A geological profile classification method, storage medium, and device
By constructing a geological profile classification method based on LightGBM model, using geological profile characteristics and new features, the problem of failure to fully utilize geological knowledge in the existing technology is solved, and the accuracy of geological profile classification and the explanatory nature of the model are improved.
Patent Information
- Application Number
- CN202411433131.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-10-15
- Publication Date
- 2024-12-17
- Estimated Expiration
- 2044-10-15
AI Technical Summary
The existing geological profile classification methods fail to make full use of geological knowledge, resulting in low classification accuracy and difficulty in capturing the nonlinear relationship between complex geological features.
Using the LightGBM model combined with on-site exploration data and remote sensing data, new features are constructed and feature selection and normalization are constructed by extracting features such as density, porosity, mineral content, particle size distribution, color index and saturation, and a geological profile classification model is constructed. , , use these and new features for training and verification.
It enhances the generalization ability and interpretability of the model, can capture complex nonlinear relationships more accurately, and improves the accuracy of geological profile classification.
Smart Images
Figure CN118965189B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of geological exploration technology, and in particular to a geological profile classification method, storage medium and equipment. Background Art
[0002] Geological profile classification is a related task in the field of geology and engineering. Geological profile classification can help geologists understand the structure of the earth's interior and the distribution of rock layers, revealing the evolutionary history of the earth. Secondly, geological profile classification can provide an important reference for mineral exploration. By analyzing the geological characteristics of geological profiles, the distribution pattern of mineral resources can be found. Geological sampling and analysis, most of the time, after collecting samples, physical and chemical analysis is performed in the laboratory to determine the properties and types of rocks. Based on spectral analysis, the spectral characteristics of different rocks and minerals are used for classification and identification.
[0003] In the prior art, machine learning and artificial intelligence models have also been gradually applied to geological profile identification and classification. The models are used to receive the geological features of the geological profile for identification and processing. However, there are complex nonlinear relationships between the geological features of the geological profile. When the models in the prior art are put into use, the complex nonlinear relationships between the geological features are not taken into account, and geological knowledge is not fully utilized. The models only rely on raw data and basic features, lack professional knowledge support, and it is difficult to fully utilize geological knowledge. The models cannot accurately identify complex geological phenomena, resulting in reduced classification accuracy. Summary of the invention
[0004] The purpose of the present invention is to solve the problem that the existing geological profile classification method does not make full use of geological knowledge, resulting in low classification accuracy, and propose a geological profile classification method, comprising the following steps:
[0005] S1, obtaining field exploration data and remote sensing data of the geological profile, as well as the category label of the geological profile, preprocessing the field exploration data and remote sensing data, and extracting features based on the preprocessed data;
[0006] S2. Based on the extracted features, new features are constructed, and the extracted features and the new features are combined into a feature set;
[0007] S3, select and normalize the features in the feature set, and divide them into a training set and a validation set;
[0008] S4. Construct a geological profile classification model, use LightGBM as the model benchmark architecture, use the training set and the category labels of the geological profile to train the model, use the validation set to verify the trained model, and input the data to be tested into the verified model to obtain the geological profile classification results.
[0009] Furthermore, the preprocessing of the field exploration data and remote sensing data is specifically as follows:
[0010] The field exploration data and remote sensing data are converted into a unified coordinate system, and the dynamically changing data are aligned in time to obtain data that is consistent in position and time.
[0011] Furthermore, the extracted features include: density, porosity, mineral content, particle size distribution, color index, and saturation.
[0012] Furthermore, based on the extracted features, a new feature representation is constructed as follows:
[0013]
[0014] Among them, C represents the rock content, a, b, and c represent the weight coefficients, represents density, represents porosity, M represents mineral content, represents the particle size distribution, Represents the rock color index, and S represents saturation.
[0015] Furthermore, the variance selection method is used to select features in the feature set, the variance of each feature is calculated, a threshold is set, and features with variance greater than the threshold are selected.
[0016] Furthermore, the initial parameters of the geological profile classification model include: learning rate, number of trees, maximum depth, minimum number of sample splits, minimum number of sample leaves, subsample ratio and feature selection ratio.
[0017] The present invention also proposes a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the above-mentioned geological profile classification method is implemented.
