A method for picking first arrival of seismic data based on SimpleNet network

Abstract

The present invention relates to the technical field of seismic data processing, and in particular to a method for picking up seismic data first arrival based on a SimpleNet network. The method comprises the following steps: acquiring seismic shot data, and performing grayscale image conversion on the seismic shot data, so as to obtain a set of enhanced grayscale images in a function set; dividing a data set according to the enhanced grayscale image set in the function set, so as to obtain a training image set and a test image set; performing low-speed body static correction on the training image set, so as to obtain a corrected training image set; constructing a seismic first arrival prediction model according to the enhanced grayscale image set in the function set, so as to obtain a seismic first arrival prediction model; performing sample training on the seismic first arrival prediction model according to the corrected training image set, so as to obtain a seismic first arrival training model; and optimizing model parameters of the seismic first arrival training model, so as to obtain an optimized seismic first arrival training model. The present invention improves the accuracy of seismic first arrival picking based on seismic data processing technology.

Description

Technical Field

[0001] The invention relates to the technical field of seismic data processing, and in particular to a seismic data first arrival picking method based on SimpleNet network. Background Art

[0002] Static correction technology in shallow seismic exploration is used to eliminate the influence of terrain, and its core lies in dealing with the influence of low-velocity bodies on deep seismic imaging. The undulation of terrain and the thickness change of low-velocity bodies will prolong the propagation time of seismic waves, resulting in distortion of deep stratum imaging and even the generation of false structures. In order to reduce the interference of low-velocity bodies, commonly used static correction techniques include elevation static correction, first-arrival wave static correction, refraction wave static correction, micro-logging, dune curve and tomographic refraction static correction. However, these technologies have limitations in practical applications: elevation static correction does not take into account the velocity and thickness changes of low-velocity bodies; first-arrival wave static correction is suitable for the case where the low-velocity body is thin and uniform, but the effect is poor when the thickness changes greatly. Therefore, the accuracy of the first-arrival picking of traditional static correction technology is not high, which has become a key challenge in the development of static correction technology. Summary of the invention

[0003] Based on this, it is necessary for the present invention to provide a method for picking up the first arrival of seismic data based on the SimpleNet network to solve at least one of the above technical problems.

[0004] To achieve the above purpose, a method for picking first arrival of seismic data based on SimpleNet network includes the following steps:

[0005] Step S1: acquiring seismic shot gather data, and performing grayscale image conversion on the seismic shot gather data, thereby obtaining an enhanced grayscale image set in a function set;

[0006] Step S2: dividing the data set according to the enhanced grayscale image set in the function set, thereby obtaining a training image set and a test image set; performing low-speed body static correction on the training image set, thereby obtaining a corrected training image set;

[0007] Step S3: constructing an earthquake first arrival prediction model according to the enhanced grayscale image set in the function set, thereby obtaining an earthquake first arrival prediction model; performing sample training on the earthquake first arrival prediction model according to the corrected training image set, thereby obtaining an earthquake first arrival training model; optimizing model parameters of the earthquake first arrival training model, thereby obtaining an optimized earthquake first arrival training model;

[0008] Step S4: Predict the first arrival of earthquakes on the test image set according to the optimized earthquake first arrival training model, so as to obtain earthquake first arrival prediction data; classify the earthquake first arrival prediction data according to image size, so as to obtain fixed-size image prediction data and arbitrary-size image prediction data; evaluate the prediction results of the fixed-size image prediction data and the arbitrary-size image prediction data, so as to obtain prediction result evaluation data, and upload them to the seismic cannon platform processing system to perform the earthquake first arrival prediction task.

[0009] The present invention converts seismic shot data into grayscale images, which helps to simplify data processing and reduce data complexity while retaining the main features of seismic waves. Through function set enhancement, the contrast and details of the image are enhanced, thereby improving the quality of the image, making subsequent image analysis and model training more accurate. The enhanced grayscale image makes the seismic signal more obvious, which helps to extract more accurate features. The grayscale image reduces the dimension of the data and simplifies subsequent processing steps, such as feature extraction and model training. The grayscale image set enhanced in the function set is divided into a training set and a test set. This ensures that the training and evaluation of the model are based on an independent data set, thereby improving the generalization ability of the model. The division of the training set and the test set ensures the performance evaluation of the model on unseen data, thereby improving the reliability of the model. The training image set is subjected to low-speed body static correction. The corrected image set can reduce the error caused by velocity change, so that the earthquake first arrival prediction model can more accurately identify the first arrival time of the seismic wave. The low-speed body static correction reduces the error caused by velocity change and improves the prediction accuracy of the first arrival time of the seismic wave. The earthquake first arrival prediction model is constructed based on the enhanced grayscale image set. The model is used to predict the first arrival time of seismic waves. The model is trained using the corrected training image set to optimize the performance of the model on actual data. Through training and optimization, the model can more accurately identify the first arrival time of seismic waves, thereby improving the prediction accuracy. The parameters of the trained model are optimized and the model parameters are adjusted to improve the prediction accuracy. The optimized model can better adapt to different seismic data sets and improve the generalization ability of the model. The optimized model is used to predict the first arrival time of earthquakes on the test set data. The optimized model can accurately predict the first arrival time of earthquakes and improve the efficiency of actual earthquake detection. The prediction results are classified by image size and divided into two categories: fixed size and arbitrary size, so as to analyze the prediction effect in more detail. The prediction results are evaluated, and the performance of the model on images of different sizes is analyzed to ensure the accuracy and reliability of the prediction results. By classifying and evaluating the prediction results of images of different sizes, the advantages and disadvantages of the model can be found, and the model can be further optimized. The evaluation data is uploaded to the seismic cannon platform for further processing and application.

[0010] Optionally, step S1 specifically includes:

[0011] Step S11: acquiring seismic shot gather data, and performing data image mapping on the seismic shot gather data, thereby obtaining a seismic shot gather grayscale image;

[0012] Step S12: performing significant marking of the first arrival function on the seismic cannon grayscale image, thereby obtaining a significant grayscale image;

[0013] Step S13: performing first arrival function distribution label clipping on the significant grayscale image to obtain a concentrated grayscale image;

[0014] Step S14: performing data enhancement based on the concentrated grayscale image to obtain a concentrated grayscale enhanced image.

[0015] The present invention collects original seismic wave data from seismic cannon equipment and converts it into a visualized grayscale image. Data mapping generally includes converting parameters such as velocity and amplitude into grayscale values ​​through a suitable algorithm. The grayscale value of each pixel reflects the physical characteristics of the underground structure. The complex seismic data is converted into a grayscale image, which is convenient for intuitive analysis and understanding of the underground structure, and provides a clear basis for subsequent processing steps, making signal processing and analysis more efficient. By identifying the first arrival wave information in the image, the image is significantly annotated, emphasizing the most important area for analysis, forming a significant grayscale image, highlighting the first arrival wave position, and facilitating the analysis of the propagation path of the seismic wave and the characteristics of the underground structure. By significantly annotating, the interference of non-critical areas is reduced, which is conducive to the accuracy of subsequent processing steps. Crop unnecessary parts of the significant grayscale image, focus on the first arrival wave information related to the underground structure, obtain a concentrated grayscale image, remove redundant information, and make the subsequent analysis centralized and simplified, reduce the amount of calculation, and focus on the analysis of specific areas, so that key geological information can be extracted more accurately. By using various data enhancement techniques (such as rotation, scaling, adding noise, etc.) to process the concentrated grayscale images, the training effect and robustness of the subsequent algorithms can be improved. The enhanced image samples are diversified, which can improve the adaptability of the deep learning model to new data. By enhancing the data, the performance in different scenarios will be more stable, thereby improving the accuracy of seismic wave detection and underground imaging.

[0016] Optionally, step S14 is specifically:

[0017] Step S141: extracting grayscale image samples according to the concentrated grayscale image, thereby obtaining concentrated grayscale image sample data;

[0018] Step S142: performing a horizontal mirror transformation on the concentrated grayscale image sample data, thereby obtaining concentrated grayscale mirror data;

[0019] Step S143: replacing the image data of the centralized grayscale image according to the centralized grayscale mirror data, thereby obtaining centralized grayscale image replacement data;

[0020] Step S144: performing image enhancement integration on the concentrated grayscale image replacement data and the concentrated grayscale image to obtain a concentrated grayscale enhanced image.

