Artificial intelligence for geoscience: Progress, challenges and perspectives

The Innovation, Год журнала: 2024, Номер 5(5), С. 100691 - 100691

Опубликована: Авг. 23, 2024

Public summary•What does AI bring to geoscience? has been accelerating and deepening our understanding of Earth Systems in an unprecedented way, including the atmosphere, lithosphere, hydrosphere, cryosphere, biosphere, anthroposphere interactions between spheres.•What are noteworthy challenges As we embrace huge potential geoscience, several arise reliability interpretability, ethical issues, data security, high demand cost.•What is future The synergy traditional principles modern AI-driven techniques holds immense promise will shape trajectory geoscience upcoming years.AbstractThis paper explores evolution geoscientific inquiry, tracing progression from physics-based models data-driven approaches facilitated by significant advancements artificial intelligence (AI) collection techniques. Traditional models, which grounded physical numerical frameworks, provide robust explanations explicitly reconstructing underlying processes. However, their limitations comprehensively capturing Earth's complexities uncertainties pose optimization real-world applicability. In contrast, contemporary particularly those utilizing machine learning (ML) deep (DL), leverage extensive glean insights without requiring exhaustive theoretical knowledge. ML have shown addressing science-related questions. Nevertheless, such as scarcity, computational demands, privacy concerns, "black-box" nature hinder seamless integration into geoscience. methodologies hybrid presents alternative paradigm. These incorporate domain knowledge guide methodologies, demonstrate enhanced efficiency performance with reduced training requirements. This review provides a comprehensive overview research paradigms, emphasizing untapped opportunities at intersection advanced It examines major showcases advances large-scale discusses prospects that landscape outlines dynamic field ripe possibilities, poised unlock new understandings further advance exploration.Graphical abstract

Язык: Английский

---

Coarse to fine-based image–point cloud fusion network for 3D object detection

Information Fusion, Год журнала: 2024, Номер 112, С. 102551 - 102551

Опубликована: Июль 2, 2024

Язык: Английский

---

Oil Spill Drift Prediction Enhanced by Correcting Numerically Forecasted Sea Surface Dynamic Fields with Adversarial Temporal Convolutional Networks

IEEE Transactions on Geoscience and Remote Sensing, Год журнала: 2025, Номер 63, С. 1 - 18

Опубликована: Янв. 1, 2025

Timely and accurate representation of sea surface dynamic fields is crucial for oil spill drift prediction. Numerically forecasted are available in a timely manner, but their accuracy limited. Conversely, reanalysis offer superior suffer from time delays. To enhance the performance prediction, we propose deep learning-based approach to correcting numerically fields, aligning them more closely with fields. Our introduces an adversarial temporal convolutional network (ATCN) framework, consisting (TCN)-based corrector discriminator. The TCN can characterize field sequences both spatially temporally. In this scenario, processes outputs corrected that approximate Adversarial training discriminator further refines corrector. This enhances prediction using We also provide dataset drifts Symphony Sanchi accidents, including related data remote sensing data, establishing baseline evaluating Experiments on validate ATCN framework's effectiveness enhancing

Язык: Английский

---

ISPDiff: Interpretable Scale-Propelled Diffusion Model for Hyperspectral Image Super-Resolution

IEEE Transactions on Geoscience and Remote Sensing, Год журнала: 2024, Номер 62, С. 1 - 14

Опубликована: Янв. 1, 2024

Hyperspectral image (HSI) super-resolution (SR) employing the denoising diffusion probabilistic model (DDPM) holds significant promise with its remarkable performance. However, existing relevant works exhibit two limitations: i) Directly applying DDPM to fusion-based HSI SR (HSI-SR) ignores physical mechanism of HSI-SR and unique characteristics HSI, resulting in less interpretability; ii) Scale-invariant suffers from a time-consuming inference. To tackle these issues, we propose an interpretable scale-propelled (ISPDiff) for HSI-SR, which combines underlying principles progressively unrolling reconstruction by learning distribution at various scales, enhancing transparency significantly reducing inference time prominently. Concretely, destroy downsample into Gaussian noise forward process ISPDiff. Then design unified scale-flexible backward iteratively refine coarse-to-fine manner through scale-matched cross-scale upsampling, can be unfolded optimization algorithms. These solved equations are one-to-one corresponding unrolled deep neural networks, called progressive perceptual model-driven restoration network (P ² MSRN) upsampling (CMUN). Through end-to-end training, proposed ISPDiff implements characterized enhanced interpretability, stronger task orientation, reduced consumption. Systematic experiments have been conducted on three public datasets, demonstrating that outperforms state-of-the-art methods. Code is available https://github.com/Jiahuiqu/ISPDiff.