[0018] The present invention also proposes an electronic device, comprising a processor and a memory, wherein the processor and the memory are interconnected, wherein the memory is used to store a computer program, the computer program comprises computer-readable instructions, and the processor is configured to call the computer-readable instructions to execute the above-mentioned geological profile classification method.
[0019] The beneficial effects brought by the technical solution provided by the present invention are:
[0020] The present invention proposes to use geological profile features to construct new features for geological profile classification, build a geological profile classification model using LightGBM as the benchmark framework, and use geological profile features and new features as model inputs. The method of the present invention constructs additional input features based on geological profile features. The model can make full use of geological knowledge, enhance feature expression capabilities, improve the generalization ability and interpretability of the model, and capture complex nonlinear relationships. BRIEF DESCRIPTION OF THE DRAWINGS
[0021] Figure 1 is a flow chart of a geological profile classification method according to an embodiment of the present invention;
[0022] Figure 2 It is a block diagram of an electronic device in an exemplary embodiment of the present invention. DETAILED DESCRIPTION
[0023] To make the objectives, technical solutions and advantages of the present invention more clear, the embodiments of the present invention will be further described below with reference to the accompanying drawings.
[0024] The flow chart of the geological profile classification method according to the embodiment of the present invention is as follows: Figure 1 , specifically including the following steps:
[0025] S1. Acquire field exploration data and remote sensing data of the geological profile, as well as the category label of the geological profile, preprocess the field exploration data and remote sensing data, and extract features based on the preprocessed data.
[0026] Among them, remote sensing data is data obtained based on high-resolution satellite images and aerial images.
[0027] Data preprocessing includes: data cleaning, removing outliers and noise; filling missing values through the mean; converting field exploration data and remote sensing data to a unified coordinate system, aligning dynamically changing data in time, and obtaining data that is consistent in position and time.
[0028] The features of field measurement data and remote sensing data are integrated to form a comprehensive feature vector, including primitive level fusion, feature level fusion and decision level fusion. Primitive level fusion is to perform preliminary data fusion at the data collection stage; feature level fusion is to perform feature fusion and integration at the feature extraction stage; decision level fusion belongs to a high-level fusion. Due to the concentration of data, the results produced by this method are relatively the least accurate, but its computational complexity and communication bandwidth requirements are the lowest.
[0029] After the above preprocessing, preprocessed data is obtained, and features are extracted from the preprocessed data. The extracted features include: density, porosity, mineral content, particle size distribution, color index, and saturation.
[0030] Density is one of the main characteristics that characterizes the physical properties of rocks, and is closely related to rock porosity, mineral content, and water content. Due to differences in formation age, environment, and material composition, the density characteristics of regional rock strata also show certain regular changes.
[0031] Porosity plays an obviously important role in geology. It controls the storage of fluids in aquifers, oil and gas fields, and geothermal systems. The extent and connectivity of the pore structure controls the flow and transport of fluids through geological formations, as well as the relationship between the properties of individual minerals and the overall properties of the formation.
[0032] Mineral content is of great significance to lithology classification, sedimentary environment judgment and source direction analysis. In the area analysis method, the ratio of the number of pixels occupied by minerals to the total number of pixels represents the mineral content.
[0033] Particle size distribution: The percentage of particles in a series of different particle size ranges in the total sample reflected by a certain method is called particle size distribution.
[0034] The color index is a quantitative evaluation of mineral color based on the principle of colorimetry in modern times. The color index includes parameters such as visual reflectance, chromaticity coordinates x, y, z, and color dominant wavelength.
[0035] Saturation determines the color component in the color space. The higher the saturation, the darker the color, and the lower the saturation, the lighter the color.
[0036] S2. Based on the extracted features, new features are constructed, and the extracted features and the new features are combined into a feature set.
[0037] Constructing new features is represented as:
[0038]
[0039] Among them, C represents the rock content, a, b, and c represent the weight coefficients, represents density, represents porosity, M represents mineral content, represents the particle size distribution, Represents the rock color index, and S represents saturation.
[0040] Indicates density and porosity The influence of the logarithm of the product of on the rock content, where the logarithmic function is used to process data of different magnitudes to make it smoother in the calculation. It means that the combination of density and porosity may affect the physical properties of rocks, specifically hardness and storage capacity.
