Search Results (1,359)

Intelligent retail cabinets require accurate and real-time detection of small retail products in complex environments, particularly under low-light conditions and large-scale variations. However, existing object detection methods often suffer from insufficient feature representation and unstable performance in small-object retail commodity recognition tasks under low illumination and complex backgrounds. To address these challenges, this paper proposes a real-time small retail product detection framework based on YOLOv26 for low-light intelligent cabinet environments, aiming to improve detection accuracy, robustness, and deployment efficiency. A C3k2-enhanced multi-scale feature extraction module is introduced to strengthen feature representation for small retail products, while a novel detection head integrates minimum-resolution feature layers and an Efficient Multi-scale Attention (EMA) mechanism to enhance feature fitting ability under low-light conditions. In addition, convolution-based downsampling and Content-Aware ReAssembly of Features module (CARAFE) is adopted to improve feature fusion efficiency and reduce computational overhead. Experimental results on the RPC commodity dataset and the 6th Commodity Recognition Competition dataset demonstrate that the proposed method achieves improved detection performance compared with baseline models, with a 0.9% increase in Recall and a 0.2% improvement in mean Average Precision at IoU threshold 0.50 (mAP@50) while maintaining competitive mean Average Precision averaged over IoU thresholds from 0.50 to 0.95 (mAP@50-95). While the GFLOPS value rose from 5.8 to 6.3, deployment on the Jetson Nano platform achieves 25 FPS, demonstrating real-time detection capability in intelligent retail environments. The proposed framework provides a reliable and deployable solution for small retail product detection in low-light intelligent cabinet systems. Full article

Accurate stockpile volume estimation is crucial for material quantification and inventory management in civil engineering, directly affecting cost assessment and on-site decision-making. Traditional manual methods suffer from subjective bias and limitations in handling irregular geometries, resulting in reduced accuracy and efficiency. This study presents a Light Detection and Ranging (LiDAR)-based workflow integrated with Robot Operating System (ROS) for point cloud processing, enabling accurate volume estimation of irregular stockpiles. The core innovation lies in the integration of multi-station scanning, point cloud registration, boundary extraction, layered slicing, and numerical integration using the trapezoidal rule, thereby enabling geometrically precise volume estimation of irregular stockpiles. The proposed system was validated through three experimental scenarios: (1) controlled experiments, showing strong agreement with theoretical volumes; (2) verification experiments, demonstrating high stability and consistency; and (3) field experiments, yielding a volume of 124.93 m³ compared to 130–135 m³ obtained by manual measurement. The results indicate that the proposed approach reduces processing time by over 80% while significantly decreasing labor requirements and improving operational safety. Overall, the proposed method provides a reliable and efficient solution for volume estimation in practical engineering applications. Full article

