A lithologic classification method based on spectral spatial residual network

Abstract

The present invention relates to a lithologic classification method based on a spectrum-space residual network, comprising the following steps: collecting multi-source geological data from a surveyed area, processing the multi-source geological data to form an image to be classified; marking the locations of lithologic samples and lithologic category codes on the image to be classified; extracting lithologic samples from the image to be classified and testing the independence of the samples; establishing a lithologic classification model based on a spectrum-space residual network architecture, setting model hyperparameters, training the model, analyzing the model's accuracy and loss value indicators, and testing the robustness of the model; extracting lithologic information from the image to be classified, delineating and correcting lithologic unit boundaries; and splicing the lithologic unit boundaries in the classification result raster image with the boundaries of non-rock exposed areas to compile a complete lithologic classification result map for the surveyed area. The present invention achieves accurate and efficient lithologic classification, which is beneficial for assisting geologists in better conducting regional geological surveys.

Description

Technical Field

[0001] The present invention relates to a method for assisting regional geological survey with multi-source data, in particular to a lithology classification method based on spectrum space residual network. Background Art

[0002] Existing lithologic classification methods can be broadly categorized into three types: unsupervised, semi-supervised, and supervised. Unsupervised classification methods do not pre-set samples; instead, they primarily classify lithologic features based on internal data feature differences. Representative algorithms include K-Means and ISODATA. Semi-supervised classification methods combine internal data feature differences with expert experience to set a small number of samples, which are then used to classify lithologic features. Representative algorithms include self-training and semi-supervised support vector machines. Supervised classification methods first set a certain number of samples based on expert experience and then classify lithologic features based on these samples. Representative algorithms include maximum likelihood and neural networks.

[0003] Existing lithologic classification methods are mostly based on mathematical statistics or shallow machine learning. Their algorithms are relatively simple, and their classification models are poorly adapted to complex lithologic features. Consequently, lithologic classification in areas with complex geological conditions currently suffers from low accuracy. Summary of the Invention

[0004] The purpose of the present invention is to provide a lithology classification method based on spectrum spatial residual network to solve the problem of low classification accuracy in existing lithology classification methods.

[0005] The object of the present invention is achieved like this:

[0006] A lithology classification method based on spectrum spatial residual network includes the following steps:

[0007] S1. Data collection: Collect multi-source geological data including geological data, aeromagnetic and aerial radio data, and hyperspectral remote sensing images in the surveyed area.

[0008] S2. Data processing: Perform routine processing on the collected multi-source geological data, then select data channels and normalize the image data to unify the format, projection, grid density and spatial range of the image data, and combine the processed images of different types into a multi-channel image; circle the non-rock exposed areas including water bodies, alluvial deposits, artificial buildings and engineering distribution areas on the multi-channel image, and remove the data of the non-rock exposed areas from the multi-channel image to form an image to be classified of the measured area.

[0009] S3. Create a lithologic sample set: Establish discriminant marks for various lithologies on the image to be classified, and mark the location of the lithologic samples and their lithologic category codes on the image to be classified; extract lithologic samples from the image to be classified based on the marked lithologic sample locations, and test the independence of the samples; perform sorting operations on the lithologic samples, including sample shuffling, expanding three-dimensional samples, and sample data normalization, and divide the sorted lithologic samples into a training set, a validation set, and a test set in proportion to form a lithologic sample set.

[0010] S4. Construct a lithology classification model: Build a lithology classification model based on the spectrum-space residual network architecture. Adjust the input port, convolution kernel size, number of convolution layers, and output port of the lithology classification model according to the sample structure, and initialize the model hyperparameters.

[0011] S5. Training model: Use the training set and validation set to train the lithology classification model, and use the test set to test the robustness of the model; comprehensively evaluate the classification performance of the lithology classification model based on the accuracy and loss value indicators of the model on the training set and validation set, as well as the confusion matrix classification accuracy on the test set. If the performance analysis results do not meet the standards, adjust the model hyperparameters item by item and repeat this step; until the performance analysis results meet the standards, determine the model hyperparameters at this time, and save the model.

