentation threshold.

9. The method for automatic extraction of alteration minerals from spaceborne hyperspectral data according to claim 8, characterized in that, In step four, the formula for calculating the curvature of the histogram curve is: Where K(x) represents curvature; L′ represents the first derivative (slope / rate of change); and L” represents the second derivative (curvature / degree of bending).

10. The method for automatic extraction of alteration minerals from spaceborne hyperspectral data according to claim 9, characterized in that, Step five includes: generating a binary image based on the step segmentation threshold, performing mathematical morphological opening operations on the binary image to remove isolated noise pixels, retaining mineralized patches with spatial continuity, converting the binary image into a vector polygon, removing fragmented patches, and outputting the final uranium ore alteration mineral distribution map.

Images