Method, system and equipment for prospecting copper deposits indicated by remote sensing identification of Polygonum capitatum plants

Abstract

The present invention discloses a prospecting method, system and equipment for remote sensing identification of copper ore indicated by Polygonum capitatum plants, and belongs to the field of remote sensing prospecting technology. It includes: acquiring remote sensing data in the target area, establishing a spectral database of Polygonum capitatum plants and altered minerals related to copper ore; using spectral data analysis and processing technology to finely identify Polygonum capitatum plants and their distribution characteristics; combining radar, multi-spectral and hyperspectral remote sensing data to identify tectonic-altered minerals and fine mapping in the target area; integrating Polygonum capitatum plants and their distribution information with tectonic-altered mineral mapping results, establishing a prospecting prediction model based on Polygonum capitatum plants indicating copper ore, delineating copper ore anomalies in the target area, and verifying the prediction results by deep learning methods; integrating multi-source data and field surveys to further verify the predicted target area and optimize the prospecting model. The present invention can reduce costs, improve efficiency and accuracy, and simplify the prospecting process.

Description

Technical Field

[0001] The invention belongs to the technical field of remote sensing prospecting, in particular to a prospecting method, system and equipment for remotely sensing and identifying copper ore indicated by Polygonum capitatum plants. Background Art

[0002] Although traditional geological, geophysical, geochemical and remote sensing prospecting techniques and their combinations have many successful examples of prospecting predictions around the world, these traditional exploration techniques and their combinations also have some significant shortcomings. First, traditional prospecting techniques and their combinations are costly, usually require a lot of manpower and material resources, and have a long prospecting application cycle. Secondly, traditional prospecting techniques and their combinations have limited coverage, and are often difficult to implement effectively in areas with complex terrain or vegetation coverage. In addition, traditional prospecting techniques and their combinations are less efficient when dealing with large areas, and are prone to missing potential mineralization areas.

[0003] As an important indicator plant for copper prospecting, Polygonum capitatum is widely used in copper prospecting prediction research. However, manual identification of Polygonum capitatum is inefficient and costly, and it is difficult to achieve in areas with complex terrain or vegetation coverage. With the development and application of remote sensing technology, remote sensing identification of plants has basically matured, and it has become possible to use remote sensing to identify plants for prospecting prediction. By analyzing the spectral characteristics of the copper indicator plant Polygonum capitatum, identifying Polygonum capitatum and its distribution characteristics, and combining with fine mapping of structural-altered minerals, a prospecting prediction model based on remote sensing identification of Polygonum capitatum is established, copper anomaly areas are delineated, and breakthroughs in prospecting are guided. Compared with traditional methods, the prospecting technical methods, systems and equipment for using remote sensing to identify Polygonum capitatum plants to indicate copper deposits have the advantages of low cost, high efficiency and wide coverage. However, the application of existing technical methods is still limited, and the main problems are as follows:

[0004] 1. Limitations of a single data source: Existing technologies mainly rely on single plant geochemical or spectral data, which is difficult to fully reflect the comprehensive mineralization characteristics, resulting in large prediction errors and low accuracy.

[0005] 2. Difficulty in identifying altered minerals and structures: In areas covered by vegetation, the identification of altered minerals and geological structures is greatly interfered. Based on the remote sensing identification of Polygonum capitatum plants, radar, multispectral and hyperspectral remote sensing data are used to identify altered minerals and structures in the target area, supplement and integrate multi-source data, and further improve the accuracy and reliability of prospecting.

[0006] 3. Difficulty in fusing multi-source data: Although some studies have attempted to combine multi-source data such as geology, geophysics and geochemistry with remote sensing data, the methods and technologies for data fusion are not yet mature, making it difficult to achieve efficient comprehensive analysis of multi-source data. Effectively fusing and processing data from different sources to improve the accuracy of prospecting predictions remains a major challenge.

[0007] In the pilot work, we found that the existing technology has the above-mentioned deficiencies in practical applications, and a new method is urgently needed to use Polygonum capitatum plants for remote sensing identification and prospecting of copper mines to make up for the deficiencies of the existing technology and improve the efficiency and accuracy of exploration. Therefore, the present invention proposes a method, system and equipment for remote sensing identification and prospecting of copper mines using Polygonum capitatum plants, which solves the deficiencies of the existing technology and improves the accuracy and efficiency of prospecting by combining radar, multi-spectral and hyperspectral remote sensing data, as well as multi-source data such as geology, geophysics and geochemistry. Summary of the invention

[0008] The present invention aims to provide a remote sensing prospecting method, system and equipment for copper mine exploration using Polygonum capitatum plants, so as to overcome the deficiencies of the prior art and improve the accuracy and efficiency of prospecting.