[0021] The present invention extracts a number of sample data from a concentrated grayscale image. These samples can be sub-regions of the image and regions selected in a specific way. Sample extraction enables image data of different regions to be analyzed and processed separately, thereby increasing the diversity of the data set. The extracted sample data can be analyzed in a targeted manner, which helps to identify and process specific features or problems in the image. The extracted grayscale image samples are horizontally mirrored, that is, the image is horizontally flipped. The horizontal mirroring increases the diversity of image samples and helps to train a more robust machine learning model. For image features with strong symmetry (such as underground structures), the mirroring can help the model better learn and identify these features. The generated concentrated grayscale mirror data is used to replace part of the data in the original concentrated grayscale image to form new image data. The replacement operation can enhance the expression of certain specific features in the image, thereby improving the sensitivity of the model to these features. Through the replacement operation, the noise or defects in the image data can be improved, making the final image data clearer and more accurate. The concentrated grayscale image replacement data and the original concentrated grayscale image are integrated, and various image enhancement techniques (such as contrast adjustment, noise removal, etc.) are used to obtain the final concentrated grayscale enhanced image. The image quality is improved through integration and enhancement processing, making the details clearer and facilitating further analysis. By integrating image data from different sources, the richness and representativeness of the data can be increased, and the effect of model training and analysis can be improved.

[0022] Optionally, step S2 specifically includes:

[0023] Step S21: dividing the data set according to the enhanced grayscale image set in the function set, so as to obtain a training image set and a test image set;

[0024] Step S22: drawing a wave velocity profile distribution map for the training image set, thereby obtaining a wave velocity profile distribution map;

[0025] Step S23: performing wave velocity statistics on the wave velocity profile distribution diagram to obtain low wave velocity data;

[0026] Step S24: performing regional positioning on the wave velocity profile distribution map according to the low wave velocity data, thereby obtaining low wave velocity regional data;

[0027] Step S25: constructing a low wave speed identification model according to the low wave speed area data, thereby obtaining a low wave speed identification model;

[0028] Step S26: performing low wave speed recognition on the training image set according to the low wave speed recognition model, thereby obtaining a low wave speed training image set;

[0029] Step S27: performing low wave velocity correction on the training image set according to the low wave velocity training image set, so as to obtain a corrected training image set.

[0030] The present invention divides the enhanced grayscale image set into a training set and a test set. Usually, the training set is used for model training, and the test set is used for model evaluation. By clearly dividing the training and test sets, the performance of the model on unseen data can be effectively evaluated to avoid overfitting. Through cross-validation or independent test sets, the evaluation of model performance is more reliable, and the adjustment and optimization of the model can be effectively guided. Using the data in the training image set, a wave velocity profile distribution map is drawn to show the change of wave velocity at different depths or lateral positions. Through the wave velocity profile distribution map, the spatial distribution characteristics of the wave velocity can be intuitively identified, which helps to understand the underground hierarchical structure. This image provides basic information for subsequent low wave velocity data statistics and regional positioning. Based on the wave velocity profile, the wave velocity is statistically analyzed to extract the relevant data of the low wave velocity area. The statistical analysis helps to reveal the low wave velocity characteristics in the entire profile, providing a quantitative basis for subsequent positioning. Through the low wave velocity data, a data foundation is laid for the construction of a low wave velocity recognition model. Based on the low wave velocity data, the low wave velocity area in the wave velocity profile is located, so as to extract the relevant regional data. The accurate regional positioning can clearly define the low wave velocity area and provide targeted regional data for subsequent model training and correction. The clear positioning of the low wave velocity area can provide valuable decision-making basis for geological exploration and related resource development. Using the obtained low wave velocity regional data, a special low wave velocity recognition model is constructed. By constructing a targeted model, the low wave velocity characteristics can be better captured, the recognition efficiency and accuracy can be improved, and useful information can be extracted from complex data to support subsequent analysis. The low wave velocity recognition model is used to analyze the training image set, identify the low wave velocity image part, and realize the automatic recognition of low wave velocity, which greatly improves the work efficiency and reduces the need for manual intervention. The identified low wave velocity training images are helpful to optimize the subsequent image analysis and processing. According to the identified low wave velocity training image set, the original training image set is corrected to eliminate the negative impact of low wave velocity. The low wave velocity correction can effectively improve the quality and accuracy of the training image, lay a solid foundation for subsequent data analysis, and the corrected data set can better reflect the actual wave velocity conditions and improve the performance of the model in practical applications.

[0031] Optionally, step S25 is specifically:

[0032] Step S251: extracting regional shape features, gradient features and variation features from the low wave velocity region data, thereby obtaining regional shape data, gradient data and variation data;

[0033] Step S252: Calculating geometric boundary smoothness of the region shape data, thereby obtaining geometric boundary smoothness data;

[0034] Step S253: performing gradient synthesis according to the gradient data and the change data, thereby obtaining gradient synthesis data;

[0035] Step S254: integrating geometric gradient data according to the geometric boundary smoothness data and the gradient synthesis data, thereby obtaining geometric gradient data;

[0036] Step S255: construct a low wave velocity identification model according to the geometric gradient data, so as to obtain a low wave velocity identification model.

[0037] The present invention can accurately describe the contour and structural characteristics of the region by analyzing the geometric shape of the low wave velocity region. This helps to identify the boundaries, shapes and sizes of the region, and provides basic data for subsequent analysis. By calculating the gradient information in the low wave velocity region, the rate and direction of the wave velocity change in the region can be understood. This helps to identify the region where the wave velocity changes, as well as the abnormal conditions that exist, and identifies the dynamic characteristics of the region by focusing on capturing the changing trends and patterns of the wave velocity in the region. These changing characteristics can reveal the evolution process and potential problems of the low wave velocity region. By calculating the smoothness of the geometric boundary, the smoothness of the boundary of the low wave velocity region can be evaluated. This helps to remove noise and irregularities, making the regional boundary more accurate, and the smoothness information of the boundary helps to better understand the geometric characteristics of the low wave velocity region and ensure the accuracy and consistency of the data in subsequent processing. Gradient synthesis can generate comprehensive gradient information by combining multiple gradient data. This helps to obtain a more comprehensive wave velocity change situation, improve the recognition accuracy of the low wave velocity region, and the synthesized gradient data can more clearly display the wave velocity gradient distribution in the region, which helps to distinguish the low wave velocity region from the normal region. Integrating the geometric boundary smoothness data with the gradient synthesis data can generate comprehensive geometric gradient data. This integration process helps to more accurately describe the overall characteristics and gradient distribution of low wave velocity areas. The geometric gradient data provides detailed characteristics of low wave velocity areas, which helps to better understand the internal structure and wave velocity variation pattern of the area. The low wave velocity identification model based on geometric gradient data can transform the data extracted in the previous steps into a model that can be used for practical applications. The model can be used to automatically identify and classify low wave velocity areas. The establishment of a low wave velocity identification model can improve the automation and accuracy of regional identification, reduce human intervention, and improve processing efficiency. Through the application of the model, large-scale data can be analyzed and processed more quickly to support decision-making and the formulation of action plans.

[0038] Optionally, step S252 is specifically:

[0039] Extracting boundary point coordinate data from the regional shape data, thereby obtaining boundary point coordinate data;

[0040] Performing curve parameterization processing on the boundary point coordinate data, thereby obtaining curve parameterization data;

[0041] Calculate the curvature of the boundary point according to the curve parameterization data, so as to obtain the curvature data of the boundary point;

[0042] Calculate the average curvature of the boundary point curvature data to obtain the average curvature data;

[0043] The smoothness index is calculated based on the mean curvature data to obtain the geometric boundary smoothness data.

[0044] The present invention can clearly define the shape features of the region through the extraction of boundary points, laying the foundation for subsequent analysis, and can be used for tasks such as object recognition and image segmentation, so that the algorithm can process the target object more accurately, which helps to simplify data, reduce the amount of information to be processed in subsequent steps, and improve calculation efficiency. Through parameterization, irregular boundary points can be converted into mathematical curve forms, which is convenient for more complex mathematical analysis. The parameterized curve provides a standardized representation method, which is convenient for subsequent calculations, including mapping, interpolation, deformation, etc., which is conducive to realizing the scaling, rotation and transformation of the curve and improving the flexibility of the graphics. Curvature is an important indicator for describing the change of boundary shape, providing information about the geometric characteristics of the curve path. The calculation of curvature can help identify the characteristic areas of the boundary, such as corners, flat sections, etc., which is crucial for shape analysis. Through curvature data, it can assist in the classification and feature extraction of boundary points, and provide support for subsequent algorithms (such as edge detection and shape matching). By calculating the average curvature, the overall smoothness and variation of the shape can be obtained, which plays a key role in shape feature analysis. This indicator can provide a basis for optimization algorithms (such as path planning and shape compression), improve the rationality and reliability of the results, and help compare between multiple shapes, so that researchers can analyze the geometric characteristics of different regions. The smoothness index provides a quantitative assessment of the smoothness of the shape boundary, which can help judge and screen the quality of the boundary, help reduce noise in numerical calculations, and improve the accuracy and stability of subsequent processing (such as graphics reconstruction and shape matching).