[0012] S6. Output lithologic classification results: Use the Gdal library in Python to load the image to be classified, extract the lithologic information and geographic parameter information including grid size, number of rows and columns, and spatial projection from the image to be classified; normalize the extracted image data information and then input it into the lithologic classification model, which outputs the lithologic classification results expressed in Roman numerals; use the Gdal library to convert the output lithologic classification results into a classification result raster image with the same geographic parameter information as the image to be classified; import the classification result raster image into ArcGIS software to outline and correct the lithologic unit boundaries.

[0013] S7. Prepare a lithologic classification result map: Splice together the boundaries of the lithologic units in the classification result raster image with the boundaries of the non-rock exposed areas to prepare a complete lithologic classification result map of the measured area. If there are areas in the lithologic classification result map of the measured area that are inconsistent with the existing geological knowledge or areas where the lithologic classification results are questionable, conduct a field survey and revise the lithologic classification result map of the measured area based on the survey results.

[0014] Based on the systematic collection of multi-source geological data from the surveyed area, the present invention first forms images to be classified based on the processed data. The images to be classified are then used to create a lithologic sample set. A lithologic classification model is then constructed based on a spectral spatial residual network. The model is trained, and the model with the best classification performance is saved. The optimal model is used to output a raster image of the lithologic classification results, and the boundaries of the lithologic units are delineated. Finally, a comprehensive lithologic classification result map of the surveyed area is compiled. This invention achieves rapid, accurate, intelligent, and efficient lithologic classification under complex geological conditions, overcoming the low accuracy of existing lithologic classification methods and helping geologists better conduct regional geological surveys. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 It is a flowchart of the lithologic classification method of the present invention.

[0016] Figure 2 It is a structural diagram of the lithologic classification model. DETAILED DESCRIPTION

[0017] Limitations of the present invention: The lithology classification involved in the present invention is limited to image data.

[0018] The present invention will be further described below in conjunction with the accompanying drawings and examples.

[0019] like Figure 1 As shown, the lithology classification method of the present invention includes the following steps:

[0020] Step 1: Collect data.

[0021] Collect multi-source geoscience data from the surveyed area, including geological data, aeromagnetic and aerial radiographic data, and hyperspectral remote sensing imagery. This example selects the Longshou Mountain area as the surveyed area, collecting geological data, aeromagnetic and aerial radiographic data, and hyperspectral remote sensing imagery from this area. Theoretically, the more diverse the data types, the richer the physical property information provided. The geological data collected in this example consist of a 1:50,000 regional geological report and related maps, the aeromagnetic and aerial radiographic data consist of 1:50,000 high-precision aeromagnetic and aerial radiographic data, and the hyperspectral remote sensing imagery is from the ZY1-02D ​​satellite. It should be noted that the geological and remote sensing data used in this example are all unprocessed raw data, while the aeromagnetic and aerial radiographic data consist of standard database files that have undergone field preprocessing.

[0022] Step 2: Process the data.

[0023] 1. Perform routine processing on the collected multi-source geological data. Routine processing operations include: vectorization of geological maps for geological data; cartographic conversion of aeromagnetic and aerial radio data; and radiometric correction, atmospheric correction, orthorectification, geometric precision correction, and abnormal band detection for hyperspectral remote sensing images.

