Mineral Anomaly Identification Method and System Based on Spectral Remote Sensing

Abstract

This invention discloses a method and system for mineral anomaly identification based on spectral remote sensing, belonging to the field of mineral anomaly identification technology. The method includes the following steps: acquiring hyperspectral remote sensing image data of the target area; performing atmospheric and geometric correction on the hyperspectral remote sensing image data to obtain surface reflectance data and extracting spectral reflectance curves; performing weak signal enhancement processing on the spectral reflectance curves to obtain enhanced spectral curves, and calculating the spectral matching degree between the enhanced spectral curves and each mineral endmember in the standard mineral endmember spectral library; analyzing the spectral matching degree, and judging the mineral anomaly situation based on the analysis results. This invention solves the technical problem of low accuracy in identifying deep ore bodies and low-grade mineralized areas by constructing a systematic weak signal enhancement technology system.

Description

Technical Field

[0001] This invention relates to the field of mineral anomaly identification technology, and in particular to a method and system for mineral anomaly identification based on spectral remote sensing. Background Technology

[0002] Spectral remote sensing technology identifies surface mineral composition by acquiring the spectral reflectance characteristics of ground features, and has become an important tool for mineral resource exploration. Different minerals exhibit unique absorption and reflectance characteristics in specific spectral bands. For example, iron-bearing minerals show characteristic absorption around 0.9 μm, while copper-bearing minerals exhibit a significant spectral response around 2.2 μm. These spectral characteristics can be used to quickly delineate mineralized alteration zones, significantly reducing the cost of field exploration. With the development of hyperspectral sensor technology, spectral resolution and signal-to-noise ratio have been continuously improved, providing a data foundation for mineral identification. Currently widely used identification methods include spectral angle mapping, hybrid pixel decomposition, and support vector machine classification. These methods have achieved good application results in shallow mineralized areas with abundant mineral types and strong spectral signals.

[0003] The spectral signals of deep ore bodies and low-grade mineralized areas are often very weak, and the spectral response of surface cover can mask the true mineral anomaly information, resulting in low accuracy of traditional identification methods. Existing technologies mainly rely on simple linear stretching or band ratio calculations when processing weak signals. These methods are prone to amplifying noise and neglect the spectral physical relationships between different bands, making it difficult to effectively separate weak mineral anomaly signals from background noise. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides a mineral anomaly identification method and system based on spectral remote sensing to solve the problems of low accuracy, easy amplification of noise, neglect of spectral physical relationships between different bands, and difficulty in effectively separating weak mineral anomaly signals from background noise in existing identification methods.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0007] In a first aspect, the present invention provides a method for identifying mineral anomalies based on spectral remote sensing, comprising the following steps: acquiring hyperspectral remote sensing image data of a target area; performing atmospheric and geometric correction on the hyperspectral remote sensing image data to obtain surface reflectance data, and extracting a spectral reflectance curve; performing weak signal enhancement processing on the spectral reflectance curve to obtain an enhanced spectral curve, and calculating the spectral matching degree between the enhanced spectral curve and each mineral endmember in a standard mineral endmember spectral library; analyzing the spectral matching degree, and determining the mineral anomaly based on the analysis results.

[0008] As a preferred solution of the mineral anomaly recognition method based on spectral remote sensing according to the present invention, the steps of atmospheric correction and geometric correction for hyperspectral remote sensing image data include: performing atmospheric correction on the hyperspectral remote sensing image data by using the FLAASH atmospheric correction model to eliminate the atmospheric scattering and absorption effects, and converting the apparent reflectance data into surface reflectance data; performing geometric fine correction on the surface reflectance data based on ground control points; extracting the reflectance values within a set wavelength range from the surface reflectance data pixel by pixel to form the spectral reflectance curve.

[0009] As a preferred solution of the mineral anomaly recognition method based on spectral remote sensing according to the present invention, the steps of weak signal enhancement processing for the spectral reflectance curve include: calculating the initial signal-to-noise ratio SNR0 of the mineral characteristic absorption band in the spectral reflectance curve, and when the initial signal-to-noise ratio SNR0 < X, determining it as a weak signal and entering the enhancement process; performing non-linear enhancement on the spectral reflectance curve by using the Gamma correction algorithm to obtain the corrected reflectance; performing piecewise histogram equalization processing on the corrected reflectance to obtain the equalized reflectance; establishing a weak anomaly signal separation model to decompose the equalized reflectance into background spectral components and anomaly spectral components; performing illumination normalization processing on the anomaly spectral components to obtain the enhanced spectral curve.

[0010] The beneficial effects of this preferred technical solution are as follows: By establishing a systematic weak signal enhancement processing flow, the combination of signal-to-noise ratio determination, non-linear enhancement, piecewise equalization, signal separation, and illumination normalization is organically combined, and multi-level enhancement is performed on the weak spectral responses of deep ore bodies and low-grade mineralized areas, improving the recognizability of weak anomaly signals.

[0011] As a preferred solution of the mineral anomaly recognition method based on spectral remote sensing according to the present invention, the Gamma correction algorithm uses the formula for non-linear enhancement; where is the input reflectance, is the corrected reflectance, is the band Gamma coefficient; the calculation of the band Gamma coefficient is , is the skewness coefficient of the reflectance of the λth band, and the value range is 0.4 - 1.0.

[0012] The beneficial effects of this preferred technical solution are as follows: By using an adaptive Gamma correction algorithm based on the skewness coefficient, the enhancement parameters can be adjusted according to the reflectance distribution characteristics of each band, and compared with the problem of over-enhancement or under-enhancement of different bands by the traditional fixed Gamma value, a more accurate non-linear enhancement effect is achieved.

[0013] As a preferred embodiment of the mineral anomaly identification method based on spectral remote sensing described in this invention, the step of performing segmented histogram equalization processing on the corrected reflectance includes: dividing the set spectral range into three sub-segments: visible-near-infrared, shortwave infrared 1, and shortwave infrared 2; and performing histogram equalization on each sub-segment to obtain the equalized reflectance. The weak anomaly signal separation model is established and solved through the following steps: constructing an overcomplete dictionary matrix, which consists of standard mineral endmember spectra and background endmember spectra; setting a first regularization parameter. Second regularization parameter , where the first regularization parameter The value range is 0.01-0.1, and this is the second regularization parameter. The value range is 0.001-0.01; an optimization objective is established to equalize the reflectivity. The reconstruction error between the product of the overcomplete dictionary matrix and the sparse coefficient vector is minimized, while L1 norm constraints are applied to the sparse coefficient vector and L2 norm constraints are applied to the background spectral components. The optimization objective is iteratively solved using an orthogonal matching pursuit algorithm to obtain the optimal sparse coefficient vector. The spectral signal is reconstructed based on the optimal sparse coefficient vector and the overcomplete dictionary matrix, and the reconstructed spectral signal is decomposed into the background spectral components. and the anomalous spectral components .

[0014] The beneficial effects of this preferred technical solution are as follows: by using the range of each spectral segment for segmented histogram equalization, and combining the signal separation model established by sparse representation theory, the background interference is effectively suppressed while preserving the mineral anomaly characteristics. The application of the orthogonal matching pursuit algorithm ensures computational efficiency and separation accuracy.