[0045] Optionally, step S253 is specifically:

[0046] Performing gradient amplitude calculation on the gradient data to obtain gradient amplitude data;

[0047] Calculating the change rate of the change data to obtain the change rate data;

[0048] Spatially aligning the change rate data according to the gradient amplitude data, thereby obtaining change space alignment data;

[0049] The gradient synthesis data is obtained by weighted synthesis based on the change space alignment data and the gradient amplitude data.

[0050] The present invention can obtain the intensity information of the changes in the data by calculating the gradient amplitude, which helps to identify the edges and features in the image or signal, making further analysis more accurate. The rate of change calculation is performed on the change data, and the speed and degree of data change can be measured, which is particularly important for the trend analysis of dynamic systems or time series data. The rate of change data is spatially aligned according to the gradient amplitude data, and the consistency of data from different data sources or different time points in space can be ensured, thereby improving the accuracy and reliability of the analysis. By weighted synthesis of the change space alignment data and the gradient amplitude data, more sophisticated and comprehensive gradient synthesis data can be generated in combination with the intensity and spatial information of the changes, which helps further analysis and processing, such as image enhancement or signal optimization.

[0051] Optionally, step S3 specifically includes:

[0052] Step S31: constructing an earthquake first arrival prediction model according to the enhanced grayscale image set in the function set, thereby obtaining an earthquake first arrival prediction model;

[0053] Step S32: extracting earthquake first arrival time data and earthquake first arrival position data from the correction training image set, thereby obtaining earthquake first arrival time data and earthquake first arrival position data;

[0054] Step S33: Calculate the probability of the first arrival time of the earthquake according to the first arrival time data, so as to obtain the probability data of the first arrival time of the earthquake;

[0055] Step S34: performing first-arrival range statistics according to the earthquake first-arrival position data, thereby obtaining earthquake first-arrival range data;

[0056] Step S35: Integrate the earthquake first arrival data according to the earthquake first arrival time probability data and the earthquake first arrival range data, so as to obtain the earthquake first arrival data;

[0057] Step S36: performing sample training on the earthquake first arrival prediction model according to the earthquake first arrival data, thereby obtaining an earthquake first arrival training model;

[0058] Step S37: Optimize the model parameters of the earthquake first-arrival training model according to the corrected training image set, so as to obtain an optimized earthquake first-arrival training model.

[0059] The present invention constructs an earthquake first arrival prediction model by enhancing the grayscale image set, which helps to improve the model's recognition ability of earthquake signals in the image, thereby improving the prediction accuracy. Extracting the earthquake first arrival time and position data provides basic data for subsequent analysis, ensuring that the training and correction of the prediction model have a reliable source of information. Calculating the first arrival time probability data helps to understand the possibility of earthquakes at different time points and enhances the ability to predict the time of earthquake occurrence. Statistical first arrival range data helps determine the spatial range of earthquake signals and provides data support for accurately predicting the location of earthquake occurrence. Integrating time probability data and range data can comprehensively consider the temporal and spatial characteristics of earthquake occurrence and improve the accuracy of the overall prediction model. The prediction model is trained by earthquake first arrival data so that the model is better adapted to actual data, thereby improving the prediction performance. Model parameter optimization further improves the accuracy and stability of the model, making the final earthquake first arrival prediction model more reliable in practical applications.

[0060] Optionally, step S4 specifically includes:

[0061] Step S41: performing earthquake first-arrival prediction on the test image set according to the optimized earthquake first-arrival training model, thereby obtaining earthquake first-arrival prediction data;

[0062] Step S42: classifying the earthquake first arrival prediction data by image size, thereby obtaining fixed-size image prediction data and arbitrary-size image prediction data;

[0063] Step S43: calculating the fixed-size image prediction accuracy rate of the fixed-size image prediction data according to the corrected training image set, thereby obtaining the fixed-size image prediction accuracy rate data;

[0064] Step S44: calculating the prediction accuracy of the image of any size on the prediction data of the image of any size according to the corrected training image set, thereby obtaining the prediction accuracy data of the image of any size;

[0065] Step S45: integrating the accuracy data according to the prediction accuracy data of the fixed-size image and the prediction accuracy data of the arbitrary-size image, thereby obtaining the earthquake first arrival prediction accuracy data;

[0066] Step S46: Evaluate the prediction results of the earthquake first arrival prediction accuracy data to obtain prediction result evaluation data, and upload it to the seismic cannon platform processing system to perform the earthquake first arrival prediction task.

[0067] The present invention can obtain earthquake first arrival prediction data by predicting the test image set according to the optimized earthquake first arrival training model. The key effect of this step is to use the optimized model to accurately predict the new data, and provide preliminary earthquake occurrence time and location prediction for actual earthquake early warning. The earthquake first arrival prediction data is classified by image size, and the data is divided into two categories: fixed size and arbitrary size. This classification can help to process different types of image data separately, so that the prediction results can obtain corresponding accuracy analysis under different conditions. According to the correction training image set, the accuracy of the fixed size image prediction data is calculated, and the prediction accuracy of the model on the fixed size image can be evaluated. The beneficial effect of this step is that the performance of the model under standardized image conditions can be revealed, providing a reference for further optimization. According to the correction training image set, the accuracy of the arbitrary size image prediction data is calculated, and the prediction performance of the model when processing images of different sizes can be evaluated. The benefit of this step is that it can ensure that the model has good prediction ability when processing various image sizes encountered in actual applications. The prediction accuracy data of fixed size and arbitrary size are integrated to obtain comprehensive earthquake first arrival prediction accuracy data. The integrated data can provide a comprehensive model performance evaluation, reflecting the overall accuracy of the model under different conditions. The results of earthquake first arrival prediction accuracy data are evaluated, and the evaluation data is obtained and uploaded to the seismic cannon platform processing system. The effect of this step is to use the evaluation results in practical applications to ensure that the prediction results can effectively provide decision support for the earthquake early warning system and enhance the timeliness and accuracy of earthquake prediction.

[0068] Optionally, step S43 is specifically:

[0069] Step S431: extracting fixed-size earthquake first-arrival prediction data from the fixed-size image prediction data, thereby obtaining fixed-size earthquake first-arrival prediction data;

[0070] Step S432: extracting seismic first-arrival correction data from the correction training image set, thereby obtaining seismic first-arrival correction data;

[0071] Step S433: performing earthquake first-arrival annotation comparison on the fixed-size earthquake first-arrival prediction data according to the earthquake first-arrival correction data, thereby obtaining fixed-size image earthquake first-arrival annotation comparison data;

[0072] Step S434: Calculate the accuracy of the fixed-size image seismic first-arrival annotation comparison data based on the corrected training image set, so as to obtain the fixed-size image prediction accuracy data.

[0073] The present invention extracts earthquake first arrival prediction data through fixed-size images, which can ensure consistency in different regions or image sizes, help reduce prediction errors caused by image size differences, and improve data comparability and processing efficiency. Extracting earthquake first arrival correction data helps to create an accurate correction model. This step provides real reference data for subsequent verification and optimization of prediction data, and enhances the reliability of the model. Comparing the first arrival annotations of earthquakes can reveal the differences between the predicted data and the actual annotations. This comparative analysis helps to identify the prediction errors and potential problems of the model, and provides a direction for further improving the model. By calculating the prediction accuracy rate, the performance of the prediction model can be quantified and its accuracy can be evaluated. This provides clear indicators for optimizing the prediction model and improves the credibility of the final prediction results. BRIEF DESCRIPTION OF THE DRAWINGS

[0074] Other features, objects and advantages of the present invention will become more apparent from the detailed description of non-limiting embodiments thereof made with reference to the following drawings:

[0075] Figure 1 The figure is a schematic flow chart of the steps of the method for picking the first arrival of seismic data based on the SimpleNet network of the present invention;

[0076] Figure 2 Detailed step flow diagram of step S1 in the present invention;

[0077] Figure 3 Detailed step flow diagram of step S14 in the present invention;

[0078] The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0079] The following is a clear and complete description of the technical method of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are 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 technicians in this field without creative work are within the scope of protection of the present invention.

