Search Results (5,093)

In natural scenes, the high resemblance between unripe tomatoes and foliage, combined with severe occlusion, challenges current semantic segmentation models, causing poor accuracy and indistinct boundaries. To overcome this, we propose CMAE-UNet, a camouflage suppression and edge enhancement model based on the UNet architecture. Specifically, a Global-Local Integrated Spatial Attention (GLISA) encoder merges dual-branch dilated convolutions, residual structures, and an Efficient Multi-scale Attention mechanism to expand receptive fields and highlight targets in complex backgrounds. Furthermore, a Frequency-Domain Feature Enhancement (FFE) module leverages the Fast Fourier Transform to separate and adaptively enhance distinct frequency components, effectively mitigating camouflage interference. Additionally, a Directional Edge Enhancement (DEE) module uses three-directional learnable convolutions and spatial attention to sharpen indistinct target contours. Evaluated on a custom Tomato dataset encompassing five complex scenarios, CMAE-UNet outperforms 12 prominent methods in mIoU, Dice, and Sen metrics, yielding smoother and more precise segmentation boundaries. The model robustly withstands field interference, providing strong technological support for automated tomato detection, intelligent harvesting, and growth monitoring. Full article

Background: Reliability-aware integration of heterogeneous data sources remains a fundamental challenge in multimodal deep learning: prevailing fusion strategies assume uniform reliability across sources and instances, limiting their responsiveness to data-dependent trustworthiness. Methods: We introduce REM-Fuse (Reliability-aware Evidential Multimodal Fusion), an evidential deep learning (EDL) framework in which per-source Dirichlet uncertainty adaptively weights each source through dual-channel weighting, asymmetric cross-scale enhancement, and Dempster–Shafer-inspired evidence accumulation. As a case study for cancer prognosis, REM-Fuse integrates multi-scale histopathology (10×, 20×) and RNA-seq on TCGA-BRCA (n = 831) via five-fold cross-validation with subtype- and stage-stratified analyses. Results: REM-Fuse attained a concordance index of 0.715 and a 60-month time-dependent AUC of 0.729, indicating moderate discrimination and significant risk separation (log-rank p < 0.001). Adaptive source weights and per-patient uncertainty varied significantly across molecular subtypes (Kruskal–Wallis p = 0.010 and p = 0.007), indicating patient-specific rather than fixed multimodal integration. Conclusions: REM-Fuse provides a compact reliability-aware fusion strategy for cancer prognosis, although external validation is needed before broader clinical or cross-cohort generalization. Full article

To reduce aggregate gradation counting errors caused by over-segmentation and under-segmentation of stacked and clustered aggregates with mixed types and diverse spatial distributions in hydraulic concrete, this study proposes BC-GECO2, a coarse and fine aggregate segmentation and counting method. Firstly, a BAHiera feature extraction network is designed to extract multi-scale deep features through edge-aware attention. In addition, a DFG-Edge module is developed to enhance the boundary features of densely distributed aggregates by integrating wavelet transform with a gated fusion mechanism, thereby alleviating the loss of small aggregate features during downsampling. Secondly, a CSFM-GFFCA module is constructed, in which a dual-branch structure is employed to adaptively fuse adjacent-scale features, strengthen the edge responses of densely distributed small aggregates, and enhance cross-layer feature interaction. Finally, a joint optimization function combining Focal loss and counting loss is established to guide the model toward hard-to-classify pixels, especially boundary pixels, thereby improving segmentation integrity and counting accuracy. Experiments conducted on an aggregate dataset collected from practical construction sites show that, compared with the baseline GECO2 model, the proposed method improves the average segmentation IoU, Dice, and BIoU by 2.92%, 5.04%, and 2.83%, respectively, while reducing the average counting MAE and RMSE by 6.92 and 15.65, respectively. Moreover, BC-GECO2 exhibits superior robustness and generalization capability under different stacking densities and blurred-boundary scenarios, providing technical support for the intelligent development of rapid concrete gradation detection. Full article