[0009] The specific technical solution is:

[0010] First, the present invention provides a method for prospecting for copper deposits indicated by remote sensing identification of Polygonum capitatum plants, comprising the following steps:

[0011] S1. Acquire remote sensing data in the target area and establish a spectral database of Polygonum capitatum plants and altered minerals related to copper mines;

[0012] S1.1. Collect Sentinel-2 MSI remote sensing data, GF2 high spatial resolution remote sensing data, GF5 hyperspectral remote sensing data, GF3 radar data, and ASTER visible light, short wave infrared, and thermal infrared remote sensing data of the study area;

[0013] S1.2, using SOC710-VP multi-spectral imager to collect the spectral information of the target Polygonum capitatum;

[0014] S1.3. Use CSD350A spectrometer to collect spectral information of copper ore alteration minerals.

[0015] S2. Use spectral data analysis and processing technology to precisely identify Polygonum capitatum plants and their distribution characteristics:

[0016] S2.1. Preprocess all remote sensing images, including image geometry correction, radiation correction, and fusion processing of multispectral data with GF2-0.8m panchromatic images to ensure data accuracy and spatial visibility;

[0017] S2.2. Use spectral analysis technology to identify the reflectance characteristics of Polygonum capitatum plants in different bands and determine the spectral characteristics related to copper ore;

[0018] S2.3. Perform spectral angle matching analysis on the fused Sentinel-2 MSI remote sensing data and the spectral data of Polygonum capitatum to preliminarily identify the possible distribution of Polygonum capitatum plants in the target area.

[0019] S3. Combine radar, multispectral and hyperspectral remote sensing data to identify and map structural-altered minerals in the target area. The steps are as follows:

[0020] S3.1. Identification of altered minerals: spectral analysis of altered minerals is performed using multispectral and hyperspectral data combined with mineral spectral data acquired by CSD350. Spectral matching algorithms are used to compare the spectral features in remote sensing data with the known mineral spectral library to accurately identify the types and distribution of altered minerals and the main altered mineral types in the target area.

[0021] S3.2. Structural feature identification: Use radar data to analyze topography and geological structures to determine the location, shape and scale of the structures.

[0022] S4. Integrate the information of Polygonum capitatum plants and their distribution with the identification and fine mapping of structural-altered minerals, establish a prospecting prediction model based on the indication of copper deposits by Polygonum capitatum plants, delineate the copper anomalies in the target area, and verify the prediction results by deep learning methods:

[0023] S4.1. Based on expert knowledge and experience, construct a comparison matrix of vegetation, structure and altered minerals and determine the indicator weights;

[0024] S4.2. Deep learning model: A deep learning model (multi-layer perceptron, MLP) was constructed. The input features included synthetic image data that were beneficial to the extraction of Polygonum capitatum plants, structures, and altered mineral bands and the preliminary prediction results of class1 (representative partial areas). The deep learning model was trained on the training dataset, and the model performance was verified on the test dataset to evaluate the prediction accuracy.

[0025] S4.3. Taking vegetation, structure and altered minerals as prediction indicators, using GIS tools, using determined weights to perform weighted overlay analysis on the indicators, and obtain preliminary prediction results.

[0026] S5. Comprehensive geological, geophysical and geochemical data and field surveys to further verify the prediction target area of ​​copper deposits indicated by Polygonum capitatum plants using remote sensing, optimize the prospecting model and improve the prediction accuracy:

[0027] S5.1. Multi-source data fusion. Use GIS tools to spatially fuse geological, geophysical and geochemical data to generate a comprehensive multi-source data layer, combined with altered mineral and structural identification results to form a comprehensive data set;

[0028] S5.2. Model verification and optimization. Use the comprehensive data set to verify the prediction results of the Polygonum capitatum plant prospecting model, evaluate the reliability and accuracy of the prediction results, and adjust and optimize the model based on the comprehensive analysis results of multi-source data to improve the prediction accuracy of the model;

[0029] S5.3. Field verification and feedback: Based on the comprehensive analysis results, select representative high-confidence areas for field verification, collect geological samples, and verify the accuracy of the prediction results.

[0030] Secondly, the remote sensing identification of Polygonum capitatum plants indicating copper ore prospecting system provided by the present invention is used to implement the remote sensing identification of Polygonum capitatum plants indicating copper ore prospecting method, and the system includes the following modules:

[0031] Remote sensing data acquisition module: responsible for acquiring remote sensing data in the target area, including multispectral, hyperspectral, radar data, etc.;

[0032] Data analysis and processing module: Process and analyze the collected remote sensing data to extract information such as indicator plants, altered minerals, structures, etc. related to copper mines;

[0033] Prospecting model building module: using the extracted feature information to build a prospecting model for Polygonum capitatum plants;

[0034] Multi-source data verification module: Integrate geological, geophysical, geochemical and other multi-source data to verify and optimize the model prediction results.