[0024] In this embodiment, Mapgis and Arcgis software were used to vectorize the geological map. According to the technical regulations for remote sensing image preprocessing, ENVI software was used to perform radiometric correction, atmospheric correction, orthorectification, geometric correction, and abnormal band detection on the ZY1-02D ​​satellite hyperspectral image. According to the aeromagnetic and airborne data processing procedures, Geosoft software was used to perform cartographic conversion processing on 1:50,000 high-precision aeromagnetic and airborne data. Seven conversion maps were produced based on the aeromagnetic data, including the aeromagnetic ΔT polar plane, the aeromagnetic ΔT polar upward extension plane, the aeromagnetic ΔT polar horizontal gradient mode plane, the aeromagnetic ΔT polar vertical first-order derivative plane, and the aeromagnetic ΔT polar oblique derivative plane. Seven conversion maps were also produced based on the aeromagnetic data, including the aeromagnetic total volume contour plane, the aeromagnetic potassium content contour plane, the aeromagnetic uranium content contour plane, the aeromagnetic thorium content contour plane, the aeromagnetic uranium / potassium contour plane, the aeromagnetic uranium / thorium contour plane, and the aeromagnetic thorium / potassium contour plane.

[0025] 2. Optimize data channels and normalize image data to unify the format, projection, grid density, and spatial range of image data; synthesize different types of processed images into a multi-channel image; circle non-rock exposed areas including water bodies, alluvial deposits, artificial buildings, and engineering distribution areas on the multi-channel image within the measured area, and remove the data of non-rock exposed areas on the multi-channel image to form an image to be classified of the measured area.

[0026] Hyperspectral remote sensing imagery has numerous bands, and information from different bands is highly correlated. This also applies to the various conversion maps from aeromagnetic and aerial radiography. To reduce the impact of inter-data correlation on analysis results, this example combines the Optimum Index Factor (OIF) method and correlation analysis to optimize data channels. Ultimately, the hyperspectral remote sensing data contributed 52 data channels, the aeromagnetic conversion maps contributed 6 data channels, and the aerial radiography conversion maps contributed 6 data channels, for a total of 64 data channels.

[0027] The parameters of the hyperspectral remote sensing and aeromagnetic / airborne data were standardized using a combination of ENVI, ArcGIS, and Geosoft software. All data were formatted as raster image files, with a WGS_1984_UTM_Zone_47N projection and a grid size of 30 m x 30 m. The spatial extent was uniformly adjusted to match the measured area. Finally, the Composite Bands tool in ArcGIS was used to overlay the hyperspectral remote sensing and aeromagnetic / airborne data channels into a single multi-channel image, totaling 64 channels.

[0028] Step 3: Prepare a lithology sample set.

[0029] 1. Establish the discriminant marks for each type of lithology on the image to be classified, and mark the location of the lithology samples and their lithology category codes on the image to be classified. The specific operation method for marking the location of lithology samples on the image to be classified is: according to the discriminant marks of each type of lithology on the image to be classified, use Arcgis software to mark the location of the lithology samples on the image to be classified, save the sample location as a vector surface file, and add an attribute field to store the lithology category code in the vector surface file. Then convert the vector file into raster form and record it as a label image. The lithology category code is represented by Roman numerals starting from "I", each numeral represents a lithology category, and the lithology category where there is no marked sample on the image to be classified is "0".

[0030] The method for establishing distinguishing markers for each type of lithology on the image to be classified is to conduct appropriate field surveys based on existing geological data for the surveyed area, summarize the main rock types in the surveyed area, and then establish distinguishing markers for each type of lithology on the image to be classified. If necessary, the survey should also collect typical rocks for rock and mineral identification experiments to assist personnel in determining the type of target lithology.

[0031] In this example, the surveyed area primarily contains six lithologies: alkaline diamictite, granite, phyllite, schist, marble, and siliceous limestone. In terms of spatial distribution, alkaline diamictite and granite are concentrated and have large outcrops, followed by phyllite, schist, and marble. Siliceous limestone, constrained by fault structures, is distributed in a banded pattern and has a smaller area. Overall, the geological conditions in the surveyed area are relatively complex.

[0032] In terms of marking the sample positions, the six rock types, namely alkaline melange, granite, phyllite, schist, marble and siliceous limestone, were coded with Roman numerals I to X, and counted according to the number of pixels. The number of rock type samples initially marked was 278, 512, 389, 123, 278 and 240, respectively, totaling 1,820.