Grassland fire is one of the major disasters threatening regional ecological security. Its occurrence, development, and spread are closely related to the spatial distribution and coverage of surface vegetation and snow cover across grassland areas. As the primary combustible fuel source, higher vegetation coverage increases fuel load and continuity, thereby directly determining grassland fire danger levels and accelerating fire spread velocity. In contrast, snow cover imposes an indirect regulatory effect on the spatiotemporal pattern of fire danger factors: it lowers surface temperature, raises near-surface humidity, and restricts the germination and growth of herbaceous vegetation in cold seasons, which effectively reduces available combustible materials and weakens regional fire hazard conditions. Therefore, accurately obtaining the coverage status of vegetation (direct combustible fuel factor) and snow cover (indirect fire-regulating factor) in complex grassland scenarios is the essential premise for reliable grassland fire danger monitoring, early warning, disaster prevention and control, and regional ecological management. Aiming at the practical problems in complex grassland scenarios (such as undulating terrain, uneven vegetation growth, large differences in snow depth, and complex lighting conditions), including difficulty in extracting vegetation and snow-covered areas, blurred and confusing boundaries, and low accuracy in coverage calculation, which seriously restrict the technical bottleneck of precise monitoring of grassland fire danger factors, this study takes near-ground images collected by grassland fire danger factor monitoring stations as the core data source, and proposes an improved UNet image segmentation model combined with image segmentation technology and deep learning methods to realize precise extraction of vegetation and snow-covered areas and efficient calculation of coverage in complex scenarios. To improve the model’s feature extraction ability, boundary localization accuracy, and reduce model parameters and computational overhead, the CBAM-ASPP (Convolutional Block Attention Module—Atrous Spatial Pyramid Pooling) module is integrated at the end of the encoding path. The attention mechanism is used to enhance the weight of key features, and the multi-scale receptive field of atrous spatial pyramid pooling is utilized to strengthen the model’s ability to fuse features of vegetation and snow areas of different scales. The residual attention mechanism is introduced in the upsampling stage to effectively alleviate the gradient disappearance problem, improve the model’s ability to accurately locate the boundaries of vegetation and snow areas, and reduce segmentation errors. In the training process, a dynamically weighted hybrid loss function is adopted to dynamically adjust the weights according to the segmentation difficulty of different types of samples during training, optimize the model training effect, and improve the segmentation accuracy and generalization ability. Experiments were conducted using near-ground images of typical complex grassland scenarios as the dataset, and the performance of the proposed model was verified through comparative experiments. The results show that in the vegetation segmentation task, the mean Intersection over Union (mIoU) of the model reaches 84.70%, and the accuracy rate is 91.28%, which are 1.48 and 1.58 percentage points higher than those of the standard UNet model, respectively. In the snow segmentation task, the mIoU of the model reaches 92.74%, and the accuracy rate is 94.19%, which are 2.39 and 2.36 percentage points higher than those of the standard UNet model, respectively. At the same time, the number of parameters of the model is reduced by 12.85% compared with the standard UNet. Also, its comprehensive performance is significantly better than that of mainstream image segmentation models such as FCN, SegNet, and DeepLabv3+. Based on the standardized time-series data retrieved by the optimized segmentation model, this study further constructs a Grassland Fire Risk Index (GFRI) using the Analytic Hierarchy Process (AHP). Pearson correlation verification confirms that the GFRI has an extremely significant positive correlation with historical fire frequency, accurately capturing the seasonal dynamic rhythm of regional grassland fire occurrence. This integrated framework of intelligent segmentation and fire risk quantification provides a reliable technical solution for grassland fire factor monitoring, dynamic fire risk assessment, early warning systems, and refined regional ecological management. Full article

Background/Objectives: The increased prevalence of multidrug-resistant Escherichia coli and carbapenemase-producing Enterobacteriaceae poses a critical threat to global health and food safety. Antimicrobial Blue Light (aBL) in the 400–470 nm spectrum has emerged as a promising, chemical-free disinfection strategy that targets intracellular porphyrins and flavins to induce oxidative stress. However, the influence of wavelength, dosimetry, and environmental stressors on endogenous photoinactivation remains poorly standardized regarding optical parameters and biological exposure protocols. This systematic review aimed to evaluate the antimicrobial efficacy of pure blue light (400–470 nm) against E. coli across various phenotypes and environmental conditions, excluding the use of exogenous photosensitizers. Methods: PubMed, Scopus, and Web of Science were searched for studies that utilized 400–470 nm light as an antimicrobial agent against E. coli. Data extraction focused on spectral efficiency, total fluence (J/cm²), and log₁₀ reduction. The Risk of Bias was assessed using an adapted Office of Health Assessment and Translation tool for in vitro studies. Results: Synthesis of 11 high-quality studies indicated that wavelengths near 405 nm have the highest germicidal efficiency due to the Soret band absorption of endogenous porphyrins. Efficacy is highly dose-dependent: significant log₁₀ reductions were achieved in planktonic cells, although biofilms required substantially higher fluences. Sub-lethal environmental stressors such as acidic pH, high salinity, and thermal fluctuations demonstrated a synergistic effect, which significantly enhanced the rate of photoinactivation. Multidrug-resistant and carbapenemase-producing Enterobacteriaceae strains showed similar susceptibility to aBL relative to antibiotic-sensitive strains, suggesting no cross-resistance between light and traditional drugs. Conclusions: Endogenous blue light is a highly effective, non-thermal technology for E. coli decontamination. Its efficacy is modulated by the interplay between optical parameters and environmental conditions. These findings provide a framework for the development of standardized protocols for applying aBL to clinical wound care and food industry use cases. They also highlight the potential of aBL as a critical tool in the post-antibiotic era. This systematic review was registered in the International prospective register of systematic reviews (PROSPERO) under protocol CRD420261331871. Full article