[0035] Finally, the prospecting equipment for remote sensing identification of Polygonum capitatum plants indicating copper ore provided by the present invention is used to implement the prospecting method for remote sensing identification of Polygonum capitatum plants indicating copper ore, and the equipment comprises:

[0036] Polygonum capitatum plant identifier: used to identify and collect distribution data of Polygonum capitatum plants in the target area. The identifier combines multispectral and hyperspectral remote sensing data to accurately locate the distribution of Polygonum capitatum plants by analyzing the spectral characteristics of the plants;

[0037] Data processor: used to process and analyze the collected remote sensing data and Polygonum capitatum plant distribution data. The data processor can perform data preprocessing, feature extraction and data fusion, providing high-quality data support for model construction and prediction;

[0038] Multi-source data fusion device: used to comprehensively process multi-source data such as Polygonum capitatum vegetation, geology, geophysics and geochemistry, and fuse these data with remote sensing data. The multi-source data fusion device can improve the accuracy and reliability of model predictions and achieve accurate predictions of copper mine distribution through multi-dimensional data analysis. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] The drawings described herein are used to provide a further understanding of the present invention and constitute a part of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0040] Figure 1 It is a flow chart of a prospecting method for remote sensing identification of copper ore indicated by Polygonum capitatum plants provided by the present invention;

[0041] Figure 2 This is a module structure diagram of a prospecting system for remote sensing identification of Polygonum capitatum plants indicating copper deposits provided by the present invention;

[0042] Figure 3 It is a schematic diagram of the structure of the prospecting equipment for remote sensing identification of Polygonum capitatum plants indicating copper ore provided by the present invention;

[0043] Figure 4 One of the multi-source remote sensing image grayscale images of the embodiment;

[0044] Figure 5 The second grayscale image of the multi-source remote sensing image in the embodiment;

[0045] Figure 6 The third grayscale image of the multi-source remote sensing image of the embodiment;

[0046] Figure 7 This is the fourth multi-source remote sensing image grayscale image of the embodiment;

[0047] Figure 8 The fifth grayscale image of the multi-source remote sensing image of the embodiment;

[0048] Fig. 9 The sixth grayscale image of the multi-source remote sensing image of the embodiment;

[0049] Fig.10 is the GF-5 fused image spectral cube map of the embodiment;

[0050] Fig.11 The vegetation distribution map of Polygonum capitatum of the embodiment;

[0051] Fig.12 The distribution map of copper ore alteration minerals in the embodiment;

[0052] Fig.13 It is the 0° direction convolution operation diagram of GF3 of the embodiment;

[0053] Fig.14 It is a 45° direction convolution operation diagram of GF3 of the embodiment;

[0054] Fig.15 90° direction convolution operation diagram of GF3 of the embodiment;

[0055] Fig.16It is the 135° direction convolution operation diagram of GF3 of the embodiment;

[0056] Fig.17 A diagram of a deep learning training area and a verification area of ​​an embodiment;

[0057] Fig.18 This is a mineralization prediction diagram for an embodiment. DETAILED DESCRIPTION

[0058] The specific technical solution of the present invention is described in conjunction with the accompanying drawings.

[0059] Example 1

[0060] like Figure 1 As shown, the method for prospecting copper deposits indicated by remote sensing identification of Polygonum capitatum plants includes the following steps:

[0061] Step 1: Acquire remote sensing data, Polygonum capitatum spectral data, and alteration mineral spectral data related to copper ore in the target area.

[0062] 1.1 Remote sensing data collection:

[0063] Sentinel-2 MSI multispectral image data of the study area were obtained from the Copernicus website https: / / www.copernicus.eu / en;

[0064] The GF2 high spatial resolution image data, GF5 hyperspectral data, GF6 multispectral data and GF3 radar data of the study area were obtained from the website of the National Space Administration https: / / www.cnsa.gov.cn / .

[0065] The ASTER visible light, shortwave infrared, and thermal infrared data of the study area were obtained from the NASA website https: / / search.earthdata.nasa.gov / .

[0066] See Figures 3 to 9 Grayscale image of multi-source remote sensing image.

[0067] 1.2 Spectral data collection:

[0068] The SOC710-VP multispectral imager was used to collect spectral data of Polygonum capitatum plants in the target area. The SOC710VP data acquisition software and SRAnal710e were used to perform data calibration, spectral correlation analysis, image processing, and data viewing.

