Intelligent extraction and identification method of geological environment information based on GCR model

Abstract

The present invention relates to the technical field of intelligent processing of geological information, and discloses a method for intelligent extraction and recognition of geological environment information based on the GCR model. A feature index set is constructed through standardization of multi-source data, a geological feature enhancement module is integrated into the OCR framework for multi-stage training, stratum feature recognition is achieved by combining morphological and color clustering algorithms, and dynamic parameter optimization is implemented based on differential analysis, forming a complete technical closed loop from data preprocessing to model self-optimization. The dynamic training control mechanism effectively suppresses model overfitting, and multi-head attention is combined to achieve multimodal fusion analysis of geological symbols, colors, and spatial features. At the same time, differentiated parameter optimization is implemented based on the importance of geological features. While improving the recognition accuracy of complex geological elements, the model's adaptability to non-uniformly distributed data is significantly enhanced, forming a full-process improvement from feature learning to training optimization.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent processing of geological information, and in particular to a method for intelligent extraction and recognition of geological environment information based on a GCR model. Background Art

[0002] With the deepening implementation of the new urbanization strategy, geological environmental factors have become a fundamental basis for decision-making in urban planning and construction. As the core carrier for recording stratigraphic structure and lithologic characteristics, the efficiency of digital analysis of geological maps directly impacts the accuracy of safety assessments for urban underground space development. Currently, the industry generally uses a combination of OCR technology and traditional image processing to extract geological elements. However, faced with the complex geological symbology and multi-source, heterogeneous data characteristics, existing technology systems have exposed significant adaptability bottlenecks.

[0003] Existing technical solutions suffer from systemic flaws when it comes to intelligently parsing geological elements. Traditional optical character recognition (OCR) models, constrained by a unimodal parsing mechanism for text features, struggle to effectively capture the correlation between geological symbols and spatial attributes, leading to coupling errors between symbol recognition rate and stratum boundary extraction accuracy. Deep learning-based solutions, while achieving breakthroughs in feature learning depth, lack geological data-specific design for parameter optimization during training, making them susceptible to overfitting due to uneven sample distribution. Furthermore, existing methods are insufficiently robust to noise interference from scanned documents, making them prone to feature misjudgment when processing historical paper maps. These issues collectively restrict the efficiency of digital transformation of geological environmental information, making it difficult to meet the multidimensional data fusion analysis requirements of urban geological safety assessments.

[0004] Taking the metamorphic rock area on the northeastern margin of the Yangtze Platform as an example, the region's complex stratigraphic contacts and multi-period tectonic features place higher demands on feature extraction algorithms. Existing technologies for processing such geological elements with strong spatial correlations often lack cross-modal feature fusion mechanisms, resulting in decoupling errors between lithologic symbol recognition and stratigraphic boundary delineation. This technical shortcoming not only affects the efficiency of geological survey results but also may result in the loss of critical data for risk assessment of urban underground space development. Therefore, developing intelligent geoinformation analysis technologies with adaptive feature fusion capabilities has become a key path to breaking through the current bottleneck of geological survey digitization. Summary of the Invention

[0005] In response to the shortcomings of the existing technology, the present invention provides an intelligent extraction and recognition method for geological environment information based on the GCR model, which solves the problems of multimodal feature decoupling, strong noise sensitivity and model overfitting in the traditional geological map digitization process.

[0006] To achieve the above objectives, the present invention is implemented through the following technical solutions: a method for intelligent extraction and identification of geological environment information based on the GCR model, comprising the following steps:

[0007] S1. Data standardization: Perform format conversion and feature extraction on multi-source geological data to construct a feature index set including symbols, colors and geometric features;

[0008] S2. Model construction and training: Based on the feature index set, a geological feature enhancement module is integrated into the OCR framework, and an optimized model is generated through multi-stage enhancement training;

[0009] S3. Stratum feature identification: applying the optimization model, combining morphological operations and color clustering algorithms to output stratigraphic boundaries and lithologic attributes;

[0010] S4. Dynamic model optimization: Based on the difference analysis between the recognition results and the verification indicators, the model parameters are updated through feature fusion and early stopping strategy.