[0015] As a preferred embodiment of the mineral anomaly identification method based on spectral remote sensing described in this invention, the illumination normalization process includes the following steps: processing the anomalous spectral components... Perform a logarithmic transformation; set three Gaussian filter functions at different scales, and use the three Gaussian filter functions at different scales to filter the anomalous spectral components. Convolution operations are performed to obtain filtering results at three different scales; logarithmic transformations are then applied to each of the three logarithmically transformed filtering results; a weighted average is then calculated from the abnormal spectral components. The enhanced spectral curve is obtained by subtracting the weighted average result from the logarithmic transformation result. .

[0016] The beneficial effects of this preferred technical solution are as follows: by capturing illumination variation information at different spatial frequencies through Gaussian filter functions of three different scales, the effects of uneven illumination caused by topographic undulations and changes in solar altitude angle can be eliminated, thereby enhancing the spectral curve to truly reflect the intrinsic spectral characteristics of the mineral.

[0017] As a preferred embodiment of the mineral anomaly identification method based on spectral remote sensing described in this invention, the calculation of spectral matching degree includes the following steps: extracting the enhanced spectral values ​​of the target pixel in each band, denoted as the target spectral vector; extracting the spectral values ​​of the reference endmember in each band from a standard mineral endmember spectral library, denoted as the reference spectral vector; calculating the sum of the products of the target spectral vector and the reference spectral vector corresponding band values; calculating the modulus of the target spectral vector and the reference spectral vector respectively, dividing the sum of the products by the product of the two modulus values ​​to obtain the cosine values ​​of the two spectral vectors; and performing an inverse cosine operation on the cosine values ​​to obtain the spectral angle.

[0018] As a preferred embodiment of the mineral anomaly identification method based on spectral remote sensing described in this invention, the step of analyzing the spectral matching degree includes: setting a spectral threshold, determining a mineral anomaly pixel when the spectral angle is less than the spectral threshold; performing spatial clustering analysis on the mineral anomaly pixel, removing isolated pixels, retaining anomaly regions with a continuous distribution area greater than a preset threshold, and generating a mineral anomaly distribution map.

[0019] In a second aspect, the present invention provides a mineral anomaly identification system based on spectral remote sensing, comprising: a data acquisition module for acquiring hyperspectral remote sensing image data of a target area;

[0020] The preprocessing module, connected to the data acquisition module, is used to perform atmospheric and geometric correction on the hyperspectral remote sensing image data to obtain surface reflectance data and extract spectral reflectance curves.

[0021] A weak signal enhancement module, connected to the preprocessing module, is used to perform weak signal enhancement processing on the spectral reflectance curve to obtain an enhanced spectral curve.

[0022] A spectral matching module, connected to the weak signal enhancement module, is used to calculate the spectral matching degree between the enhanced spectral curve and each mineral endmember in the standard mineral endmember spectral library.

[0023] An anomaly detection module, connected to the spectral matching module, is used to analyze the spectral matching degree, determine the mineral anomaly based on the analysis results, and generate a mineral anomaly distribution map.

[0024] As a preferred embodiment of the mineral anomaly identification system based on spectral remote sensing described in this invention, it further includes:

[0025] The storage module is connected to the data acquisition module, the preprocessing module, the weak signal enhancement module, the spectral matching module, and the anomaly detection module respectively, and is used to store hyperspectral remote sensing image data, surface reflectance data, enhanced spectral curves, spectral matching degrees, and mineral anomaly distribution maps.

[0026] The display module, connected to the anomaly detection module, is used to display a mineral anomaly distribution map.

[0027] Compared with existing technologies, the beneficial effects of this invention are as follows: By constructing a systematic weak signal enhancement technology system, it solves the technical problem of low accuracy in identifying deep ore bodies and low-grade mineralized areas using traditional methods. The Gamma correction algorithm dynamically adjusts the enhancement parameters according to the band characteristics, the segmented histogram equalization fully explores the information potential of each spectral sub-segment, the signal separation model of sparse representation theory extracts anomalous components, and illumination normalization eliminates environmental interference. The synergistic effect of multiple technologies enables even weak signals with low signal-to-noise ratios to be accurately identified.

[0028] This invention's method does not rely on a large number of labeled samples; it only requires a standard mineral endmember spectral library to complete the identification task, thus reducing application costs. It employs an orthogonal matching pursuit algorithm to solve the sparse optimization problem, achieving higher computational efficiency than traditional convex optimization methods, making it suitable for rapid processing of large-scale remote sensing data. Through a discrimination strategy combining spectral angle thresholding and spatial clustering analysis, isolated noise pixels can be eliminated, resulting in a mineral anomaly distribution map with clear boundaries and accurate positioning, which can then be directly used to guide the deployment of field exploration work. Attached Figure Description

[0029] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0030] Figure 1 This is a schematic diagram of the overall process of a mineral anomaly identification method based on spectral remote sensing according to an embodiment of the present invention.

[0031] Figure 2 This is a spectral curve diagram of a mineral anomaly identification method based on spectral remote sensing according to an embodiment of the present invention. Detailed Implementation

[0032] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the protection scope of the present invention.

[0033] Example 1, referring to Figure 1 As an embodiment of the present invention, a method for identifying mineral anomalies based on spectral remote sensing is provided, comprising the following steps S1 to S4:

[0034] S1. Acquire hyperspectral remote sensing image data of the target area.

[0035] S2. Perform atmospheric and geometric corrections on the hyperspectral remote sensing image data to obtain surface reflectance data and extract the spectral reflectance curve.

[0036] S3. Perform weak signal enhancement processing on the spectral reflectance curve to obtain the enhanced spectral curve, and calculate the spectral matching degree between the enhanced spectral curve and each mineral endmember in the standard mineral endmember spectral library.

[0037] S4. Analyze the spectral matching degree and determine the mineral anomaly based on the analysis results.

[0038] It should be noted that the spectral signals of deep ore bodies and low-grade mineralized areas are often very weak. Factors such as surface vegetation cover, soil background, and atmospheric interference can mask the true mineral anomaly information, leading to a decrease in the accuracy of traditional identification methods. In actual exploration, weak signal areas with a signal-to-noise ratio below 3 are often missed, resulting in the omission of potential mineral resources. Furthermore, changes in lighting conditions, topographic relief, and seasonal vegetation changes further interfere with the extraction of weak mineral spectral features, resulting in a high false positive rate in anomaly identification. In addition, traditional methods, which use fixed enhancement parameters and simple linear processing, are ill-suited to the differences in spectral responses across different bands and cannot effectively separate background noise from true anomaly signals, leading to insufficient reliability of the identification results. Therefore, accurate identification of deep ore bodies and enhancement of anomaly signals are also crucial.

[0039] Therefore, to address the aforementioned problems of weak signal identification and anomaly detection, a systematic weak signal enhancement technology process is established through steps S1-S4. This process acquires hyperspectral data of the target area and performs standardized preprocessing, enabling multi-level enhancement and accurate extraction of weak mineral anomaly signals. By correcting and segmented histogram equalization, enhancement parameters are dynamically adjusted based on the spectral characteristics of each band, improving the contrast of weak anomaly signals. A signal separation model is established using sparse representation theory to decompose the enhanced spectrum into background and anomaly components, thereby removing background interference. Multi-scale illumination normalization is employed to eliminate the influence of topographic and illumination variations, ensuring the authenticity of spectral features. Simultaneously, based on spectral angle matching algorithms and spatial clustering analysis, the location and reliable identification of mineral anomaly areas are achieved, improving the exploration success rate of deep ore bodies and low-grade mineralized areas.