[0033] 2. Based on the locations of the marked lithologic samples, extract lithologic samples from the image to be classified and test their independence to ensure that the sample quality meets the requirements. Specifically, use ENVI software to calculate the transformed divergence of the marked lithologic samples and test their independence based on the calculated results. If the independence test fails to meet the requirements, reselect the lithologic samples and repeat the independence test until all lithologic samples meet the independence requirements. Specifically, use the Compute ROI Separability tool in ENVI software to calculate the transformed divergence (Transformed Divergence) between the marked lithologic samples. The results show that the transformed divergence is greater than 1.9, indicating that the sample independence meets the requirements.

[0034] 3. Perform sorting operations on the lithologic samples, such as sample shuffling, expanding three-dimensional samples, and sample data normalization. By dividing the samples, a lithologic sample set consisting of a training set, a validation set, and a test set is formed.

[0035] 3.1 Sample shuffling operation for lithologic samples. Since there may be a certain spatial correlation between adjacent samples when labeling samples, the order of lithologic samples is shuffled to avoid interference of spatial correlation on the analysis results. The specific operation method for shuffling lithologic samples is: use the shuffle() function of the random module in Python to shuffle the order of the lithologic sample positions. Specifically, use the Gdal library in Python to load the label image and convert it into array form using the Numpy library. The values ​​in the array correspond one-to-one to the values ​​on the label image. Traverse the points in the array whose values ​​are not 0 and save their position information in the form of a list. Then use the shuffle() function of the random module to shuffle the list. In this way, the order of the samples is shuffled.

[0036] 3.2 The operation of expanding and selecting three-dimensional samples for lithologic samples aims to fully utilize the spectral and spatial information of the selected data to achieve better lithologic classification results. Specifically, a three-dimensional lithologic sample is generated according to a three-dimensional data structure of "one-dimensional spectrum × two-dimensional space." Based on the sample shuffle, assuming the new position of a sample is A, a three-dimensional range is defined, centered on the pixel at A, to cover all data channels spectrally and spatially expand to an appropriate size. All data within this range is recorded as the data for that sample. In other words, the sample data structure can be represented as C × L × L, where C represents the number of spectral channels and L represents the spatial size. Specifically, the image to be classified is loaded using the Gdal library in Python and converted to array form using the Numpy library. The new position of the sample is used to match and extract the corresponding information in the array. Based on the data characteristics and hardware and software requirements, the data structure of a single lithologic sample in this embodiment is set to 64 × 5 × 5, meaning that the spectral dimension has 64 data channels and the spatial dimension is a 5 × 5 grid. Considering that the total number of samples is 1820, the structure of the lithologic sample set prepared in this embodiment can be expressed as 1820×64×5×5.

[0037] 3.3 The normalization operation of the sample data of lithologic samples is because the dimensions and ranges of the data of each channel often vary greatly, which usually has a significant impact on the analysis efficiency and accuracy. Therefore, the sample data should be normalized. Normalization can reduce the range of the sample data to a reasonable range and eliminate dimensional differences, significantly improving the analysis effect.

[0038] To ensure the analysis effect, the present invention recommends the use of the maximum-minimum normalization method, which can convert the data range into the [0,1] interval. The calculation formula is:

[0039]

[0040] Where C is a channel in the sample data, Cx is the initial data of the channel, Cmin is the minimum value in the initial data, and Cmax is the maximum value in the initial data. is the result value after normalization.

[0041] The specific method of sample data normalization is to use the max() function and min() function of the Numpy library in Python to calculate the maximum and minimum values ​​of a single data channel, and then perform maximum-minimum value normalization on this basis. Finally, the values ​​are converted to the [0,1] interval.

[0042] 3.4 The organized lithologic samples are divided into a training set, a validation set, and a test set according to a certain ratio. The training set and validation set are used to train the lithologic classification model, and the test set is used to test the robustness of the model. In this example, the split() function in Python is used to divide the samples according to the ratio of "training set: validation set: test set = 6:2:2". The specific number of lithologic samples in the training set, validation set, and test set is: 1092, 364, and 364 respectively. This completes the process of creating the lithologic sample set, forming the lithologic sample set.

