Search Results (433)

Accurate crop mapping is important for agricultural monitoring and land management; yet, identifying robust and compact feature subsets from high-dimensional multi-sensor remote sensing data remains challenging, particularly in heterogeneous agricultural landscapes affected by spatial autocorrelation. Although combining multi-sensor data provides complementary spectral and structural information, traditional workflows often neglect spatial dependence during feature evaluation, leading to over-optimistic validation metrics and spatially unstable feature subsets. To address this issue, this study proposes a hierarchical feature selection and subset optimization framework for crop mapping by integrating Sentinel-1 Synthetic Aperture Radar (SAR) and Sentinel-2 optical imagery within the Google Earth Engine (GEE) platform. A total of 135 multi-sensor features were constructed, including spectral bands, vegetation indices, SAR metrics, texture descriptors, and phenological statistics. To improve feature compactness and spatial robustness, a multi-stage selection strategy combining correlation-based redundancy removal, spatial cross-validation (SCV) control, Boruta, recursive feature elimination (RFE), L1 regularization, SHapley Additive exPlanations (SHAP), and Non-dominated Sorting Genetic Algorithm II (NSGA-II) was developed. Results showed that temporal and phenological features contributed more strongly to crop discrimination than static spectral or SAR features, while multi-sensor integration further improved classification stability. Notably, the proposed framework reduced the feature space from 135 to 12 variables while slightly improving classification performance. The final optimized model achieved an overall accuracy (OA) of 96.98% under SCV and generated spatially consistent crop maps at 10 m resolution. The framework provides an efficient and scalable solution for fine-scale crop mapping in complex agricultural regions and demonstrates the practical potential of incorporating spatial dependence control into feature selection for large-scale agricultural monitoring applications. Full article

Decoding mental states from non-invasive neural recordings is central to brain-computer interface research. Multimodal acquisition that combines electroencephalography (EEG) and functional near-infrared spectroscopy (fNIRS) couples the high temporal resolution of EEG with the spatial specificity of fNIRS, compensating for the individual limitations of each modality. While such hybrid systems achieve strong intra-subject performance, cross-subject generalization remains constrained by inter-individual variability in neural responses. This study introduces the Eigenvalue-Guided Positional Encoding Classification Transformer (EPECT), an architecture that integrates eigenvalue-aware multi-head self-attention with sinusoidal positional encoding to capture both the spectral structure of the learned feature representations and the temporal ordering of multimodal sequences. Stacked one-dimensional convolutions extract local patterns prior to transformer encoding, and global average pooling aggregates the final representation for classification. EPECT was evaluated on two publicly available EEG-fNIRS datasets covering motor imagery (MI), n-back, discrimination/selection response (DSR), and word generation (WG) paradigms under a cross-subject protocol. The model achieved classification accuracies of 97.3%, 96.3%, 98.1%, and 97.9% on the MI, n-back, DSR, and WG tasks, respectively. Ablation studies quantified the contribution of each architectural component, and integrated gradients analysis revealed structured modality-specific attribution patterns aligned with task-relevant cortical regions. Additional experiments with synthetic cortical perturbations demonstrate the sensitivity of EPECT to subtle activity changes, indicating potential utility for tracking neurorehabilitation outcomes in future clinical applications. Full article

Glacier boundary extraction on the Tibetan Plateau (TP) faces persistent challenges due to rugged terrain, seasonal snow, extensive debris cover, and topographic shadows. Traditional methods utilizing single-source or single-temporal data often yield limited accuracy. Thus, we propose an automated Double Random Forest (Double-RF) framework integrating single- and multi-temporal features from Sentinel-1 (SAR) and Sentinel-2 (Optical) data within the Google Earth Engine. We established a multidimensional feature space comprising spectral, textural, polarimetric, and topographic attributes. Feature optimization was performed using importance metrics and out-of-bag (OOB) error. A hierarchical classification strategy was employed: the first RF identifies clean glaciers and glaciers in shadow, while the second RF executes refined boundary extraction of debris-covered glaciers to mitigate spectral confusion. The results indicate that the Double-RF method significantly achieves an overall accuracy exceeding 0.84 across all sub-basins and reaching above 0.95 at best. The derived glacier inventory reveals a distinct spatial pattern: higher concentrations in the western and peripheral regions compared to the eastern and interior TP. Glaciers are predominantly distributed on shaded aspects with gentle-to-moderate slopes, highlighting the combined influence of climatic gradients and topographic controls. This multi-source, multi-temporal fusion strategy provides a robust methodological foundation for long-term glacier monitoring over the TP. Full article

