Search Results (123)

Accurate classification of indoor scenes remains a challenging problem in computer vision, particularly when datasets contain diverse room types and varying levels of contamination. We propose a novel method, CLIP-Guided Clustering, which introduces archetype-based similarity as a semantic feature space. Instead of directly using raw image embeddings, we compute similarity scores between each image and predefined textual archetypes (e.g., “clean room,” “cluttered room with dry debris,” “moldy bathroom,” “room with workers”). These scores form low-dimensional semantic vectors that enable interpretable clustering via K-Means. To evaluate clustering robustness, we systematically explored UMAP parameter configurations (n_neighbors, min_dist) and identified the optimal setting (n_neighbors = 5, min_dist = 0.0) with the highest silhouette score (0.631). This objective analysis confirms that archetype-based representations improve separability compared with conventional visual embeddings. In addition, we developed a hybrid segmentation pipeline combining the Segment Anything Model (SAM), DeepLabV3, and pre-processing techniques to accurately extract floor regions even in low-quality or cluttered images. Together, these methods provide a principled framework for semantic classification and segmentation of residential environments. Beyond application-specific domains, our results demonstrate that combining vision–language models with segmentation networks offers a generalizable strategy for interpretable and robust scene understanding. Full article

Municipal sewer networks span across large areas in cities around the world and require regular inspection to identify structural failures, blockages, and other issues that pose public health risks. Traditional inspection methods rely on remote-controlled robotic cameras or CCTV surveys performed by skilled inspectors. These processes are time-consuming, expensive, and often inconsistent; for example, the United States alone has more than 1.2 million miles of underground sewer pipes, and up to 75,000 failures are reported annually. Manual CCTV inspections can only cover a small fraction of the network each year, resulting in delayed discovery of defects and costly repairs. To address these limitations, this paper proposes a scalable and low-power fault detection system that integrates embedded machine vision and Tiny Machine Learning (TinyML) on resource-constrained microcontrollers. The system uses transfer learning to train a lightweight TinyML model for defect classification using a dataset of sewer pipe images and deploys the model on battery-powered devices. Each device captures images inside the pipe, performs on-device inference to detect cracks, intrusions, debris, and other anomalies, and communicates inference results over a long-range LoRa radio link. The experimental results demonstrate that the proposed system achieves 94% detection accuracy with sub-hundred-millisecond inference time and operates for extended periods on battery power. The research contributes a template for autonomous, scalable, and cost-effective sewer condition assessment that can help municipalities prioritize maintenance and prevent catastrophic failures. Full article

Exosomes, which are extracellular vesicles measuring 30–150 nm, are becoming a promising new target from cellular debris classification to a recognized biomarker with the potential to transform diagnostics. They have a fundamental role in intercellular communication, with selective molecular cargo that can reflect the pathophysiological state of parent cells. Exosomes are particularly advantageous for non-invasive liquid biopsies, as they provide continuous monitoring of disease progression or response to treatment. We detail the most recent diagnostic proteins, nucleic acids, and lipids in the context of different diseases. Here, we show the potential of exosomes as non-invasive biomarkers across diverse diseases, which may transcend the sensitivity of conventional biomarkers. The potential of exosome-based liquid biopsies to transform clinical laboratory practice will be determined by their ability to overcome challenges. Limitations comprise preanalytical variability, absence of standardized protocols, and heterogeneity in exosome isolation, which limit their diagnostic potential. The implementation is limited by isolation and analytical processes; however, many advanced platforms may offer multiplexed detection, which is accelerating their implementation process in clinical laboratories. Finally, we provide an overview of the clinical applications and preclinical advancements of exosomes to provide a perspective on the significance of exosomes for their use in biomarker study, as well as therapeutic monitoring in different diseases. Future initiatives must emphasize coordinated validation, economical scalability, and incorporation into clinical workflows to fulfill the potential of exosomes as advanced diagnostics. Full article

Windborne debris generated during tornadoes and hurricanes plays a critical role in building damage. This damage occurs either through direct impact on structural and nonstructural components or indirectly by increasing internal pressure when debris penetrates openings (e.g., windows and doors) or creates new ones. These breaches can significantly raise internal pressure, even at lower wind speeds compared to debris-free conditions. Current provisions in ASCE 7, the nationally adopted standard for wind load calculations in the United States, account for factors such as building geometry, location, and exposure category. However, they do not consider the effects of windborne debris on internal pressure coefficients. This study proposes an enhancement to ASCE 7 by incorporating debris effects through the use of a more conservative enclosure classification. Real-world damage observations from three tornado-impacted residential buildings are presented, followed by a failure mechanism analysis, supporting analytical fragility data, and numerical simulations of debris effects on building damage. The findings suggest that treating buildings as Partially Enclosed under ASCE 7 can more accurately reflect debris-induced internal pressures and improve building resilience under extreme wind events. Full article