Image compression has evolved from human-centric perceptual coding toward support for diverse machine vision applications, requiring modern codecs to serve both human viewing and downstream tasks in closed-set settings (where target tasks are incorporated during training) and open-set settings (where previously unseen tasks arise at test time). While recent learned compression methods jointly optimize perceptual quality and closed-set task performance, they often fail to generalize to unseen open-set tasks due to fixed training assumptions and objectives. Our prior work, LoRA-comp (Low-Rank Adaptation Compression), effectively addresses open-set challenges via instance-specific test-time fine-tuning (TTFT) without requiring task-specific pre-training. Nevertheless, its fixed LoRA architecture, which assigns a uniform rank across all layers, often leads to suboptimal instance-level performance. Moreover, allocating the same number of training epochs to every instance introduces unnecessary encoding-time overhead. To address these challenges, we propose Instance-Specific Early Stopping with Dynamic Rank Adaptation (ISER), which extends LoRA-comp. Building upon the LoRA-comp–based instance-specific adaptation framework, ISER introduces (i) instance-specific early stopping (ISES) combined with a multi-scale training strategy (MSTS) to reduce TTFT overhead and (ii) instance-specific dynamic rank adaptation (ISRA) to tailor the LoRA architecture per instance. Experiments demonstrate that ISER consistently outperforms competing methods. Compared to LoRA-comp, ISER achieves up to a 7% BD-Rate improvement and up to a 44% reduction in encoding time. Moreover, ISER achieves up to a 98% relative improvement in BD-Rate gain and up to a 19.2% reduction in decoding time over TransTIC. Full article

Forest fires are among the most critical disturbance agents reshaping Mediterranean ecosystems under accelerating climate change. This study employs a multi-scale remote sensing approach to examine the relationship between post-fire vegetation dynamics and climate variability in high-fire-risk areas of southern İzmir, Türkiye. Burned areas were delineated using the Burned Area Index (BAI) and differenced Normalized Burn Ratio (dNBR) applied to Landsat imagery (1990–2024) and Sentinel-2 imagery (2017–2024). Post-fire vegetation recovery was quantified through the Normalized Difference Vegetation Index (NDVI), Soil-Adjusted Vegetation Index (SAVI), Vegetation Condition Index (VCI), Leaf Area Index (LAI), and Land Surface Temperature (LST) derived from the Moderate Resolution Imaging Spectroradiometer (MODIS) products. Climate variables, including soil moisture, precipitation, and maximum, minimum, and mean air temperature, were derived from the TerraClimate dataset. Long-term spatiotemporal trends were assessed using the non-parametric Mann–Kendall (MK) test and Sen’s slope estimator for the 2000–2023 period. Results indicate a statistically significant increase in mean temperature (p < 0.05) and a concurrent decline in soil moisture over the past three decades, consistent with progressive atmospheric aridification. Vegetation indices exhibited marked seasonal asymmetry: significant declines in NDVI, SAVI, and LAI were recorded during summer months, whereas partial recovery was confined to the winter–spring wet season. A pronounced warm-dry shift was identified in the post-2015 period, characterized by positive Land Surface Temperature anomalies and compressed vegetation recovery windows. These findings highlight that increasing thermal stress and diminishing soil moisture collectively constrain post-fire ecosystem resilience in the Mediterranean climatic zone (MCZ). The integrated remote sensing framework developed here provides a robust and transferable basis for fire ecosystem monitoring and the formulation of climate adaptation strategies in fire-prone dryland regions. Full article