Sweet cherry cracking severely impairs its commercial value and causes huge economic losses, and the accurate real-time detection of fine cracking defects remains a challenging small-target detection task. Traditional manual sorting and conventional machine vision methods suffer from low efficiency and poor robustness, while existing YOLO-based models have limitations in multi-scale feature fusion, local feature discrimination and spatial information retention for cherry cracking detection, and their effectiveness in natural production environments has not been statistically validated. To address these issues, this study proposes YOLO-CY for cherry cracking defect detection. Three key modules were optimized: the C3k2_AdditiveBlock was designed to enhance multi-scale feature extraction, the C2PSA_CGLU module improved the discriminability of local crack features via refined channel attention, and the Efficient Up-Convolution Block replaced traditional upsampling to reduce spatial information loss. Experiments were conducted on a self-constructed dataset of 3662 cherry images acquired on a real sorting line under natural ambient light. The results showed that YOLO-CY achieved an mAP50 of 94.88% and an mAP50-95 of 64.92%, with precision and recall reaching 93.90% and 90.81%, respectively, significantly outperforming mainstream lightweight YOLO models and two-stage detectors. Ablation experiments verified the synergistic effect of the three improved modules, and the model only had a marginal increase in parameters (2.62 M) and GFLOPs (6.60), maintaining lightweight characteristics. YOLO-CY can accurately detect fine, low-contrast and pedicel-overlapping cracks and is suitable for real-time detection on automated cherry-sorting lines, providing a technical solution for intelligent cherry quality inspection. Full article

The fresh-cut avocado processing generates significant amounts of by-products, mainly peel and seed, with the peel representing a valuable source of phenolic compounds. In this context, the growing demand for sustainable technologies encourages the use of green solvents for bioactive compound recovery. In this study, natural deep eutectic solvents (NaDES) were evaluated as environmentally friendly solvents for the extraction of phenolic compounds from Hass avocado peels through ultrasound-assisted extraction and for their potential application in fresh-cut avocado. Phenolics were extracted using acidic water, ethanol, and NaDES based on choline chloride as a fixed hydrogen bond acceptor (HBA) and hydrogen bond donors (HBDs; lactic acid, glycerol, and citric acid) with the ultrasound-assisted system. The stability of the extracts was monitored for eight weeks (four weeks in darkness followed by four weeks under light exposure). Among the tested formulations, the lactic-acid-based NaDES showed the highest extraction efficiency and the best stability of phenolic compounds during storage (≥20 mg GAE g⁻¹ dw during the storage period). The lactic-acid-based extract was then applied to fresh-cut avocado to evaluate its potential for antioxidant enrichment and browning prevention during refrigerated storage. The treatment increased phenolic content and contributed to improved color stability (during seven days of storage). Overall, lactic-acid-based NaDES represent a promising green solvent system for recovering phenolics from avocado peel and for their functional application in fresh-cut avocado within a circular valorization approach. Full article

The Automatic Identification System (AIS) generates massive volumes of real-world ship trajectory data, providing a critical foundation for maritime ship-type classification. However, existing methods often struggle to simultaneously capture long-range temporal dependencies, maintain computational efficiency, and ensure model interpretability, making accurate multi-class classification challenging in real-world maritime environments. To address these limitations, this study proposes a robust and efficient hybrid framework that integrates a Feature Transformer module for deep temporal feature extraction with a LightGBM model for ensemble classification. The multi-head self-attention within the Feature Transformer captures long-range dependencies in preprocessed AIS sequences to generate compact 64-dimensional trajectory fingerprints. These deep representations are concatenated with 103 carefully designed kinematic, geometric, statistical, frequency-domain, and segment-level features and fed into a LightGBM classifier for final ship-type identification. We evaluate the framework on a real-world AIS dataset of 2196 trajectories collected between 2019 and 2023, covering 14 ship types under a natural long-tail distribution. Across five random seeds, the proposed hybrid model achieves 78.06% ± 1.15% accuracy (95% CI) and 74.09% ± 1.82% Macro-F1 (95% CI), significantly outperforming Transformer-only (65.09% accuracy) and LightGBM-only (66.85%) baselines, with paired statistical tests confirming the improvement (McNemar χ² = 172.07, p < 10⁻³⁹ vs. Transformer; χ² = 92.24, p < 10⁻²¹ vs. LightGBM). The hybrid model offers ultra-fast inference at 0.051 ms per trajectory on GPU at batch size 128 (≈19,500 samples/s), and provides instance-level interpretability via SHapley Additive exPlanations (SHAP) analysis. These properties make the framework practical for near-real-time maritime traffic monitoring and decision support. Full article