[0011] Preferably, the data standardization process in step S1 includes:

[0012] Convert scanned geological maps into standard image formats and perform denoising;

[0013] Verify the coordinate system and metadata of GIS files;

[0014] Establish a mapping relationship library between geological symbols and lithologic properties.

[0015] Preferably, in the model construction and training of step S2:

[0016] The geological feature enhancement module includes a convolution optimization layer and a geological rule verification unit;

[0017] The multi-stage enhancement training consists of three stages of gradually increasing the complexity of image noise.

[0018] Preferably, the geological rule verification unit is configured as follows:

[0019] When the sequence of layers violates the geological time scale, a correction mechanism is triggered;

[0020] Detect symbolic logic conflicts and perform rule-based error correction.

[0021] Preferably, the formation feature identification in step S3 includes:

[0022] S3-1. Boundary extraction: Based on morphological gradient operation, a combination of a 5×5 rectangular kernel and a 7×7 elliptical kernel is used to extract the initial stratum boundary line;

[0023] S3-2, color clustering: performing K-means clustering on the pixel points within the area enclosed by the dividing line in the LAB color space to generate a stratum main color classification result;

[0024] S3-3, noise filtering: performing spatial continuity analysis on the classification results, and removing abnormal pixels whose discreteness exceeds a preset spatial distribution density threshold;

[0025] S3-4, logical verification: matching the filtered classification results with the symbol information recognized by OCR in step S2 for spatial position and semantic consistency.

[0026] Preferably, in the dynamic optimization of the model in step S4:

[0027] The early stopping strategy is automatically triggered by monitoring the fluctuation of the F1 value of the validation set;

[0028] The feature fusion strategy dynamically adjusts the weight ratio of text features and color features.

[0029] Preferably, the three stages of the multi-stage enhanced training include:

[0030] In the first stage, Gaussian noise is added to simulate low-quality scanned images;

[0031] In the second stage, motion blur is applied to simulate the fold deformation of geological maps;

[0032] The third stage performs downsampling processing to improve the generalization ability of the model.

[0033] Preferably, the triggering condition of the early stopping strategy satisfies the following formula:

[0034] ;

[0035] in, is the ratio of iterations with performance degradation, is the number of samples in the validation set, Indicates the The F1 value of the validation set of the iteration, is the indicator function. When there are 5 consecutive iterations >0.7 triggers the parameter rollback mechanism.

[0036] Preferably, the parameter rollback mechanism uses an adaptive momentum algorithm to update model parameters:

[0037] ;

[0038] Where, For the The model parameters of the iteration, is the adaptive learning rate factor, according to Dynamic attenuation, and Update according to the following rules:

[0039] ;

[0040] ;

[0041] in, 、 is the momentum decay coefficient, is a numerical stability constant, is the current parameter gradient.

[0042] The present invention provides a method for intelligent extraction and identification of geological environment information based on the GCR model. It has the following beneficial effects:

[0043] 1. This invention integrates a differential analysis module with an early stopping mechanism to achieve real-time optimization and adjustment of the model training process. Compared with traditional training methods with a fixed number of iterations, this solution can autonomously identify overfitting trends and adjust the learning process, addressing the model's lack of stability when learning complex geological features.

[0044] 2. This invention uses a multi-head attention mechanism to integrate geological symbols, color attributes, and spatial features to establish cross-modal feature associations. Compared to OCR models that process text or images alone, this design enhances the ability to collaboratively analyze multiple elements of geological maps and improves the integrity of feature recognition in complex scenarios.

