Geological stratum lithology regional prediction method, system, terminal device and storage medium

Abstract

The present application provides a kind of geologic stratum lithology regional prediction method, system, terminal equipment and storage medium, it is related to computer system field.The system includes data management module, sample management module, model training module, model management module, lithology prediction module and the like.The present application utilizes geophysical, geochemical, remote sensing and other multi-source data, based on deep learning image segmentation algorithm, forms regional shallow surface stratum lithology identification model, carries out high-precision rapid stratum lithology prediction and regional segmentation to the bedrock exposed area and shallow cover area with lower exploration degree, supports geological mapping and mineral exploration work.

Description

Technical Field

[0001] This invention relates to the field of computer systems, and more specifically, to a method, system, terminal device, and storage medium for predicting geological stratigraphy and lithology regions. Background Technology

[0002] Currently, various land cover classification methods based on remote sensing or multi-source information rely on manual judgment by experts in the field or traditional statistical calculations. Data processing, spectral feature extraction, sample selection, and classification methods are all affected by prior knowledge, staff experience, and theoretical deficiencies, which to some extent impacts classification accuracy. Land cover objects are mostly things with obvious spectral and shape characteristics, such as buildings, roads, vehicles, water bodies, forests, and farmland. The geological strata and rock types of the Earth's shallow surface are diverse, complex in composition, and have unclear boundaries. Identification of geological strata and rocks and regional segmentation based solely on remote sensing data often requires manual on-site investigation, as the accuracy of automatic identification using traditional computer data models cannot meet production requirements. Summary of the Invention

[0003] This invention provides a method, system, terminal equipment, and storage medium for predicting geological stratigraphy and lithology in a region.

[0004] A method for regional prediction of geological stratigraphy and lithology, comprising the following steps:

[0005] Step 1: Normalize the geophysical, geochemical, and remote sensing data of the known area to ensure that the resolution and spatial location of various multi-source data are consistent.

[0006] Step two involves merging the multi-source geophysical and geochemical data (after normalization in Step one) into image files, categorized by different layers. The multi-source geophysical, geochemical, and remote sensing data are merged in a unified layer order using the GeoTIFF file format. Several methods can be selected for geophysical data based on the actual exploration level of the target area, while geochemical data can be selected based on the lithological type of the target area, choosing a planar map with significantly different concentrated elemental contents. Topographic maps and hydrogeological maps can be added to enhance the model's robustness. The exploration accuracy of each type of data will affect the accuracy of the final identification results.

[0007] Step 3: Slice the multi-source image files into slices of fixed size to form a sample set;

[0008] Step 4: Slice the geological map according to the slice location of the multi-source image file and match it one-to-one with the sample set to form lithological labels for the sample set;

[0009] Step 5: Based on the deep learning image segmentation algorithm, train the model on a multi-source geophysical, geochemical, and remote sensing data sample set. During model training, adopt the training strategy of resuming training at breakpoints and dynamically adjusting the learning rate, and update the saved weight file at any time.

[0010] Step 6: Following the methods in Step 1 and Step 2, normalize the geophysical, geochemical, and remote sensing data of the unknown area according to a unified standard and merge the layers to obtain multi-source image files of the unknown area.

[0011] Step 7: Slice the multi-source image file of the unknown region, and when segmenting the image, perform window-style slicing with a step size smaller than the slice side length to obtain the prediction set;

[0012] Step 8: Use the trained regional shallow surface lithology identification model to predict the prediction set images respectively, and obtain the lithology prediction image segmentation result set;

[0013] Step nine: The lithology prediction image segmentation result set is stitched together according to the corresponding positions. When stitching the slices, an edge-ignoring strategy is adopted, that is, the edge part of each slice is ignored, forming unknown areas for stratigraphic lithology prediction and segmentation images.

[0014] Furthermore: the adjacent slices mentioned in step seven contain overlapping areas.

[0015] Furthermore, the prediction steps of the trained regional shallow surface lithology identification model for the prediction set images include extracting the trained GeoU-net network model, weight file, index file, and network structure used for segmentation. Then, the segmentation program is initialized according to these parameters. Subsequently, the prediction set images are loaded. If the acquired image scale is large, the image is scaled before calculation. Then, the target image is used as input for segmentation calculation to obtain the category of the image pixels.