With the widespread adoption of cloud computing technology, modern cloud platforms have become increasingly complex and dynamic, posing significant challenges for efficient resource management. Accurate forecasting of cloud resource loads has therefore become essential for improving service quality, optimizing resource utilization, and reducing operational costs. To address the intrinsic characteristics of cloud load time series, including nonlinear fluctuations, multi-scale temporal dependencies, and redundant high-dimensional features, this paper proposes the MST-iTransformer model, which integrates multi-scale temporal encoding, sparse attention, and adaptive feature selection mechanisms. Specifically, a multi-scale temporal encoding module is developed to capture and fuse temporal dependencies across multiple periodic scales. Furthermore, an adaptive feature selection module is introduced to dynamically assign importance weights to resource features, enhancing informative variables while suppressing redundant ones. Meanwhile, a sparse attention mechanism is incorporated to reduce computational overhead while maintaining forecasting accuracy. The proposed model is evaluated on the Alibaba Cluster Trace dataset. Experimental results demonstrate that MST-iTransformer achieves MSE, RMSE, and MAE values of 0.5559, 0.7456, and 0.4968, respectively. Compared with the original iTransformer, the proposed model achieves simultaneous reductions in prediction errors and inference latency, validating the effectiveness of the multi-scale temporal encoding, sparse attention mechanism, and adaptive feature selection modules in improving forecasting accuracy and computational efficiency. These improvements provide reliable prediction support for resource scheduling and elastic scaling in cloud data centers. Full article

Fixed delays and single-state thresholds are difficult to adapt to dependent event sequences, short authorization windows, and multi-parameter constraints in rocket-sled ejection tests. This study presents PMT-TC, a programmable multi-event triggering and timing coordination system. It provides multi-source state validation, mission-table-driven closed-window and Boolean-predicate evaluation, predecessor-gated one-shot output, and time-correlated event recording. Compared with fixed-delay or single-speed triggering, PMT-TC changes event variables and logic between missions while preserving order, data-validity and safety gates, and records for post-test verification. An onboard–ground cooperative architecture links pre-test configuration, autonomous onboard execution, buffered recording, and post-test reconstruction on a common mission-relative time base. Validation comprised vibration and shock tests, navigation and integration tests, development-stage strategy comparisons, and two mission-scale tests. Across the two tested configurations, all six $E_1$–$E_3$ events were issued in order. In mission B, Hall-effect and BDS integer speed fields agreed at six time-aligned samples across three 500 ms windows; $E_3$ was authorized at 6.92762240 s and 211 m/s by an AND predicate combining dual-source equality, speed, and time conditions. These results support programmable mission execution and traceable reconstruction within the tested configurations. Full article

Climate change and variability pose serious threats to natural and human systems. The Mediterranean and North Africa (MNA) are among the world’s climate change hotspots. An in-depth understanding of the decadal climate variability in this region is critical to support planning and management, as well as adaptation in important sectors such as water resources. Therefore, in this study, fifth-generation ECMWF atmospheric reanalysis (ERA5) precipitation data and the outputs of the Coordinated Regional Downscaling Experiment-COmmon Regional Experiment (CORDEX-CORE) regional models were used to characterize the decadal precipitation variability in MNA and its sub-regions (Western North Africa: WNA, Sahara: SAH, southern Mediterranean: SMED, and northern Mediterranean: NMED). The models showed overestimation in most areas and underestimation in a few areas relative to the ERA5 data, with the magnitude varying by region and season. The positive biases obtained from the regional climate models (RCMs) were higher than the positive biases obtained from the general circulation models (GCMs). The wet biases were dominant during the annual, summer, and autumn seasons over MNA and its sub-regions. Negative biases were mostly associated with GCMs, mainly HadGEM2-ES and/or NorESM1-M; meanwhile, they were linked with RCMs such as CCLM5-0-15 and/or RegCM4_v7 and were mostly obtained in winter and spring. The multi-model mean (MME) was better at reproducing the decadal precipitation patterns over MNA, SMED, and NMED at all time scales, while REMO2015-NorESM1-M and the MME performed better than the remaining models at the annual time scale over WNA and SAH. These findings are useful for improving climate modeling, the water resources management and related sectors, and climate adaptation strategies in the region, especially in North Africa. The short-period coverage of the simulated data available for this study constitutes a limitation to the findings. Full article