[0045] 3. This invention uses a momentum adjustment algorithm based on the importance of geological elements to achieve differentiated learning of different feature channels. Traditional homogenized parameter update methods are difficult to adapt to the non-uniform distribution characteristics of geological data. This solution improves model convergence efficiency and feature learning accuracy by dynamically adjusting the parameter update amplitude. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 is a flow chart of the method of the present invention;

[0047] Figure 2 This is a flowchart of the key feature image recognition of the present invention. DETAILED DESCRIPTION

[0048] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the drawings in the present specification. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0049] Please see the attached Figure 1The embodiment of the present invention provides a method for intelligent extraction and identification of geological environment information based on the GCR model, comprising the following steps:

[0050] S1. Data standardization: Perform format conversion and feature extraction on multi-source geological data to construct a feature index set including symbols, colors and geometric features;

[0051] Data standardization is achieved through the collaboration of the scan conversion unit, coordinate verification unit, and symbol mapping unit. The scan conversion unit is connected to the image input interface and configured to perform the following processing flow: first, the geological map scan is converted into a standard image format, and a bilateral filter is used to eliminate noise. The pixel update formula is:

[0052] ;

[0053] Where, is the spatial weight kernel function, is the range weight kernel function, and the filter window size k is dynamically adjusted according to the image resolution. The input image is at coordinates The pixel value at is the grayscale value difference between the neighboring pixels and the central pixel. Then the morphological opening operation is performed, and the structural element uses a rectangular kernel.

[0054] The coordinate verification unit is connected to the GIS data interface and configured to perform coordinate system transformation and solve the transformation parameters using the least squares method:

[0055] ;

[0056] in, and For the The source coordinate system coordinates of the control points, and For the The target coordinate system coordinates of the control points, 、 、 、 is the affine transformation parameter matrix element, is the number of control points.

[0057] The number of control points meets the preset threshold, the parameter matrix is ​​solved using matrix decomposition, and the planar residuals meet the engineering standard tolerances. The converted spatial data is written to the spatial database and a metadata integrity check is performed.

[0058] The symbol mapping unit includes a feature extraction subunit and a relationship construction subunit. The feature extraction subunit uses the target detection model to locate the geological symbol area and output the symbol boundary coordinates. The color feature is calculated in the standard color space Histogram distribution:

[0059] ;

[0060] in, is the color component value at pixel (x,y), For the The base value of the color interval, is the impulse function.

[0061] Geometric feature calculation compactness parameter:

[0062] ;

[0063] in, is the pixel area of ​​the geological symbol area, The boundary perimeter of the geological symbol area.

[0064] The relationship construction subunit establishes an index relationship between the multidimensional feature vector and the lithology code, and the feature dimension conforms to the preset engineering specifications.

[0065] Example: When processing geological map scans, the resolution is first normalized to the project's required value, and bilateral filtering is used to eliminate scanning noise. Coordinate transformation selects control points that meet regulatory requirements, and affine transformation parameters are calculated. Symbol detection uses a preset confidence threshold, and feature matching establishes mapping relationships based on a similarity algorithm.

[0066] S2. Model construction and training: Based on the feature index set, the geological feature enhancement module is integrated into the OCR framework, and the optimized model is generated through multi-stage enhancement training;

[0067] Model building and training are achieved through the collaboration of the feature enhancement module and the multi-stage training framework. The feature enhancement module includes a convolution optimization part and a rule verification part, where the convolution optimization part is configured to perform deformable convolution operations:

[0068] ;

[0069] Where, is the learnable offset, To preset the convolution kernel size, The output feature map is at position The eigenvalue at ; For the The weight parameters of the convolution kernel; is the input feature map tensor; is the first in the standard convolution kernel The rule verification unit is connected to the time series analysis unit. When the stratigraphic sequence violates the geological time scale, the correction function is triggered:

[0070] ;

[0071] in, is the formation timing constraint threshold, set according to engineering standards, is the formation transition probability matrix, is the stratigraphic age code currently detected, is the code for the age of the preceding stratigraphic layer, is the probability of stratigraphic transition, which comes from the geological time scale database.

[0072] The multi-stage augmentation training framework consists of a three-stage data augmentation pipeline. The first stage adds Gaussian noise to the input image, with the noise intensity meeting the standards of the scanned image degradation model. The second stage applies a motion blur kernel with parameters that match the fold deformation characteristics of the geological map. The third stage performs downsampling, with the sampling rate set according to engineering specifications.