Agricultural production in arid regions is strongly constrained by water stress, making timely evaluation of crop water conditions increasingly important. However, conventional measurements of plant moisture content (PMC) primarily rely on destructive oven-drying methods, which are not only labor-intensive and time-consuming but also constrained by limited sample size and spatial coverage. These shortcomings make it difficult to capture the spatial heterogeneity of crop water status across large agricultural regions, thereby restricting regional-scale water diagnosis and precision irrigation decision-making. Focusing on silage maize cultivated in the arid region of Gansu Province, China, this work develops a regional PMC estimation approach by combining multi-source remote sensing data. High-resolution unmanned aerial vehicle (UAV) observations were integrated with Sentinel-2 and Sentinel-3 imagery, while radiometric and temperature corrections were applied to improve data consistency. A set of spectral, textural, and thermal features was derived from multispectral, visible, and thermal infrared datasets. Feature selection based on Pearson correlation was then carried out, followed by the construction of three models, namely Random Forest (RF), Support Vector Machine (SVM), and Partial Least Squares Regression (PLSR). Among them, the RF model performed more reliably, achieving a validation R² of 0.92 with relatively low prediction error. In addition, calibration using UAV data led to a clear improvement in satellite-based estimates, with R² increasing from 0.52–0.62 to 0.71–0.74. The generated PMC maps captured both the temporal decline during the growing season and the spatial variability across the study area. Overall, the proposed approach offers a practical option for large-scale monitoring of crop water status and can support irrigation management in water-limited environments. Full article

To track land use and land cover transformation in Makkah, techniques that allow steep relief, spectral confusion, and dense sacred–commercial mosaics, and can be justified in terms of planning, should be used. Satellite images are tedious and prone to uneven labeling on mixed-pixel boundaries, particularly in urban regions and Haram borders. Using multi-temporal Landsat-8 data (2013 and 2024), a hybrid deep learning architecture comprising U-Net, DenseNet201, and a Vision Transformer was trained. U-Net retained the geometry of the boundaries, DenseNet201 reinforced feature transfer across heterogeneous textures, and the transformer modeled long-range context. Gradient-weighted Class Activation Mapping (Grad-CAM) was used to incorporate interpretability during spatial attention mapping, and Shapley Additive exPlanations (SHAP) during spectral topographic attribution, after which paired class-level statistical tests were performed. Modern residential increased from 15% to 20% (180 million to 240 million m²); roads from 5% to 10% (60 million to 120 million m²); industrial facilities from 3% to 5% (36 million to 60 million m²). The vegetation expanded by 1 to 5% (an addition of 48 million m²), and agriculture declined by 2 to 1% (a loss of 12 million m²). Its tension with urban development and preservation of productive land was growing. The proposed U-Net–DenseNet201–ViT hybrid system achieved over 98% overall accuracy on the test data for both study years, with kappa coefficients of 0.978 and 0.981 for 2013 and 2024, respectively. Grad-CAM identified attention focused on development fronts and transport corridors, whereas SHAP identified SWIR, thermal response, and slope as the main drivers. Significant class-level gains were statistically validated (p < 0.01), confirming an interpretable and auditable account of land transformation in Makkah. Full article

Gas pipeline leakages in urban underground utility tunnels pose a severe threat to public safety. Leakages of varying aperture sizes trigger differentiated risks of diffusion and explosion; thus, achieving precise identification of leakage hole size has become a critical issue in safety management. To address the difficulty of traditional methods in effectively separating the acoustic features of different leakage levels within complex utility tunnel environments, this paper proposes a gas pipeline leakage risk level identification method based on a lightweight Spatial–Temporal Graph Neural Network (ST-GNN). First, relying on a real utility tunnel simulation platform, acoustic signals under different pressures and leakage hole size are collected, and time-frequency magnitude features are constructed through Short-Time Fourier Transform (STFT). Furthermore, each acoustic sample is independently converted into a graph with STFT time frames as nodes, where temporal neighborhood edges and K-nearest neighbor edges jointly encode local dynamics and non-local spectral similarities. This transforms unstructured acoustic signals into graph-structured data that embodies spatial–temporal coupling relationships. Building upon this, a lightweight Chebyshev graph convolutional network is designed to progressively extract discriminative features strongly correlated with leakage levels using multi-layer convolution. Experimental results on the actual utility tunnel simulation platform dataset demonstrate that the proposed method achieves excellent performance in a three-level leakage classification task. The t-SNE visualization reveals the effective separation of features, progressing from complete mixing in the input layer to distinct separation in the output layer. Through multiple training statistics and ablation experiments, the impact of dataset size and the number of network layers on the identification performance is analyzed, validating the robustness of the proposed model under limited samples and the effectiveness of its lightweight structure. This provides a feasible solution for the automated and refined identification of gas pipeline leakage levels in underground utility tunnels. Full article