[0041] It represents the influence of the square root of the product of mineral content M and particle size distribution D on rock content, where the square root is used to smooth data. It means that the combination of mineral content and particle size reflects the structural characteristics of the rock, and the structural characteristics include particle size and mineral distribution.
[0042] It represents the influence of the ratio of color index CI to saturation S on rock content, where 1 is added to avoid the situation where the denominator is zero. The relationship between color index and saturation may affect the optical properties and reservoir performance of rocks.
[0043] Indicates particle size distribution D and porosity The influence of the sine value of the product of on the rock content, where the sine function is used to introduce periodic changes, specifically reflecting the changes in rock properties in different sedimentary environments that may be caused by periodic changes in grain size and porosity.
[0044] Indicates density The influence of the exponential value of the ratio to saturation S on rock content, where the exponential function is used to process rapidly changing data, representing that the ratio of density to saturation may affect the mechanical properties and storage capacity of rocks.
[0045] By calculating the C value using the above improved formula, for the geological profile identification level, by integrating multiple geological features and estimating the content of a certain rock type, it is helpful to identify different rock layers in the geological profile; for the lithology classification level, based on the calculated rock content C, the rock types in the geological profile are classified, which effectively helps geologists and engineers to quickly understand the geological structure; for the resource exploration level, identifying and classifying the rock types in the geological profile is of great significance to the exploration of mineral resources and groundwater resources; for the disaster prediction level, understanding the rock types in the geological profile is helpful to predict and prevent geological disasters, which specifically refer to landslides and mud-rock flows.
[0046] The characteristic values of sample No. 1 collected in the embodiment of the present invention are shown in Table 1.
[0047] Table 1
[0048]
[0049] In a specific embodiment of the present invention, a is 0.5, b is 0.8, and c is 0.3. The C value is calculated based on the values in the above table:
[0050]
[0051] Construct the feature vector of sample No. 1 , calculate the C value of each sample and add the C value to the feature vector of each sample. All feature vectors constitute the feature set.
[0052] S3. Select and normalize the features in the feature set and divide it into a training set and a validation set.
[0053] The variance selection method is used to select features in the feature set, calculate the variance of each feature, set a threshold, and select features with variance greater than the threshold. If the threshold is set to 0.1, features with variance greater than the threshold are selected, and features that contribute significantly to the model are retained. Based on the set threshold, features with variance greater than the threshold are screened out to form a new feature set.
[0054] Feature normalization: Normalize all features to have the same dimension and range. Use Min-Max normalization, such as Min-Max normalization for density: ,in, represents the normalized density, represents the minimum value of density, Indicates the maximum value of density.
[0055] S4. Construct a geological profile classification model, use LightGBM as the model benchmark architecture, train the model with the training set and the category labels of the geological profile, use the validation set to validate the trained model, and input the data to be tested into the validated model to obtain the geological profile classification result. The data to be tested are the feature data (density, porosity, mineral content, particle size distribution, color index, saturation) obtained from the geological profile to be tested and the newly constructed feature data (rock content).
[0056] LightGBM (Light Gradient Boosting Machine) is a distributed gradient boosting framework based on the decision tree algorithm. It uses a histogram-based decision tree algorithm, supports efficient parallel training, has lower memory usage and higher accuracy. The main features of LightGBM include histogram difference acceleration, Leaf-wise leaf growth strategy, category feature support and parallel learning. By adjusting parameters such as num_leaves (the number of leaf nodes) and min_data_in_leaf (the minimum number of samples per leaf node), overfitting can be effectively prevented and model performance can be improved.
[0057] The initial parameters of the geological profile classification model include: learning rate, number of trees, maximum depth, minimum sample split number, minimum sample leaf number, subsample ratio and feature selection ratio. Using 5-fold cross validation, the grid search will traverse all possible hyperparameter combinations, perform grid search, evaluate the performance of each hyperparameter combination through cross validation, select the best combination, and ensure the stability of the model's performance on different data sets based on cross validation. Retrain the model with the optimal hyperparameters, and use the test data set to evaluate the performance of the final model. Record the accuracy, precision, recall and F1 score of the final model to ensure the best performance of the model in geological profile recognition.