[0040] Example 2, refer to Figure 1 and Figure 2 This is one embodiment of the present invention. Based on the above embodiment, a method for identifying mineral anomalies based on spectral remote sensing is provided. It includes the following steps:

[0041] S1. Acquire hyperspectral remote sensing image data of the target area.

[0042] In this embodiment, the target area is a copper mine exploration area, covering approximately 50 square kilometers. Hyperspectral remote sensing image data was acquired using an airborne hyperspectral imaging system. This hyperspectral remote sensing image data contains spectral reflectance information within the 400nm-2500nm wavelength range, comprising 224 continuous spectral bands with a spectral resolution of 10nm and a spatial resolution of 2m. Data acquisition was conducted during a clear, cloudless morning period from 10:00 AM to 2:00 PM, when the solar altitude angle is high and lighting conditions are stable.

[0043] S2. Perform atmospheric and geometric corrections on the hyperspectral remote sensing image data to obtain surface reflectance data and extract the spectral reflectance curve.

[0044] The steps for atmospheric and geometric correction of hyperspectral remote sensing image data include A1~A3:

[0045] A1. The hyperspectral remote sensing image data is atmospherically corrected using the FLAASH atmospheric correction model to eliminate atmospheric scattering and absorption effects, and the apparent reflectance data is converted into surface reflectance data.

[0046] The FLAASH (Fast Line-of-sight Atmospheric Analysis of Spectral Hypercubes) atmospheric correction model was used to perform atmospheric correction on hyperspectral remote sensing image data. First, sensor parameters, flight altitude (1500m), imaging time, and geographical location information were extracted from the image header file. Based on the imaging date and geographical location, a mid-latitude summer atmospheric model was set, a rural aerosol model was selected, and the initial visibility was set to 40km. The FLAASH model converts apparent reflectance data into surface reflectance data by calculating parameters such as atmospheric transmittance and path radiance, eliminating Rayleigh scattering, Mie scattering, and absorption effects of gases such as water vapor, oxygen, and carbon dioxide. The corrected surface reflectance values ​​range from 0 to 1, the spectral curves are continuous and smooth across bands, and atmospheric absorption bands (such as water vapor absorption near 1400nm and 1900nm) are effectively restored.

[0047] A2. Perform geometric fine correction on the surface reflectance data based on ground control points.

[0048] Specifically, geometric correction was performed on surface reflectance data based on ground control points. Twenty-five ground control points were evenly selected within the study area. These control points were chosen from easily identifiable landmarks such as road intersections, building corners, and river confluences. Precise geographic coordinates were obtained using GPS field measurements, with an accuracy better than 0.5m. The correspondence between image coordinates and geographic coordinates was established in ENVI software. Geometric correction was performed using a quadratic polynomial transformation model, and bilinear interpolation was used for resampling. The correction accuracy was verified through residual analysis. The average correction error for the control points was 0.8 pixels, and the maximum error did not exceed 1.5 pixels, meeting the requirements for subsequent analysis. The geometrically corrected image was projected using the UTM coordinate system, ensuring the spatial positioning accuracy of the image.

[0049] A3. Extract the reflectance values ​​within a set band range from the surface reflectance data pixel by pixel to form the spectral reflectance curve.

[0050] Extract the reflectance values within the wavelength range of 400 nm - 2500 nm from the surface reflectance data pixel by pixel to form the spectral reflectance curve. For each pixel (2m × 2m), sequentially read the surface reflectance values of 224 bands and arrange them in wavelength order to form a one-dimensional spectral vector. Plot the spectral curve with wavelength as the abscissa and reflectance as the ordinate. The spectral reflectance curve completely records the spectral response characteristics of ground objects in the visible light, near-infrared, and short-wave infrared bands. For typical ground objects (such as vegetation, soil, rocks) in the study area, their spectral curves show characteristic differences: vegetation has a green peak reflection near 550 nm and a sharp increase in the red edge near 750 nm; iron-bearing minerals have a characteristic absorption valley near 900 nm; copper-bearing mineralized altered rocks show hydroxyl absorption characteristics near 2200 nm. The extracted spectral reflectance curve provides basic data for subsequent weak signal enhancement processing and mineral anomaly identification.

[0051] S3. Perform weak signal enhancement processing on the spectral reflectance curve to obtain an enhanced spectral curve, and calculate the spectral matching degree between the enhanced spectral curve and each mineral endmember in the standard mineral endmember spectral library.

[0052] The steps of performing weak signal enhancement processing on the spectral reflectance curve include B1 - B5:

[0053] B1. Calculate the initial signal-to-noise ratio SNR0 of the mineral characteristic absorption band in the spectral reflectance curve. When the initial signal-to-noise ratio SNR0 < X, it is determined as a weak signal and enters the enhancement process.

[0054] Specifically, for copper ore exploration, pay attention to the hydroxyl absorption characteristic band (2180 - 2220 nm) near 2200 nm and the iron-bearing mineral absorption band (880 - 920 nm) near 900 nm. Select the mean value of the reflectance within these characteristic bands as the signal intensity S, and select the standard deviation of the reflectance in the non-absorption band (such as 2300 - 2350 nm) as the noise level N, and calculate the initial signal-to-noise ratio SNR0 = S / N.

[0055] In this embodiment, the signal-to-noise ratio is calculated for 50,000 pixels in the study area. The results show that about 35% of the pixels have an SNR0 less than 3 in the mineral characteristic absorption band. Among them, the SNR0 in the deep mineralization area and the vegetation-covered area is generally between 1.5 - 2.8, belonging to typical weak signal areas. Set the threshold X = 3. When SNR0 < 3, it is determined as a weak signal and enters the enhancement process. The statistical results show that 17,500 pixels need to be subjected to weak signal enhancement processing.

[0056] B2. Use the Gamma correction algorithm to perform non-linear enhancement on the spectral reflectance curve to obtain the corrected reflectance;

[0057] The Gamma correction algorithm uses the formula... Perform nonlinear enhancement;

[0058] For input reflectivity, The corrected reflectivity, This refers to the band Gamma coefficient;

[0059] Band Gamma Coefficient The calculation is as follows , The skewness coefficient of the reflectivity in the λ-th band is... The value range is 0.4-1.0.

[0060] Specifically, Gamma correction can focus on enhancing low reflectivity regions while preventing high reflectivity regions from becoming oversaturated.

[0061] First, calculate the skewness coefficient for the reflectance data of each band λ. The skewness coefficient reflects the degree of skewness in the reflectance distribution. A positive skewness coefficient indicates a right-skewed distribution (more low-reflectance pixels), while a negative skewness coefficient indicates a left-skewed distribution (more high-reflectance pixels). In this embodiment, the calculated skewness coefficient for the mineral characteristic absorption band (such as the 2200nm band) is Skew(2200nm) = 1.85, indicating that this band contains a large number of low-reflectance pixels, requiring strong enhancement.

