Metallization environment analysis method and system based on aviation hyperspectral remote sensing

Abstract

The invention discloses a metallogenic environment analysis method and system based on aviation hyperspectral remote sensing. The method comprises the steps that an existing stratigraphic map of a rift valley environment, an existing stratigraphic map of a sea ditch environment, an existing stratigraphic map of an island arc environment, an existing stratigraphic map of a mainland marginal sea environment, an existing stratigraphic map of a ground suture line type metallogenic environment and an existing stratigraphic map of a plate internal environment are collected and preprocessed, and then a metallogenic feature map is extracted; constructing a metallogenic characteristic stratigraphic analysis model based on a deep learning architecture, and training the metallogenic characteristic stratigraphic analysis model by using a metallogenic characteristic map; and using aviation hyperspectral remote sensing SASI data of a to-be-detected area and performing preprocessing, inputting the preprocessed data into the trained metallogenic characteristic stratum analysis model to obtain the metallogenic characteristics of the to-be-detected area, and analyzing the metallogenic environment according to the obtained metallogenic characteristics to obtain a metallogenic environment analysis result.

Description

Technical Field

[0001] The present invention relates to the technical field of mineralization environment analysis, and in particular to a mineralization environment analysis method and system based on aerial hyperspectral remote sensing. Background Art

[0002] Aerial hyperspectral remote sensing technology is a cutting-edge technology in the field of remote sensing. It can obtain objects in multiple bands and directly identify objects based on the characteristics of spectral curves, playing an important role in remote sensing geological applications. Aerial hyperspectral remote sensors can simultaneously obtain two-dimensional geometric spatial information and one-dimensional spectral information of the target area, which is three-dimensional data. Therefore, hyperspectral data has the form and structure of an "image cube", reflecting the characteristics and advantages of "graph-spectrum integration", which can well characterize the spectral characteristics and spatial texture information of objects, and has a good recognition effect on mineralization elements such as lithology, alteration, uranium mineralization and surface structure in bedrock exposed areas. However, hyperspectral remote sensing data, like multispectral remote sensing data, also has some limitations: limited penetration depth of the surface; partially or completely covered dead branches and leaves, vegetation coverage increases the difficulty of identifying minerals and rocks; minerals with similar spectral characteristics are not easy to distinguish. Therefore, the effect of directly extracting geological information from hyperspectral data is sometimes affected by surface coverage. Summary of the invention

[0003] In order to solve the above technical problems, the present invention provides a method for analyzing the mineralization environment based on aerial hyperspectral remote sensing, the method comprising:

[0004] Step S1, collecting existing environmental stratigraphic maps, pre-processing them, and then extracting mineralization characteristic maps;

[0005] Step S2, constructing a mineralization characteristic stratum analysis model based on a deep learning architecture, and training the mineralization characteristic stratum analysis model using a mineralization characteristic map;

[0006] Step S3, using the aerial hyperspectral remote sensing SASI data of the area to be measured and preprocessing it, inputting the preprocessed data into the trained mineralization characteristic stratigraphic analysis model to obtain the mineralization characteristics of the area to be measured, analyzing the mineralization environment according to the obtained mineralization characteristics, and obtaining the mineralization environment analysis results.

[0007] Optionally, in step S1, the environmental stratigraphic map includes a stratigraphic map of a rift environment, a stratigraphic map of a trench environment, a stratigraphic map of an island arc environment, a stratigraphic map of a continental marginal sea environment, a stratigraphic map of a suture-type mineralization environment, and a stratigraphic map of an internal plate environment.

[0008] Optionally, in step S2, the mineralization characteristic stratigraphic analysis model includes:

[0009] Step S21, using the PCA algorithm to reduce the dimension of the mineralization feature map, and then inputting it into the ResNet network to extract multi-scale spatial features;

[0010] Step S22, for the spectral features in the mineralization characteristic map, use wavelet transform to perform denoising and then input the spectral features into LSTM to extract the spectral features;

[0011] Step S23: obtaining the metallogenic characteristics of each stratigraphic map based on the spectral characteristics and the multi-scale spatial characteristics.

