Search Results (10,725)

This paper presents the development of a robot-oriented Distributed Acoustic Sensing (DAS) system designed for environmental monitoring and hazard detection on ground robotic platforms. Unlike conventional DAS solutions primarily intended for stationary or quasi-stationary infrastructures, the proposed approach explicitly accounts for robot-induced mechanical vibrations, mobility constraints, and limited onboard resources. A dedicated anti-jitter signal processing pipeline combined with edge-based data processing is introduced to suppress motion-induced strain components while preserving weak external acoustic signals. The system integrates optical fiber deployment along the robot structure using flexible guides and vibration-isolated clamps, ensuring stable mechanical coupling under continuous motion. Experimental validation, including laboratory tests and preliminary outdoor field trials, demonstrates reliable detection of acoustic events in the 10–200 Hz frequency range, with reduced processing latency of 80–100 ms and a detection reliability of up to 95%. Comparative analysis with conventional sensors confirms the advantages of the proposed DAS-based approach in terms of sensitivity, spatial coverage, and robustness. The results demonstrate the feasibility and effectiveness of DAS technology for real-time sensing applications on mobile robotic platforms. Full article

Vehicle-to-grid (V2G) services are gaining attention as a strategy to integrate electric vehicles (EVs) into sustainable energy systems. Although technological aspects have been widely studied, methodologies for identifying optimal V2G hubs and forecasting the energy available for grid transfer remain limited. This study introduces a data-driven approach to (i) identify the optimal V2G region based on the aggregated parking duration using floating car data (FCD; collected from GPS-enabled vehicles); (ii) estimate the surplus battery capacity of electric vehicles in that region; and (iii) forecast the energy transferable to the grid. The methodology applies spatial k-means clustering to define candidate zones, computes aggregated parking durations, and selects the optimal hub. The surplus energy is estimated considering the daily mobility needs of users, 20% reserve, and transfer rates. For forecasting, autoregressive integrated moving average (ARIMA) and long short-term memory (LSTM) models are implemented and compared. The proposed methodology has been applied to a real case study, using 58 days of FCD observations. The empirical findings of this study show the goodness of the proposed methodology, and the opportunity offered V2G technology to support the sustainable use of energy. The ARIMA model demonstrated a superior forecasting performance with an RMSE of 52.424, MAE of 36.05, and MAPE of 12.98%, outperforming LSTM (RMSE of 99.09, MAE of 80.351, and MAPE of 53.20%) under the current data conditions. The results of this study suggest that for supporting grid needs of a medium-sized city, V2G plays a key role, and at the current status of the EV penetration, the use of FCD and predictive approaches is paramount for making an informed decision. Full article

Performance diagnostics in alpine skiing increasingly rely on integrated biomechanical and kinematic assessments to support technique optimization under real training conditions; however, many existing approaches address trajectory geometry or biomechanical variables separately, limiting their explanatory power. This study evaluates an integrated analysis framework combining OptiPath, an AI-assisted video-based trajectory analysis tool, with XSensDOT wearable inertial sensors to identify technical inefficiencies during giant slalom skiing. Thirty competitive youth athletes (n = 30; 14–16 years) performed controlled runs with predefined lateral offsets from the gates, enabling systematic examination of the relationship between spatial trajectory deviations, biomechanical execution, and performance outcomes. Skier trajectories were extracted using computer vision-based methods, while lower-limb kinematics, trunk motion, and tri-axial acceleration were recorded using inertial measurement units. Deviations from mathematically defined ideal trajectories were quantified through regression-based calibration and arc-based modeling. The results show that although OptiPath reliably detected trajectory variations, shorter skiing paths did not consistently produce faster run times. Instead, superior performance was associated with more efficient biomechanical execution, reflected by coordinated trunk–lower limb motion, controlled vertical loading, reduced lateral corrections, and higher forward acceleration, even when longer trajectories were followed. These findings indicate that trajectory geometry alone is insufficient to explain performance outcomes and support the integration of wearable biomechanics with trajectory modeling as a practical, low-cost, and field-deployable tool for alpine skiing performance diagnostics. Full article