[0058] In the embodiment of the present invention, the learning rate is set to 0.1. The smaller the learning rate, the slower the model learns, but the performance is more accurate; the number of trees is set to 100. The more trees there are, the higher the complexity of the model; the maximum depth is set to 7. The larger the maximum depth, the more complex the model is and the more feature details can be captured; the minimum sample split number is set to 2. This parameter can prevent overfitting when the number of samples is small; the minimum sample leaf number is set to 1. If the minimum sample leaf number is too large, the model learning noise problem may occur easily; the subsample ratio and the feature selection ratio are both set to 0.8.
[0059] Calculate the error between the model prediction result and the actual label, analyze the source of the error and the performance of the model in different data categories, generate error distribution graphs and residual graphs, display the error distribution, and identify samples and categories with large errors. Among them, conduct in-depth analysis on samples with large errors to find the cause of the error, which may be data noise, underfitting or overfitting of the model.
[0060] Export the trained LightGBM model and save it in a file format, ensure that the model can be loaded in the production environment, deploy the model to the production environment, set up API interfaces or batch scripts, realize the automatic prediction and identification of geological profile data, monitor the operation of the model in the production environment, and regularly check the model performance and prediction results. The regularity is 7 to 15 days as a large cycle and 2 to 3 days as a small cycle.
[0061] Import the actual geological profile data into the model and ensure that the data format is consistent with the data format during model training. Based on the verified model, predict the actual geological profile data, obtain the model's prediction results, generate visualization charts, and combine the prediction results with the actual geological profile data. The visualization charts are specifically profile diagrams and thermal maps.
[0062] Provide continuous feedback and optimization, collect feedback from the deployment end, record problems encountered in actual applications and improvement suggestions, analyze the collected feedback, identify improvement points and optimization directions of the model, optimize the model structure and parameters based on feedback and analysis results, perform incremental training and updates, and improve model performance.
[0063] Through the above steps, actual geological profile data can be effectively imported to test the model, verify the model's automated recognition effect, and ensure the best performance of the model in practical applications through continuous feedback and optimization.
[0064] In an exemplary embodiment, a computer-readable storage medium is included, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the above-mentioned geological profile classification method is implemented.
[0065] See also Figure 2 In an exemplary embodiment, an electronic device is also included, including at least one processor, at least one memory, and at least one communication bus.
[0066] Wherein, a computer program is stored in the memory, and the computer program includes computer-readable instructions. The processor calls the computer-readable instructions stored in the memory through the communication bus to execute the above-mentioned geological profile classification method.
[0067] The above description of the disclosed embodiments enables one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.
Claims
1. A geological profile classification method, characterized in that: The following steps are involved: S1, obtaining field exploration data and remote sensing data of the geological profile, as well as the category label of the geological profile, preprocessing the field exploration data and remote sensing data, and extracting features based on the preprocessed data; S2. Based on the extracted features, new features are constructed, and the extracted features and the new features are combined into a feature set; The extracted features include: density, porosity, mineral content, particle size distribution, color index, saturation; Constructing new features is represented as: Among them, C represents the rock content, a, b, and c represent the weight coefficients, represents density, represents porosity, M represents mineral content, represents the particle size distribution, It represents the rock color index, S represents the saturation; S3, select and normalize the features in the feature set, and divide them into a training set and a validation set; S4. Construct a geological profile classification model, use LightGBM as the model benchmark architecture, use the training set and the category labels of the geological profile to train the model, use the validation set to verify the trained model, and input the data to be tested into the verified model to obtain the geological profile classification results.
2. A geological profile classification method according to claim 1, characterized in that: The preprocessing of field exploration data and remote sensing data is as follows: The field exploration data and remote sensing data are converted into a unified coordinate system, and the dynamically changing data are aligned in time to obtain data that is consistent in position and time.
3. A geological profile classification method according to claim 1, characterized in that: The variance selection method is used to select features in the feature set, calculate the variance of each feature, set a threshold, and select features whose variance is greater than the threshold.
4. A geological profile classification method according to claim 1, characterized in that: The initial parameters of the geological profile classification model include: learning rate, number of trees, maximum depth, minimum number of sample splits, minimum number of sample leaves, subsample ratio, and feature selection ratio.
5. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 4 is implemented.
6. An electronic device, characterized in that: The method comprises a processor and a memory, wherein the processor and the memory are interconnected, wherein the memory is used to store a computer program, the computer program comprises computer-readable instructions, and the processor is configured to call the computer-readable instructions to execute the method according to any one of claims 1 to 4.