[0080] In addition, the accompanying drawings are only schematic illustrations of the present invention and are not necessarily drawn to scale. The same reference numerals in the figures represent the same or similar parts, and their repeated description will be omitted. Some of the block diagrams shown in the accompanying drawings are functional entities and do not necessarily correspond to physically or logically independent entities. The functional entities can be implemented in software form, or implemented in one or more hardware modules or integrated circuits, or implemented in different networks and / or processor methods and / or microcontroller methods.

[0081] It should be understood that, although the terms "first", "second", etc. may be used herein to describe various units, these units should not be limited by these terms. These terms are used only to distinguish one unit from another unit. For example, without departing from the scope of the exemplary embodiments, the first unit may be referred to as the second unit, and similarly the second unit may be referred to as the first unit. The term "and / or" used herein includes any and all combinations of one or more of the listed associated items.

[0082] To achieve this, please refer to Figures 1 to 3 The present invention provides a method for picking up first arrivals of seismic data based on SimpleNet network, the method comprising the following steps:

[0083] Step S1: acquiring seismic shot gather data, and performing grayscale image conversion on the seismic shot gather data, thereby obtaining an enhanced grayscale image set in a function set;

[0084] In this embodiment, seismic shot data are collected, which include raw seismic signals obtained from seismic detection instruments. These signals are usually recorded in the form of images (for example, acoustic reflection images of seismic waves). Before processing these image data, they are converted into grayscale images, which can be achieved by converting the color channels of RGB images into a single brightness channel. This process helps to reduce the complexity of the image and focus on the intensity information of the seismic waves. Next, image enhancement technology is used to improve the contrast and clarity of the image. Enhancement techniques include histogram equalization, contrast-limited adaptive histogram equalization (CLAHE) and other methods. These techniques can make the characteristics of seismic waves more prominent, thereby improving the accuracy of subsequent processing and analysis. In the end, a set of enhanced grayscale image sets is obtained, preparing for the next step of data set division.

[0085] Step S2: dividing the data set according to the enhanced grayscale image set in the function set, thereby obtaining a training image set and a test image set; performing low-speed body static correction on the training image set, thereby obtaining a corrected training image set;

[0086] In this embodiment, based on the enhanced grayscale image set, these image data are divided into a training image set and a test image set. The common division ratio is 80% for training and 20% for testing. Ensure the representativeness and balance of the data during division so that various seismic features can be covered when training the model. The training image set is subjected to low-velocity static correction, which usually includes correcting the seismic waves in the image to eliminate the influence of low-velocity bodies (such as crustal inhomogeneities) on wave propagation. Existing correction algorithms can be used, for example, correction techniques based on wave velocity models. The corrected image set is called a corrected training image set, which ensures that the images used for training have more accurate seismic features, thereby improving the prediction ability of the model.

[0087] Step S3: constructing an earthquake first arrival prediction model according to the enhanced grayscale image set in the function set, thereby obtaining an earthquake first arrival prediction model; performing sample training on the earthquake first arrival prediction model according to the corrected training image set, thereby obtaining an earthquake first arrival training model; optimizing model parameters of the earthquake first arrival training model, thereby obtaining an optimized earthquake first arrival training model;

[0088] In this embodiment, an earthquake first arrival prediction model is constructed according to a grayscale image set enhanced in a function set. This can be achieved by selecting an appropriate machine learning algorithm, such as a convolutional neural network (CNN), and designing an appropriate network architecture (such as a combination of convolutional layers, pooling layers, and fully connected layers). The goal of the model is to predict the first arrival time of the seismic wave from the input image, that is, the time when the seismic wave arrives at the detection point from the source. The initially constructed prediction model is trained using a correction training image set. This includes training the model for multiple rounds (epochs) and adjusting the weights and biases of the model to minimize the prediction error. Cross-validation techniques are used during training to ensure that the model performs consistently on different data subsets. After model training, parameter optimization is required. Hyperparameter tuning techniques, such as Grid Search or Bayesian Optimization, can be used to adjust hyperparameters such as learning rate and batch size to obtain optimal model performance. The optimized model is called an optimized earthquake first arrival training model.

[0089] Step S4: Predict the first arrival of earthquakes on the test image set according to the optimized earthquake first arrival training model, so as to obtain earthquake first arrival prediction data; classify the earthquake first arrival prediction data according to image size, so as to obtain fixed-size image prediction data and arbitrary-size image prediction data; evaluate the prediction results of the fixed-size image prediction data and the arbitrary-size image prediction data, so as to obtain prediction result evaluation data, and upload them to the seismic cannon platform processing system to perform the earthquake first arrival prediction task.

[0090] In this embodiment, the optimized earthquake first arrival training model is used to predict the first arrival of earthquakes on the test image set. The test set images are predicted to obtain the earthquake first arrival time data of each image. The prediction results are classified according to the image size. Assume that the image has two sizes: fixed size and arbitrary size. The fixed size image prediction data and the arbitrary size image prediction data are evaluated separately, and indicators such as prediction accuracy, recall rate and F1 score are used to measure the quality of the prediction results. Finally, the prediction results are uploaded to the seismic cannon platform processing system. This platform can be used for further data analysis and visualization to help earthquake researchers understand the characteristics of seismic wave propagation and make decisions. The upload process includes storing the prediction results and evaluation data in a standard format and ensuring the integrity and accuracy of the data.

[0091] Optionally, step S1 specifically includes:

[0092] Step S11: acquiring seismic shot gather data, and performing data image mapping on the seismic shot gather data, thereby obtaining a seismic shot gather grayscale image;

[0093] In this embodiment, obtaining seismic shot data includes collecting raw seismic wave data from seismic detection equipment. These data are usually in the form of continuous time series, including amplitude information of seismic waves. The data image mapping process converts these time series data into image format, and a common technique is to map the amplitude value of the seismic wave to the pixel value of the grayscale image. For example, a normalization technique can be used to map the amplitude value to a range of 0-255 to generate a corresponding grayscale image, in which the intensity of the seismic wave is represented as different grayscale levels in the image.

[0094] Step S12: performing significant marking of the first arrival function on the seismic cannon grayscale image, thereby obtaining a significant grayscale image;

[0095] In this embodiment, the first arrival function is to identify the characteristics of the first arrival wave by processing the grayscale image. The first arrival wave characteristics in the image are marked prominently using an image processing algorithm, such as edge detection (e.g., Canny edge detection) or a threshold segmentation method. For example, by applying an adaptive threshold method, the obvious area of ​​the first arrival wave can be highlighted, and a significant grayscale image can be generated to clearly mark the position and intensity of the first arrival wave.

[0096] Step S13: performing first arrival function distribution label clipping on the significant grayscale image to obtain a concentrated grayscale image;

[0097] In this embodiment, the significant grayscale image is cropped by the first arrival function distribution label, with the goal of removing unnecessary background noise and irrelevant image parts. This process usually includes using an image cropping algorithm to crop the image according to the spatial distribution of the first arrival wave. For example, by setting a region of interest (ROI), only the area with obvious first arrival waves is retained to generate a concentrated grayscale image, which can effectively focus on the main feature area.

[0098] Step S14: performing data enhancement based on the concentrated grayscale image to obtain a concentrated grayscale enhanced image.

[0099] In this embodiment, data enhancement of the centralized grayscale image is used to improve the image quality and the robustness of the model. Common data enhancement techniques include rotation, translation, scaling, and color jittering. For example, rotation and scaling transformations are applied to the centralized grayscale image to generate multiple enhanced images, simulating different observation angles and conditions to enhance the generalization ability of the model and the ability to recognize seismic wave characteristics.

[0100] Optionally, step S14 is specifically:

[0101] Step S141: extracting grayscale image samples according to the concentrated grayscale image, thereby obtaining concentrated grayscale image sample data;

[0102] In this embodiment, in image processing, grayscale image sample extraction based on the concentrated grayscale image generally involves the following steps. First, by defining the sampling window size and position, the pixel values ​​of a specific area are extracted from the original grayscale image as sample data. This can be achieved by extracting by sliding window or area selection. For example, in face recognition, sample data is formed by extracting feature areas such as eyes and mouth, and then feature learning and recognition are performed.