Italy is estimated to host thousands of abandoned mines, many of which contain large volumes of mine residues that negatively affect land and aquatic ecosystems, also posing a risk to human health. This study evaluates the effectiveness of spaceborne imaging spectroscopy combined with laboratory spectroscopy for characterizing the mineralogy and geochemistry of residues from the abandoned Montevecchio sulfide mine in southwestern Sardinia, a site recognized as a significant source of environmental pollution. Mine tailings and their downstream dispersion along the Rio Irvi River were systematically studied and sampled in the field. Collected samples were analyzed in the lab using an Analytical Spectral Device (ASD) spectroradiometer, complemented by powder X-ray Diffraction (XRD) for mineralogical characterization. Affected zones were subsequently mapped using the Environmental Mapping and Analysis Program (EnMAP) hyperspectral satellite data at a 30 m spatial resolution, by applying a polynomial fitting technique to the image spectra. The results reveal the presence of Fe- and Zn-bearing sulfates and oxy/hydroxides, indicative of acidic-to-circum-neutral drainage conditions in the mine tailings and along affected streams. Specifically, EnMAP was able to detect jarosite and subtle chemical and physical variations in Fe-hydroxides. This integrated approach enabled the delineation of environmental conditions and zones with varying acidity based on the spectral characteristics of secondary minerals. Overall, the study demonstrates the potential of EnMAP data for mapping acid mine drainage and assessing environmental impacts in legacy mining areas. Full article

Remote sensing semantic segmentation encounters several challenges, including scale variation, the coexistence of class similarity and intra-class diversity, difficulties in modeling long-range dependencies, and shadow occlusions. Slender structures and complex boundaries present particular segmentation difficulties, especially in high-resolution imagery acquired by satellite and aerial cameras, UAV-borne optical sensors, and other imaging payloads. These sensing systems deliver large-area coverage with fine ground sampling distance, which magnifies domain shifts between different sensors and acquisition conditions. This work builds upon DeepLabV3+ and proposes complementary improvements at three stages: input, context, and decoder fusion. First, to mitigate the interference of complex and heterogeneous data distributions on network optimization, a feature-mapping network is introduced to project raw images into a simpler distribution before they are fed into the segmentation backbone. This approach facilitates training and enhances feature separability. Second, although the Atrous Spatial Pyramid Pooling (ASPP) aggregates multi-scale context, it remains insufficient for modeling long-range dependencies. Therefore, a routing-style global modeling module is incorporated after ASPP to strengthen global relation modeling and ensure cross-region semantic consistency. Third, considering that the fusion between shallow details and deep semantics in the decoder is limited and prone to boundary blurring, a fusion module is designed to facilitate deep interaction and joint learning through cross-layer feature alignment and coupling. The proposed model improves the mean Intersection over Union (mIoU) by 8.83% on the LoveDA dataset and by 6.72% on the ISPRS Potsdam dataset compared to the baseline. Qualitative results further demonstrate clearer boundaries and more stable region annotations, while the proposed modules are plug-and-play and easy to integrate into camera-based remote sensing pipelines and other imaging-sensor systems, providing a practical accuracy–efficiency trade-off. Full article

Semi-supervised joint classification of Hyperspectral Images (HSIs) and LiDAR-derived Digital Surface Models (DSMs) remains challenging due to scarcity of labeled pixels, strong intra-class variability, and the heterogeneous nature of spectral and elevation features. In this work, we propose a Hybrid Mamba–Graph Fusion Network (HMGF-Net) with Multi-Stage Pseudo-Label Refinement (MS-PLR) for semi-supervised hyperspectral–LiDAR classification. The framework employs a spectral–spatial HSI backbone combining 3D–2D convolutions, a compact LiDAR CNN encoder, Mamba-style state-space sequence blocks for long-range spectral and cross-modal dependency modeling, and a graph fusion module that propagates information over a heterogeneous pixel graph. Semi-supervised learning is realized via a three-stage pseudolabeling pipeline that progressively filters, smooths, and re-weights pseudolabels based on prediction confidence, spatial–spectral consistency, and graph neighborhood agreement. We validate HMGF-Net on three benchmark hyperspectral–LiDAR datasets. Compared with a set of eight state-of-the-art (SOTA) baselines, including 3D-CNNs, SSRN, HybridSN, transformer-based models such as SpectralFormer, multimodal CNN–GCN fusion networks, and recent semi-supervised methods, the proposed approach delivers consistent gains in overall accuracy, average accuracy, and Cohen’s kappa, especially in low-label regimes (10% labeled pixels). The results highlight that the synergy between sequence modeling and graph reasoning in combination with carefully designed pseudolabel refinement is essential to maximizing the benefit of abundant unlabeled samples in multimodal remote sensing scenarios. Full article