[0012] Optionally, in step S21, the method of using the PCA algorithm to reduce the dimension of the mineralization feature map and then inputting the map into a Res Net network for multi-scale spatial feature extraction specifically includes:

[0013] S211, using PCA to reduce the dimension of the mineralization characteristic map to obtain reduced dimension data;

[0014] S212, performing regional segmentation on the dimension-reduced data, and downsampling the segmented data using a Gaussian pyramid algorithm to obtain multi-scale and multi-resolution image data;

[0015] S213, inputting the multi-scale and multi-resolution image data into the ResNet network to perform feature extraction, and obtaining dimension space features.

[0016] Optionally, in S212, the dimension reduction data is segmented into regions, and the segmented data is downsampled using a Gaussian pyramid algorithm to obtain the multi-scale and multi-resolution image data, specifically including:

[0017] The segmented image is taken as the bottom layer of the pyramid. The segmented image is convolved and then down-sampled to obtain the image of the previous layer. The convolution and down-sampling are repeated to obtain multi-scale and multi-resolution image data processed by the Gaussian pyramid.

[0018] The present invention also discloses a mineralization environment analysis system based on aerial hyperspectral remote sensing, which includes:

[0019] Feature extraction module, used to collect existing environmental stratigraphic maps and extract mineralization feature maps after preprocessing;

[0020] A model building module, used to build a mineralization characteristic stratum analysis model based on a deep learning architecture, and train the mineralization characteristic stratum analysis model using a mineralization characteristic map;

[0021] The mineralization environment analysis module is used to use the aerial hyperspectral remote sensing SASI data of the area to be tested and preprocess it, input the preprocessed data into the trained mineralization characteristic stratigraphic analysis model to obtain the mineralization characteristics of the area to be tested, analyze the mineralization environment according to the obtained mineralization characteristics, and obtain the mineralization environment analysis results.

[0022] Optionally, in the feature extraction module, the environmental stratigraphic map includes a stratigraphic map of a rift environment, a stratigraphic map of a trench environment, a stratigraphic map of an island arc environment, a stratigraphic map of a continental marginal sea environment, a stratigraphic map of a suture-type mineralization environment, and a stratigraphic map of an internal plate environment.

[0023] Optionally, the model building module further includes a spatial feature extraction submodule, a spectral feature extraction submodule and a feature fusion submodule;

[0024] The spatial feature extraction submodule is used to reduce the dimension of the mineralization feature map using the PCA algorithm and then input it into the ResNet network to extract multi-scale spatial features;

[0025] The spectral feature extraction submodule is used to extract spectral features from the mineralization feature map by denoising the spectral features using wavelet transform and then inputting them into LSTM;

[0026] The feature fusion submodule is used to obtain the metallogenic characteristics of each stratigraphic map based on the spectral characteristics and the multi-scale spatial characteristics.

[0027] Optionally, the workflow of the spatial feature extraction submodule specifically includes:

[0028] Use PCA to reduce the dimension of the mineralization characteristic map to obtain reduced-dimensional data;

[0029] Performing regional segmentation on the dimension-reduced data, and downsampling the segmented data using a Gaussian pyramid algorithm to obtain multi-scale and multi-resolution image data;

[0030] The multi-scale and multi-resolution image data are input into the ResNet network for feature extraction to obtain dimension space features.

[0031] Optionally, the process of obtaining multi-scale and multi-resolution image data specifically includes:

[0032] The segmented image is taken as the bottom layer of the pyramid. The segmented image is convolved and then down-sampled to obtain the image of the previous layer. The convolution and down-sampling are repeated to obtain multi-scale and multi-resolution image data processed by the Gaussian pyramid.