[0062] Calculate the band Gamma coefficient based on the skewness coefficient. For the 2200nm band, For the 900nm band, Skew(900nm) = 1.62, and γ(900nm) = 0.68 is calculated. For non-mineralized characteristic bands such as 1600nm, Skew(1600nm) = 0.45, and γ(1600nm) = 0.91 is calculated. It can be seen that the weak signal band achieves a smaller γ value, thus realizing a stronger nonlinear enhancement effect.

[0063] After calculating the Gamma coefficient for each of the 224 bands, the formula was used. Nonlinear enhancement is performed. For example, the original reflectivity R(2200nm) of a pixel in the 2200nm band is 0.15, which is obtained after Gamma correction. The reflectivity was increased by 110%. High-reflectivity pixels (e.g., R=0.75) were corrected to... The increase was only 9%, effectively avoiding oversaturation. After correction, the contrast in weak signal areas was improved, and the mineral absorption characteristics became more obvious.

[0064] B3. Perform segmented histogram equalization on the corrected reflectance to obtain the equalized reflectance.

[0065] Specifically, the set spectral range is divided into three sub-segments: the visible-near-infrared band, the short-wave infrared 1 band, and the short-wave infrared 2 band. Histogram equalization is performed on each sub-segment to obtain the equalized reflectance. .

[0066] The corrected reflectance is then subjected to piecewise histogram equalization to obtain the equalized reflectance. Since the reflectance range varies greatly across different spectral regions, a segmented processing strategy can fully utilize the information potential of each segment.

[0067] The 400-2500 nm spectral range is divided into three sub-bands: the visible-near-infrared band (400-1000 nm, corresponding to bands 1-60), the shortwave infrared band 1 (1000-1800 nm, corresponding to bands 61-140), and the shortwave infrared band 2 (1800-2500 nm, corresponding to bands 141-224). This division is based on atmospheric window characteristics and mineral spectral response features, ensuring that each sub-band has relatively independent spectral physical meaning.

[0068] Histogram equalization was performed on each sub-segment. Taking the shortwave infrared band 2 as an example, the corrected reflectance of all pixels in all bands of this sub-segment was statistically analyzed, and a cumulative distribution function was constructed. The original reflectance distribution was concentrated in the range of 0.1-0.3. Through equalization mapping, the reflectance was redistributed to a wider range of 0.05-0.95. After equalization, a pixel with a reflectance of 0.15 in the 2200nm band was mapped to Re(2200nm)=0.32, and a pixel with a reflectance of 0.18 in the 2180nm band was mapped to Re(2180nm)=0.41. The absorption valley depth was increased from 0.03 to 0.09, an increase of 3 times.

[0069] After equalizing the three sub-segments separately, they are recombined to form a complete equalized reflectivity curve. The equalization process makes the reflectance distribution of each segment more uniform, highlights weak anomaly characteristics, and maintains the continuity and physical interpretability of the spectral curve.

[0070] B4. Establish a weak anomalous signal separation model to decompose the equalized reflectance into background spectral components and anomalous spectral components.

[0071] This weak anomaly signal separation model is based on sparse representation theory, assuming that the real mineral anomaly signal is sparsity under a suitable dictionary representation, while the background signal is represented by a smooth substrate.

[0072] The weak anomaly signal separation model is established and solved using the following steps:

[0073] B4.1 Construct an overcomplete dictionary matrix, which consists of standard mineral endmember spectra and background endmember spectra.

[0074] Specifically, an overcomplete dictionary matrix D was constructed, consisting of standard mineral endmember spectra and background endmember spectra. Standard mineral endmember spectra related to the mineralization types of the study area were extracted from the spectral library, including: copper-bearing mineral spectra such as chalcopyrite, chalcocite, and malachite (12 types in total); iron-bearing altered mineral spectra such as hematite and goethite (8 types in total); and clay-altered mineral spectra such as kaolinite, illite, and chlorite (15 types in total), for a total of 35 mineral endmember spectra.

[0075] Meanwhile, background endmember spectra were extracted from the non-mineralized background areas of the study area, including: different types of vegetation spectra (coniferous forest, broad-leaved forest, shrubs, grassland, a total of 8 types), different types of soil spectra (red soil, yellow soil, brown soil, a total of 6 types), and bedrock spectra (granite, sandstone, shale, a total of 5 types), for a total of 19 background endmember spectra.

[0076] The endmember spectra of 35 minerals and 19 background endmember spectra are combined to form a 54×224-dimensional overcomplete dictionary matrix D, where 54 represents the number of endmembers and 224 represents the number of bands. The overcompleteness property (the case where the number of endmembers > the number of bands does not hold, but the endmember library is larger than the actual number of mixed endmembers) makes the model more expressive and robust.

[0077] B4.2 Setting the first regularization parameter Second regularization parameter , where the first regularization parameter The value range is 0.01-0.1, and this is the second regularization parameter. The value range is 0.001-0.01.

[0078] It is important to know the first regularization parameter. Used to control the strength of the L1 norm constraint on the sparse coefficient vector. The larger the value, the sparser the resulting sparse coefficient vector, meaning fewer endmembers participate in the mixing. In this embodiment, considering that actual mineralized areas typically contain 2-5 main minerals, the following settings are used: =0.05, this value was determined through cross-validation, which can ensure sparsity while avoiding information loss caused by excessive sparsity.

[0079] Second regularization parameter The L2 norm constraint strength is used to control the background spectral composition. Its function is to suppress high-frequency fluctuations in background components, thus maintaining their smoothness. (Settings) =0.005, this value ensures that the background components exhibit a smooth base morphology, while classifying abrupt uptake features as anomalous components. Parameter and The value of was determined by evaluating the separation effect of 10 known mineralized samples. Under this parameter combination, the separation accuracy reached the optimal level.

[0080] B4.3 Establish optimization goals to achieve uniform reflectivity The reconstruction error between the overcomplete dictionary matrix and the product of the sparse coefficient vector is minimized, while the L1 norm constraint is applied to the sparse coefficient vector and the L2 norm constraint is applied to the background spectral components.

[0081] Specifically, the optimization objective includes three terms: the first term is the reconstruction error term, which requires that the product of the dictionary D and the sparse coefficients X can reconstruct the original equalized reflectivity well. The reconstruction error is measured using the L2 norm, which is the square of the Euclidean distance. The second term is a sparsity constraint term, which applies an L1 norm constraint to the sparse coefficient vector X. The L1 norm is calculated as the sum of the absolute values ​​of all coefficients. This constraint causes most coefficients to tend to zero, retaining only a few significant non-zero coefficients, corresponding to the actual endmembers. The third term is a smoothness constraint term, which applies an L2 norm constraint to the background spectral component B(λ). This constraint keeps the background component smooth and continuous, avoiding misjudging high-frequency noise as anomalies.

[0082] Three items passed the regularization parameter and A weighted combination is performed. In this embodiment, for a certain pixel, its equalized reflectance Re(λ) is a 224-dimensional vector. The value of the reconstruction error term is 0.0023, the value of the sparsity constraint term is 0.18, and the value of the smoothness constraint term is 0.012. The total objective function value after weighted combination is 0.0023 + 0.05 × 0.18 + 0.005 × 0.012 = 0.0113. The optimization process is to find the optimal sparse coefficient vector X that minimizes the objective function value.