Traditional particle swarm optimization (PSO) easily falls into premature convergence, while differential evolution (DE) is highly sensitive to fixed control parameters. Existing PSO-DE hybrid frameworks suffer from static population sizes and insufficient cross-population information exchange. This paper proposes PSO-DE-ADP, a hybrid optimizer with sinusoidal adaptive parameters, elite-guided mutation, ring neighborhood-weighted PSO and fitness-driven dynamic dual-population allocation. Four complementary mechanisms are integrated: (i) sine-wave perturbation superimposed on linear decay adaptively adjusts PSO inertia weight, acceleration factors and DE scaling/crossover coefficients to balance search stages; (ii) global elite individuals are embedded into DE mutation to reduce blind random search; (iii) ring topology with weighted learning realizes bidirectional information interaction between PSO and DE subpopulations; (iv) the proportion of PSO/DE individuals is dynamically adjusted according to elite ratio to allocate computing resources. Experiments adopt the CEC2017 30-dimensional benchmark with 30 test functions covering unimodal, multimodal, hybrid and composite landscapes. Compared with 8 state-of-the-art metaheuristics, PSO-DE-ADP achieves the lowest Friedman rank (1.08 vs. 2.23–4.90 for PSO variants; 1.53 vs. 2.07–5.00 for non-PSO algorithms). Ablation tests prove each component significantly boosts accuracy; The algorithm only costs 0.172 s average runtime, superior to all competitors. Statistical Wilcoxon and Friedman tests verify its significant superiority. Future work extends this method to multi-objective, constrained and real engineering optimization tasks. Full article

Periodic manual oversight is difficult to scale for large portfolios of funded research projects. Progress delays, budget irregularities, and superficial reporting often go undetected until final acceptance. Many conventional detection methods also generate false positives when contextually supported schedule adjustments resemble anomalous patterns. We present the Intelligent Project Monitoring System (IPMS), which couples a feature-decoupled multi-task LSTM-Attention network with an Adaptive Isolation Forest. The LSTM-Attention component models project workflows through finite state machines and predicts milestone deviations. The Adaptive Isolation Forest then flags records after a context gate screens cases meeting the study’s legacy legitimate-deviation criteria before final alerting. A multi-head attention module tracks how execution performance evolves over the project lifecycle, and the system includes a loss-ratio signal for candidate-shift review and feedback-gated controlled recalibration; its response was evaluated only under one researcher-designed synthetic global policy-change injection. On a real-world dataset from a provincial management platform, in which approximately 7% of legacy-labeled records carried an anomalous reference label, IPMS achieved an AUC of 0.924 and a false-positive rate of 4.2% against the available legacy binary reference labels, a 76.9% relative reduction in observed FPR compared with standard Isolation Forest. Milestone deviation prediction reached an MAE of 1.85 days, 34.2% lower than standard LSTM. Execution profiling achieved an MAE of 0.082. Removing the deviation filter alone degraded F1 by 12.3%, and removing multi-scale fusion increased the miss rate for long-duration stalls by 23%. Full article