Spatial resolution is a critical parameter for optical frequency domain reflectometry (OFDR). Phase-sensitive OFDR (Φ-OFDR) measures strain by detecting phase variations between adjacent sampling points, having the potential to achieve the theoretical limitation of spatial resolution. However, the results of Φ-OFDR suffer from large fluctuations due to multiple types of noise, including coherent fading and system noise. This work presents an OFDR-based strain sensing method that combines phase demodulation with cross-correlation analysis to achieve high spatial resolution. In the phase demodulation, the frequency-shift averaging (FSAV) and rotating vector summation (RVS) algorithms are first employed to suppress coherent fading noise and achieve accurate strain localization. Then the cross-correlation approach with an adaptive window is proposed. Guided by the accurate strain boundary obtained from phase demodulation, the length and position of the cross-correlation window are automatically adjusted to fit for continuous and uniform strain regions. As a result, an accurate and complete strain distribution along the entire fiber is finally obtained. The experimental results show that, within a strain range of 100–700 με, the method achieves a spatial resolution of 0.27 mm for the strain boundary, with a root-mean-square error approaching 0.94%. The processing time reaches approximately 0.035 s, with a demodulation length of 1.6 m. The proposed approach offers precise spatial localization of the strain boundary and stable strain measurement, demonstrating its potential for high-resolution OFDR-based sensing applications. Full article

Background/Objectives: Hepatocellular carcinoma (HCC) shows marked molecular and histopathological heterogeneity. Among the alterations most strongly associated with clinical outcome are mutations in TP53 and CTNNB1, as well as the presence of microvascular invasion (MVI). Although these factors are well established as prognostic indicators, how their molecular effects relate to tumor morphology remains unclear. In this work, we studied transcriptomic changes linked to TP53 and CTNNB1 mutational status and to MVI, and examined whether these changes are reflected in routine histology. Methods: RNA sequencing data from HCC samples annotated for mutations and vascular invasion were analyzed using differential expression analysis combined with machine learning-based feature selection to characterize the underlying transcriptional programs. In parallel, we trained a weakly supervised multitask deep learning model on hematoxylin and eosin-stained whole-slide images using slide-level labels only, without spatial annotations, to assess whether these features could be inferred from global histological patterns. Results: Distinct gene expression profiles were observed for TP53-mutated, CTNNB1-mutated, and MVI-positive tumors, involving pathways related to proliferation, metabolism, and invasion. Image-based models were able to capture morphological patterns associated with these states, achieving above-random discrimination with variable performance across tasks. Conclusions: Taken together, these results support the existence of coherent biological programs underlying key risk determinants in HCC and indicate that their phenotypic effects are, at least in part, detectable in routine histopathology. This provides a rationale for integrative morpho-molecular approaches to risk assessment in HCC. Full article

Black carbon (BC) is a short-lived climate pollutant with substantial warming and health impacts, yet its contribution from maritime activities in data-limited regions remains poorly constrained. This study conducts a methodological and uncertainty-focused evaluation of tier-based emission inventory approaches from the European Monitoring and Evaluation Programme/European Environment Agency (EMEP/EEA) Guidebook, examining fuel-based (Tier I) and activity-based (Tier III) methodologies using national fuel statistics, port call activity, vessel registry data, and an operational Philippine Coast Guard dataset. Monte Carlo uncertainty analysis, spatial mapping, and hotspot intensity analysis are applied to evaluate how each tier responds to data limitations and parameter uncertainty rather than to reconcile absolute emission magnitudes. Results indicate that Tier I provides scalability for national reporting but exhibits substantial uncertainty for gasoline-dominated segments due to reliance on particulate matter-based proxies, underscoring the role of Tier II as a targeted refinement option. Tier III applies an activity-based formulation using fuel consumption resolved by operational phase and phase-specific emission factors, consistent with EMEP/EEA Tier III guidance. These findings are integrated into a decision-oriented synthesis to support informed selection and combination of tiered emission approaches under data-limited maritime conditions aligned with national and international climate commitments. Full article