To address yield uncertainties from contrasting hydrothermal events in black soil regions, this study developed a universal estimation framework integrating multi-dimensional features. The universal yield estimation framework leveraged data from contrasting flood (2024) and drought (2025) scenarios in Youyi Farm in the Northeast Black Soil Region. And we fused multi-dimensional environmental features, including remote sensing, soil, and micro-topography factors, to identify “Regime Shifts” in yield-driving mechanisms across contrasting years. We evaluated four ML algorithms (RF, XGBoost, MLP, and TabNet) using Recursive Feature Elimination with Cross-Validation (RFECV) for variable optimization. Results showed the following: (1) The Universal RF model achieved superior robustness (R² = 0.80), overcoming inter-annual fluctuations. (2) Mechanistic analysis identified a “Regime Shift” in yield drivers, transitioning from micro-topography-governed “drainage limitation” during flooding to soil-texture-dominant (SAND) “linear limitation” during drought. (3) Dynamic growth-stage differential features successfully corrected asymmetric spectral responses, resolving slope inversion and overestimation driven by “non-productive greenness” during flooding. (4) Spatio-temporal yield mapping revealed a transition from topography-constrained linear distributions (2024) to soil-moisture-driven “patchy mosaic” structures (2025). Moran’s I increased from 0.21 to 0.45, reflecting intensified yield clustering and intensified spatial clustering under drought. This study provides a robust tool for food security monitoring and site-specific management in climate-vulnerable intensive agricultural zones. Full article

Accurate and up-to-date mapping of built-up areas is of great importance for sustainable urban planning, disaster management, and the monitoring of environmental changes. In this study, a residual U-Net-based deep learning architecture named FiveBandTTA is proposed for built-up area segmentation from Sentinel-2 multispectral satellite imagery. The proposed model aims to simultaneously learn spatial and spectral features by jointly processing RGB, NIR (B8), and SWIR (B11) bands within the same encoder–decoder structure. The model incorporates standard residual blocks following the conventional residual learning principle, multi-level skip connection mechanisms, and TTA-based inference strategies. Within the scope of the study, a multi-temporal built-up area dataset was constructed from Sentinel-2 imagery acquired over Kocaeli Province. The performance of the proposed model was comparatively evaluated against RGB Baseline, FiveBand Single, DeepLabV3+, and SegFormer models. Experimental results demonstrated that the proposed model achieved the highest segmentation performance among all compared approaches, obtaining 0.8447 IoU, 0.9124 Dice, and 0.9249 Precision scores. It was observed that the use of multispectral bands together with the residual encoder–decoder structure may contribute to improved representation of small-scale built-up regions and complex boundary structures. Furthermore, the comparative experiments indicated that the NIR and SWIR bands provide complementary spectral information for distinguishing built-up areas, while the TTA-based inference strategy may contribute to improved segmentation stability and prediction consistency. Overall, the obtained results demonstrate that the proposed approach is an effective and robust method for built-up area segmentation from medium-resolution Sentinel-2 imagery. Full article

Suspended sediment concentration (SSC) is a key indicator of river sediment transport processes and water environmental change. For medium-width rivers, continuous-reach SSC monitoring remains constrained by the spatial discontinuity of station observations and the temporal or consistency limitations of single-source satellite imagery. To improve multi-year SSC characterization in the middle and lower reaches of the Liaohe River, this study integrated Harmonized Landsat and Sentinel-2 (HLS) surface reflectance imagery from 2016 to 2022 with SSC observations from five hydrological stations and developed a random forest retrieval model using multi-band reflectance and sediment-related spectral features. The trained model was applied to valid HLS images to examine SSC spatial distribution, interannual variation, and inter-station reach differences. The model achieved a test-set R² of 0.641, an RMSE of 0.083 kg·m⁻³, and an MAE of 0.067 kg·m⁻³. The median composite of 52 retrieval images showed a lower SSC in the Tieling–Mahushan and Mahushan–Pinganbao reaches and a higher SSC in the Pinganbao–Liaozhong and Liaozhong–Liujianfang reaches. SSC was generally higher in 2016 and 2022 and lower in 2018. These findings indicate that HLS-based retrieval can support continuous-reach SSC monitoring and regional water–sediment dynamic assessment in medium-width rivers, although the accurate quantification of extreme high-SSC events still requires additional in situ samples and higher-frequency observations. Full article