[0033] Compared with the prior art, the present invention has the following beneficial effects:

[0034] The aerial hyperspectral remote sensing used in the present invention can quickly collect data in a large range, covers a wide area, and is helpful for systematic analysis of the entire mineralization area or potential mineralization belt, and its efficiency is much higher than that of traditional ground exploration methods. For aerial hyperspectral images, there are many bands, the correlation between adjacent bands is large, and there is a high degree of information redundancy. In addition, the amount of aerial hyperspectral image data is large, requiring a large amount of computing resources, and because of the correlation between adjacent pixels of the same type, classification is also difficult. The present invention combines spectral features with spatial features to achieve feature extraction of mineralization spectral images, thereby improving the detection accuracy of mineralization environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] In order to more clearly illustrate the technical solution of the present invention, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0036] Figure 1 The figure is a schematic diagram of a method flow of an embodiment of the present invention. DETAILED DESCRIPTION

[0037] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0038] Embodiment 1

[0039] A method for analyzing the mineralization environment based on aerial hyperspectral remote sensing, such as Figure 1 As shown, the method includes:

[0040] Step S1, collect existing environmental stratigraphic maps, perform preprocessing, and then extract mineralization characteristic maps. Environmental stratigraphic maps include stratigraphic maps of rift environments, stratigraphic maps of trench environments, stratigraphic maps of island arc environments, stratigraphic maps of continental marginal sea environments, stratigraphic maps of suture-type mineralization environments, and stratigraphic maps of plate internal environments.

[0041] For rift environments, including oceanic crust rift environments, copper-zinc pyrite deposits, chromite deposits and copper-nickel sulfide deposits are easily formed; continental crust rift environments are prone to forming layered Pb, Zn, Fe, Mn, Ba deposits, and young continental crust rift environments are prone to forming greenstone belt gold deposits. Trench environments are prone to producing Cr, Pt, Cu-Ni, Ti, Fe deposits and pyrite deposits. Island arc environments are prone to generating porphyry Cu-Mo-Au deposits and magmatic deposits related to potassium-rich basic rocks. Continental marginal sea environments are prone to generating non-ferrous and rare metal minerals. Suture-type mineralization environments are prone to generating chromite deposits, copper-nickel sulfides and porphyry copper deposits. Plate internal environments generally produce layered and layer-controlled Cu, Pb, and Zn sulfide deposits.

[0042] Step S2: construct a mineralization characteristic stratigraphic analysis model based on a deep learning architecture, and use a mineralization characteristic map to train the mineralization characteristic stratigraphic analysis model.

[0043] The mineralization characteristic stratum analysis model includes: step S21, using the PCA algorithm to reduce the dimension of the mineralization characteristic map, and then inputting it into the ResNet network for multi-scale spatial feature extraction; S211, using PCA to reduce the dimension of the mineralization characteristic map to obtain reduced dimension data; specifically, PCA projects the high-dimensional data of the mineralization characteristic map into a new coordinate system, and the coordinate axes of the new coordinate system are arranged in descending order of the variance of the data. These principal components capture the main information and variability in the data, so as to find the most representative features while reducing the dimension. S212, performing regional segmentation on the reduced dimension data, using the Gaussian pyramid algorithm to downsample the segmented data, and obtaining multi-scale multi-resolution image data; using the segmented image as the bottom layer of the pyramid, convolving the segmented image and then downsampling it to obtain the upper layer image, repeating the convolution and downsampling, and obtaining multi-scale multi-resolution image data processed by the Gaussian pyramid. S213, inputting the multi-scale multi-resolution image data into the ResNet network for feature extraction to obtain the size space feature. The image after dimensionality reduction and the image after Gaussian pyramid cutting are input into the ResNet network for feature extraction. Although the image has undergone PCA dimensionality reduction, due to the large number of image dimensions, the number of layers after dimensionality reduction is not particularly small, so it is more appropriate to use the ResNet network to extract spatial features. The multi-scale image is obtained by downsampling the processed image by executing the same Gaussian pyramid steps, and the same processing is applied to the mineralization feature image so that HSI can be analyzed at multiple scales, and different spatial features can be obtained at different scales. Step S22, for the spectral features in the mineralization feature map, use wavelet transform to denoise and then input into LSTM to extract spectral features; use the spectral curve data after wavelet transform as sequence data for training the LSTM network, and improve the accuracy of the model by using the forget gate. Step S23, obtain the mineralization characteristics of each stratigraphic map based on the spectral features and the multi-scale spatial features.

[0044] Step S3, using the aerial hyperspectral remote sensing SASI data of the area to be measured and preprocessing it, inputting the preprocessed data into the trained mineralization characteristic stratigraphic analysis model to obtain the mineralization characteristics of the area to be measured, analyzing the mineralization environment according to the obtained mineralization characteristics, and obtaining the mineralization environment analysis results.