Quality of life, as an essential component of sustainable development, is particularly relevant in transnational contexts characterized by deep inequalities in human development, equity, and social well-being. The objective of this paper is to analyze the temporal and spatial changes in transnational patterns of quality of life observed between 2018 and 2025, taking into account levels of human development. To this end, multivariate statistical techniques were applied: partial triadic analysis, which allows the identification of both the common structure of the data and the temporal evolution of the indicators, together with the HJ-Biplot and cluster analysis, which provide a multidimensional and interpretable visualization of country profiles. The results reveal consistent configurations of quality of life, largely aligned with levels of human development, and highlight persistent inequalities in environmental quality, economic accessibility, and objective well-being. These findings are relevant for the formulation of policies aimed at enhancing population well-being, particularly in countries facing structural constraints despite their high levels of development. The contribution of this research lies in its three-dimensional, dynamic, and reproducible approach, which makes it possible to identify regional contrasts that are not visible through traditional methods based on unidimensional indicators or cross-sectional analyses. Full article

The categorization of real-time defects in heterogeneous domains is a long-standing challenge in the field of industrial visual inspection systems, primarily due to significant visual variations and the lack of labelled information in real-world inspection settings. This work presents the Adaptive Attention DropBlock (AADB) framework, a lightweight deep learning framework that was developed to promote cross-domain defect detection using attention-guided regularization. The proposed architecture integrates the Convolutional Block Attention Module (CBAM) and an organized DropBlock-based regularization scheme, creating a unified and robust framework. Although CBAM-based approaches improve localization of defect-related areas and traditional DropBlock provides a generic spatial regularization, neither of them alone is specifically designed to reduce domain overfitting. To address this limitation, AADB combines attention-directed feature refinement with a progressive, transfer-aware dropout policy that promotes the learning of domain-invariant representations. The proposed model is built on a MobileNetV2 base and trained through a two-phase transfer learning regime, where the first phase consists of pretraining on a source domain and the second phase consists of adaptation to a visually dissimilar target domain with constrained supervision. The overall analysis of a metal surface defect dataset (source domain) and an aircraft surface defect dataset (target domain) shows that AADB outperforms CBAM-only, DropBlock-only, and conventional MobileNetV2 models, with an overall accuracy of 91.06%, a macro-F1 of 0.912, and a Cohen’s k of 0.866. Improved feature separability and localization of error are further described by qualitative analyses using Principal Component Analysis (PCA) and Grad-CAM. Overall, the framework provides a practical, interpretable, and edge-deployable solution to the classification of cross-domain defects in the industrial inspection setting. Full article

Fine-grained food image segmentation is essential for accurate dietary assessment and nutritional analysis, yet remains highly challenging due to ambiguous boundaries, inter-class similarity, and dense layouts of meals containing many different ingredients in real-world settings. Existing methods based solely on CNNs, Transformers, or Mamba architectures often fail to simultaneously preserve fine-grained local details and capture contextual dependencies over long distances. To address these limitations, we propose HDF (Hybrid Decoder for Food Image Segmentation), a novel decoding framework built upon the MambaVision backbone. Our approach first employs a convolution-based feature pyramid network (FPN) to extract multi-stage features from the encoder. These features are then thoroughly fused across scales using a Cross-Layer Mamba module that models inter-level dependencies with linear complexity. Subsequently, an Attention Refinement module integrates global semantic context through spatial–channel reweighting. Finally, a Food Co-occurrence Module explicitly enhances food-specific semantics by learning dynamic co-occurrence patterns among categories, improving segmentation of visually similar or frequently co-occurring ingredients. Evaluated on two widely used, high-quality benchmarks, FoodSeg103 and UEC-FoodPIX Complete, which are standard datasets for fine-grained food segmentation, HDF achieves a 52.25% mean Intersection-over-Union (mIoU) on FoodSeg103 and a 76.16% mIoU on UEC-FoodPIX Complete, outperforming current state-of-the-art methods by a clear margin. These results demonstrate that HDF’s hybrid design and explicit co-occurrence awareness effectively address key challenges in food image segmentation, providing a robust foundation for practical applications in dietary logging, nutritional estimation, and food safety inspection. Full article