Interpreting chemical analysis results to identify ignitable liquid (IL) residues in fire debris samples is challenging, owing to the complex chemical composition of ILs and the diverse sample matrices. This work investigated a transfer learning approach with convolutional neural networks (CNNs), pre-trained for image recognition, to classify gas chromatography and mass spectrometry (GC/MS) data transformed into scalogram images. A small data set containing neat gasoline samples with diluted concentrations and burned Nylon carpets with varying weights was prepared to retrain six CNNs: GoogLeNet, AlexNet, SqueezeNet, VGG-16, ResNet-50, and Inception-v3. The classification tasks involved two classes: “positive of gasoline” and “negative of gasoline.” The results demonstrated that the CNNs performed very well in predicting the trained class data. When predicting untrained intra-laboratory class data, GoogLeNet had the highest accuracy (0.98 ± 0.01), precision (1.00 ± 0.01), sensitivity (0.97 ± 0.01), and specificity (1.00 ± 0.00). When predicting untrained inter-laboratory class data, GoogLeNet exhibited a sensitivity of 1.00 ± 0.00, while ResNet-50 achieved 0.94 ± 0.01 for neat gasoline. For simulated fire debris samples, both models attained sensitivities of 0.86 ± 0.02 and 0.89 ± 0.02, respectively. The new deep transfer learning approach enables automated pattern recognition in GC/MS data, facilitates high-throughput forensic analysis, and improves consistency in interpretation across various laboratories, making it a valuable tool for fire debris analysis. Full article

Landslides pose persistent threats to mountainous regions in Taiwan, particularly in areas such as Nanfeng Village, Nantou County, where steep terrain and concentrated rainfall contribute to chronic slope instability. This study investigates spatiotemporal patterns of vegetation change as a proxy for identifying potential landslide-prone zones, with a focus on the Tung-An tribal settlement in the eastern part of the village. Using high-resolution satellite imagery from SPOT 6/7 (2013–2023) and Pléiades (2019–2023), we derived annual NDVI layers to monitor vegetation dynamics across the landscape. Long-term vegetation trends were evaluated using the Mann–Kendall test, while spatiotemporal clustering was assessed through Emerging Hot Spot Analysis (EHSA) based on the Getis-Ord Gi* statistic within a space-time cube framework. The results revealed statistically significant NDVI increases in many valley-bottom and mid-slope regions, particularly where natural regeneration or reduced disturbance occurred. However, other valley-bottom zones—especially those affected by recurring debris flows—still exhibited declining or persistently low vegetation. In contrast, persistent low or declining NDVI values were observed along steep slopes and debris-flow-prone channels, such as the Nanshan and Mei Creeks. These zones consistently overlapped with known landslide paths and cold spot clusters, confirming their ecological vulnerability and geomorphic risk. This study demonstrates that integrating NDVI trend analysis with spatiotemporal hot spot classification provides a robust, scalable approach for identifying slope hazard areas in data-scarce mountainous regions. The methodology offers practical insights for ecological monitoring, early warning systems, and disaster risk management in Taiwan and other typhoon-affected environments. By highlighting specific locations where vegetation decline aligns with landslide risk, the findings can guide local authorities in prioritizing slope stabilization, habitat conservation, and land-use planning. Such targeted actions support the Sustainable Development Goals, particularly SDG 11 (Sustainable Cities and Communities), SDG 13 (Climate Action), and SDG 15 (Life on Land), by reducing disaster risk, enhancing community resilience, and promoting the long-term sustainability of mountain ecosystems. Full article