[0073] Example: When training the GCR model, clean samples, noisy samples, and downsampled samples are loaded in stages. The rule verification unit loads a standard stratigraphic chronology database, and a correction mechanism is automatically triggered when the stratigraphic sequence is inverted. The convolutional optimization layer uses a progressive parameter unfreezing strategy, initially freezing the underlying network weights.

[0074] Please see the attached Figure 2 ,S3, Stratum feature recognition: Apply the optimization model, combine morphological operations and color clustering algorithm to output stratigraphic boundaries and lithologic attributes;

[0075] Stratum feature recognition is achieved through a multi-stage processing pipeline, including a boundary extraction module, a color clustering module, a noise filtering module, and a logic check module. The boundary extraction module is configured to perform morphological gradient operations:

[0076] ;

[0077] Where, is a rectangular structural element, It is an elliptical structural element, and its size parameters are dynamically adjusted according to the scale of the geological map. is the input geological image matrix, is a morphological dilation operation used to expand bright areas. This is a morphological erosion operation used to shrink bright areas. The extracted boundary lines are optimized by the edge detector and then output to the color analysis unit.

[0078] The color clustering module is connected to the color space conversion component to perform K-means clustering on the boundary area:

[0079] ;

[0080] in, Pixel The feature vector in LAB color space, No. The characteristic vectors of cluster centers, is the number of dynamic clusters, which is automatically determined by the regional lithology type. Belong to A set of pixels that are clustered.

[0081] The number of clusters is automatically determined based on the regional lithology, and an improved sampling strategy is used for initialization. The main color classification results are transmitted to the verification unit after spatial distribution density analysis.

[0082] The noise filtering module is configured to calculate the pixel neighborhood connectivity index:

[0083] ;

[0084] in, Pixel The number of similar pixels in the eight-neighborhood, is the total number of neighborhood pixels (fixed to 8), It is a spatial distribution density index that reflects the reliability of pixel attribution.

[0085] When the indicator value falls below the preset threshold, the rejection mechanism is triggered. The spatial distribution density threshold is dynamically calculated based on regional characteristics and is positively correlated with the complexity of the formation.

[0086] The logic verification module connects to the OCR recognition database and performs spatial position matching:

[0087] ;

[0088] in, It is the overlapping area between the OCR recognition symbol area and the color cluster area. is the union area of ​​the symbol area and the cluster area, It is a spatial semantic consistency indicator, and the threshold is set according to engineering standards.

[0089] If the matching degree meets the engineering standard, it is considered semantically consistent; otherwise, the review process is triggered. The verification results are linked to the formation attribute database and updated.

[0090] Example: When processing a typical geological map, stratigraphic boundaries are first extracted to generate topological boundaries. Adaptive clustering is then performed on the target area to identify key lithologic color features. A spatial filtering mechanism automatically eliminates discrete noise points, and a logic verification unit spatially associates color features with OCR-recognized symbols to correct three lithologic labeling conflicts.

[0091] S4. Dynamic model optimization: Based on the difference analysis between the recognition results and the verification indicators, the model parameters are updated through feature fusion and early stopping strategy.

[0092] The model dynamic optimization module includes a difference analysis unit, a feature fusion unit, and a parameter update unit. The difference analysis unit is configured to calculate the multi-dimensional differences between the recognition results and the verification indicators:

[0093] ;

[0094] Where, is the feature weight coefficient, is the distribution difference weight, KL divergence measures the predicted distribution and the true distribution The difference, For the Class feature verification index benchmark value, For the The actual recognition results of class features, is the distribution difference adjustment factor, which controls the KL divergence item weight, The model predicts the feature distribution probability, Geological prior knowledge feature distribution probability. Feature fusion unit connected to multi-head attention mechanism:

[0095] ;

[0096] in The projection matrices from geological features, symbol features, and color features, respectively, is the scaling factor.

[0097] The early stopping strategy module monitors the F1 value fluctuation of the validation set:

[0098] ;

[0099] in, is the ratio of iterations with performance degradation, is the number of samples in the validation set, No. The F1 value of the validation set of the iteration, is an indicator function, which takes the value 1 when the condition is met.