[0069] Copper ore alteration mineral samples were collected in the study area, and the sample spectral data were collected using CSD350A spectrometer and analyzed using MSAV3.6 mineral spectral analysis expert system MSA.

[0070] Step 2: Preprocessing of remote sensing data and establishment of the Polygonum capitatum plant prospecting model.

[0071] 2.1 Data preprocessing:

[0072] The ENVI software geometric correction module RPC-Orthorectification was used to correct the data of GF2, GF5, and GF6, with reference to Landsat8 OLI data and ALOS-DEM data.

[0073] Radiometric Correction software ENVI is used to perform radiometric calibration (Radiometric Calibration) and FLAASH atmospheric correction on all remote sensing images.

[0074] Using the Gram-Schmidt Pan Sharpening tool in the Image Sharpening module of ENVI software, the panchromatic band uses GF2-PAN images, and the hyperspectral GF5, multispectral GF6, Sentinel-2 MSI, and ASTER data are fused with GS. The fused images not only have improved spatial resolution, but also richer details. Fig.10 GF5 fused image spectral cube map. Due to the fusion of panchromatic image information, the spectral curve of each pixel will change slightly. This change depends on the algorithm used. The Gram-Schmidt method can usually maintain the spectral information better. The key steps are Gram-Schmidt orthogonal transformation and inverse Gram-Schmidt transformation.

[0075] Gram-Schmidt orthogonal transformation: simulate a panchromatic band (PAN simulated band) from the low spatial resolution hyperspectral band of GF5, and combine the simulated panchromatic band and hyperspectral band into a data set. Perform a Gram-Schmidt orthogonal transformation on this combined data set to generate a set of orthogonalized bands. For each original multispectral band, subtract its projection on all previous bands to make it orthogonal to the previous band. The formula is as follows:

[0076]

[0077] in, is the orthogonalized band, is the original band, yes exist projection on

[0078] Inverse Gram-Schmidt transform: Replace the simulated panchromatic band with the GF2 high-resolution panchromatic band, that is, the first band is the high-resolution panchromatic band, and the remaining bands are still orthogonalized hyperspectral bands. Starting from the last orthogonal band, gradually reverse the orthogonalization process and restore each band in turn:

[0079]

[0080] in, is the original band after reconstructing. is the orthogonal band, yes exist Projection on.

[0081] 2.2 Feature Extraction:

[0082] Using spectral analysis technology: Compare the spectral curves of various plants in the study area, the reflectance characteristics in different bands, and determine the spectral characteristics of Polygonum capitatum plants related to copper anomalies. Analyze the mineral spectral data obtained by CSD350, compare it with the known mineral spectral library, establish the altered mineral spectral library in the study area, accurately identify the types and distribution of altered minerals, and identify the main altered mineral types in the target area.

[0083] The specific steps are as follows: The spectral data of the measured Polygonum capitatum and the spectral data of the area of ​​interest in the image are analyzed with the fused Sentinel-2 MSI remote sensing data and GF6 wide-band remote sensing data for spectral angle matching, and the possible distribution of Polygonum capitatum plants in the target area is preliminarily identified. Fig.11 The distribution map of Polygonum capitatum vegetation. The threshold is set according to the actual situation of the study area.

[0084] The distribution of main alteration mineral types is obtained by using GF5 hyperspectral data and spectral unmixing technology Spectral Hourglass Wizard model. The implementation methods include:

[0085] (1) Input the remote sensing data after radiation correction.

[0086] (2) Use MNF transformation to reduce spectral dimension: MNF transformation separates and equalizes the noise in the data, reduces the data dimension for target detection processing, and the data bands after MNF transformation are sorted in descending order of spatial coherence:

[0087]

[0088] Among them C N is the noise covariance matrix, X is the original data, and Y is the denoised data;

[0089]

[0090] Where W is the transformation matrix, the combined weight matrix after PCA and noise removal; Z is the transformed MNF component; the size of W is p×k, p is the feature dimension of the original data, and k is the number of principal components.

[0091] (3) Review the MNF results by visualizing and analyzing the images of each MNF component. The MNF bands are sorted in descending order of importance, and the first MNF component usually contains the most useful information.

[0092] (4) Determine data dimension using spatial coherence measurement:

[0093]

[0094] Where: C is the spatial coherence coefficient; is the characteristic value of each pixel; is the mean.

[0095] (5) Use pixel purity index (PPI) to reduce spatial dimensions:

[0096]

[0097] Where PPI is the pixel purity index, I is the indicator function, indicating whether the pixel exceeds the threshold T, N is the total number of random vectors, is the pixel value after projection.