[0016] Furthermore, during prediction, the prediction program needs to load the previously trained optimal GeoU-net network model file, including the weight file, index file, and network structure used by the segmentation network. After running the prediction program, the semantic segmentation prediction program is initialized based on these saved parameters, segmentation operations are performed, the category of the image pixels is obtained, the image segmentation result is obtained on the original image, and a complete prediction process is completed.

[0017] Furthermore: A geological stratigraphy and lithology regional prediction system based on deep learning includes a data management module for the comprehensive management of geological, geophysical, geochemical, and remote sensing data, data normalization, regional extraction, and GeoTIFF synthesis; the aforementioned data normalization is to maintain consistency in resolution and spatial location for all known regions and various multi-source data.

[0018] The sample management module is used to implement functions such as stratigraphic lithology classification and labeling, GeoTIFF format slicing, data augmentation, and sample slice management.

[0019] The model training module manages GPU server resources and trains image segmentation models; the model management module manages the trained stratigraphic lithology segmentation models for different types of regions.

[0020] The lithology prediction module performs regional prediction and result evaluation of the lithology of geological strata in unknown areas with low exploration levels.

[0021] Furthermore, the terminal device may include a processor, a storage medium, and a bus. The storage medium stores machine-readable instructions that can be executed by the processor. When the terminal device is running, the processor communicates with the storage medium via the bus, and the processor executes the machine-readable instructions to perform the steps of the deep learning model training method as described in the foregoing embodiments.

[0022] Further: a storage medium storing a computer program, which, when executed by a processor, performs the steps of the method described above.

[0023] The beneficial effects of this invention are as follows: This invention utilizes multi-source data such as geophysics, geochemistry, and remote sensing, and based on deep learning image segmentation algorithms, to form a regional shallow surface lithology identification model. It can perform high-precision and rapid lithology prediction and regional segmentation for bedrock exposed areas and shallow overburden areas with low exploration levels, supporting geological mapping and mineral exploration work. The accuracy rate of lithology identification and regional division reaches more than 90%. Attached Figure Description

[0024] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0025] Figure 1 A flowchart illustrating the method of the present invention is shown.

[0026] Figure 2 This diagram illustrates the distribution of granite and metamorphic rock areas covering 5,000 square kilometers predicted by the model applied in this invention.

[0027] Figure 3 A schematic diagram of the system of the present invention is shown.

[0028] Figure 4 A schematic diagram of the components of the terminal device of the present invention is shown. Detailed Implementation

[0029] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. It should be understood that the accompanying drawings in the present invention are for illustrative and descriptive purposes only and are not intended to limit the scope of protection of the present invention. Furthermore, it should be understood that the schematic drawings are not drawn to scale. The flowcharts used in this invention illustrate operations implemented according to some embodiments of the present invention. It should be understood that the operations in the flowcharts may not be implemented in sequence, and steps without logical contextual relationships may be reversed or implemented simultaneously. In addition, those skilled in the art, guided by the content of this invention, may add one or more other operations to the flowcharts, or remove one or more operations from the flowcharts.

[0030] Furthermore, the embodiments described herein are merely some, not all, of the embodiments of the invention. The components of the embodiments of the invention described and illustrated herein can typically be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the invention without inventive effort are within the scope of protection of the invention.

[0031] It should be noted that the term "comprising" will be used in the embodiments of the present invention to indicate the presence of a feature subsequently declared, but does not preclude the addition of other features. It should also be noted that similar reference numerals and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. In the description of the present invention, it should also be noted that the terms "first," "second," "third," etc., are used only for distinguishing descriptions and should not be construed as indicating or implying relative importance.

[0032] Figure 1 A flowchart illustrating the steps of the method of the present invention is shown.

[0033] The specific steps of the method of the present invention include:

[0034] Step 1: Normalize the geophysical, geochemical, and remote sensing data of the known areas. In this step, the resolution and spatial location of each known area and various types of multi-source data are kept consistent, which is called normalization.

[0035] Step two involves merging multi-source geophysical and geochemical data into image files based on remote sensing data, organized by different layers. Multi-source geophysical, geochemical, and remote sensing data are merged in a unified layer order using the GeoTIFF file format. For geophysical data, several methods can be selected based on the actual exploration progress of the target area. For geochemical data, a planar map with significantly different concentrated elemental contents can be selected based on the lithological type of the target area. Topographic maps, hydrogeological maps, and other layers can be added to enhance the model's robustness. The exploration accuracy of each type of data will affect the accuracy of the final identification results.

[0036] Step 3: Slice the multi-source image files into slices of a fixed size to form a sample set. The fixed size can be adjusted according to the configuration of the model training server; this example uses a resolution of 256*256.