Object detection in underground coal mines is plagued by critical challenges, including low illumination, high dust-induced noise, extensive target scale variation, frequent occlusion, and fragmented target feature representation, which commonly result in severe missed detections and insufficient detection confidence. To tackle these bottlenecks, this study proposes DLR-YOLO, a high-performance lightweight object detector built upon the YOLOv11n baseline, with three core optimized modules. Specifically, a dynamic multi-scale global perception enhancement module (DMGPEM) is embedded in the backbone to realize adaptive multi-scale feature extraction under low-light conditions; a lightweight cross-attention (LCA) module is integrated into the neck to achieve efficient fusion of shallow detail features and deep semantic features while suppressing dust-related noise; and a Reparameterized stem (RepStem) module is developed for initial feature extraction to minimize critical information loss during downsampling. Experimental results on our self-collected and annotated in-house underground coal mine dataset demonstrate that DLR-YOLO achieves 94.4% mAP@50 and 66.7% mAP@50–95, corresponding to 3.5 and 5.7 percentage point improvements over the YOLOv11n baseline, respectively. Ablation studies further validate the independent effectiveness of each proposed module. Meanwhile, the detector maintains a lightweight architecture with only 2.7M parameters and 6.6 GFLOPs, and reaches an inference speed of 157.1 FPS, outperforming several state-of-the-art real-time detectors, including YOLOv12, YOLOv13, and RT-DETR, on the same dataset. These findings confirm that DLR-YOLO provides a robust, high-performance technical foundation for real-time safety monitoring systems in complex underground coal mine environments. Full article

Sustainability challenges in the wine sector have intensified the need for alternatives to conventional oak barrel maturation, a practice associated with high wood consumption, long maturation periods, and considerable economic and environmental cost. This study evaluates a resource-efficient maturation strategy for white wine using an experimental design comparing conventional oak alternatives with ultrasound-assisted extraction. Experiments were conducted in triplicate (n = 3) considering oak type (French chips vs. granules), dosage, toasting level (fresh, light, medium), and contact time (10 vs. 20 days). To enhance mass transfer, a 15 min ultrasound treatment (35 kHz) was applied. Statistical analysis (ANOVA One Way) indicated that oak fragment type and contact time significantly governed phenolic extraction (p < 0.05). Gallic acid concentrations increased significantly from 1.54 ± 0.03 mg L⁻¹ in the control to 4.41 ± 0.12 mg L⁻¹ in the most intensive ultrasound-assisted extraction treatment (p < 0.05). Syringaldehyde concentrations also showed a significant rise (1.13 to 1.44 mg L⁻¹; p < 0.05). Ultrasound significantly accelerated extraction kinetics while mitigating the loss of flavan-3-ols (≤28%) compared to conventional oak treatments (up to 34%). Economic assessment demonstrated a substantial reduction in production costs, from 0.21–0.56 € L⁻¹ range for standard fragment treatments to 0.05–0.07 € L⁻¹ when ultrasound was applied. Cost-efficiency metrics (<0.03 € mg⁻¹ gallic acid) confirmed that the combination of ultrasound and alternative oak materials provides an optimal, statistically significant balance between phenolic yield and economic viability. Full article