[0103] Step S142: performing a horizontal mirror transformation on the concentrated grayscale image sample data, thereby obtaining concentrated grayscale mirror data;

[0104] In this embodiment, horizontal mirroring of grayscale image sample data is a common data enhancement technique that can be used to expand the training data set to enhance the generalization ability of the model. This involves flipping the original image along the vertical central axis to generate mirror data. For example, for traffic sign recognition, the left-right structure of the original sign can be transformed through horizontal mirroring to increase the diversity of training data.

[0105] Step S143: replacing the image data of the centralized grayscale image according to the centralized grayscale mirror data, thereby obtaining centralized grayscale image replacement data;

[0106] In this embodiment, in the image data replacement stage, the grayscale image is replaced or merged at the pixel level using the centralized grayscale mirror data. This may include pixel-level difference calculation, application of a blending mode, or pixel replacement based on conditional judgment. For example, the content filling operation in image restoration uses a replacement technique to fill in the missing area so that it is coordinated with the surrounding environment.

[0107] Step S144: performing image enhancement integration on the concentrated grayscale image replacement data and the concentrated grayscale image to obtain a concentrated grayscale enhanced image.

[0108] In this embodiment, image enhancement integration involves the integration of various enhancement processing methods by replacing the grayscale image data and the original grayscale image. This includes operations such as histogram equalization, filtering, and sharpening. For example, in medical image processing, the image data is fused and the enhancement technology is used to improve the contrast and clarity of the image, thereby better showing the details of the lesion or tissue structure.

[0109] Optionally, step S2 specifically includes:

[0110] Step S21: dividing the data set according to the enhanced grayscale image set in the function set, so as to obtain a training image set and a test image set;

[0111] In this embodiment, a grayscale image set is obtained from a function set, and these images represent different scenes or objects. These images are preprocessed, such as standardization and denoising. Then, the image set is randomly divided into a training set and a test set according to a certain ratio (such as 80%:20%). The training set is used for model training, and the test set is used to evaluate the model performance. In a specific implementation, data enhancement techniques (such as rotation, translation, and scaling) can be used to expand the training set to improve the generalization ability of the model.

[0112] Step S22: drawing a wave velocity profile distribution map for the training image set, thereby obtaining a wave velocity profile distribution map;

[0113] In this embodiment, for each image in the training image set, a wave velocity profile distribution map is generated using an appropriate algorithm (such as gradient boosting, convolutional neural network). The map depicts the wave velocity changes in different areas of the image. The generation process includes image feature extraction and wave velocity calculation, and these wave velocity information can be obtained through a set calculation model. In specific implementations, different wave velocity calculation methods, such as Fourier transform, can be selected to generate profiles of different forms.

[0114] Step S23: performing wave velocity statistics on the wave velocity profile distribution diagram to obtain low wave velocity data;

[0115] In this embodiment, the velocity profile distribution diagram is statistically analyzed to extract the velocity data. Low velocity values ​​are identified by statistical methods (such as mean, median, standard deviation). Specifically, threshold segmentation or histogram analysis methods can be used to distinguish low velocity areas. During implementation, appropriate threshold standards are set to ensure the accuracy of low velocity data, and post-processing (such as filtering) is required to remove noise.

[0116] Step S24: performing regional positioning on the wave velocity profile distribution map according to the low wave velocity data, thereby obtaining low wave velocity regional data;

[0117] In this embodiment, based on the low wave velocity data, the region is located in the wave velocity profile distribution map, and the low wave velocity region is marked. This process includes using a region growing algorithm or a threshold segmentation technique to define the low wave velocity region. During implementation, image segmentation techniques such as K-means clustering can be combined to accurately locate these regions and generate a labeled map of the low wave velocity region data.

[0118] Step S25: constructing a low wave speed identification model according to the low wave speed area data, thereby obtaining a low wave speed identification model;

[0119] In this embodiment, a low-velocity area data is used to construct a low-velocity recognition model. A suitable machine learning model (such as a support vector machine, a deep neural network) is selected and trained to identify and classify low-velocity areas in the image. In the specific implementation, it is necessary to design a training process, including model selection, hyperparameter tuning, and cross-validation, to ensure the accuracy and robustness of the model.

[0120] Step S26: performing low wave speed recognition on the training image set according to the low wave speed recognition model, thereby obtaining a low wave speed training image set;

[0121] In this embodiment, the constructed low wave velocity recognition model is used to perform low wave velocity recognition on the training image set to generate a low wave velocity training image set. This step involves model prediction and post-processing to determine which image areas are classified as low wave velocity. During implementation, it is necessary to evaluate the recognition accuracy of the model and perform multiple iterations to optimize the results.

[0122] Step S27: performing low wave velocity correction on the training image set according to the low wave velocity training image set, so as to obtain a corrected training image set.

[0123] In this embodiment, the low wave velocity area in the training image set is corrected to generate a corrected training image set. This step uses the output results of the low wave velocity recognition model to adjust the image to correct the errors or deviations in the recognition. In specific implementation, it is necessary to apply image processing technology (such as image fusion and enhancement) to improve the image quality and ensure that the corrected image set can more accurately reflect the low wave velocity characteristics.

[0124] Optionally, step S25 is specifically:

[0125] Step S251: extracting regional shape features, gradient features and variation features from the low wave velocity region data, thereby obtaining regional shape data, gradient data and variation data;

[0126] In this embodiment, the data of the low wave velocity area is processed to extract multiple features. First, the regional shape feature extraction involves identifying the shape features of the low wave velocity area from the data, such as the boundary curvature, aspect ratio and area of ​​the area. Specifically, the regional boundary can be captured by image processing technology, such as edge detection algorithm (such as Canny edge detection), and then the regional shape can be quantified by shape descriptors (such as Hu moments). Gradient feature extraction focuses on the direction and intensity of the wave velocity change in the region. For example, by calculating the gradient vector field in the region, the spatial change trend of the wave velocity is obtained. Change feature extraction includes analyzing the time or space change pattern of the wave velocity, such as obtaining the rate of change of the wave velocity through time series analysis or spatial filtering technology. These extracted features will form regional shape data, gradient data and change data.

[0127] Step S252: Calculating geometric boundary smoothness of the region shape data, thereby obtaining geometric boundary smoothness data;

[0128] In this embodiment, the geometric boundary smoothness calculation is performed on the regional shape data, the purpose is to reduce boundary noise and improve the accuracy of feature extraction. In specific implementation, a smoothing algorithm, such as Gaussian filtering or median filtering, can be applied to smooth the geometric shape of the regional boundary. These algorithms eliminate noise by adjusting the curvature and shape of the boundary to ensure that the boundary is smoother. The smoothed boundary can be used to calculate boundary smoothness indicators, such as the root mean square error of the boundary curve or the smoothness ratio of the boundary. These indicators are used to evaluate the smoothness of the regional boundary and generate geometric boundary smoothness data.

[0129] Step S253: performing gradient synthesis according to the gradient data and the change data, thereby obtaining gradient synthesis data;

[0130] In this embodiment, gradient synthesis is to merge the gradient data obtained from different sources into a comprehensive gradient data set. In the specific implementation, first, the extracted gradient data are fused according to certain rules. For example, a weighted average method can be used, in which the weight of each gradient data source is set according to its importance in practical applications. Then, the fused gradient data is standardized to ensure that the gradient information of different data sources has consistent dimensions and distribution after synthesis. The goal of this process is to generate a gradient synthesis data that comprehensively reflects the wave velocity changes in the region.

[0131] Step S254: integrating geometric gradient data according to the geometric boundary smoothness data and the gradient synthesis data, thereby obtaining geometric gradient data;

[0132] In this embodiment, the geometric boundary smoothness data is integrated with the gradient synthesis data. The specific implementation method includes first performing feature matching and alignment on the geometric boundary smoothness data and the gradient synthesis data, and then merging the two through a fusion algorithm (such as weighted sum or principal component analysis). The integration method can be adjusted according to the characteristics of the data. For example, the geometric gradient characteristics of each region are determined through comprehensive analysis by combining the boundary smoothness and gradient synthesis data. The final result of this step is geometric gradient data, which comprehensively considers the smoothness of the regional shape and the gradient information of the wave velocity change.

[0133] Step S255: construct a low wave velocity identification model according to the geometric gradient data, so as to obtain a low wave velocity identification model.