Tropical rainforests are characterized by high species diversity and complex canopy structure, making accurate forest stand classification important for ecosystem assessment, biodiversity monitoring, and forest carbon estimation. However, single-source remote sensing data lacks sufficient discrimination ability to address the issue of spectral similarity among classes, and conventional convolutional neural networks often struggle to extract discriminative features and integrate heterogeneous data in highly complex forests. To address these challenges, this study developed a dual-branch deep learning framework that integrates DenseNet and ConvNeXt for classification in Hainan Tropical Rainforest National Park. The framework combines sub-meter Google Earth imagery to capture spatial–textural detail with multi-temporal Sentinel-2 imagery to represent phenological variation. The results showed that multi-temporal Sentinel-2 data outperformed single-date imagery by capturing phenological patterns, and that the fusion of high-resolution spatial information and multi-temporal spectral information yielded higher accuracy than either data source alone. The dual-branch model achieved an overall accuracy of 94.47% and a Kappa coefficient of 0.94, outperforming all benchmark models. These findings indicate that branch-specific feature extraction and adaptive fusion can improve fine-scale classification in complex tropical rainforest environments. The proposed framework provides a practical approach for fine-scale forest stand mapping and may support biodiversity monitoring, ecological assessment, and sustainable forest management. Full article

Rapid identification of maize waterlogging is essential for post-disaster agricultural assessment, but most existing methods rely on multi-temporal imagery that is often unavailable immediately after extreme rainfall events. This study proposes SAB-DeepLabV3+, a semantic segmentation model for mapping waterlogging-affected maize from single-date multispectral imagery within pre-extracted maize planting areas. Built on DeepLabV3+, the model integrates three task-specific modules: a Spectral-Spatial Information Enhancement Module to improve feature discrimination under spectral mixing, an Adaptive Multi-Scale Pooling Module to capture heterogeneous patch sizes, and a Boundary Enhancement Module to refine transition zones. A pixel-level dataset containing 12,198 image patches was constructed from 62 multispectral scenes collected across five major maize-producing cities in Heilongjiang Province, China, during 2022–2024. On the test set, SAB-DeepLabV3+ achieved a waterlogged-class IoU of 68.30%, mIoU of 80.37%, mF1 of 88.62%, and OA of 93.49%, outperforming DeepLabV3+. Leave-one-city-out evaluation further produced an average mIoU of 76.56% and a waterlogged-class IoU of 63.45%. These results indicate that single-date high-resolution multispectral imagery can support rapid and reliable maize waterlogging mapping. Full article

Emotion recognition through electroencephalogram (EEG) signals is crucial for brain–computer interfaces (BCIs), yet existing methods often struggle with heterogeneous feature fusion and capturing long-range temporal dependencies. To address these challenges, we propose MAF-TransNet, a novel unified spatiotemporal framework. Specifically, parallel Fully Connected Neural Network (FCNN) modules first non-linearly align heterogeneous differential entropy (DE) and power spectral density (PSD) features. Subsequently, an Adaptive Channel-wise Feature Encoder (ACFE) recalibrates spatial–spectral responses to highlight emotion-relevant cortical activations. Finally, a Transformer encoder dynamically models the global temporal evolution of emotional states. Evaluated on the SEED-IV and DEAP datasets, MAF-TransNet achieves superior subject-dependent (SD) accuracies of 88.80% and 96.58%, respectively, alongside robust subject-independent (SI) performance. Furthermore, Granger causality analysis reveals distinct emotion-dependent prefrontal asymmetry, while t-SNE visualizations confirm the formation of a highly discriminative, linearly separable feature manifold. Ultimately, MAF-TransNet effectively unifies local spatial–spectral extraction with global temporal modeling, providing an accurate and robust approach, while offering preliminary insights into the spatiotemporal dynamics of emotion for future affective BCI applications. Full article