[0037] Step 4: Slice the geological map according to the slice location of the multi-source image file and match it one-to-one with the sample set to form lithological labels for the sample set.

[0038] Step 5: Based on the deep learning image segmentation algorithm, train the model on a multi-source geophysical, geochemical, and remote sensing data sample set. During model training, adopt the training strategy of resuming training at breakpoints and dynamically adjusting the learning rate, and update the saved weight file at any time.

[0039] This example uses the GeoU-net algorithm model to obtain a regional shallow surface lithology identification model through calculation iteration. This step is for training the regional shallow surface lithology identification model.

[0040] GeoU-net is an improved U-net model that incorporates several optimization strategies for segmenting multi-source geophysical, geochemical, and remote sensing data. First, to avoid overfitting caused by insufficient edge pixel features in sliced ​​sample images, it uses the edge cross-entropy function as the loss function, which improves classification accuracy in the prediction area to some extent. Second, it adds batch normalization after each convolution, reducing the network's insensitivity to initial weights and accelerating model training convergence. Third, it adds a Dropout layer to the model, which also helps prevent overfitting to some extent.

[0041] Subsequently, stratigraphic and lithological predictions and regional segmentation were carried out for unknown areas with low exploration levels. The specific steps are as follows:

[0042] Step 6: Following the methods in Steps 1 and 2, normalize the geophysical, geochemical, and remote sensing data of the unknown area according to a unified standard and merge the layers to obtain multi-source image files of the unknown area.

[0043] Step 7: Slice the multi-source image files of the unknown region;

[0044] To ensure the accuracy of the prediction results, adjacent slices must have a certain degree of overlap. The image is segmented using a window-like slicing method with a specific step size to obtain the prediction set.

[0045] Step 8: Use the trained regional shallow surface lithology identification model to predict the prediction set images respectively, and obtain the lithology prediction image segmentation result set. The prediction steps of the trained regional shallow surface lithology identification model for the prediction set images include extracting the trained GeoU-net network model, weight file, index file and network structure used for segmentation, then initializing the segmentation program according to these parameters, and then taking the target image as input to perform segmentation operation and obtain the category to which the image pixels belong.

[0046] Step nine: The lithology prediction image segmentation result set is stitched together according to the corresponding positions. When stitching, an edge ignoring strategy is adopted. Overlapping areas are cropped and discarded. Finally, the images are stitched together to form unknown areas for stratigraphic lithology prediction and segmentation.

[0047] When making predictions, the prediction program needs to load the previously trained optimal GeoU-net network model file, including the weight file, index file, and network structure used by the segmentation network. After running the prediction program, the semantic segmentation prediction program is initialized based on these saved parameters, the segmentation operation is performed, the category of the image pixels is obtained, the image segmentation result is obtained on the original image, and a complete prediction process is completed.

[0048] This method and system were applied to conduct stratigraphic lithology prediction and regional segmentation experiments in northern Gansu Province. An image sample set was created by synthesizing satellite remote sensing, geophysical gravity and magnetic data, and geochemical data. A 1:200,000 geological map generated through manual surveying and mapping was used as the standard result to train a regional granite and metamorphic rock lithology identification model. The model was applied to predict the distribution of granite and metamorphic rocks in a 2000 square kilometer area. Figure 2 The results, when compared with standard geological maps, show that the distribution areas of the rock masses largely match. Calculations indicate that the accuracy of the predictions reached 91.2%.

[0049] like Figure 3 As shown, a deep learning-based geological stratigraphy and lithology regional prediction system includes:

[0050] The data management module is used for the comprehensive management of geological, geophysical, geochemical, and remote sensing data, as well as data normalization, regional extraction, and GeoTIFF synthesis. The aforementioned data normalization process maintains consistency in resolution and spatial location for all known regions and various multi-source data.

[0051] The sample management module is used to implement functions such as stratigraphic lithology classification and labeling, GeoTIFF format slicing, data augmentation, and sample slice management.

[0052] The model training module manages GPU server resources and trains image segmentation models.

[0053] The model management module manages the trained stratigraphic lithology segmentation models for different types of regions.

[0054] The lithology prediction module predicts, segments, and evaluates the lithology of geological strata in unknown areas with low exploration levels.