[0098] (6) Review PPI results: Determine the distribution of pure pixels by visualizing and analyzing the PPI result image. Pixels with higher PPI values ​​represent pure pixels.

[0099] (7) n-dimensional visualization, including automatic clustering, selection and extraction of individual endmembers.

[0100] (8) Spectral mapping using spectral angle classifier (SAM) and / or hybrid tuned matched filter (MTMF):

[0101]

[0102] Where: t is the target spectrum vector, r is the reference spectrum vector (measured mineral spectrum characteristics), Represents the dot product of two vectors. are the magnitudes of the two vectors (i.e. the Euclidean lengths of the vectors).

[0103] when The smaller the value, the more similar the two spectral vectors are, reflecting the similarity between the target spectrum and the reference spectrum. Set a threshold threshold = 0.1, if threshold If it is less than 0.1, the sample is considered to belong to the reference spectrum category.

[0104] (9) Study the mapping results. Compare the relationship between the identified minerals and the minerals in the established spectral library to obtain the distribution location of alteration minerals related to copper ore.

[0105] ASTER multispectral data is used to extract characteristic spectra, and the CEM algorithm is used to obtain the distribution position of altered minerals. CEM (Constrained Energy Minimization) achieves detection by constraining the response of the target and minimizing the energy of other background noise. The formula of CEM is as follows:

[0106]

[0107] in is the target detector weight vector, R is the background covariance matrix, which is usually calculated from the background data in the image, and d is the target spectral vector, that is, the spectral characteristics of the target to be detected. The inverse of the background covariance matrix.

[0108] Target detection: weight vector Applied to each pixel spectral vector x, the target response value of each pixel is calculated:

[0109]

[0110] y is the detection result, that is, the similarity value between the pixel and the target spectrum (usually the larger the y, the more similar it is); X is the spectral vector of each pixel in the input multispectral data.

[0111] See Fig.12 Distribution map of copper-related alteration minerals.

[0112] Structural feature recognition, using radar data to analyze geological structures and determine the location, morphology and scale of the structures. Use convolution and morphological filters to apply convolution filters or morphological filters to image data, and select directional filters: a first-order differential edge enhancement filter that selectively enhances image features with specific directional components (gradients). The sum of the directional filter convolution kernel factors is zero. The result is that in the output image, areas with uniform pixel values ​​are zeroed, while those with changes appear as bright edges, which can significantly enhance the display of fault structures in the study area, using 0°, 90°, 45°, and 135° respectively. The convolution kernels are as follows:

[0113] The horizontal filter (0 degrees) emphasizes the east-west edges and features in the image. The convolution kernel is:

[0114]

[0115] The vertical filter (90 degrees) emphasizes the north-south edges and features in the image. The convolution kernel is:

[0116]

[0117] The diagonal filter (45 degrees) emphasizes the diagonal edges and features in the image from the northeast direction. The convolution kernel is:

[0118]

[0119] The anti-diagonal filter (135 degrees) emphasizes the diagonal edges and features from the northwest in the image. The convolution kernel is:

[0120]

[0121] Finally, filters in multiple directions are combined and used to extract features according to the ROI threshold to construct a classification layer, and then the results are merged to enhance all directional edges in the image.

[0122] See Figure 13 to Figure 16 The results of GF3 convolution operations in different directions.

[0123] 2.3 Model construction:

[0124] Construct a multimodal deep learning network to capture the correlation between different modes. Use an interactive multimodal fusion network to obtain efficient representations from images to Polygonum capitatum plants, structures, and altered minerals. The specific steps are as follows:

[0125] (1) Input processed optical, infrared, SAR, multispectral, and hyperspectral images.

[0126] (2) Multi-domain feature extraction: The data of each modality is mapped to different geometric spaces through a dedicated processing channel, where Euclidean space processes optical data and conventional geometric space is suitable for common remote sensing imaging;

[0127]

[0128] Complex space, for processing complex data such as SAR, used to process signals with phase information (such as interferometric synthetic aperture radar);

[0129]

[0130] Hyperbolic space,processes high-dimensional complex data (such as hyperspectral data).,Hyperbolic space is more suitable for processing hierarchical or multi-level features.

[0131]

[0132] Finally, the features in each space are converted into super-pixel feature tags to represent the features of the local area of ​​the space, which plays a role in feature compression and dimensionality reduction.

[0133]

[0134] is from The extracted superpixel tags, Different modal data - optical, infrared, SAR, hyperspectral.