[0100] When the performance of consecutive iterations degrades, the parameter rollback mechanism is triggered. The parameter update adopts the adaptive momentum algorithm:

[0101] ;

[0102] in, No. The model parameters of the iteration, is the adaptive learning rate factor, is the first-order momentum estimator, is the second-order momentum estimator, is the numerical stability coefficient, the momentum term and second-order moment Update by exponential decay, specifically according to Dynamic attenuation, and Update according to the following rules:

[0103] ;

[0104] ;

[0105] in, 、 is the momentum decay coefficient, is a numerical stability constant, is the current parameter gradient.

[0106] Example: When processing a geological map, the difference analysis module detected a sudden increase in the symbol recognition error rate, and the feature fusion unit automatically adjusted the geometric feature weights. After the early stopping strategy triggered parameter rollback, the model maintained stable accuracy on the validation set.

[0107] This implementation fully discloses the optimization mechanism through the aforementioned formula, with each module achieving state synchronization via a control bus. A multi-head attention mechanism improves feature fusion accuracy, an adaptive momentum algorithm accelerates convergence, and a parameter rollback mechanism reduces the risk of overfitting. Difference thresholds are strictly aligned with engineering standards, ensuring that the model's dynamic optimization complies with geological feature identification specifications.

[0108] Test Example 1:

[0109] This experiment used 168 1:50,000 geological maps of the northeastern margin of the Yangtze Platform as test samples. These maps contain 1,235 limestone symbols (■) for the Permian Qixia Formation and 892 sandstone symbols (▲) for the Triassic Yinkeng Formation. The experimental environment was configured with an Intel Xeon Gold 6248R processor, an NVIDIA A100 graphics card, Python 3.8, and the PaddleOCR 2.5 framework.

[0110] Experimental methods:

[0111] Establish four sets of comparison models:

[0112] Control group A: Tesseract 4.0 standard OCR model;

[0113] Control group B: PaddleOCRv2.5 pre-trained model;

[0114] Control group C: ResNet-50 architecture customized geological model;

[0115] Experimental group: GCR model (integrated multi-head attention mechanism + dynamic early stopping strategy).

[0116] Evaluation Metrics:

[0117] Symbol recognition rate: ,in, The number of correctly identified geological symbols (such as limestone symbol ■, sandstone symbol ▲), is the total number of geoscience symbols marked in the test sample, is the symbol recognition accuracy percentage;

[0118] Noise immunity F1 value: (Gaussian noise σ = 0.2 is added), where, is the precision, which is the percentage of correctly identified symbols in high-noise data. is the recall rate, which is the proportion of actual symbols recognized in high-noise data. It is a comprehensive evaluation indicator;

[0119] Training efficiency: average time per epoch

[0120] Overfitting index: ,in, is the accuracy of the training set (e.g., the GCR model training set reaches 98%), is the validation set accuracy (e.g., 95.2% for the GCR validation set), A quantitative indicator of the degree of overfitting.

[0121] Experimental results:

[0122] Model Symbol recognition rate High noise F1 value Training time Overfitting fluctuations Tesseract 4.0 34% 0.52 PaddleOCR 63% 0.68 7h50m 8% ResNet 63% 0.73 9h12m 5% GCR 75% 0.81 5h24m ＜3%

[0123] As shown in the table above, the GCR model significantly outperformed the control group in key metrics:

[0124] The symbol recognition rate is 41 percentage points higher than Tesseract (75% vs 34%) and 12 percentage points higher than ResNet.

[0125] The F1 value in high-noise scenarios reaches 0.81, which is 19.1% higher than PaddleOCR.

[0126] Training took 5 hours and 24 minutes, 42% shorter than ResNet, and the performance fluctuation on the validation set was less than 3%.

[0127] The overfitting index was controlled at 2.8%, which was 5-8 percentage points better than the control group.