Recently, artificial neural networks (ANNs) have shown impressive performance in the compressive hyperspectral imaging (CHI) reconstruction task, but the high energy consumption limits their deployment on energy-constrained devices. This paper develops a novel spiking neural network (SNN), termed spiking spectral-weighting reconstruction network (SSWR-Net), to significantly improve the energy–efficiency ratio in CHI reconstruction. Firstly, a spiking spectral-weighting convolution block is proposed to adaptively modulate the spiking signals, enabling the SNN to fit continuous spectral correlation curves. Secondly, a residual feature reuse module with more direct connections is designed to achieve efficient and lightweight spatial–spectral feature extraction. Thirdly, customized feature scaling architectures are introduced to resolve the dimensional mismatch issue and enhance information flow. Finally, we propose a novel temporal-wise progressive training method to optimize the multi-timestep SSWR-Net, which can significantly improve both training efficiency and reconstruction quality. Both simulation and real experiments demonstrate the superiority of the proposed method in both CHI reconstruction performance and energy efficiency. Specifically, SSWR-Net outperforms its ANN-based counterpart by 0.87 dB at a 19.74% energy cost. Full article

Motor imagery (MI)-based brain–computer interfaces (BCIs) have led to great interest as a result of their potential use in neurorehabilitation, assistive robotics, and human–computer interaction. However, decoding electroencephalographic (EEG) signals with high accuracy continues to be a difficult task due to the weak signal-to-noise ratio, differences among subjects, and the complicated temporal–spectral–spatial neural dynamics. Deep learning methods recently developed, such as convolutional neural networks, recurrent architectures, graph neural networks, and adversarial transfer learning, have enhanced MI decoding performance, yet many models are still concentrating on a single representation domain or they need costly adaptation phases in terms of computation. To tackle these shortcomings, we present NeuroCrossNet, a unified tri-modal deep learning model that is able to learn the temporal, spectral, and spatial EEG features jointly for robust and calibration-free MI decoding. The suggested network combines a Temporal HyperMixer Block for capturing long-range temporal dependencies, a wavelet transformer for learning localized time–frequency representation, and a Graph Attention Network for EEG topology-aware spatial reasoning. Additionally, a Dynamic Residual Attention Gate (DRAG) has been developed to adaptively merge heterogeneous feature streams, and a compact subject-aware normalization (SAN) method enhances cross-subject generalization without the use of labeled target-domain calibration data. Our proposed model was tested following the rigorous leave-one-subject-out (LOSO) approach on BCI Competition IV-2a and High-Gamma datasets. NeuroCrossNet reached a classification accuracy of 91.30%, surpassing several strong benchmark methods, including CNN-LSTM, EEGNet, DeepConvNet, spectral CNN, and graph-based EEG decoding frameworks. Furthermore, a large number of ablation studies reveal that the integration of temporally, spectrally, and spatially complementary representations considerably boosts robustness and inter-subject consistency. Full article

With the widespread adoption of multiple-input–multiple-output (MIMO) technology, channel state information (CSI) prediction has become a crucial technique for enhancing the performance of wireless communication systems. Traditional channel prediction methods face performance bottlenecks under high-speed mobility and complex channel conditions, making it difficult to meet the requirements of modern communication systems. To address this issue, this paper proposes a fully complex-valued cross-domain modeling framework, termed a complex-valued multi-scale transformer with time–frequency cross-attention network (CMTF-Net), for MIMO CSI prediction. CMTF-Net integrates a learnable multi-scale short-time Fourier transform (LMS-STFT), complex-valued multi-head self-attention (C-MHSA), and bidirectional cross-domain attention for complex-valued sequences (BCDA-CVS). These modules are designed to preserve amplitude–phase consistency, adapt time–frequency representations to CSI evolution, and enable information interaction between temporal and spectral features. On the simulated Overall Test set, CMTF-Net achieves the lowest MAE of 0.000032 and the highest Corr. ($\rho$) of 0.8230 among the compared methods, while maintaining competitive SE and BER values of 0.4240 and 0.2411 at SNR = 10 dB. On the DICHASUS measured datasets, CMTF-Net also shows favorable Test-ID and Test-OOD performance. For example, on DICHASUS-2186, it obtains Corr. ($\rho$)/SE/BER values of 0.8367/0.4935/0.2243 on Test-ID and 0.8061/0.4697/0.2351 on Test-OOD. These results indicate that CMTF-Net provides a balanced performance profile across prediction accuracy, spatial alignment, and communication-oriented evaluation. Full article