[0055] like Figure 4 As shown, the terminal device 6 may include a processor 601, a storage medium 602, and a bus 603. The storage medium 602 stores machine-readable instructions executable by the processor 601. When the terminal device is running, the processor 601 communicates with the storage medium 602 via the bus 603. The processor 601 executes the machine-readable instructions to perform the steps of the deep learning model training method described in the foregoing embodiments. The specific implementation and technical effects are similar and will not be repeated here.

[0056] For ease of explanation, only one processor is described in the terminal device described above. However, it should be noted that in some embodiments, the terminal device of the present invention may also include multiple processors, and therefore the steps performed by one processor described in the present invention may also be performed jointly by multiple processors or individually.

[0057] The above are merely specific embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for regional prediction of geological stratigraphy and lithology, characterized in that, Includes the following steps: Step 1: Normalize the geophysical, geochemical, and remote sensing data of the known area to ensure that the resolution and spatial location of various multi-source data are consistent. Step 2: Based on the geophysical, geochemical, and remote sensing data after normalization in Step 1, the multi-source geophysical and geochemical data of various types are merged into a GeoTIFF format multi-source image file according to different layers. Step 3: Slice the multi-source image files into slices of fixed size to form a sample set; Step 4: Slice the geological map according to the slice location of the multi-source image file and match it one-to-one with the sample set to form lithological labels for the sample set; Step 5: Based on the deep learning image segmentation algorithm, train the multi-source geophysical, geochemical, and remote sensing data sample sets. Use the GeoU-net algorithm model to obtain the regional shallow surface lithology identification model through calculation iteration. The GeoU-net algorithm model uses the edge cross-entropy function as the loss function to avoid overfitting caused by insufficient edge pixel features of slice sample images. Step 6: Following the methods in Step 1 and Step 2, normalize the geophysical, geochemical, and remote sensing data of the unknown area according to a unified standard and merge the layers to obtain multi-source image files of the unknown area. Step 7: Slice the multi-source image files of the unknown region, and perform window-style slicing according to a certain step size when segmenting the image to obtain the prediction set; Step 8: Use the trained regional shallow surface lithology identification model to predict the prediction set images respectively, and obtain the lithology prediction image segmentation result set; Step nine: The lithology prediction image segmentation result set is stitched together according to the corresponding positions to form an unknown area for stratigraphic lithology prediction and segmentation image. When stitching the slices, an edge-ignoring strategy is adopted, that is, the edge part of each slice is ignored.

2. The method according to claim 1, characterized in that, The prediction steps of the trained regional shallow surface lithology identification model for the prediction set images include extracting the trained GeoU-net network model, weight file, index file, and network structure used for segmentation, then initializing the segmentation program according to these parameters, and then loading the prediction set images.

3. The method according to claim 1 or 2, characterized in that, When making predictions, the prediction program needs to load the previously trained optimal GeoU-net network model file, including the weight file, index file, and network structure used by the segmentation network. After running the prediction program, the semantic segmentation prediction program is initialized based on these saved parameters, the segmentation operation is performed, the category of the image pixels is obtained, the image segmentation result is obtained on the original image, and a complete prediction process is completed.

4. The method according to claim 1, characterized in that, During model training, a training strategy of resuming training at breakpoints and dynamically adjusting the learning rate is adopted, and the saved weight file is updated at any time.

5. The method according to claim 2, characterized in that, If the acquired image is large in scale, the image is scaled before the calculation is performed. Then, the target image is used as input to perform segmentation calculations to obtain the category to which the image pixels belong.

6. A geological stratigraphy and lithology regional prediction system based on deep learning, characterized in that, For performing the method as described in any one of claims 1 to 5, comprising: The data management module is used for the comprehensive management of geological, geophysical, geochemical, and remote sensing data, as well as the implementation of data normalization functions. The sample management module is used for the implementation of functions such as stratigraphic lithology classification and labeling, GeoTIFF format slicing, data augmentation, and sample slice management. The model training module manages GPU server resources and trains image segmentation models; the model management module manages the trained stratigraphic lithology segmentation models for different types of regions. The lithology prediction module performs regional prediction and result evaluation of the lithology of geological strata in unknown areas with low exploration levels.

7. A terminal device, characterized in that, include: The device includes a processor, a storage medium, and a bus, wherein the storage medium stores machine-readable instructions executable by the processor, and when the terminal device is running, the processor communicates with the storage medium via the bus, and the processor executes the machine-readable instructions to perform the steps of the method as described in any one of claims 1 to 5.

8. A storage medium, characterized in that, The storage medium stores a computer program, which, when executed by a processor, performs the steps of the method as described in any one of claims 1 to 5.

Images