Nutating mills offer intense comminution dynamics without the gravitational constraints of conventional tumbling mills; however, their structural response and charge–structure interaction mechanisms remain insufficiently characterized. This work examines the dynamic behavior of a laboratory-scale nutating mill (NuMill) with granular charge through combined experimental characterization and a two-way coupled numerical framework integrating multi-body dynamics (MBD) with the discrete element method (DEM). This study expands on previous work, extending the characterization of the NuMill to include mount stiffness, damping, and charge–structure coupling. The NuMill was adapted with vibration isolation mounts and internal chamber ribs to more closely emulate the operating behavior of industrial Hicom mills. Measurements of forces, torques, and accelerations were obtained across a range of mounting, charge, and chamber geometry configurations. Results show that approximating the granular charge in a ribbed chamber as a rigid body leads to substantial predictive error, overestimating crank-pin forces by 21% and underestimating driveshaft torque by 82% at 700 RPM. Incorporating experimentally characterized stiffness into the coupled MBD–DEM model showed good prediction accuracy for granular charge at 700 RPM. The simulation overestimated crank-pin force by 34%, underestimated driveshaft torque by 25%, and reproduced rigid-body natural frequencies within 1%. These findings demonstrate that structural compliance and charge–structure coupling play a central role in determining operational loads in nutating mills. The validated modeling framework developed here provides a more reliable basis for design assessment and parameter selection in industrial nutating milling applications and extends existing experimental foundations for laboratory-scale systems. Full article

Urban public spaces in cold-climate cities commonly face the challenge of seasonal vitality attenuation. Taking the central urban area of Shenyang as a case study, this paper constructs a HEVA four-dimensional framework based on “spatial diagnosis, mechanism analysis, and design response.” By integrating Multiscale Geographically Weighted Regression (MGWR) with GIS spatial analysis, the study reveals the spatial heterogeneity and driving mechanisms of winter vitality attenuation in linear green spaces. The findings are threefold: (1) Winter wind fields induce significant local wind-chill effects that severely suppress winter stay vitality in green spaces; (2) MGWR results show that greenery coverage exhibits a predominantly positive effect (approximately 81% of sample points have positive coefficients), bus stop density exhibits a “bifurcated spatial pattern”, with positive coefficients at approximately 53% of sample points and negative at the remaining 47%, indicating a highly localized effect., commercial service density is positive across the entire study area (100% positive), building density is predominantly negative (approximately 87%), and road network density has a relatively weak effect; (3) Based on the above mechanisms, a three-pronged elastic design strategy system is proposed, encompassing “climate-adaptive regulation, functional composite implantation, and regional cultural response.” This study explores a data-driven, mechanism-guided geospatial design pathway for the renewal of underperforming spaces in cold-climate cities. The generalizability of the methodological framework warrants further validation through multi-city and multi-seasonal data. Full article

The quantitative measurement of TCC at tourist cities involves considerable methodological complexities. Conventional evaluation methods based on fixed numerical criteria are frequently insufficient since tourism systems are characterized by unclear information, gradual shifts, and subjective experiences. To address these methodological constraints, this study develops and implements a hybrid FAHP-weighted Fuzzy Comprehensive Evaluation (FCE) framework to estimate TCC. The framework incorporates four critical indicators, visitor density, waste generation, infrastructure load, visitor perception, reflecting the physical, environmental, infrastructural, and socio-cultural dimensions of tourism development. These indicators are assessed using a structured analytical procedure integrating fuzzification, expert-derived weighting through the fuzzy analytic hierarchy process, fuzzy comprehensive evaluation, and centroid defuzzification. This study presents a hybrid SEM–FAHP–FCE framework, which is a multi-method integrated decision model that combines SEM (structural equation modeling) for visitor perception and behavioral drivers, FAHP (fuzzy analytic hierarchy process) for expert-based indicator weighting and FCE (fuzzy comprehensive evaluation) for final carrying capacity assessment under uncertainty. This combined methodology offers a more adaptive and empirically grounded assessment of TCC by synthesizing quantitative conditions with qualitative stakeholder perceptions within a unified evaluative model. The framework is validated through empirical application to Vrindavan, Uttar Pradesh, one of India’s most visited destinations, across eight study sites and two seasonal conditions (normal and festival-peak). Primary data collection comprises 300 visitor count observation units and field data, i.e., visitor count, waste generation and infrastructure details. The defuzzified FAHP–FCE results for each site (W = 3.94, 3.33, 2.44, 1.95, 3.66, 1.87, 1.86 and 2.71) reveal that major pilgrimage nodes exhibit scores approaching the upper bound of the evaluation scale, indicating that tourism pressure has significantly exceeded carrying capacity during peak periods. This study establishes proof-of-concept for the FAHP–FCE framework as a standardized, replicable instrument for evidence-based TCC governance at tourism cities. Full article

