查看更多>>摘要：Microscopic rock thin section image recognition is crucial in rock mineral analysis. Typically, deep learning models are used to automate expert knowledge, but the scarcity of samples in certain categories limits the available training data, affecting the performance of traditional deep learning models. This paper proposes a novel few-shot learning model to address the challenge of classifying rock thin section images under limited sample conditions. Based on advanced few-shot learning processes involving pre-training and meta-training, we first introduce a Cross Attention Feature Fusion (CAFF) module. This module generates new features by combining plane polarized light images (PPL) and cross-polarized light images (XPL) of rock thin sections under a microscope, integrating these with the original features through autonomous learning to obtain more comprehensive features. Secondly, we propose a Feature Selection (FS) module based on the prototypical network (ProtoNet). This module enhances the model's classification capability by extracting key feature dimensions from two perspectives: intra-class representativeness and inter-class distinctiveness. Finally, using the pre-trained ResNet50 and Swim-Transformer on ImageNet-1000k as the backbone network, simulation experiments were conducted on the Nanjing University Rock Thin Section Teaching Dataset. Under the 5-Way 5-Shot few-shot learning task standard, the proposed ProtoNet-CAFF-FS model achieved an average classification accuracy of 96.70% and 99.16%, outperforming traditional modeling methods and demonstrating the effectiveness of the newly added modules.
查看更多>>摘要：Persistent Scatterer Interferometry Synthetic Aperture Radar (PS-InSAR) provides high-precision ground deformation measurements over wide areas. However, analyzing PS time series remains challenging due to complex temporal patterns and the need to consider comprehensive displacement fields to fully characterize ground deformation processes. This study evaluates and compares unsupervised clustering approaches for PS time series analysis, contrasting feature extraction techniques against raw time series methods. We developed an online optimization algorithm for cluster number determination and introduced a custom density-based score (MLRD) for evaluating clustering quality in sparse geospatial datasets. The approaches were tested on Sentinel-1-derived PS data from the landslide-prone Offida municipality (Marche region, Italy), where feature-based methodologies demonstrated superior performance, achieving improvements of one to two orders of magnitude in clustering quality metrics compared to conventional approaches. The multivariate analysis notably outperformed univariate methods, with optimal MLRD (2.59 & sdot; 10-5) and Calinski-Harabasz scores (194.73) at 50% explained variance, while preserving the physical interpretability of the results. This comprehensive analysis identified coherent deformation clusters extending beyond previously mapped landslide boundaries, demonstrating the effectiveness of multivariate clustering in detecting potentially unstable areas. This methodological framework advances PS time series analysis through robust pattern recognition while enhancing geohazard assessment capabilities, offering a robust foundation for identifying unstable areas and providing quantitative support for improving our understanding of complex ground deformation mechanisms.
查看更多>>摘要：The longest flow path is one of the key features of a catchment, commonly considered in hydrological analysis and modeling. Recent literature highlights that identifying the longest flow paths using existing software tools is time-consuming. Over the last few years, attempts have been made to develop more computationally efficient algorithms for this particular task. This paper extends previously published research and presents a new GPU algorithm designed for fast identification of the longest flow paths using DEM-derived flow directions. Performance measurements show significantly shorter execution times compared to other existing algorithms for the same task. Additionally, this algorithm is able to efficiently process multiple catchments in the same run, offering further performance improvements.
查看更多>>摘要：Uncertainty quantification is a critical component in the interpretation of spatial phenomena, particularly within the geosciences, where incomplete subsurface data leads to various possible scenarios, making it crucial for risk assessment and decision-making. Traditional geostatistical methods have served as the cornerstone for uncertainty analysis; however, the incorporation of machine learning, particularly ensemble methods, offers a compelling augmentation, especially in handling complex and noisy datasets. Building on our previous work, which introduced a spatial bagging technique for enhancing prediction accuracy, this study extends the method to uncertainty quantification by applying a widely-used UQ metric from geostatistics. Our approach employs a bootstrap method adjusted for effective sample size derived from spatial statistics, addressing the common issue of overfitting when dealing with dependent data. We demonstrate, through a series of synthetic datasets with varied noise levels and spatial structures, that our spatial bagging method not only outperforms standard bagging techniques in prediction accuracy but also provides superior uncertainty quantification. The robustness of the method against noise and its computational efficiency, particularly in spatially correlated data, positions it as a promising tool for geoscientists and others who require reliable uncertainty measures in spatial analysis.
查看更多>>摘要：Smooth anisotropic media are often met when implementing effective medium theory, full waveform inversion or seismic imaging. However, computational overburden is often a recurring problem when working with high frequencies or when quantifying uncertainties. In this context, adaptive meshes constitute, in principle, an attractive representation to maximize simulation accuracy while minimizing the computational cost. However, such meshes are difficult to create in the context of smooth anisotropic media as the optimal local size of the elements is not clearly defined. In this work, we present a two-step algorithm to efficiently mesh these media for spectral element method (SEM) simulation in the 2D elastic case. Our algorithm yields quadrangular only meshes which adapt the size of the element to the local and directional S-wave velocity. It relies on a quadtree division introduced by Mar & eacute;chal (2009) to divide the mesh until the size of each element edge is adapted to the local minimum wavelength that will be propagated. Then, a Laplacian smoothing is applied to further optimize the size of the elements, increasing the global time step and makes the SEM simulation faster while keeping a good accuracy and even improving it in some cases. An application of our method on a 2D section of the homogenized Groningen area shows that simulation time can be reduced by a factor up to 7.
NETL
NSTL
Elsevier