Debris flows are rapid mass movements of water-laden debris that flow down mountainsides into valley channels and eventually settle on valley floors. The risk of debris flows can be significantly increased after wildfires. Following the destructive 2021 debris flows in Glenwood Canyon, the Colorado Geological Survey (CGS) initiated a LiDAR-Based Alluvial Fan Mapping Project to improve geologic hazard delineation of alluvial and high-angle fans in response to developing wildfire-ready watersheds. These landforms, shaped by episodic sediment-laden flows, pose significant risks and are often misrepresented on conventional geologic maps. CGS delineated fan-shaped landforms with improved precision using 1-m resolution LiDAR-based DEMs, DEM-derived terrain metrics, hydrologic analysis, and geospatial analysis tools within the ArcGIS Pro platform. Our results reveal previously unmapped or misclassified alluvial or high-angle fans in areas undergoing increasing development pressure, where low-gradient terrain indicates a high hazard potential. Through this study, over 3200 alluvial and high-angle fan polygons were delineated across six Colorado counties, encompassing approximately 81 km² of alluvial fans and 54 km² of high-angle fans. High-resolution LiDAR data, geospatial analytical techniques, and systematic QA/QC protocols were used to support refined hazard awareness. The resulting dataset enhances proactive land-use planning and wildfire resilience by identifying areas prone to debris flow and flood hazards. These maps are intended for regional screening and planning purposes and are not intended for site-specific design. These maps also serve as a critical resource for prioritizing geologic evaluations and guiding mitigation planning across Colorado’s wildfire-affected landscapes. Full article

Simulated data created in silico using a previously reported method were sampled by bootstrapping to generate data sets for training multiple copies of an ensemble learner (i.e., a machine learning (ML) method). The posterior probabilities of class membership obtained by applying the ensemble of ML models to previously unseen validation data were fitted to a beta distribution. The shape parameters for the fitted distribution were used to calculate the subjective opinion of sample membership into one of two mutually exclusive classes. The subjective opinion consists of belief, disbelief and uncertainty masses. A subjective opinion for each validation sample allows identification of high-uncertainty predictions. The projected probabilities of the validation opinions were used to calculate log-likelihood ratio scores and generate receiver operating characteristic (ROC) curves from which an opinion-supported decision can be made. Three very different ML models, linear discriminant analysis (LDA), random forest (RF), and support vector machines (SVM) were applied to the two-state classification problem in the analysis of forensic fire debris samples. For each ML method, a set of 100 ML models was trained on data sets bootstrapped from 60,000 in silico samples. The impact of training data set size on opinion uncertainty and ROC area under the curve (AUC) were studied. The median uncertainty for the validation data was smallest for LDA ML and largest for the SVM ML. The median uncertainty continually decreased as the size of the training data set increased for all ML.The AUC for ROC curves based on projected probabilities was largest for the RF model and smallest for the LDA method. The ROC AUC was statistically unchanged for LDA at training data sets exceeding 200 samples; however, the AUC increased with increasing sample size for the RF and SVM methods. The SVM method, the slowest to train, was limited to a maximum of 20,000 training samples. All three ML methods showed increasing performance when the validation data was limited to higher ignitable liquid contributions. An ensemble of 100 RF ML models, each trained on 60,000 in silico samples, performed the best with a median uncertainty of 1.39x${10}^{-2}$ and ROC AUC of 0.849 for all validation samples. Full article

The quantification of solid debris in used lubricating oil is essential for assessing transmission system wear and optimizing maintenance strategies. This study introduces a low-cost capacitive proximity sensor for monitoring total solid particle contamination in lubricants, with a focus on ferrous (Fe), non-ferrous (Al), and non-metallic (SiO₂) debris. Controlled tests were performed using five mixing ratios of large-to-small particles (100:0, 75:25, 50:50, 25:75, and 0:100) at a fixed debris mass of 0.5 g per 25 mL of SAE 85W-140 automotive gear oil. Cubic regression analysis yielded high predictive accuracy, with average R² values of 0.994 for Fe, 0.943 for Al, and 0.992 for SiO₂. Further dimensionality reduction using Principal Component Analysis (PCA), along with Linear Discriminant Analysis (LDA) of multivariate statistical analysis, effectively classifies debris types and enhances interpretability. These results demonstrate the potential of capacitive sensing as an offline, non-invasive alternative to traditional techniques for wear debris monitoring in transmission systems. These results confirm the potential of capacitive sensing, supported by statistical modeling, as a non-invasive, cost-effective technique for offline classification and monitoring of wear debris in transmission systems. Full article