[0135] (3) Cross-domain fusion: The features extracted from the above different modalities are integrated (such as combining spectrum, morphology, spatial structure, etc.), and the information of different modalities interact and fuse to extract unified features with stronger expressive power to adapt to the characteristics of remote sensing data under different modalities.

[0136]

[0137] in They are spectral, morphological and spatial structure characteristics respectively.

[0138] (4) Self-supervised pre-training: First, construct a layer stack of the fused Sentinal-2-MSI, GF3-SAR, and GF5 hyperspectral remote sensing MNF bands, all sampled to 0.8m. Then adjust the preliminary prediction results based on the field investigation of the mining area, and select representative areas, one part for training and one part for verification. Use self-supervised learning to perform pre-training, and use unlabeled data sets to improve the model's feature extraction capabilities. By comparing the features of different modalities, the model can learn the common information and differences between multimodal data without clear labels. At the same time, the recognition tasks of Polygonum capitate plants, altered minerals, and structures are carried out, and the model is used to identify and classify different targets. See the training area and verification area. Fig.17 .

[0139]

[0140] in is the data embedding of different modalities, is the thermal radiation parameter and Sim is the similarity measure (cosine similarity).

[0141] (5) Deep Learning: Deep learning model training tools include: building deep learning grids, building label gratings from ROIs, extracting attributes and metadata, initializing ENVINet5 Multiclass models, training TensorFlow mask models, and TensorFlow mask classification.

[0142] Here are the steps:

[0143] Data preparation: Select an existing binary classification image or use ROI to collect training samples from the image, or use the ENVI feature counting tool to identify features in the image; Build label grid: Use an existing binary classification image or build a label grid from ROI;

[0144] Model initialization: Initialize ENVI Net5 single-class model, or initialize ENVI Net5 multi-class model;

[0145] Training: Train the TensorFlow Mask model;

[0146] Classification: TensorFlow Mask classification;

[0147] According to the above steps, the entire area is classified and processed after classification to complete the prediction of the copper mineralization area. In the same way, the accuracy is verified using the verification data.

[0148] (6) Weight setting: Based on expert knowledge or experience, each category is compared pairwise to construct a comparison matrix, and the relative importance between the two categories in the comparison matrix is ​​compared. In this way, the mineralization indicator weights between the distribution, structure, and altered minerals of Polygonum capitatum are obtained. The matrix is ​​as follows, and the calculated weight values ​​are 0.3, 0.4, and 0.3 respectively:

[0149]

[0150] (7) Comprehensive prediction and result visualization: Weighted superposition of various factors is performed according to weights to obtain a preliminary comprehensive prediction of mineralization potential areas and visualize the output. Fig.18 Mineralization prediction map.

[0151] Step 3: Integrate geological, geophysical, and geochemical data from multiple sources to further verify the prediction results and improve prediction accuracy.

[0152] 3.1 Multi-source data fusion:

[0153] Using GIS (Geographic Information System) tools, the geological, geophysical and geochemical data are spatially integrated to generate a comprehensive layer of multi-source data. Combined with the altered mineral and structural identification results, a comprehensive data set is formed.

[0154] 3.2 Model validation and optimization:

[0155] Using a comprehensive data set, the prediction results of the Polygonum capitatum plant prospecting model were evaluated and verified for accuracy. The model parameters and fusion methods were adjusted according to the results, and the model was adjusted and optimized to improve the model's prediction accuracy.

[0156] (1) Confusion Matrix:

[0157]

[0158] Among them, TP (True Positives): the number of samples correctly identified as target categories by the model; FP (False Positives): the number of samples incorrectly identified as target categories by the model; FN (False Negatives): the number of samples that the model failed to correctly identify as target categories; TN (True Negatives): the number of samples correctly identified as non-target categories by the model.

[0159] Overall Accuracy (OA) represents the ratio of the number of samples correctly identified by the model to the total number of samples:

[0160]

[0161] User's Accuracy (UA) indicates the proportion of samples correctly identified as the target category:

[0162]

[0163] Producer's Accuracy (PA): Indicates the proportion of samples in the target category that are correctly identified as the target category:

[0164]

[0165] Kappa Coefficient: measures the random consistency of classification results. The Kappa coefficient formula is:

[0166]

[0167]

[0168] in is the actual classification accuracy, that is, the overall accuracy; is the expected random agreement rate.

[0169] (2) ROC curve and AUC (Area Under Curve): The ROC curve is used to describe the performance of the model under different thresholds. The horizontal axis is the false positive rate (FPR) and the vertical axis is the true positive rate (TPR):

[0170]

[0171] AUC is the area under the ROC curve, which indicates the comprehensive classification performance of the model. The closer the AUC is to 1, the better the classification performance of the model.