[0043] Step 4: Construct a lithology classification model.

[0044] A lithology classification model was established based on the spectral spatial residual network architecture. The input port, convolution kernel size, number of convolution layers, and output port of the lithology classification model were adjusted according to the sample structure, and the model's hyperparameters were initialized. The classic spectral spatial residual network architecture has three core components: a three-dimensional convolution block, a spectral residual block, and a spatial residual block. This paper constructed the lithology classification model using the PyTorch development framework.

[0045] like Figure 2 As shown, the lithology classification model constructed by the present invention refers to the classic spectral spatial residual network structure and is appropriately adjusted according to the structure of the lithology sample. From front to back, the constructed lithology classification model includes an input layer, a convolutional layer, two spectral residual blocks, two convolutional layers, two spatial residual blocks, a pooling layer, and a fully connected layer. Its specific structure is: the input port structure is 5×5×64, the spectral convolution kernel size is 1×1×7×28, the spatial convolution kernel size is 3×3×28, and the output port structure is 1×1×6. At the same time, the model hyperparameters including learning rate, batch size, number of iterations, optimizer, activation function, and loss function are initialized and set according to conventional methods.

[0046] Step 5: Train the model.

[0047] The lithology classification model is trained using sample data from the training and validation sets, and the model's robustness is tested using the test set. The lithology classification model's classification performance is comprehensively evaluated based on the model's accuracy and loss metrics on the training and validation sets, as well as the confusion matrix classification accuracy on the test set. If the performance analysis results do not meet the requirements, adjust the model hyperparameters one by one and repeat this step until the performance analysis results meet the requirements. The model hyperparameter combination at this point is determined, and the best-performing model is saved.

[0048] The operation method for testing the robustness of the model is as follows: input the samples of the test set into the lithology classification model, and then extract the sample data and label part of the test set; use the sample data to perform lithology classification processing, and calculate the classification accuracy of the lithology classification model for the lithology samples in the test set based on the classification results and the corresponding actual labels, and test the robustness of the model based on the classification accuracy; if the robustness of the model does not meet the standard, reselect the lithology samples and repeat steps 3 to 5 until the robustness of the model meets the standard.

[0049] A confusion matrix is ​​constructed based on the classification results and the corresponding actual labels. The classification accuracy of the lithology classification model on the test set is evaluated based on the confusion matrix. Evaluation metrics include accuracy, precision, recall, and F1 score.

[0050] 1. The calculation formula of the accuracy index is:

[0051]

[0052] Among them, ACC represents accuracy, TP represents true positive, TN represents true negative, FP represents false positive, and FN represents false negative.

[0053] 2. The calculation formula of the precision rate indicator is:

[0054]

[0055] Among them, PPV represents precision, TP represents true positive, and FP represents false positive.

[0056] 3. The calculation formula of recall rate is:

[0057]

[0058] Among them, TRP represents recall rate, TP represents true positive examples, and FN represents false negative examples.

[0059] 4. The calculation formula of F1 score indicator is:

[0060]

[0061] Among them, F1 represents F1 score, PPV represents precision, and TRP represents recall.

[0062] Dynamic adjustments were performed on six key hyperparameters: learning rate, batch size, number of iterations, optimizer, activation function, and loss function. This adjustment method employed a controlled variable approach, whereby only one hyperparameter was modified at a time. The parameter combination with the best classification performance was then selected for the next round of adjustments. The model's classification performance was comprehensively evaluated based on the accuracy and loss values ​​on the training and validation sets, as well as the confusion matrix classification accuracy on the test set. Generally, higher accuracy, smaller fluctuations, lower loss values, and faster convergence on the training and validation sets indicate better model training. Higher confusion matrix classification accuracy on the test set indicates greater model robustness.