[0134] In this embodiment, a low wave velocity recognition model is constructed. Based on the geometric gradient data extracted above, the model is trained by a machine learning algorithm. Common methods include support vector machines (SVM), decision trees, or neural networks. The training data set includes labeled low wave velocity area samples and their corresponding geometric gradient data. During the model training process, cross-validation technology is used to optimize the model parameters to ensure the generalization ability of the model. After the training is completed, the model can identify and classify the low wave velocity areas of the newly input data. Model evaluation can verify its performance through indicators such as accuracy and recall to ensure its effectiveness in practical applications.

[0135] Optionally, step S252 is specifically:

[0136] Extracting boundary point coordinate data from the regional shape data, thereby obtaining boundary point coordinate data;

[0137] In this embodiment, the boundary point coordinate data is extracted from the regional shape data. For example, in urban planning, GIS software is used to import the vector layer data of the region, which is usually stored in the form of polygons. The boundary point coordinates of the region are exported through the boundary extraction tool of the software. These coordinate points will be arranged in order to form a point list, such as [(0,0), (0,10), (10,10), (10,0)]. It is ensured that all boundary points are extracted in the actual order of the boundary so that the curve fitting and analysis in the subsequent processing are accurate.

[0138] Performing curve parameterization processing on the boundary point coordinate data, thereby obtaining curve parameterization data;

[0139] In this embodiment, the extracted boundary point coordinate data is subjected to curve parameterization. The arc length parameterization method is selected, that is, parameter mapping is performed based on the distance between boundary points. The total distance from each boundary point to the starting point is calculated, and these distances are normalized to generate a parameter value sequence. For example, for a rectangular area, the length of each side is calculated, and the parameter value of each point is set to the total length accumulated along the boundary. This parameterization can convert a discrete set of points into a continuous curve description, laying the foundation for subsequent curvature calculations.

[0140] Calculate the curvature of the boundary point according to the curve parameterization data, so as to obtain the curvature data of the boundary point;

[0141] In this embodiment, the curvature of each boundary point is calculated using the curve parameterization data. An appropriate numerical method is selected to calculate the curvature, such as using the second-order derivative of the curve. If the curve is discrete, the curvature can be estimated using the difference method of adjacent points. For example, for a rectangle, the curvature of its corner points is usually very large, so the curvature calculation can be concentrated on the midpoint of the curve segment. In this way, a series of curvature values ​​are obtained, which can reflect the changing characteristics of the boundary.

[0142] Calculate the average curvature of the boundary point curvature data to obtain the average curvature data;

[0143] In this embodiment, all calculated curvature data are averaged to obtain average curvature data. For example, the curvature values ​​of all boundary points are summarized and their arithmetic mean is calculated. Assuming that the calculated curvature data includes {0.1, 0.2, 0.15, 0.25}, the average curvature is the average of these values. The purpose of this step is to provide an overall curvature feature to help understand the overall smoothness of the boundary.

[0144] The smoothness index is calculated based on the mean curvature data to obtain the geometric boundary smoothness data.

[0145] In this embodiment, the smoothness index of the boundary is further calculated based on the calculated average curvature data. This index is used to measure the smoothness of the boundary, and is usually the inverse of the average curvature or some transformation thereof. For example, if the calculated average curvature is 0.2, the smoothness index can be 1 / 0.2=5. This smoothness index provides a quantitative indicator to evaluate the smoothness of the boundary, and the higher the value, the smoother the boundary. This result can be used to determine whether the regional design meets the expected geometric standards.

[0146] Optionally, step S253 is specifically:

[0147] Performing gradient amplitude calculation on the gradient data to obtain gradient amplitude data;

[0148] In this embodiment, the gradient amplitude of image data or terrain data is calculated. Assume that the operation is performed in a digital elevation model (DEM). The gradient amplitude of each pixel is calculated using image processing software or a geographic information system (GIS) tool. The specific steps are to apply a gradient operator (such as a Sobel operator) to each pixel to calculate the gradient components in the horizontal direction (x direction) and the vertical direction (y direction). Then, the gradient amplitude is calculated using these components, which is essentially the square root of the sum of the squares of these two components. For example, for a pixel, the horizontal gradient is 3 and the vertical gradient is 4, then the gradient amplitude of the pixel is 5. This amplitude value provides the intensity of each point change.

[0149] Calculating the change rate of the change data to obtain the change rate data;

[0150] In this embodiment, the change rate of the change data is calculated. The change data comes from a time series image, such as surface temperature data at different time points. Calculating the change rate is to measure the speed of change over time. For example, if there are data at two time points, t1 and t2, and the corresponding values ​​are V1 and V2, then the change rate is (V2-V1) / Δt, where Δt is the time interval. For the surface temperature data, assuming that the temperature of a certain place is 15°C at t1 and 18°C ​​at t2, and the time interval is 1 year, then the change rate is 3°C / year. This step is used to measure the speed at which data changes over time.

[0151] Spatially aligning the change rate data according to the gradient amplitude data, thereby obtaining change space alignment data;

[0152] In this embodiment, the gradient amplitude data and the rate of change data are spatially aligned. Assuming that the gradient amplitude data and the rate of change data originate from different sensors or data sets, they need to be spatially aligned so that they can be analyzed in the same coordinate system. Specifically, the image registration technology is used to align the two data sets through affine transformation or other spatial transformation. For example, between satellite images and ground sensor data, feature matching methods (such as SIFT) are used to align data points to ensure that the gradient amplitude of each point matches the corresponding rate of change. This step ensures that subsequent weighted synthesis can be processed at the same spatial position.

[0153] The gradient synthesis data is obtained by weighted synthesis based on the change space alignment data and the gradient amplitude data.

[0154] In this embodiment, weighted synthesis is performed based on the aligned data to generate gradient synthetic data. A suitable weighting method is selected to combine the gradient amplitude data and the change space alignment data. For example, the gradient amplitude can be used as a weight to weight the rate of change data, that is, the rate of change of each point is multiplied by its corresponding gradient amplitude, and then normalized to obtain a weighted synthetic value. For example, if the gradient amplitude of a certain area is 5 and the rate of change is 3°C / year, the weighted synthetic value of the area is 15. This synthetic data can be used for further analysis, such as identifying key areas of surface change or evaluating the impact of terrain on change.

[0155] Optionally, step S3 specifically includes:

[0156] Step S31: constructing an earthquake first arrival prediction model according to the enhanced grayscale image set in the function set, thereby obtaining an earthquake first arrival prediction model;

[0157] In this embodiment, first, a representative set of earthquake grayscale images is collected and integrated, which should cover different earthquake events and geological backgrounds. Then, these grayscale images are preprocessed using image enhancement techniques (such as histogram equalization and Gaussian filtering) to improve the contrast and clarity of the images. Next, a deep learning model (such as a convolutional neural network) is used to train an earthquake first arrival prediction model, which will learn the characteristics of seismic wave propagation from the enhanced images, thereby accurately predicting the first arrival time and location of the earthquake.

[0158] Step S32: extracting earthquake first arrival time data and earthquake first arrival position data from the correction training image set, thereby obtaining earthquake first arrival time data and earthquake first arrival position data;

[0159] In this embodiment, the correction training image set is analyzed to extract the data of the earthquake first arrival time and position. These data should include the actual recorded earthquake first arrival time and the corresponding seismic wave position. The extraction method can be through image recognition technology or manual annotation to ensure that the obtained data set has high accuracy and representativeness.

[0160] Step S33: Calculate the probability of the first arrival time of the earthquake according to the first arrival time data, so as to obtain the probability data of the first arrival time of the earthquake;

[0161] In this embodiment, the probability distribution of the first arrival time is calculated using the extracted earthquake first arrival time data. This process involves statistical analysis of historical earthquake data to determine the probability distribution of the first arrival time of different earthquake events. The probability data of the first arrival time of an earthquake is obtained by constructing a probability model (such as a Bayesian model) to describe the changing trend of these times.

[0162] Step S34: performing first-arrival range statistics according to the earthquake first-arrival position data, thereby obtaining earthquake first-arrival range data;

[0163] In this embodiment, a statistical analysis of the first arrival range is performed based on the earthquake first arrival position data. By analyzing the first arrival position data of different earthquake events, the spatial distribution range of the first arrival wave is calculated. The statistical method may include calculating the mean, standard deviation and spatial distribution characteristics of the position data to obtain the geographical range data of the earthquake first arrival.