[0172] 3.3 Field validation and feedback: Based on the comprehensive analysis results, select representative high-confidence areas for field validation, collect geological samples, and verify the accuracy of the prediction results. Compare the field validation results with the model prediction results to further optimize the model. Use cross-validation to adjust the optimization parameters to reduce the possibility of model overfitting and provide a reliable estimate of model performance.

[0173]

[0174] in is the optimization parameter, is the loss function, k is the number of cross validations, are the K-th training and validation datasets, respectively. The model will be trained K times in K iterations, each time using a different training set and validation set. Finally, the results of all K validations will be averaged to get a comprehensive evaluation of the model performance.

[0175] Example 2

[0176] like Figure 2 As shown, the remote sensing identification of Polygonum capitatum plant indicating copper ore prospecting system is used to implement the remote sensing identification of Polygonum capitatum plant indicating copper ore prospecting method, and the system includes the following modules:

[0177] Remote sensing data acquisition module: responsible for acquiring remote sensing data in the target area, including multispectral, hyperspectral, radar data, etc.;

[0178] Data analysis and processing module: Process and analyze the collected remote sensing data to extract information such as indicator plants, altered minerals, structures, etc. related to copper mines;

[0179] Prospecting model building module: using the extracted feature information to build a prospecting model for Polygonum capitatum plants;

[0180] Multi-source data verification module: Integrate geological, geophysical, geochemical and other multi-source data to verify and optimize the model prediction results.

[0181] Example 3

[0182] like Figure 3 As shown, the prospecting equipment for remote sensing identification of Polygonum capitatum plants indicating copper ore is used to implement the prospecting method for remote sensing identification of Polygonum capitatum plants indicating copper ore, and the equipment includes:

[0183] Polygonum capitatum plant identifier: used to identify and collect distribution data of Polygonum capitatum plants in the target area. The identifier combines multispectral and hyperspectral remote sensing data to accurately locate the distribution of Polygonum capitatum plants by analyzing the spectral characteristics of the plants;

[0184] Data processor: used to process and analyze the collected remote sensing data and Polygonum capitatum plant distribution data. The data processor can perform data preprocessing, feature extraction and data fusion, providing high-quality data support for model construction and prediction;

[0185] Multi-source data fusion device: used to comprehensively process multi-source data such as Polygonum capitatum vegetation, geology, geophysics and geochemistry, and fuse these data with remote sensing data. The multi-source data fusion device can improve the accuracy and reliability of model predictions and achieve accurate predictions of copper mine distribution through multi-dimensional data analysis.

Claims

1. A method for prospecting for copper deposits using remote sensing to identify Polygonum capitatum plants, characterized in that: The following steps are involved: S1. Acquire remote sensing data in the target area and establish a spectral database of Polygonum capitatum plants and altered minerals related to copper mines; S2. Use spectral data analysis and processing technology to precisely identify Polygonum capitatum plants and their distribution characteristics; S3. Combine radar, multispectral and hyperspectral remote sensing data to identify and map structural-altered minerals in the target area; S4. Integrate the information of Polygonum capitatum plants and their distribution with the identification and fine mapping of structural-altered minerals, establish a prospecting prediction model based on the indication of copper deposits by Polygonum capitatum plants, delineate the copper anomalies in the target area, and verify the prediction results by deep learning methods; It includes the following sub-steps: S4.

1. Based on expert knowledge and experience, construct a comparison matrix of vegetation, structure and altered minerals and determine the indicator weights; S4.2, Deep learning model: Construct a deep learning model, with input features including synthetic image data that are beneficial to the extraction of Polygonum capitatum plants, structures, and altered mineral bands and preliminary prediction results class1. Train the deep learning model on the training data set, verify the model performance on the test data set, and evaluate the prediction accuracy; Deep learning model training tools include: building deep learning grids, building label gratings from ROIs, extracting attributes and metadata, initializing ENVINet5 Multiclass models, training TensorFlow mask models, and TensorFlow mask classification; Through expert knowledge or experience, each category is compared pairwise, and a comparison matrix is ​​constructed to compare the relative importance of the two categories in the matrix; in this way, the mineralization indication weights between the distribution, structure, and altered minerals of Polygonum capitatum are obtained. The matrix is ​​as follows, and the calculated weight values ​​are 0.3, 0.4, and 0.3 respectively: ， S4.

3. Taking vegetation, structure and altered minerals as prediction indicators, using GIS tools, using determined weights to perform weighted overlay analysis on the indicators, and obtaining preliminary prediction results; S5. Integrate geological, geophysical and geochemical multi-source data and field surveys to further verify the use of remote sensing to identify the predicted target areas of copper deposits indicated by Polygonum capitatum plants and optimize the prospecting model.