[0128] Typical embodiments:

[0129] Taking the geological map H50E003015 as an example, the GCR model processing flow is as follows:

[0130] Extract stratum boundaries through morphological gradient operation:

[0131] ;

[0132] in, is a 5×5 rectangular structural element, An ellipse element with a major axis of 15px;

[0133] K-means clustering (k=6) was used to segment the lithologic regions, and the cluster centers were matched with the symbol library;

[0134] The dynamic early stopping mechanism is triggered at the 1,200th iteration and rolls back to the 1,150th parameter snapshot.

[0135] The final output includes recognition results for 28 strata, with a symbol-color matching accuracy of 92.3%, a 17.6 percentage point improvement over the ResNet model. This test case demonstrates the significant technical advantages of the GCR model in terms of feature fusion accuracy, training stability, and noise immunity.

[0136] Test Case 2: Verification of the Effectiveness of Multi-stage Training of the GCR Model

[0137] Experimental setup:

[0138] Model architecture: A dual-module structure is built based on the PaddleOCR framework. The text detection module uses the DB algorithm (ResNet50 backbone network), and the text recognition module uses the CRNN+Attention mechanism.

[0139] Training data: Contains 800 geological maps (including histograms, stratigraphic boundaries, and other elements), divided into training / validation / test sets at a ratio of 7:2:1

[0140] Hardware configuration: NVIDIA RTX3090 GPU, PyTorch1.10 environment

[0141] Experimental steps:

[0142] Three-stage iterative training:

[0143] Basic training phase (600 iterations): fixed backbone network parameters, learning rate 0.001;

[0144] Enhanced training phase (1200 iterations): unfreeze the top network and reduce the learning rate to 0.0003;

[0145] Fine-tuning phase (2000 iterations): All network parameters are adjustable, and the learning rate is 0.0001.

[0146] Dynamic early stop monitoring:

[0147] Calculate the best evaluation index of the validation set every 100 iterations;

[0148] Early stop is triggered when the indicator fluctuates <0.5% for 5 consecutive times;

[0149] Optimal model saving condition: indicator improvement >1%.

[0150] Experimental results:

[0151] Evaluation Metrics 600 iterations 1200 iterations 2000 iterations Best evaluation indicators 0.63 0.72 0.78 Precision 0.59 0.63 0.65 Recall 0.94 0.96 0.97 Inference speed (FPS) 0.12 0.15 0.16 Optimal training cycle 600 1200 1250 Training duration 2h40m 5h24m 11h38m

[0152] Conclusion analysis:

[0153] The model performance shows diminishing marginal returns as the number of iterations increases:

[0154] The increase in the best evaluation index decreased from 14.3% (0.63 → 0.72) in the 600 → 1200 time interval to 8.3% (0.72 → 0.78) in the 1200 → 2000 time interval;

[0155] The accuracy improvement narrowed from 6.8% (0.59→0.63) to 3.2% (0.63→0.65);

[0156] Dynamic early stopping mechanism trigger analysis:

[0157] At the 2000 iteration stage, the optimal training cycle was advanced to 1250 (actually saving 37.5% of the iterations);

[0158] A comparison of training times shows that completing all 2,000 iterations takes 11.6 hours, while the early stopping mechanism only takes 7.2 hours for 1,250 iterations (saving 38% time cost).

[0159] Precision-recall balance verification:

[0160] The recall rate is stable at above 94%, indicating that the model has a strong ability to capture geological elements;

[0161] The accuracy increased by 6 percentage points (0.59→0.65), proving that multi-stage training effectively reduces the false positive rate;

[0162] The experimental data verified the effectiveness of the dynamic early stopping mechanism in controlling overfitting and optimizing training resources, and also proved that the multi-stage iterative strategy can significantly improve the model's recognition accuracy of geological symbols.