Underwater towed survey systems are widely used for marine observation, resource exploration, and target identification. While high-speed towing is increasingly required to improve operational efficiency, conventional single-stage towed systems face a critical trade-off: active heave compensation systems are complex and costly, whereas purely passive configurations lack sufficient disturbance rejection at higher speeds. To address this gap, this study proposes a two-stage towing system consisting of a vessel, heavy cable, depressor, light cable, and detection towed body, where the depressor functions as a mechanical low-pass filter. The depressor reduces vessel-induced heave motion transmission by approximately 79% compared with a conventional single-stage system. CFD simulations are conducted to evaluate hydrodynamic performance and extract key coefficients. A lumped-mass dynamic model is established for time-domain motion simulations. An integral sliding-mode controller with vessel heave feedforward compensation is designed to enhance depth-tracking capability. The active controller eliminates step response overshoot and provides robust depth regulation under wave disturbances. Sea trials under real ocean conditions validate the system’s motion stability, demonstrating satisfactory depth-keeping performance at high towing speeds. The simulation results show good agreement with experimental data, confirming the effectiveness of the proposed system and dynamic model. This work offers a practically validated towing platform solution for high-precision underwater survey operations. Full article

We perform three-dimensional general relativistic magnetohydrodynamic (GRMHD) simulations of a near-maximally spinning black hole (spin parameter $a=0.998$) with varying initial magnetic field geometries, systematically exploring the parameter space connecting magnetically arrested disk (MAD), intermediate (INT), and standard and normal evolution (SANE) accretion states. The magnetic flux threading the black hole horizon emerges as the fundamental state variable controlling jet efficiency, flow magnetization, and radiative output across all three states. We introduce complementary diagnostics—broadband spectral energy distributions spanning radio through hard X-ray frequencies and time-resolved X-ray light curves—that together connect simulation dynamics directly to multiwavelength observables. The radiative output follows a clear MAD > INT > SANE hierarchy in time-averaged luminosity, mean X-ray emission, as well as variability. Furthermore, MAD exhibits the highest fractional variability through quasi-periodic magnetic flux eruption events, and INT and SANE show moderate variability driven by episodic reconnection and stochastic MRI turbulence, respectively. Scaling to GRS 1915+105, Cyg X-1, and HLX-1, we demonstrate that all twelve temporal classes of GRS 1915+105 map naturally onto our three magnetic states, Cyg X-1’s persistent hard state is reproduced by a sustained INT configuration, and HLX-1’s extreme luminosities arise through efficient Blandford–Znajek extraction in MAD states scaled to higher black hole mass. Full article

Although Non-Wood Forest Products can offer interesting economic opportunities for rural communities, little is known about their commercial harvesters. Our work aims to shed light on the labour profiles, their accessibility to new entrants, and attractiveness for future green jobs. Through in-depth interviews, we explored the five-capitals profile of commercial resin, cork, mastic foliage, chestnut, pine nut, and wild mushroom harvesters in Spain. We found either freelance harvesters or entrepreneurs with a small gang. Our data show a typical male collector, who started the activity through his social networks (Social Capital), and whose origin depends on the product and Spanish region. Some commercial female harvesters were found in mushroom, chestnut and resin harvesting. Social constructs around the masculinization of these activities may explain their limited attractiveness for women. The ratio of non-Spanish commercial harvesters correlates with the weight of migrants in the analysed regions. Only a subgroup of resin harvesters devotes most of their year to this single activity. The rest complement NWFP income with a main forestry (cork and pinenut) or non-forestry occupation (mushroom, chestnut and mastic). For the latter products, access to Natural Capital was found to be crucial for job progress, as non-landowners require administrative and/or negotiation capacities to secure harvesting permits. Human Capital differs across NWFPs, from simpler skills such as recognising marketable produce and handling easy tools (mushroom, chestnuts, pine nut ground gathering and mastic), to complex abilities needed to balance efficiency with minimising tree damage (in resin tapping, pinenut shaking, and cork extraction). Such specialised tools and machinery (Built Capital) typically act as a barrier to entry and advancement. These profiles are expected to help decision-makers to design instruments promoting and regulating commercial harvesting, and tackle their risks: local landowners in allocating harvesting rights to external collectors; regional policymakers as competent authorities in forest legislation; and state-level administration concerning cultural, fiscal and labour-permit aspects. Full article