[0164] Step S35: Integrate the earthquake first arrival data according to the earthquake first arrival time probability data and the earthquake first arrival range data, so as to obtain the earthquake first arrival data;

[0165] In this embodiment, the earthquake first arrival time probability data is integrated with the first arrival range data. This process involves combining the time probability distribution with the spatial range data to create a comprehensive earthquake first arrival data set. This data set will help understand the time and spatial distribution characteristics of the earthquake first arrival and provide more comprehensive information for the prediction model.

[0166] Step S36: performing sample training on the earthquake first arrival prediction model according to the earthquake first arrival data, thereby obtaining an earthquake first arrival training model;

[0167] In this embodiment, the integrated earthquake first arrival data is used to train the earthquake first arrival prediction model initially constructed. The integrated data set is used as a training sample, and the model parameters are optimized through the training process to improve the accuracy and robustness of the model. The ultimate goal of this step is to obtain a trained earthquake first arrival prediction model that can predict the first arrival time and location of an earthquake based on the input image data.

[0168] Step S37: Optimize the model parameters of the earthquake first-arrival training model according to the corrected training image set, so as to obtain an optimized earthquake first-arrival training model.

[0169] In this embodiment, the earthquake first arrival training model is optimized. The parameters of the model are adjusted using the data in the correction training image set to improve the prediction accuracy of the model. The optimization method includes adjusting hyperparameters such as learning rate, batch size, and number of training rounds to ensure that the model has the best performance in practical applications. The optimized earthquake first arrival training model finally obtained should have high prediction accuracy and reliability.

[0170] Optionally, step S4 specifically includes:

[0171] Step S41: performing earthquake first-arrival prediction on the test image set according to the optimized earthquake first-arrival training model, thereby obtaining earthquake first-arrival prediction data;

[0172] In the present embodiment, the test image set is predicted using the optimized earthquake first arrival training model. These test image sets come from the earthquake observation system, and they are pre-processed to ensure the quality and consistency of their data. The optimized earthquake first arrival training model is usually a deep learning model trained based on historical earthquake data, which can identify and predict the first arrival time of seismic waves. The optimization of the model includes adjusting the network structure, training parameters, and adding the latest seismic wave feature data. The predicted result is the predicted time of the first arrival of the earthquake in each test image, and the predicted data will be used for subsequent evaluation and analysis.

[0173] Step S42: classifying the earthquake first arrival prediction data by image size, thereby obtaining fixed-size image prediction data and arbitrary-size image prediction data;

[0174] In this embodiment, the earthquake first arrival prediction data is classified by image size. Specifically, according to the actual size of the image, the prediction data is divided into fixed-size image prediction data and arbitrary-size image prediction data. Fixed-size image prediction data refers to images of fixed size (for example, 512x512 pixels) specified during training, and these images remain unchanged in size during the prediction process. Arbitrary-size image prediction data includes images whose specific size is not specified during the training process, and the size may change. During the classification process, the image size needs to be detected and recorded so that targeted accuracy calculations can be performed in subsequent steps.

[0175] Step S43: calculating the fixed-size image prediction accuracy rate of the fixed-size image prediction data according to the corrected training image set, thereby obtaining the fixed-size image prediction accuracy rate data;

[0176] In this embodiment, the accuracy of the fixed-size image prediction data is calculated. First, the correction training image set (i.e., the seismic image data set that has been manually annotated or corrected) is used as a reference to compare the difference between the first arrival time predicted by the model and the actual first arrival time. The calculation of the accuracy includes calculating statistical indicators such as the error and standard deviation of the predicted time to evaluate the performance of the model on the fixed-size image. The accuracy is usually expressed as the relative difference between the predicted value and the actual value, which provides the reliability of the model prediction on the fixed-size image data.

[0177] Step S44: calculating the prediction accuracy of the image of any size on the prediction data of the image of any size according to the corrected training image set, thereby obtaining the prediction accuracy data of the image of any size;

[0178] In this embodiment, the accuracy rate is calculated for the prediction data of images of any size. The image data of any size in the correction training image set is used for comparison to evaluate the performance of the model on images of different sizes. The calculation method is similar to that of fixed-size images, including error calculation, statistical analysis, etc. Since images of any size have more variability, preprocessing operations such as image scaling or cropping are required to ensure consistency in the comparison of prediction results.

[0179] Step S45: integrating the accuracy data according to the prediction accuracy data of the fixed-size image and the prediction accuracy data of the arbitrary-size image, thereby obtaining the earthquake first arrival prediction accuracy data;

[0180] In this embodiment, the accuracy data is integrated. The specific method includes summarizing the accuracy data of fixed-size and arbitrary-size images, and obtaining the overall earthquake first arrival prediction accuracy data through weighted average or other statistical methods. The purpose of the integration is to provide a comprehensive evaluation to reflect the overall performance of the model on different types of images. This step involves calculating indicators such as the overall prediction accuracy and error range to provide an evaluation of the overall performance of the model.

[0181] Step S46: Evaluate the prediction results of the earthquake first arrival prediction accuracy data to obtain prediction result evaluation data, and upload it to the seismic cannon platform processing system to perform the earthquake first arrival prediction task.

[0182] In this embodiment, the integrated earthquake first arrival prediction accuracy data is evaluated. This includes analyzing the reliability of the prediction results, the performance of the model in practical applications, etc. The evaluation data can be presented in the form of statistical reports, charts, etc. After the evaluation is completed, the prediction results are uploaded to the seismic cannon platform processing system, which will further analyze the data and apply it to the actual earthquake prediction task. The upload process needs to ensure the integrity and accuracy of the data, usually including steps such as data formatting and verification, for subsequent processing and analysis.

[0183] Optionally, step S43 is specifically:

[0184] Step S431: extracting fixed-size earthquake first-arrival prediction data from the fixed-size image prediction data, thereby obtaining fixed-size earthquake first-arrival prediction data;

[0185] In this embodiment, a convolutional neural network (CNN) is used to analyze seismic data. Considering the spatial distribution characteristics of seismic data, the input seismic data image is first cropped or scaled to a uniform fixed size (e.g., 256x256 pixels). The first arrival of an earthquake is predicted by a pre-trained deep learning model, and a predicted image is output, which contains the first arrival prediction value corresponding to each pixel. In order to ensure the uniformity and comparability of the data, the resolution and size of the predicted data must be consistent with the training data.

[0186] Step S432: extracting seismic first-arrival correction data from the correction training image set, thereby obtaining seismic first-arrival correction data;

[0187] In this embodiment, the actual observation data of the earthquake first arrival is extracted from the correction training image set. First, each image in the correction training image set is cropped or scaled to the same fixed size as the predicted data. For example, for each correction image, the actual arrival time of its earthquake first arrival is determined and marked as the first arrival correction data. The correction data should be strictly aligned with the predicted data in space to ensure that the labeled data of each pixel accurately reflects the real earthquake first arrival information.

[0188] Step S433: performing earthquake first-arrival annotation comparison on the fixed-size earthquake first-arrival prediction data according to the earthquake first-arrival correction data, thereby obtaining fixed-size image earthquake first-arrival annotation comparison data;

[0189] In this embodiment, the fixed-size earthquake first arrival prediction data is compared with the earthquake first arrival correction data. Using a pixel-level comparison method, each predicted value is matched with the corresponding correction data to identify the difference between the prediction and the actual annotation. Through comparative analysis, the earthquake first arrival annotation comparison data of the fixed-size image is generated. This data set shows the error distribution of each predicted point and the actual correction point, which helps to further evaluate and improve the prediction model.

[0190] Step S434: Calculate the accuracy of the fixed-size image seismic first-arrival annotation comparison data based on the corrected training image set, so as to obtain the fixed-size image prediction accuracy data.

[0191] In this embodiment, the prediction accuracy is calculated based on the comparison data of the first earthquake arrival annotations. The accuracy is defined as the ratio of correctly predicted pixels in the prediction data to the total predicted pixels. The matching of the accurately annotated prediction points and the actual value points in the comparison data is compared, and the ratio between the number of correctly predicted pixels and the total number of predicted pixels is counted. The prediction accuracy data of the fixed-size image is obtained, which reflects the prediction ability of the model under fixed size. In order to improve the calculation accuracy, different types of error analysis can be considered, such as error distribution and deviation measurement.

[0192] Therefore, the embodiments should be regarded as illustrative and non-restrictive from all points, and the scope of the present invention is limited by the appended claims rather than the above description, and it is intended that all changes falling within the meaning and range of equivalent elements of the application documents are included in the present invention.