[0045] Embodiment 2

[0046] A mineralization environment analysis system based on aerial hyperspectral remote sensing includes: a feature extraction module, a model building module and a mineralization environment analysis module;

[0047] The feature extraction module is used to collect the existing environmental stratigraphic map and extract the mineralization feature map after preprocessing.

[0048] In the feature extraction module, the environmental stratigraphic map includes the stratigraphic map of the rift environment, the stratigraphic map of the trench environment, the stratigraphic map of the island arc environment, the stratigraphic map of the continental marginal sea environment, the stratigraphic map of the suture-type mineralization environment, and the stratigraphic map of the internal environment of the plate.

[0049] The model building module is used to build a mineralization characteristic stratigraphic analysis model based on a deep learning architecture, and to train the mineralization characteristic stratigraphic analysis model using a mineralization characteristic map.

[0050] The model construction module also includes a spatial feature extraction submodule, a spectral feature extraction submodule and a feature fusion submodule; the spatial feature extraction submodule is used to use the PCA algorithm to reduce the dimension of the mineralization feature map, and then input it into the ResNet network for multi-scale spatial feature extraction; the spectral feature extraction submodule is used to use wavelet transform to denoise the spectral features in the mineralization feature map, and then input it into the LSTM to extract spectral features; the feature fusion submodule is used to obtain the mineralization characteristics of each stratigraphic map based on the spectral features and the multi-scale spatial features.

[0051] The working process of the spatial feature extraction submodule specifically includes: using PCA to reduce the dimension of the mineralization feature map to obtain reduced dimension data; performing regional segmentation on the reduced dimension data, and using the Gaussian pyramid algorithm to downsample the segmented data to obtain multi-scale and multi-resolution image data; inputting the multi-scale and multi-resolution image data into the ResNet network for feature extraction to obtain dimension space features.

[0052] The mineralization environment analysis module is used to use the aerial hyperspectral remote sensing SASI data of the area to be tested and preprocess it, input the preprocessed data into the trained mineralization characteristic stratigraphic analysis model to obtain the mineralization characteristics of the area to be tested, analyze the mineralization environment according to the obtained mineralization characteristics, and obtain the mineralization environment analysis results.

[0053] The embodiments described above are only descriptions of the preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Without departing from the design spirit of the present invention, various modifications and improvements made to the technical solutions of the present invention by ordinary technicians in this field should all fall within the protection scope determined by the claims of the present invention.

Claims

1. A method for analyzing mineralization environment based on aerial hyperspectral remote sensing, characterized in that: The method comprises: Step S1, collecting existing environmental stratigraphic maps, pre-processing them and extracting mineralization characteristic maps; Step S2, constructing a mineralization characteristic stratum analysis model based on a deep learning architecture, and training the mineralization characteristic stratum analysis model using a mineralization characteristic map; Step S3, using the aerial hyperspectral remote sensing SASI data of the area to be measured and preprocessing it, inputting the preprocessed data into the trained mineralization characteristic stratigraphic analysis model to obtain the mineralization characteristics of the area to be measured, analyzing the mineralization environment according to the obtained mineralization characteristics, and obtaining the mineralization environment analysis results.

2. The method for analyzing the mineralization environment based on aerial hyperspectral remote sensing according to claim 1, characterized in that: In step S1, the environmental stratigraphic map includes a stratigraphic map of a rift environment, a stratigraphic map of a trench environment, a stratigraphic map of an island arc environment, a stratigraphic map of a continental marginal sea environment, a stratigraphic map of a suture-type mineralization environment, and a stratigraphic map of an internal plate environment.

3. The method for analyzing the mineralization environment based on aerial hyperspectral remote sensing according to claim 1, characterized in that: In step S2, the mineralization characteristic stratum analysis model includes: Step S21, using the PCA algorithm to reduce the dimension of the mineralization feature map, and then inputting it into the ResNet network to extract multi-scale spatial features; Step S22, for the spectral features in the mineralization characteristic map, use wavelet transform to perform denoising and then input the spectral features into LSTM to extract the spectral features; Step S23: obtaining the metallogenic characteristics of each stratigraphic map based on the spectral characteristics and the multi-scale spatial characteristics.