[0083] B4.4. The optimization objective is solved iteratively using the orthogonal matching pursuit algorithm to obtain the optimal sparse coefficient vector.

[0084] The Orthogonal Matching Pursuit algorithm is a greedy iterative algorithm that selects the dictionary atom most relevant to the current residual in each iteration, gradually approaching the optimal solution.

[0085] During algorithm initialization, the reflectivity will be equalized. As the initial residual, the sparse coefficient vector X is initialized to zero, and the selected atom set is empty. In the first iteration, the inner product of the residual and all 54 atoms in the dictionary D is calculated, and the atom with the largest inner product, i.e., the endmember spectrum most relevant to the residual, is selected. In this embodiment, the malachite endmember spectrum is selected in the first iteration, with a correlation coefficient of 0.87 with the residual. This endmember is added to the selected set, and the sparse coefficients are updated to minimize the fitting error of the selected endmembers to the residual. After the update, a new residual is calculated, and the second iteration begins.

[0086] The iterative process continues until the stopping condition is met: the L2 norm of the residual is less than a preset threshold (set to 0.001), or the number of iterations reaches the upper limit (set to 10). In this embodiment, most pixels converge after 4-6 iterations. For example, after 5 iterations, a mineralized pixel selects 5 endmembers in sequence: malachite, chalcopyrite, goethite, kaolinite, and vegetation. The residual is finally reduced to 0.0008, obtaining the optimal sparse coefficient vector. .

[0087] B4.5. Reconstruct the spectral signal based on the optimal sparse coefficient vector and the overcomplete dictionary matrix, and decompose the reconstructed spectral signal into the background spectral components. and the anomalous spectral components .

[0088] Reconstructed spectral signal That is, the product of the dictionary matrix D and the sparse coefficient vector X. The coefficients of the corresponding mineral endmembers in the sparse coefficient vector X are extracted and denoted as... The coefficients corresponding to the background endmembers are denoted as Abnormal spectral components That is, the product of the mineral end-member dictionary and the mineral coefficient; background spectral components That is, the product of the background end-member sub-dictionary and the background coefficients.

[0089] In the example of mineralized pixels, the anomalous spectral component A(λ) is composed of a weighted combination of malachite (weight 0.32), chalcopyrite (weight 0.18), goethite (weight 0.15), and kaolinite (weight 0.09). This anomalous component exhibits a significant hydroxyl absorption feature at 2200 nm (absorption depth 0.12) and an iron absorption feature at 900 nm (absorption depth 0.08), which is highly consistent with the typical spectral characteristics of copper ore alteration. The background spectral component B(λ) is mainly composed of vegetation endmembers (weight 0.11), exhibiting a smooth base spectrum with a weak green peak at 550 nm and a slow red edge rise at 750 nm, without significant mineral absorption features.

[0090] Reference Figure 2As shown in Figure B5, the abnormal spectral components are subjected to illumination normalization processing to obtain the enhanced spectral curve.

[0091] Illumination normalization includes the following steps B5.1~B5.5:

[0092] B5.1 Perform a logarithmic transformation on the abnormal spectral components.

[0093] A logarithmic transformation is performed on the anomalous spectral component A(λ) to obtain log[A(λ)]. The purpose of the logarithmic transformation is to convert the multiplicative illumination effect into an additive effect, which is convenient for subsequent elimination by subtraction. In this embodiment, the anomalous spectral value A(2200nm) = 0.32 for a certain pixel in the 2200nm band, and after the logarithmic transformation, log[A(2200nm)] = log(0.32) = -0.495. Logarithmic transformations are performed on each of the 224 bands to form anomalous spectral curves in the logarithmic domain.

[0094] B5.2 Set up three Gaussian filter functions with different scales, and use the three Gaussian filter functions with different scales to perform convolution operations on the abnormal spectral components to obtain three filtering results with different scales.

[0095] The standard deviations of the Gaussian filter functions at the three scales are respectively set as follows: Pixel Pixel A pixel corresponds to a small scale (local illumination variation), a medium scale (the influence of terrain undulation), and a large scale (the influence of atmospheric scattering).

[0096] Gaussian filter function For a pixel at spatial location (x, y), a weighted average is performed within its neighborhood of radius 3σ, centered at x. The weights are determined by a Gaussian function. Convolution operation. Essentially, it calculates the Gaussian weighted average of the anomalous spectral values ​​of all pixels in the neighborhood of the given pixel in the λ band, and the resulting filtered result represents the illumination composition at that scale.

[0097] In this embodiment, for the aforementioned mineralized pixel (located in row 120 and column 85), small-scale filtering ( The results mainly reflect the local illumination changes within a 30×30 pixel area around it, with a value of 0.29 for the 2200nm band after filtering; mesoscale filtering ( The results reflect the terrain lighting effect within a 160×160 pixel area around it, with a filter value of 0.27; large-scale filtering ( The results reflect the atmospheric illumination effects within a 500×500 pixel area around the image, with a filter value of 0.25. The filtering results at the three scales collectively capture illumination variation information at different spatial frequencies.

[0098] B5.3 Perform logarithmic transformation on the filtering results of the three different scales respectively.

[0099] Logarithmic transformations were performed on the filtering results at the three different scales to obtain... , , For the aforementioned mineralized pixels in the 2200nm band, the logarithm of the small-scale filtering result is log(0.29) = -0.538, the logarithm of the medium-scale filtering result is log(0.27) = -0.569, and the logarithm of the large-scale filtering result is log(0.25) = -0.602. The logarithmic transformation converts the illumination components at the three scales to the logarithmic domain.

[0100] B5.4. Perform a weighted average of the filtering results after the three logarithmic transformations.

[0101] In this embodiment, the weights are all set to 1 / 3. The purpose of weighted averaging is to integrate illumination information from three scales to obtain the overall illumination composition of the pixel. In this embodiment, for the aforementioned mineralized pixel in the 2200nm band, the weighted average result is -0.570. Weighted averaging is performed on each of the 224 bands to obtain the complete illumination composition curve.

[0102] B5.5. Subtract the weighted average result from the logarithmic transformation result of the abnormal spectral components to obtain the enhanced spectral curve.

[0103] The enhanced spectral curve A*(λ) is obtained by subtracting the weighted average result from the logarithmic transformation result log[A(λ)] of the anomalous spectral component A(λ). The subtraction operation eliminates the illumination component while preserving the inherent reflectivity of the ground object.

[0104] In this embodiment, for the aforementioned mineralized pixels in the 2200nm band, the enhanced spectral curve A*(2200nm) = log[A(2200nm)] - weighted average = -0.495 - (-0.570) = 0.075. Subtraction is performed on each of the 224 bands to obtain the complete enhanced spectral curve A*(λ). The enhanced spectral curve eliminates the influence of uneven illumination, making the characteristic absorption peaks of the minerals more regular and significant.

[0105] The enhanced spectral curve A*(λ) is transformed back from the logarithmic domain to the linear domain (by performing the antilogarithmic operation exp[A*(λ)]) to obtain the final weak signal enhancement result.