Manual assembly remains critical in Industry 5.0 high-mix/low-volume manufacturing, but it introduces resilience challenges due to cognitive load, training demands and frequent product changes. While AI-based supervision can mitigate errors, deploying such systems is often hindered by the cost of collecting and labelling thousands of real images for each product variant. This paper presents a Human-in-the-Loop semantic-twin pipeline that generates approximately 45,000 labelled synthetic images from a single CAD-based configuration and uses them to train an object detection model for real-time assembly supervision. Experiments on seven virtual environment configurations show that removing realistic lighting or camera motion reduces F1-score on real images from 0.87 to 0.46, confirming their critical role for synthetic-to-real transfer. A controlled laboratory study on a single bicycle chainring assembly task with 10 participants and 100 monitored cycles demonstrates the feasibility of automatic KPI extraction, with error events associated with a 25.6% increase in average cycle time (from 58.4 s to 73.3 s) under the tested conditions. Compared to manual annotation, where labelling 3000 images required approximately 4 h, the semantic-twin configuration takes around 4 to 6 h including image generation that enables rapid creation of large labelled datasets for new product variants without additional human annotation. These results provide a proof-of-concept foundation for resilient, data-efficient supervision of high-mix manual workstations, with full industrial validation across multiple products, stations and operator demographics identified as the necessary next step. Full article

To address accuracy and reliability challenges in simultaneous localization and mapping (SLAM) systems under extreme conditions, this paper presents LIVE-SLAM, a tightly-coupled LiDAR–inertial–visual framework. The technical core integrates a LiDAR Probabilistic Feature Extraction (LPFE) module to reduce frontend overhead by retaining high-confidence features, an adaptive confidence-based weighting strategy in the backend optimization to dynamically balance multi-modal residuals during sensor degradation, and a Visual Redundancy Removal (VRR) based hybrid loop closure mechanism to mitigate perceptual aliasing. Evaluation on the KITTI benchmark and challenging real-world datasets demonstrates that our multi-sensor fusion effectively prevents tracking failures typical of single-sensor systems. Specifically, compared to the LVI-SAM framework, the frontend runtime is reduced by 49% and backend efficiency is improved by 25% in complex urban sequences. Furthermore, our approach achieves an average RMSE improvement of 35.3% over FAST-LIO2 and LIO-SAM in diverse real-world scenarios, particularly in environments with geometric degradation and lighting variations. These findings confirm the system’s superior real-time efficiency and global localization precision in both standard benchmarks and complex practical applications. Full article

The growing demand for sustainable water treatment technologies requires photocatalysts that combine low environmental impact, energy efficiency, and mechanistic robustness. In this work, $Ag/Zn$ nanocomposites were green-synthesized using Chlorella vulgaris polar extract as a bio-mediated reducing and stabilizing agent, eliminating hazardous reagents and high-energy processing steps. Structural characterization (XRD, FTIR, SEM, UV–Vis) confirmed the coexistence of crystalline wurtzite $ZnO$ with metallic $Ag$ and ${Ag}_{2}O$ phases. Photocatalytic activity was evaluated through Congo Red degradation under a sequential dark–light protocol, enabling clear separation of adsorption and photoactivated pathways. During the $60 \mathrm{m}\mathrm{i}\mathrm{n}$ dark stage, removal remained limited ($~9–11\%$), consistent with adsorption–desorption equilibration. Upon UV irradiation, a distinct kinetic transition occurred, leading to final removal efficiencies of $44–49\%$ after $180 \mathrm{m}\mathrm{i}\mathrm{n}$. Notably, performance remained stable across the investigated photon flux range, indicating operation beyond a strictly photon-limited regime and highlighting an intrinsically energy-resilient catalytic response. A mechanistic kinetic model integrating reversible adsorption with light-dependent degradation accurately reproduced all experimental profiles ($NRMSE=3.14\%$) and successfully predicted an independent dark-control experiment without additional fitting. By coupling green synthesis with quantitative kinetic validation, this study proposes a sustainability-oriented framework for designing photocatalysts that align low-impact fabrication with energy-conscious water remediation. Full article