[0063] After tuning parameters in this example, the optimal hyperparameter combination was determined to be: learning rate = 0.001, batch size = 32, iterations = 200, optimizer = RMSProp, activation function = ReLU, and loss function = Cross Entropy. The model achieved optimal classification accuracy of 97.50% for the training set and 93.93% for the validation set. The overall accuracy of the model for lithology classification on the test set was 92.31%, with a precision of no less than 88.00% for all six lithologies, a recall of no less than 90.24%, and an F1 score of no less than 0.89 (see Table 1). Finally, the optimal model hyperparameters were saved as a model file with the suffix pth using the save() function in the Torch library in Python.

[0064] Table 1. Lithology classification accuracy calculation results

[0065]

[0066] Step 6: Output the lithology classification results.

[0067] The GDAL library in Python was used to load the image to be classified. Lithologic information and geographic parameter information, including grid size, number of rows and columns, and spatial projection, were extracted from the image. The extracted image data was normalized and then input into a lithologic classification model, which outputs lithologic classification results expressed in Roman numerals. Using the GDAL library, the output lithologic classification results were converted into a classification result raster image with the same geographic parameter information as the image to be classified. The classification result raster image was then imported into ArcGIS software to delineate and correct lithologic unit boundaries.

[0068] In this embodiment, the data in the image to be classified is input into the optimal lithologic classification model to obtain a classification result raster image of the six lithologies in the measured area, and then the accurate lithologic unit boundaries are outlined in the Arcgis software.

[0069] Step seven, make a lithology classification result map.

[0070] The boundaries of the lithologic units in the classification grid image are spliced ​​together with the boundaries of the non-rock exposed areas to compile a complete lithologic classification map of the surveyed area. If the lithologic classification map of the surveyed area contains areas that are inconsistent with existing geological knowledge or areas where the lithologic classification results are questionable, a field survey is conducted and the lithologic classification map of the surveyed area is revised based on the survey results.

[0071] The lithology classification method of the present invention was compared and verified with the lithology classification method based on the maximum likelihood method, that is, the sample data of the test set in this embodiment were subjected to lithology classification processing, and the accuracy values ​​are shown in Table 2.

[0072] Table 2. Comparison of classification accuracy of test set sample data

[0073]

[0074] From the comparative data in Table 2, it can be seen that in terms of accuracy, the value of the lithology classification method of the present invention is 92.31%, and the value of the maximum likelihood method is 87.91%; in terms of the F1 scores of the six lithologies, the calculation results of the lithology classification method of the present invention are all greater than those of the maximum likelihood method. It can be seen that compared with the existing lithology classification method, the lithology classification method of the present invention has higher classification accuracy and better performance.