[0106] In this embodiment of the application, calculating the spectral matching degree includes the following steps C1 to C5:

[0107] C1. Extract the enhanced spectral values ​​of the target pixel in each band, and denote them as the target spectral vector.

[0108] After weak signal enhancement processing in steps B1-B5, each pixel in the study area obtained an enhanced spectral curve A*(λ), which contains spectral values ​​of 224 bands.

[0109] In this embodiment, the mineralized pixel mentioned above (located in row 120 and column 85) is selected as an example. After the enhanced spectral curve of this pixel is transformed back to the linear domain using an antilogarithmic method, the spectral values ​​in each band are as follows: The enhanced spectral values ​​of these 224 bands are arranged in wavelength order to form the target spectral vector. ,in, The corresponding enhanced spectral value at 400nm is 0.142. The corresponding enhanced spectral value for the 2500nm band is 0.318.

[0110] To ensure the accuracy of subsequent calculations, the target spectral vector undergoes normalization preprocessing to remove the influence of the envelope. In this embodiment, a continuum removal method is used. Endpoints and key inflection points (such as 1000nm and 1800nm) are selected within the 400-2500nm range and connected to form an envelope. The spectral values ​​of each band are then divided by the corresponding envelope value to obtain the normalized target spectral vector. After normalization, the absorption characteristics of minerals are more prominent. For example, the normalized spectral value of the 2200nm band is 0.88 (indicating an absorption depth of 12% relative to the envelope), and the normalized spectral value of the 900nm band is 0.91 (indicating an absorption depth of 9%).

[0111] C2. Extract the spectral values ​​of the reference endmembers in each band from the standard mineral endmember spectral library, and denote them as the reference spectral vector.

[0112] In this embodiment, the standard mineral endmember spectral library used contains the 35 mineral endmembers mentioned in step B4.1. The spectral resolution was obtained by measuring with a high-precision spectrometer in the laboratory, and it is consistent with the hyperspectral remote sensing image, which is 10 nm.

[0113] Extract the spectral values ​​of each mineral endmember from the standard mineral endmember spectral library one by one. For example, the spectral values ​​of the malachite endmember in each band are: , ,..., Arrange the spectral values ​​of these 224 bands in the same order to form a reference spectral vector. .

[0114] Similarly, the reference spectral vector is also subjected to continuum removal and normalization to obtain... Malachite's normalized spectrum exhibits a significant hydroxyl absorption characteristic in the 2200 nm band, with a normalized spectral value of 0.75 (absorption depth 25%), which is typical of copper-bearing altered minerals. Reference spectral vectors for 35 mineral end-members, including chalcopyrite, chalcocite, hematite, goethite, and kaolinite, were extracted sequentially for matching calculations with the target spectral vector.

[0115] C3. Calculate the sum of the products of the target spectral vector and the corresponding band values ​​of the reference spectral vector.

[0116] Calculate the target spectral vector With the reference spectral vector The sum of the products of the corresponding band values. This sum of products is essentially the inner product of two vectors, reflecting the degree of similarity between the two spectral curves.

[0117] Taking the matching of the target pixel with the malachite endmember as an example, calculate the inner product: .

[0118] The specific calculation process is as follows: The product of the first band (400nm) is 0.152 × 0.095 = 0.01444, the product of the second band (410nm) is 0.167 × 0.102 = 0.01703, ..., the product of the 124th band (2200nm) is 0.88 × 0.75 = 0.66, ..., the product of the 224th band (2500nm) is 0.342 × 0.265 = 0.09063. Summing up the products of all 224 bands, the sum of the products is: .

[0119] In this embodiment, the target pixel is multiplied by 35 mineral endmembers, resulting in 35 numerical values. Besides the sum of the products with malachite (156.83), the sum of the products with chalcopyrite is 142.76, with kaolinite is 138.92, with hematite is 115.48, and with the vegetation background is 89.23. A larger sum of products indicates a higher similarity between the two spectral curves.

[0120] C4. Calculate the magnitude of the target spectral vector and the reference spectral vector respectively, and divide the sum of the products by the product of the two magnitudes to obtain the cosine values ​​of the two spectral vectors.

[0121] The formula for calculating the modulus of the target spectral vector is:

[0122] ;

[0123] For the aforementioned target pixel, calculate the modulus: ;

[0124] The modulus of the reference spectral vector (malachite) is calculated as follows: .

[0125] Dividing the sum of the products by the product of the two moduli gives the cosine value: .

[0126] The cosine value ranges from [-1, 1]. The closer the cosine value is to 1, the more similar the two spectral vectors are, and the closer their directions are. In this embodiment, the cosine value of the target pixel and the malachite endmember is 0.816, indicating that they have a high degree of similarity.

[0127] The cosine values ​​of the target pixel with other mineral endmembers were calculated sequentially: 0.762 with chalcopyrite, 0.731 with kaolinite, 0.645 with hematite, and 0.512 with the vegetation background. The cosine value ranking shows that this pixel has the highest similarity to copper-bearing mineral endmembers (malachite and chalcopyrite), followed by clay alteration minerals (kaolinite), and a lower similarity to background endmembers. Preliminary judgment suggests that this pixel exhibits copper mineralization anomaly characteristics.

[0128] C5. Perform an inverse cosine operation on the cosine value to obtain the spectral angle θ.

[0129] The spectral angle is the core parameter of the Spectral Angle Mapper (SAM) algorithm. It represents the angle between two spectral curves in radians or degrees. The smaller the spectral angle, the more similar the two spectral curves are.

[0130] The formula for calculating the inverse cosine is: ;

[0131] For the aforementioned target pixel and malachite endmember, the spectral angle is calculated as follows: .

[0132] The spectral angles of the target pixel and other mineral endmembers were calculated sequentially. The 35 spectral angles were sorted from smallest to largest, with the three smallest angles being: malachite (35.98°), chalcopyrite (41.08°), and kaolinite (43.55°). The smaller the spectral angle, the higher the matching degree. The spectral characteristics of this pixel are closest to the malachite endmember, indicating that the main mineral component of this pixel is malachite, belonging to a copper mineralization alteration anomaly.

[0133] In this embodiment, the spectral matching degree of all 50,000 pixels in the study area was calculated step C1-C5 one by one. The spectral angle of each pixel was calculated with 35 mineral endmembers, forming a 50,000×35 spectral angle matrix. For each pixel, the mineral endmember with the smallest spectral angle was selected as the best matching mineral type for that pixel, and the minimum spectral angle value was recorded as the basis for subsequent anomaly judgment. Statistical results show that the minimum spectral angle of approximately 18,500 pixels is less than 50°, indicating that these pixels have a high matching degree with the standard mineral endmembers and may have mineralization anomalies.

[0134] S4. Analyze the spectral matching degree and determine the mineral anomaly based on the analysis results.

[0135] The steps for analyzing the spectral matching degree include D1~D2:

[0136] D1. Set a spectral threshold. When the spectral angle is less than the spectral threshold, it is determined to be a mineral anomalous pixel.

[0137] Setting the spectral threshold requires comprehensive consideration of the spectral variability of different mineral types, the geological background of the study area, and the requirements for controlling the misjudgment rate.