[0193] The above description is only a specific embodiment of the present invention, so that those skilled in the art can understand or implement the present invention. Various modifications to these embodiments will be apparent to those 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 should conform to the widest scope consistent with the principles and novel features invented herein.

Claims

1. A method for picking first arrivals of seismic data based on SimpleNet network, characterized in that: The following steps are involved: Step S1: obtaining seismic shot data and performing grayscale image conversion on the seismic shot data, thereby obtaining an enhanced grayscale image set in a function set; Step S1 is specifically as follows: Step S11: acquiring seismic shot gather data, and performing data image mapping on the seismic shot gather data, thereby obtaining a seismic shot gather grayscale image; Step S12: performing significant marking of the first arrival function on the seismic cannon grayscale image, thereby obtaining a significant grayscale image; Step S13: performing first arrival function distribution label clipping on the significant grayscale image to obtain a concentrated grayscale image; Step S14: performing data enhancement according to the concentrated grayscale image to obtain a concentrated grayscale enhanced image; Step S2: divide the data set according to the enhanced grayscale image set in the function set to obtain a training image set and a test image set; perform low-speed body static correction on the training image set to obtain a corrected training image set; Step S2 is specifically as follows: Step S21: dividing the data set according to the enhanced grayscale image set in the function set, so as to obtain a training image set and a test image set; Step S22: drawing a wave velocity profile distribution map for the training image set, thereby obtaining a wave velocity profile distribution map; Step S23: performing wave velocity statistics on the wave velocity profile distribution diagram to obtain low wave velocity data; Step S24: performing regional positioning on the wave velocity profile distribution map according to the low wave velocity data, thereby obtaining low wave velocity regional data; Step S25: constructing a low wave velocity identification model according to the low wave velocity area data, thereby obtaining a low wave velocity identification model; Step S25 is specifically as follows: Step S251: extracting regional shape features, gradient features and variation features from the low wave velocity region data, thereby obtaining regional shape data, gradient data and variation data; Step S252: Calculating geometric boundary smoothness of the region shape data, thereby obtaining geometric boundary smoothness data; Step S253: performing gradient synthesis according to the gradient data and the change data, thereby obtaining gradient synthesis data; Step S254: integrating geometric gradient data according to the geometric boundary smoothness data and the gradient synthesis data, thereby obtaining geometric gradient data; Step S255: constructing a low wave velocity identification model according to the geometric gradient data, thereby obtaining a low wave velocity identification model; Step S26: performing low wave speed recognition on the training image set according to the low wave speed recognition model, thereby obtaining a low wave speed training image set; Step S27: performing low wave velocity correction on the training image set according to the low wave velocity training image set, thereby obtaining a corrected training image set; Step S3: constructing an earthquake first arrival prediction model according to the enhanced grayscale image set in the function set, thereby obtaining an earthquake first arrival prediction model; performing sample training on the earthquake first arrival prediction model according to the corrected training image set, thereby obtaining an earthquake first arrival training model; optimizing model parameters of the earthquake first arrival training model, thereby obtaining an optimized earthquake first arrival training model; Step S4: performing earthquake first arrival prediction on the test image set according to the optimized earthquake first arrival training model, thereby obtaining earthquake first arrival prediction data; The earthquake first arrival prediction data is classified by image size to obtain fixed-size image prediction data and arbitrary-size image prediction data; the prediction result evaluation data is evaluated on the fixed-size image prediction data and arbitrary-size image prediction data to obtain prediction result evaluation data, and uploaded to the seismic cannon platform processing system to perform the earthquake first arrival prediction task.

2. The method for picking first arrivals of seismic data based on SimpleNet network according to claim 1, characterized in that: Step S14 is specifically as follows: Step S141: extracting grayscale image samples according to the concentrated grayscale image, thereby obtaining concentrated grayscale image sample data; Step S142: performing a horizontal mirror transformation on the concentrated grayscale image sample data, thereby obtaining concentrated grayscale mirror data; Step S143: replacing the image data of the centralized grayscale image according to the centralized grayscale mirror data, thereby obtaining centralized grayscale image replacement data; Step S144: performing image enhancement integration on the concentrated grayscale image replacement data and the concentrated grayscale image to obtain a concentrated grayscale enhanced image.

3. The method for picking first arrivals of seismic data based on SimpleNet network according to claim 1, characterized in that: Step S252 is specifically as follows: Extracting boundary point coordinate data from the regional shape data, thereby obtaining boundary point coordinate data; Performing curve parameterization processing on the boundary point coordinate data, thereby obtaining curve parameterization data; Calculate the curvature of the boundary point according to the curve parameterization data, so as to obtain the curvature data of the boundary point; Calculate the average curvature of the boundary point curvature data to obtain the average curvature data; The smoothness index is calculated based on the mean curvature data to obtain the geometric boundary smoothness data.

4. The method for picking first arrivals of seismic data based on SimpleNet network according to claim 1, characterized in that: Step S253 is specifically as follows: Performing gradient amplitude calculation on the gradient data to obtain gradient amplitude data; Calculating the change rate of the change data to obtain the change rate data; Spatially aligning the change rate data according to the gradient amplitude data, thereby obtaining change space alignment data; The gradient synthesis data is obtained by weighted synthesis based on the change space alignment data and the gradient amplitude data.

5. The method for picking first arrivals of seismic data based on SimpleNet network according to claim 1, characterized in that: Step S3 is specifically as follows: Step S31: constructing an earthquake first arrival prediction model according to the enhanced grayscale image set in the function set, thereby obtaining an earthquake first arrival prediction model; Step S32: extracting earthquake first arrival time data and earthquake first arrival position data from the correction training image set, thereby obtaining earthquake first arrival time data and earthquake first arrival position data; Step S33: Calculate the probability of the first arrival time of the earthquake according to the first arrival time data, so as to obtain the probability data of the first arrival time of the earthquake; Step S34: performing first-arrival range statistics according to the earthquake first-arrival position data, thereby obtaining earthquake first-arrival range data; Step S35: integrating earthquake first arrival data according to earthquake first arrival time probability data and earthquake first arrival range data, thereby obtaining earthquake first arrival data; Step S36: performing sample training on the earthquake first arrival prediction model according to the earthquake first arrival data, thereby obtaining an earthquake first arrival training model; Step S37: Optimize the model parameters of the earthquake first-arrival training model according to the corrected training image set, so as to obtain an optimized earthquake first-arrival training model.

6. The method for picking first arrivals of seismic data based on SimpleNet network according to claim 1, characterized in that: Step S4 is specifically as follows: Step S41: performing earthquake first-arrival prediction on the test image set according to the optimized earthquake first-arrival training model, thereby obtaining earthquake first-arrival prediction data; Step S42: classifying the earthquake first arrival prediction data by image size, thereby obtaining fixed-size image prediction data and arbitrary-size image prediction data; Step S43: calculating the fixed-size image prediction accuracy rate of the fixed-size image prediction data according to the corrected training image set, thereby obtaining the fixed-size image prediction accuracy rate data; Step S44: calculating the prediction accuracy of the image of any size on the prediction data of the image of any size according to the corrected training image set, thereby obtaining the prediction accuracy data of the image of any size; Step S45: integrating the accuracy data according to the prediction accuracy data of the fixed-size image and the prediction accuracy data of the arbitrary-size image, thereby obtaining the earthquake first arrival prediction accuracy data; Step S46: Evaluate the prediction results of the earthquake first arrival prediction accuracy data to obtain prediction result evaluation data, and upload it to the seismic cannon platform processing system to perform the earthquake first arrival prediction task.

7. The method for picking first arrivals of seismic data based on SimpleNet network according to claim 6, characterized in that: Step S43 is specifically as follows: Step S431: extracting fixed-size earthquake first-arrival prediction data from the fixed-size image prediction data, thereby obtaining fixed-size earthquake first-arrival prediction data; Step S432: extracting seismic first-arrival correction data from the correction training image set, thereby obtaining seismic first-arrival correction data; Step S433: performing earthquake first-arrival annotation comparison on the fixed-size earthquake first-arrival prediction data according to the earthquake first-arrival correction data, thereby obtaining fixed-size image earthquake first-arrival annotation comparison data; Step S434: Calculate the accuracy of the fixed-size image seismic first-arrival annotation comparison data based on the corrected training image set, so as to obtain the fixed-size image prediction accuracy data.

Images