4. The method for analyzing the mineralization environment based on aerial hyperspectral remote sensing according to claim 3 is characterized in that: Step S21, after reducing the dimension of the mineralization characteristic map by using the PCA algorithm, the method of inputting the map into the ResNet network for multi-scale spatial feature extraction specifically includes: S211, using PCA to reduce the dimension of the mineralization characteristic map to obtain reduced dimension data; S212, performing regional segmentation on the dimension-reduced data, and downsampling the segmented data using a Gaussian pyramid algorithm to obtain multi-scale and multi-resolution image data; S213, inputting the multi-scale and multi-resolution image data into the ResNet network to perform feature extraction, and obtaining dimension space features.

5. The method for analyzing the mineralization environment based on aerial hyperspectral remote sensing according to claim 4 is characterized in that: In S212, the dimension reduction data is segmented into regions, and the segmented data is downsampled using a Gaussian pyramid algorithm to obtain multi-scale and multi-resolution image data, which specifically includes: The segmented image is taken as the bottom layer of the pyramid. The segmented image is convolved and then down-sampled to obtain the image of the previous layer. The convolution and down-sampling are repeated to obtain multi-scale and multi-resolution image data processed by the Gaussian pyramid.

6. A mineralization environment analysis system based on aerial hyperspectral remote sensing, the system applying the environmental analysis method according to any one of claims 1 to 5, characterized in that: The system includes: Feature extraction module, used to collect existing environmental stratigraphic maps and extract mineralization feature maps after preprocessing; A model building module, used to build a mineralization characteristic stratum analysis model based on a deep learning architecture, and train the mineralization characteristic stratum analysis model using a mineralization characteristic map; The mineralization environment analysis module is used to use the aerial hyperspectral remote sensing SASI data of the area to be tested and preprocess it, input the preprocessed data into the trained mineralization characteristic stratigraphic analysis model to obtain the mineralization characteristics of the area to be tested, analyze the mineralization environment according to the obtained mineralization characteristics, and obtain the mineralization environment analysis results.

7. The mineralization environment analysis system based on aerial hyperspectral remote sensing according to claim 6 is characterized in that: In the feature extraction module, the environmental stratigraphic map includes the stratigraphic map of the rift environment, the stratigraphic map of the trench environment, the stratigraphic map of the island arc environment, the stratigraphic map of the continental marginal sea environment, the stratigraphic map of the suture line mineralization environment and the stratigraphic map of the plate internal environment.

8. The mineralization environment analysis system based on aerial hyperspectral remote sensing according to claim 6 is characterized in that: The model building module also includes a spatial feature extraction submodule, a spectral feature extraction submodule and a feature fusion submodule; The spatial feature extraction submodule is used to reduce the dimension of the mineralization feature map using the PCA algorithm and then input it into the ResNet network to extract multi-scale spatial features; The spectral feature extraction submodule is used to extract spectral features from the mineralization feature map by denoising the spectral features using wavelet transform and then inputting them into LSTM; The feature fusion submodule is used to obtain the metallogenic characteristics of each stratigraphic map based on the spectral characteristics and the multi-scale spatial characteristics.

9. The mineralization environment analysis system based on aerial hyperspectral remote sensing according to claim 8 is characterized in that: The workflow of the spatial feature extraction submodule specifically includes: Use PCA to reduce the dimension of the mineralization characteristic map to obtain reduced-dimensional data; Performing regional segmentation on the dimension-reduced data, and downsampling the segmented data using a Gaussian pyramid algorithm to obtain multi-scale and multi-resolution image data; The multi-scale and multi-resolution image data are input into the ResNet network for feature extraction to obtain dimension space features.

10. The mineralization environment analysis system based on aerial hyperspectral remote sensing according to claim 9 is characterized in that: The process of obtaining multi-scale and multi-resolution image data specifically includes: The segmented image is taken as the bottom layer of the pyramid. The segmented image is convolved and then down-sampled to obtain the image of the previous layer. The convolution and down-sampling are repeated to obtain multi-scale and multi-resolution image data processed by the Gaussian pyramid.

Images