[0138] In this embodiment, differentiated spectral thresholds are set according to different mineral types. For copper-bearing minerals (malachite, chalcopyrite, chalcocite, etc.), due to their distinct spectral characteristics, the hydroxyl absorption near 2200 nm and the iron absorption near 900 nm are relatively stable, so a stricter spectral threshold is set to ensure the reliability of the identification results. For iron-bearing altered minerals (hematite, goethite, etc.), because their spectral characteristics are greatly affected by the degree of oxidation and have strong variability, a relatively lenient spectral threshold is set. For clay-altered minerals (kaolinite, illite, chlorite, etc.), considering that clay minerals are easily confused with the soil background, a moderately strict spectral threshold is set.

[0139] D2. Perform spatial clustering analysis on the mineral anomaly pixels, remove isolated pixels, retain anomaly regions with a continuous distribution area greater than a preset threshold, and generate a mineral anomaly distribution map.

[0140] The purpose of spatial clustering analysis is to eliminate random noise and isolated false positives, and to extract the real mineralization anomaly areas.

[0141] Spatial clustering is performed using a connected component labeling algorithm. First, the mineral anomaly pixels identified in step D1 are spatially labeled: anomaly pixels are assigned a value of 1, and non-anomaly pixels are assigned a value of 0, forming a binarized anomaly distribution image. Then, an 8-neighborhood connectivity rule is used to group adjacent anomaly pixels into the same connected component. An 8-neighborhood refers to the eight adjacent pixels around each pixel (top, bottom, left, right, top-left, top-right, bottom-left, bottom-right). If two anomaly pixels are adjacent within their 8-neighborhood, they are grouped into the same cluster region.

[0142] In this embodiment, a depth-first search algorithm is used to label connected components. Starting from the first anomalous cell, all connected anomalous cells are recursively searched along the 8-neighborhood direction, and they are assigned the same cluster number. After the search is completed, a new search continues from the next unlabeled anomalous cell until all anomalous cells are labeled. After connected component labeling, the 8720 copper mine anomalous cells are divided into 327 independent cluster regions, numbered from 1 to 327.

[0143] Calculate the area of ​​each cluster region. The area of ​​a cluster region is equal to the number of pixels contained in the region multiplied by the area of ​​a single pixel (2m × 2m = 4m). 2 Statistical results show that the area distribution of the 327 cluster regions varies greatly: the largest cluster region (number 23) contains 1856 pixels and has an area of ​​7424 m². 2 (Approximately 0.74 hectares); the smallest cluster region (number 312) contains only 1 cell and has an area of ​​4m². 2 The median cluster region contains 6 pixels and has an area of ​​24m². 2 .

[0144] A preset area threshold is set, and clusters with areas smaller than this threshold are removed. In this embodiment, based on the mineralization characteristics of the study area and field verification experience, the preset area threshold is set to 100m². 2 (Equivalent to 25 pixels). Area less than 100m² 2 Clustering regions are often misjudged due to spectral noise, land cover effects, or local vegetation shading, and do not have actual mineralization significance. After area threshold filtering, 58 of the 327 clustered regions have an area greater than 100m². 2 269 ​​regions were retained as candidate anomaly regions; the area of ​​each region was less than 100m². 2 It was removed as isolated noise.

[0145] In summary, by constructing a systematic weak signal enhancement technology system, the technical challenge of low accuracy in identifying deep ore bodies and low-grade mineralized areas using traditional methods has been solved. The Gamma correction algorithm dynamically adjusts enhancement parameters based on band characteristics, segmented histogram equalization fully exploits the information potential of each spectral sub-segment, anomalous components are extracted using a signal separation model based on sparse representation theory, and illumination normalization eliminates environmental interference. The synergistic effect of these multiple technologies enables accurate identification of weak signals with low signal-to-noise ratios.

[0146] This invention's method does not rely on a large number of labeled samples; it only requires a standard mineral endmember spectral library to complete the identification task, thus reducing application costs. It employs an orthogonal matching pursuit algorithm to solve the sparse optimization problem, achieving higher computational efficiency than traditional convex optimization methods, making it suitable for rapid processing of large-scale remote sensing data. Through a discrimination strategy combining spectral angle thresholding and spatial clustering analysis, isolated noise pixels can be eliminated, resulting in a mineral anomaly distribution map with clear boundaries and accurate positioning, which can then be directly used to guide the deployment of field exploration work.

[0147] Example 3 illustrates a schematic scheme for a mineral anomaly identification method based on spectral remote sensing. It should be noted that the technical solution of this spectral remote sensing-based mineral anomaly identification system belongs to the same concept as the technical solution of the aforementioned spectral remote sensing-based mineral anomaly identification method. Details not described in detail in the technical solution of the spectral remote sensing-based mineral anomaly identification system in this embodiment can be found in the description of the technical solution of the aforementioned spectral remote sensing-based mineral anomaly identification method.

[0148] This embodiment also provides a mineral anomaly identification system based on spectral remote sensing, including:

[0149] The data acquisition module is used to acquire hyperspectral remote sensing image data of the target area;

[0150] The preprocessing module, connected to the data acquisition module, is used to perform atmospheric and geometric correction on the hyperspectral remote sensing image data to obtain surface reflectance data and extract spectral reflectance curves.

[0151] A weak signal enhancement module, connected to the preprocessing module, is used to perform weak signal enhancement processing on the spectral reflectance curve to obtain an enhanced spectral curve.

[0152] A spectral matching module, connected to the weak signal enhancement module, is used to calculate the spectral matching degree between the enhanced spectral curve and each mineral endmember in the standard mineral endmember spectral library.

[0153] An anomaly detection module, connected to the spectral matching module, is used to analyze the spectral matching degree, determine the mineral anomaly based on the analysis results, and generate a mineral anomaly distribution map.

[0154] Also includes:

[0155] The storage module is connected to the data acquisition module, the preprocessing module, the weak signal enhancement module, the spectral matching module, and the anomaly detection module respectively, and is used to store hyperspectral remote sensing image data, surface reflectance data, enhanced spectral curves, spectral matching degrees, and mineral anomaly distribution maps.

[0156] The display module, connected to the anomaly detection module, is used to display a mineral anomaly distribution map.

[0157] This embodiment also provides an electronic device suitable for mineral anomaly identification based on spectral remote sensing, comprising: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the mineral anomaly identification method based on spectral remote sensing as proposed in the above embodiment.

[0158] This embodiment also provides a storage medium on which a computer program is stored. When the program is executed by a processor, it implements the mineral anomaly identification method based on spectral remote sensing as proposed in the above embodiments.

[0159] The storage medium proposed in this embodiment and the mineral anomaly identification method based on spectral remote sensing proposed in the above embodiments belong to the same inventive concept. Technical details not described in detail in this embodiment can be found in the above embodiments, and this embodiment has the same beneficial effects as the above embodiments.

[0160] Based on the above description of the implementation methods, those skilled in the art can clearly understand that the present invention can be implemented using software and necessary general-purpose hardware, and of course, it can also be implemented using hardware. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as a computer floppy disk, read-only memory (ROM), random access memory (RAM), flash memory, hard disk, or optical disk, etc., including several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods of the various embodiments of the present invention.