Background: The precise segmentation and classification of dermoscopic images remain prominent obstacles in automated skin cancer evaluation due, in part, to variability in lesions, low-contrast borders, and additional artifacts in the background. There have been recent developments in foundation models, with a particular emphasis on the Segment Anything Model (SAM)—these models exhibit strong generalization potential but require domain-specific adaptation to function effectively in medical imaging. The advent of new architectures, particularly Vision Transformers (ViTs), expands the means of implementing robust lesion identification; however, their strengths are limited without spatial priors. Methods: The proposed study lays out an integrated foundation-model-based framework that utilizes SAM-Adapter-fine-tuning for lesion segmentation and a ViT-based classifier that incorporates lesion-specific cropping derived from segmentation and cross-attention fusion. The SAM encoder is kept frozen while lightweight adapters are fine-tuned only, to introduce skin surface-specific capacity. Segmentation priors are incorporated during the classification stage through fusion with patch-embeddings from the images, creating lesion-centric reasoning. The entire pipeline is trained using a joint multi-task approach using data from the ISIC 2018, HAM10000, and PH2 datasets. Results: From extensive experimentation, the proposed method outperforms the state-of-the-art segmentation and classification across the dataset. On the ISIC 2018 dataset, it achieves a Dice score of 94.27% for segmentation and an accuracy of 95.88% for classification performance. On PH2, a Dice score of 95.62% is achieved, and for HAM10000, an accuracy of 96.37% is achieved. Several ablation analyses confirm that both the SAM-Adapters and lesion-specific cropping and cross-attention fusion contribute substantially to performance. Paired t-tests are used to confirm statistical significance for all the previously stated measures where improvements over strong baselines indicate a $p<0.01$ for most comparisons and with large effect sizes. Conclusions: The results indicate that the combination of prior segmentation from foundation models, plus transformer-based classification, consistently and reliably improves the quality of lesion boundaries and diagnosis accuracy. Thus, the proposed SAM-ViT framework demonstrates a robust, generalizable, and lesion-centric automated dermoscopic analysis, and represents a promising initial step towards clinically deployable skin cancer decision-support system. Next steps will include model compression, improved pseudo-mask refinement and evaluation on real-world multi-center clinical cohorts. Full article

The increasing need for accurate prediction of geochemical anomalies requires methods capable of capturing complex spatial patterns that traditional approaches often fail to represent adequately. For N datasets of the form (X_i,Y_i) representing the geographic coordinates of sampling points and C_i denoting the geochemical measurement, training multilayer perceptrons (MLPs) presents a challenge. The low informativeness of the input features and their weak correlation with the target variable result in excessively simplified predictions. Analysis of a baseline model trained only on geographic coordinates showed that, while the loss function converges rapidly, the resulting values become overly “compressed” and fail to reflect the actual concentration range. To address this, a preprocessing method based on anisotropy was developed to enhance the correlation between input and output variables. This approach constructs, for each prediction point, a structured informational model that incorporates the direction and magnitude of spatial variability through sectoral and radial partitioning of the nearest sampling data. The transformed features are then used in a dual-MLP architecture, where the first network produces sectoral estimates, and the second aggregates them into the final prediction. The results show that anisotropic feature transformation significantly improves neural network prediction capabilities in geochemical analysis. Full article

In video scene graph prediction, the aim is to capture structured object interactions that occur over time in dynamic visual content. While recent spatiotemporal attention-based models have improved performance, they often suffer from high computational costs and limited structural consistency across long sequences. Therefore, we developed a Fourier transformer mixer network (FTM-Net), a modular, frequency-aware architecture that integrates spatial and temporal modeling via spectral operations. It incorporates a resolution-invariant Fourier Mixer for global spatial encoding and a Fast Fourier Transform (FFT)-Net-based temporal encoder that efficiently represents long-range dependencies with less complexity. To improve structural integrity, we introduce a spectral consistency loss function that synchronizes high-frequency relational patterns between frames. Experiments conducted utilizing the Action Genome dataset demonstrate that FTM-Net surpasses previous methodologies in terms of both Recall@K and mean Recall@K while markedly decreasing parameter count and inference duration, providing an efficient, interpretable, and generalizable approach for structured video comprehension. Full article