Язык: Английский

---

RustQNet: Multimodal deep learning for quantitative inversion of wheat stripe rust disease index

Computers and Electronics in Agriculture, Год журнала: 2024, Номер 225, С. 109245 - 109245

Опубликована: Июль 26, 2024

Язык: Английский

---

Iterative Saliency Aggregation and Assignment Network for Efficient Salient Object Detection in Optical Remote Sensing Images

IEEE Transactions on Geoscience and Remote Sensing, Год журнала: 2024, Номер 62, С. 1 - 13

Опубликована: Янв. 1, 2024

Язык: Английский

---

Advancing soybean biomass estimation through multi-source UAV data fusion and machine learning algorithms

Smart Agricultural Technology, Год журнала: 2025, Номер 10, С. 100778 - 100778

Опубликована: Янв. 7, 2025

Язык: Английский

---

Revolutionizing hyperspectral image classification for limited labeled data: unifying autoencoder-enhanced GANs with convolutional neural networks and zero-shot learning

Earth Science Informatics, Год журнала: 2025, Номер 18(2)

Опубликована: Янв. 30, 2025

Язык: Английский

---

How to Learn More? Exploring Kolmogorov–Arnold Networks for Hyperspectral Image Classification

Remote Sensing, Год журнала: 2024, Номер 16(21), С. 4015 - 4015

Опубликована: Окт. 29, 2024

Convolutional neural networks (CNNs) and vision transformers (ViTs) have shown excellent capability in complex hyperspectral image (HSI) classification. However, these models require a significant number of training data are computational resources. On the other hand, modern Multi-Layer Perceptrons (MLPs) demonstrated great classification capability. These MLP-based significantly less compared with CNNs ViTs, achieving state-of-the-art accuracy. Recently, Kolmogorov–Arnold (KANs) were proposed as viable alternatives for MLPs. Because their internal similarity to splines external MLPs, KANs able optimize learned features remarkable accuracy, addition being learn new features. Thus, this study, we assessed effectiveness HSI Moreover, enhance accuracy obtained by KANs, developed hybrid architecture utilizing 1D, 2D, 3D KANs. To demonstrate KAN architecture, conducted extensive experiments on three newly created benchmark datasets: QUH-Pingan, QUH-Tangdaowan, QUH-Qingyun. The results underscored competitive or better KAN-based model across datasets over several CNN- ViT-based algorithms, including 1D-CNN, 2DCNN, CNN, VGG-16, ResNet-50, EfficientNet, RNN, ViT.

Язык: Английский

---

A Spectral Diffusion Prior for Unsupervised Hyperspectral Image Super-Resolution

IEEE Transactions on Geoscience and Remote Sensing, Год журнала: 2024, Номер 62, С. 1 - 13

Опубликована: Янв. 1, 2024

Fusion-based hyperspectral image (HSI) superresolution aims to produce a high-spatial-resolution HSI by fusing low-spatial-resolution and multispectral image.Such super-resolution process can be modeled as an inverse problem, where the prior knowledge is essential for obtaining desired solution.Motivated success of diffusion models, we propose novel spectral fusion-based super-resolution.Specifically, first investigate spectrum generation problem design model data distribution.Then, in framework maximum posteriori, keep transition information between every two neighboring states during reverse generative process, thereby embed trained into fusion form regularization term.At last, treat each step final optimization its subproblem, employ Adam solve these subproblems sequence.Experimental results conducted on both synthetic real datasets demonstrate effectiveness proposed approach.The code approach will available https://github.com/liuofficial/SDP.

Язык: Английский

---