Claims

1. A lithology classification method based on spectrum spatial residual network, characterized by: The following steps are involved: S1. Data collection: Collect multi-source geoscientific data including geoscientific data, aeromagnetic and aerial radio data, and hyperspectral remote sensing images in the surveyed area; S2. Data processing: Perform routine processing on the collected multi-source geological data, then select data channels and normalize the image data to unify the format, projection, grid density, and spatial range of the image data. The processed images of different types are combined into a multi-channel image. Non-rock exposed areas, including water bodies, alluvial deposits, artificial structures, and engineering distribution areas, are delineated on the multi-channel image. The data of the non-rock exposed areas are removed from the multi-channel image to form an image to be classified of the measured area. Routine processing operations for multi-source geological data include: vectorization of geological maps for geological data; cartographic conversion of aeromagnetic and aerial radio data; radiometric correction, atmospheric correction, orthorectification, geometric precision correction and abnormal band detection for hyperspectral remote sensing images; S3. Create a lithologic sample set: Establish discriminant markers for each type of lithology on the image to be classified, and mark the locations of lithologic samples and their lithologic category codes on the image to be classified; extract lithologic samples from the image to be classified based on the marked lithologic sample locations, and verify the independence of the samples; perform sorting operations on the lithologic samples, including sample shuffling, expanding 3D samples, and sample data normalization, and divide the sorted lithologic samples into a training set, a validation set, and a test set in proportion to form the lithologic sample set; S4. Build a lithology classification model: Build a lithology classification model based on the spectral spatial residual network architecture. Adjust the input port, convolution kernel size, number of convolution layers, and output port of the lithology classification model according to the sample structure, and initialize the model's hyperparameters. S5. Training model: Use the training set and validation set to train the lithology classification model, and use the test set to test the robustness of the model. Comprehensively evaluate the classification performance of the lithology classification model based on the accuracy and loss value indicators of the model on the training set and validation set, as well as the confusion matrix classification accuracy on the test set. If the performance analysis results do not meet the standards, adjust the model hyperparameters one by one and repeat this step until the performance analysis results meet the standards. Determine the model hyperparameters at this time and save the model. S6. Output lithologic classification results: Use the Gdal library in Python to load the image to be classified, extract lithologic information and geographic parameter information including grid size, number of rows and columns, and spatial projection from the image to be classified; normalize the extracted image data and then input it into the lithologic classification model, which outputs lithologic classification results expressed in Roman numerals; use the Gdal library to convert the output lithologic classification results into a classification result raster image with the same geographic parameter information as the image to be classified; import the classification result raster image into ArcGIS software to outline and correct the lithologic unit boundaries; S7. Prepare a lithologic classification result map: Splice together the boundaries of the lithologic units in the classification result raster image with the boundaries of the non-rock exposed areas to prepare a complete lithologic classification result map of the measured area. If there are areas in the lithologic classification result map of the measured area that are inconsistent with the existing geological knowledge or areas where the lithologic classification results are questionable, conduct a field survey and revise the lithologic classification result map of the measured area based on the survey results.

2. The lithology classification method based on spectrum spatial residual network according to claim 1 is characterized in that: The operation method for marking the locations of lithologic samples on the image to be classified in step S3 is as follows: based on the discriminant marks of each type of lithology on the image to be classified, the locations of lithologic samples are marked on the image to be classified using ArcGIS software, the sample locations are saved as a vector surface file, and an attribute field for storing the lithologic category code is added to the vector surface file. The vector surface file is then converted into a raster format and recorded as a label image; the lithologic category code is represented by Roman numerals starting with "I", each numeral represents a lithologic category, and the lithologic category of the location without a marked sample on the image to be classified is "0".

3. The lithology classification method based on spectrum spatial residual network according to claim 1 is characterized in that: The operation method for testing the independence of samples in step S3 is: use ENVI software to calculate the conversion separation of the marked lithological samples, and analyze the independence of the lithological samples based on the calculation results; if the analysis results do not meet the standards, reselect the lithological samples and repeat the independence test until the independence of all lithological samples meets the standards.

4. The lithology classification method based on spectrum spatial residual network according to claim 1 is characterized in that: In step S3, the operation method for scrambling the lithologic samples is to use the shuffle() function of the random module in Python to shuffle the order of the lithologic samples.

5. The lithology classification method based on spectrum spatial residual network according to claim 1 is characterized in that: The structure of the lithology classification model constructed in step S4 includes an input layer, a convolutional layer, two spectral residual blocks, two convolutional layers, two spatial residual blocks, a pooling layer and a fully connected layer from front to back; the model hyperparameters initialized include learning rate, batch size, number of iterations, optimizer, activation function and loss function.

6. The lithology classification method based on spectrum spatial residual network according to claim 1 is characterized in that: The operation method for testing the robustness of the model in step S5 is: input the samples of the test set into the lithology classification model, and then extract the sample data and label part in the test set; use the sample data to perform lithology classification processing, and calculate the classification accuracy of the lithology classification model for the lithology samples in the test set based on the classification results and the corresponding actual labels, and test the robustness of the model based on the classification accuracy; if the robustness of the model does not meet the standard, reselect the lithology samples and repeat steps S3 to S5 until the robustness of the model meets the standard.

Images