[0163] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. The intelligent extraction and identification method of geological environment information based on the GCR model is characterized by: The following steps are involved: S1. Data standardization: Format conversion and feature extraction of multi-source geological data to construct a feature index set including symbols, colors and geometric features; S2. Model construction and training: Based on the feature index set, a geological feature enhancement module is integrated into the OCR framework, and an optimized model is generated through multi-stage enhancement training; Model building and training are achieved through the collaboration of a feature enhancement module and a multi-stage training framework. The feature enhancement module includes a convolution optimization unit and a rule verification unit. The convolution optimization unit is configured to perform deformable convolution operations: ； Where, is the learnable offset, To preset the convolution kernel size, The output feature map is at position The eigenvalue at ; For the The weight parameters of the convolution kernel; is the input feature map tensor; is the first in the standard convolution kernel The preset spatial offset of each position is connected to the rule verification unit. When the stratigraphic sequence violates the geological time scale, the correction function is triggered: ； in, is the formation timing constraint threshold, set according to engineering standards, is the formation transition probability matrix, is the stratigraphic age code currently detected, is the code for the age of the preceding stratigraphic layer, is the probability of stratigraphic temporal transition, derived from the geological time scale database; The multi-stage augmentation training framework consists of a three-stage data augmentation pipeline. In the first stage, Gaussian noise is added to the input image, with the noise intensity meeting the standards of the scanned image degradation model. In the second stage, a motion blur kernel is applied, with parameters set to meet the characteristics of fold deformation in geological maps. In the third stage, downsampling is performed, with the sampling rate set according to engineering specifications. S3. Stratum feature identification: applying the optimization model, combining morphological operations and color clustering algorithms to output stratigraphic boundaries and lithologic attributes; S4. Dynamic model optimization: Based on the difference analysis between the recognition results and the verification indicators, the model parameters are updated through feature fusion and early stopping strategy.

2. The method for intelligent extraction and identification of geological environment information based on the GCR model according to claim 1 is characterized in that: The data standardization process in step S1 includes: Convert scanned geological maps into standard image formats and perform denoising; Verify the coordinate system and metadata of GIS files; Establish a mapping relationship library between geological symbols and lithologic properties.

3. The method for intelligent extraction and identification of geological environment information based on the GCR model according to claim 1 is characterized in that: The geological rule verification unit is configured as follows: When the sequence of layers violates the geological time scale, a correction mechanism is triggered; Detect symbolic logic conflicts and perform rule-based error correction.

4. The method for intelligent extraction and identification of geological environment information based on the GCR model according to claim 1 is characterized in that: The formation feature identification step S3 includes: S3-1. Boundary extraction: Based on morphological gradient operation, a combination of a 5×5 rectangular kernel and a 7×7 elliptical kernel is used to extract the initial stratum boundary line; S3-2, color clustering: performing K-means clustering on the pixel points within the area enclosed by the dividing line in the LAB color space to generate a stratum main color classification result; S3-3, noise filtering: performing spatial continuity analysis on the classification results, and removing abnormal pixels whose discreteness exceeds a preset spatial distribution density threshold; S3-4, logical verification: matching the filtered classification results with the symbol information recognized by OCR in step S2 for spatial position and semantic consistency.

5. The method for intelligent extraction and identification of geological environment information based on the GCR model according to claim 1 is characterized in that: In the dynamic optimization of the S4 step model: The early stopping strategy is automatically triggered by monitoring the fluctuation of the F1 value of the validation set; The feature fusion strategy dynamically adjusts the weight ratio of text features and color features.

6. The method for intelligent extraction and identification of geological environment information based on the GCR model according to claim 5 is characterized in that: The triggering condition of the early stopping strategy satisfies the following formula: ； in, is the ratio of iterations with performance degradation, is the number of samples in the validation set, Indicates the The F1 value of the validation set of the iteration, is the indicator function. When there are 5 consecutive iterations >0.7 triggers the parameter rollback mechanism.

7. The method for intelligent extraction and identification of geological environment information based on the GCR model according to claim 5 is characterized in that: The parameter rollback mechanism uses an adaptive momentum algorithm to update model parameters: ； Where, For the The model parameters of the iteration, is the adaptive learning rate factor, according to Dynamic attenuation, and Update according to the following rules: ； ； in, 、 is the momentum decay coefficient, is a numerical stability constant, is the current parameter gradient.

Images