Akerlof’s classical theory of the “Market for Lemons,” which conceptualizes adverse selection driven by information asymmetry, established the foundation of information economics. While the traditional model assumes static equilibria among a limited number of agents, analyzing its behavioral dynamics within large-scale, complex network environments remains a highly relevant task in computational social science. This study extends the classical lemon market model into a nonlinear dynamical system on adaptive networks. We mathematically elucidate macro-level social phase transitions—specifically structural exclusion such as scapegoating and collective ostracism—induced by computational cognitive limits, and evaluate optimal intervention strategies to mitigate these systemic failures. Multi-agent simulations utilizing large-scale tensor operations demonstrate that autonomous edge rewiring under incomplete information does not merely result in the uniform displacement of high-quality goods as predicted by static theory. Instead, the network self-organizes into an irreversible structural division: a core group of influential agents monopolizes high-quality information, while marginalized agents are isolated into a peripheral “lemon echo chamber” where only low-quality information circulates. To address this structural pathology under a resource constraint limiting intervention to 10% of the total agents, we evaluated two distinct approaches. The results indicate that providing informational support to influential hubs functions as a trap that exacerbates systemic inequality, superficially elevating the overall market evaluation but permanently fixing the exclusion gap. Conversely, the forced maintenance and protection of “weak ties” bridging disconnected clusters constitutes the mathematically optimal solution to dissolve fragmentation, effectively eliminating the price gap and facilitating social inclusion. Furthermore, this study demonstrates that the mechanism of social exclusion exhibits strong hysteresis effects. A distinct tipping point governs the progression toward a fragmented lemon echo chamber. Interventions implemented after crossing this critical threshold fail to restore the system to its baseline state despite identical resource expenditure, confirming the presence of an irreversible phase transition. These findings establish that the collapse dynamics outlined in the classical lemon market serve as a generalized model for explaining contemporary collective ostracism driven by information cascades. Consequently, the analysis highlights the necessity of early intervention prior to critical thresholds and the systemic preservation of structural bypasses rather than post-hoc remediation. Full article

Regenerative aluminum melting serves as a core process in recycled aluminum production. In the regenerative aluminum melting process, the furnace temperature is a key variable which affects product performance and energy costs. The extreme in-furnace temperature necessitates sensors equipped with protective jackets, which increases measurement costs and severely compromises real-time monitoring capability. Accordingly, accurate furnace temperature prediction is highly valuable for regenerative aluminum melting. In regenerative aluminum melting furnaces, periodic burner nozzle commutation and frequent material charging and discharging lead to complex and time-varying operating conditions, posing considerable challenges to high-precision furnace temperature prediction. To address these issues, a condition-adaptive multi-scale convolution-enhanced Transformer (CA-MC-Transformer) model is proposed for furnace temperature prediction. Firstly, an agglomerative hierarchical clustering algorithm based on the weighted dynamic time warping (WDTW) distance is designed to perform unsupervised clustering on historical process data, thereby extracting physically interpretable prior labels for macroscopic operating conditions. Secondly, multi-scale dilated causal convolutions are utilized to capture local dynamic features at diverse temporal resolutions. A soft attention mechanism is further introduced to dynamically assign fusion weights to condition embeddings and local features, enabling condition-adaptive feature reconstruction. Finally, the fused adaptive features are fed into an encoder-only Transformer network to capture the global long-range temporal dependencies and achieve accurate furnace temperature prediction. Comparative experiments conducted on real operational datasets from an aluminum plant verify that the proposed method effectively eliminates the inherent tracking lag of conventional deep learning models, and substantially improves prediction accuracy and anti-noise robustness under complex and variable operating conditions. Full article