2. The method for prospecting for copper deposits indicated by remote sensing identification of Polygonum capitatum plants according to claim 1, characterized in that: S1 specifically includes the following sub-steps: S1.

1. Collect Sentinel-2 MSI remote sensing data, GF2 high spatial resolution remote sensing data, GF5 hyperspectral remote sensing data, GF3 radar data, and ASTER visible light, short wave infrared, and thermal infrared remote sensing data of the study area; S1.2, using SOC710-VP multi-spectral imager to collect the spectral information of the target Polygonum capitatum; S1.

3. Use CSD350A spectrometer to collect spectral information of copper ore alteration minerals.

3. The method for prospecting for copper deposits indicated by remote sensing identification of Polygonum capitatum plants according to claim 1, characterized in that: S2 specifically includes the following sub-steps: S2.

1. Preprocess all remote sensing images, including image geometry correction, radiation correction, and fusion processing of multispectral data with GF2-0.8m panchromatic images to ensure data accuracy and spatial visibility; S2.

2. Use spectral analysis technology to identify the reflectance characteristics of Polygonum capitatum plants in different bands and determine the spectral characteristics related to copper ore; S2.

3. Perform spectral angle matching analysis on the fused Sentinel-2 MSI remote sensing data and the spectral data of Polygonum capitatum to preliminarily identify the possible distribution of Polygonum capitatum plants in the target area.

4. The method for prospecting for copper deposits indicated by remote sensing identification of Polygonum capitatum plants according to claim 1, characterized in that: S3 specifically includes the following sub-steps: S3.

1. Identification of altered minerals: spectral analysis of altered minerals is performed using multispectral and hyperspectral data combined with mineral spectral data acquired by CSD350. Spectral matching algorithms are used to compare the spectral features in remote sensing data with the known mineral spectral library to accurately identify the types and distribution of altered minerals and the main altered mineral types in the target area. S3.

2. Structural feature identification: Use radar data to analyze topography and geological structures to determine the location, shape and scale of the structures.

5. The method for prospecting for copper deposits indicated by remote sensing identification of Polygonum capitatum plants according to claim 1, characterized in that: S5 specifically includes the following sub-steps: S5.

1. Multi-source data fusion: Use GIS tools to spatially fuse geological, geophysical and geochemical data to generate a comprehensive multi-source data layer, combined with altered mineral and structural identification results to form a comprehensive data set; S5.

2. Model verification and optimization: Use comprehensive data sets to verify the prediction results of the Polygonum capitatum plant prospecting model, evaluate the reliability and accuracy of the prediction results, and adjust and optimize the model based on the comprehensive analysis results of multi-source data to improve the prediction accuracy of the model; S5.

3. Field verification and feedback: Based on the comprehensive analysis results, select representative high-confidence areas for field verification, collect geological samples, and verify the accuracy of the prediction results.

6. A remote sensing system for identifying Polygonum capitatum plants as indicators of copper deposits, characterized in that: The system is used to implement the prospecting method of remote sensing identification of Polygonum capitatum plants indicating copper deposits as described in any one of claims 1 to 5, and the system comprises the following modules: Remote sensing data acquisition module: responsible for acquiring remote sensing data in the target area, including multispectral, hyperspectral, and radar data; Data analysis and processing module: process and analyze the collected remote sensing data to extract indicator plants, altered minerals, and structural information related to copper mines; Prospecting model building module: using the extracted feature information to build a prospecting model for Polygonum capitatum plants; Multi-source data verification module: Integrate geological, geophysical and geochemical multi-source data to verify and optimize the model prediction results.

7. A prospecting device for remote sensing identification of Polygonum capitatum plants indicating copper ore, characterized in that: The device is used to implement the prospecting method for remote sensing identification of Polygonum capitatum plants indicating copper deposits as described in any one of claims 1 to 5, and comprises: Polygonum capitatum plant identifier: used to identify and collect distribution data of Polygonum capitatum plants in the target area; the identifier combines multispectral and hyperspectral remote sensing data to accurately locate the distribution of Polygonum capitatum plants by analyzing the spectral characteristics of plants; Data processor: used to process and analyze the collected remote sensing data and Polygonum capitatum plant distribution data; the data processor can perform data preprocessing, feature extraction and data fusion, providing high-quality data support for model construction and prediction; Multi-source data fusion device: used to comprehensively process multi-source data of Polygonum capitatum vegetation, geology, geophysics and geochemistry, and fuse these data with remote sensing data; the multi-source data fusion device realizes accurate prediction of copper mine distribution through multi-dimensional data analysis.

Images