[0161] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for identifying mineral anomalies based on spectral remote sensing, characterized in that, It includes the following steps: Obtain hyperspectral remote sensing image data of the target area; Perform atmospheric correction and geometric correction on the hyperspectral remote sensing image data to obtain surface reflectance data, and extract spectral reflectance curves; Perform weak signal enhancement processing on the spectral reflectance curve to obtain an enhanced spectral curve, and calculate the spectral matching degree between the enhanced spectral curve and each mineral endmember in the standard mineral endmember spectral library. The steps for performing weak signal enhancement processing on the spectral reflectance curve include: Calculate the initial signal-to-noise ratio SNR0 of the mineral characteristic absorption band in the spectral reflectance curve. When the initial signal-to-noise ratio SNR0 < X, it is determined as a weak signal and enters the enhancement process; Use the Gamma correction algorithm to perform nonlinear enhancement on the spectral reflectance curve to obtain the corrected reflectance; Perform piecewise histogram equalization processing on the corrected reflectance to obtain the equalized reflectance; Establish a weak anomaly signal separation model and decompose the equalized reflectance into background spectral components and anomaly spectral components; Perform illumination normalization processing on the anomaly spectral components to obtain the enhanced spectral curve; Analyze the spectral matching degree and judge the mineral anomaly situation according to the analysis results.

2. The mineral anomaly identification method based on spectral remote sensing as described in claim 1, characterized in that, The steps for performing atmospheric correction and geometric correction on the hyperspectral remote sensing image data include: Use the FLAASH atmospheric correction model to perform atmospheric correction on the hyperspectral remote sensing image data, eliminate atmospheric scattering and absorption effects, and convert the apparent reflectance data into surface reflectance data; Perform geometric fine correction on the surface reflectance data based on ground control points; Extract the reflectance values within the set wavelength range of the surface reflectance data pixel by pixel to form the spectral reflectance curve.

3. The mineral anomaly identification method based on spectral remote sensing as described in claim 1, characterized in that, The Gamma correction algorithm uses the formula... Perform nonlinear enhancement; in, For input reflectivity, The corrected reflectivity, This refers to the band Gamma coefficient; The band Gamma coefficient The calculation is as follows , Let be the skewness coefficient of the reflectivity in the λ-th band. The value range is 0.4-1.

0.

4. The mineral anomaly identification method based on spectral remote sensing as described in claim 3, characterized in that, The steps for performing piecewise histogram equalization processing on the corrected reflectance include: The set spectral range is divided into three sub-segments: visible-near infrared, short-wave infrared 1, and short-wave infrared 2. Histogram equalization is performed on each sub-segment to obtain the equalized reflectance. ; The weak anomaly signal separation model is established and solved through the following steps: Construct an overcomplete dictionary matrix, which consists of standard mineral endmember spectra and background endmember spectra; Set the first regularization parameter Second regularization parameter , where the first regularization parameter The value range is 0.01-0.1, and this is the second regularization parameter. The value range is 0.001-0.01; Establish optimization goals to achieve equalized reflectivity The reconstruction error between the overcomplete dictionary matrix and the product of the sparse coefficient vector is minimized, while the L1 norm constraint is applied to the sparse coefficient vector and the L2 norm constraint is applied to the background spectral components. Use the orthogonal matching pursuit algorithm to iteratively solve the optimization objective to obtain the optimal sparse coefficient vector; The spectral signal is reconstructed based on the optimal sparse coefficient vector and the overcomplete dictionary matrix, and then decomposed into the background spectral components. and the anomalous spectral components .

5. The mineral anomaly identification method based on spectral remote sensing as described in claim 4, characterized in that, The illumination normalization processing includes the following steps: For the anomalous spectral components Perform a logarithmic transformation; Three Gaussian filter functions with different scales are set up, and the three Gaussian filter functions with different scales are used to filter the abnormal spectral components. Perform convolution operations to obtain filtering results at three different scales; Perform logarithmic transformation on the filtering results of the three different scales respectively; Perform weighted average on the three logarithmically transformed filtering results; From the abnormal spectral components The enhanced spectral curve is obtained by subtracting the weighted average result from the logarithmic transformation result. .

6. The mineral anomaly identification method based on spectral remote sensing as described in claim 5, characterized in that, Calculating the spectral matching degree includes the following steps: Extract the enhanced spectral values of the target pixel in each band, denoted as the target spectral vector; Extract the spectral values of the reference endmember in each band from the standard mineral endmember spectral library, denoted as the reference spectral vector; Calculate the sum of the products of the corresponding band values of the target spectral vector and the reference spectral vector; Calculate the norms of the target spectral vector and the reference spectral vector respectively, and divide the sum of the products by the product of the two norms to obtain the cosine value of the two spectral vectors; Perform inverse cosine operation on the cosine value to obtain the spectral angle.

7. The mineral anomaly identification method based on spectral remote sensing as described in claim 6, characterized in that, The steps for analyzing the spectral matching degree include: Set a spectral threshold. When the spectral angle is less than the spectral threshold, it is determined as a mineral anomaly pixel; Perform spatial clustering analysis on the mineral anomaly pixels, eliminate isolated pixels, retain the anomaly areas with a continuous distribution area greater than a preset threshold, and generate a mineral anomaly distribution map.

8. A mineral anomaly identification system based on spectral remote sensing, employing the method described in any one of claims 1-7, characterized in that, Including: A data acquisition module for acquiring hyperspectral remote sensing image data of a target area; A preprocessing module connected to the data acquisition module for performing atmospheric correction and geometric correction on the hyperspectral remote sensing image data to obtain surface reflectance data and extract spectral reflectance curves; A weak signal enhancement module connected to the preprocessing module for performing weak signal enhancement processing on the spectral reflectance curves to obtain enhanced spectral curves. The steps of performing weak signal enhancement processing on the spectral reflectance curves include: Calculating the initial signal-to-noise ratio SNR0 of the mineral characteristic absorption band in the spectral reflectance curve. When the initial signal-to-noise ratio SNR0 < X, it is determined as a weak signal and enters the enhancement process; Performing non-linear enhancement on the spectral reflectance curve using the Gamma correction algorithm to obtain the corrected reflectance; Performing segmented histogram equalization processing on the corrected reflectance to obtain the equalized reflectance; Establishing a weak anomaly signal separation model to decompose the equalized reflectance into background spectral components and anomaly spectral components; Performing illumination normalization processing on the anomaly spectral components to obtain the enhanced spectral curves; A spectral matching module connected to the weak signal enhancement module for calculating the spectral matching degrees of the enhanced spectral curves with each mineral endmember in the standard mineral endmember spectral library; An anomaly judgment module connected to the spectral matching module for analyzing the spectral matching degrees, judging the mineral anomaly situation according to the analysis results, and generating a mineral anomaly distribution map.

9. The mineral anomaly identification system based on spectral remote sensing as described in claim 8, characterized in that, It further includes: A storage module respectively connected to the data acquisition module, the preprocessing module, the weak signal enhancement module, the spectral matching module and the anomaly judgment module for storing hyperspectral remote sensing image data, surface reflectance data, enhanced spectral curves, spectral matching degrees and mineral anomaly distribution maps; A display module connected to the anomaly judgment module for displaying the mineral anomaly distribution map.

Images