Located on the northern side of the China–Pakistan Highway in the Pamir Plateau, Laqi Gully represents a typical rainfall–meltwater coupled debris flow gully. During 2020–2024, seven debris flow events occurred in this area, four of which disrupted traffic and posed significant threats to the China–Pakistan Economic Corridor (CPEC). The hazard assessment of debris flows constitutes a crucial component in disaster prevention and mitigation. However, current research presents two critical limitations: traditional models primarily focus on single precipitation-driven debris flows, while low-resolution digital elevation models (DEMs) inadequately characterize the topographic features of alpine narrow valleys. Addressing these issues, this study employed GF-7 satellite stereo image pairs to construct a 1 m resolution DEM and systematically simulated debris flow propagation processes under 10–100-year recurrence intervals using a coupled rainfall–meltwater model. The results show the following: (1) The mudslide develops rapidly in the gully section, and the flow velocity decays when it reaches the highway. (2) At highway cross-sections, maximum velocities corresponding to 10-, 20-, 50-, and 100-year recurrence intervals measure 2.57 m/s, 2.75 m/s, 3.02 m/s, and 3.36 m/s, respectively, with maximum flow depths of 1.56 m, 1.78 m, 2.06 m, and 2.52 m. (3) Based on the hazard classification model of mudslide intensity and return period, the high-, medium-, and low-hazard sections along the highway were 58.65 m, 27.36 m, and 24.1 m, respectively. This research establishes a novel hazard assessment methodology for rainfall–meltwater coupled debris flows in narrow valleys, providing technical support for debris flow mitigation along the CPEC. The outcomes demonstrate significant practical value for advancing infrastructure sustainability under the United Nations Sustainable Development Goals (SDGs). Full article

Suyukou Gully, located on the eastern slope of the Helan Mountains in northwest China, is a typical debris-flow-prone catchment characterized by a steep terrain, fractured bedrock, and abundant loose colluvial material. The area is subject to intense short-duration convective rainfall events, which often trigger destructive debris flows that threaten the Suyukou Scenic Area. To investigate the dynamics and risks associated with such events, this study employed the FLO-2D two-dimensional numerical model to simulate debris flow propagation, deposition, and hazard distribution under four rainfall return periods (10-, 20-, 50-, and 100-year scenarios). The modeling framework integrated high-resolution digital elevation data (original 5 m DEM resampled to 20 m grid), land-use classification, rainfall design intensities derived from regional storm atlases, and detailed field-based sediment characterization. Rheological and hydraulic parameters, including Manning’s roughness coefficient, yield stress, dynamic viscosity, and volume concentration, were calibrated using post-event geomorphic surveys and empirical formulations. The model was validated against field-observed deposition limits and flow depths, achieving a spatial accuracy within 350 m. Results show that the debris flow mobility and hazard intensity increased significantly with rainfall magnitude. Under the 100-year scenario, the peak discharge reached 1195.88 m³/s, with a maximum flow depth of 20.15 m and velocities exceeding 8.85 m·s⁻¹, while the runout distance surpassed 5.1 km. Hazard zoning based on the depth–velocity (H × V) product indicated that over 76% of the affected area falls within the high-hazard zone. A vulnerability assessment incorporated exposure factors such as tourism infrastructure and population density, and a matrix-based risk classification revealed that 2.4% of the area is classified as high-risk, while 74.3% lies within the moderate-risk category. This study also proposed mitigation strategies, including structural measures (e.g., check dams and channel straightening) and non-structural approaches (e.g., early warning systems and land-use regulation). Overall, the research demonstrates the effectiveness of physically based modeling combined with field observations and a GIS analysis in understanding debris flow hazards and supports informed risk management and disaster preparedness in mountainous tourist regions. Full article

Electric discharge machining (EDM) is widely employed for machining hard, conductive materials. Tool rotation has emerged as an effective strategy to enhance debris flushing and improve stability during deep-hole EDM drilling. This study proposes a machine learning-based approach to evaluate the influence of tool rotation and predict the unstable machining conditions in EDM of ultrafine grained tungsten carbide. A structured analytical workflow, combining Taguchi–Grey optimization, regression analysis, and classification models, was designed to capture discharge dynamics across macro- and micro-timescales. Classification models trained on raw and processed electrical signal features achieved over 88% accuracy and 90% recall. SHAP analysis revealed that the relationship between key discharge events such as sparks and short circuits varied significantly across stable and unstable machining phases, underscoring the importance of phase-specific modeling. While correlation analysis showed weak global associations, phase-dependent SHAP values revealed opposing feature effects, allowing the context-informed interpretation of model behavior. Phase segmentation revealed that, compared to 1000 RPM, short circuits were reduced by about 40% during stable machining at 8000–9000 RPM. Conversely, during unstable phases, spark effectiveness dropped by nearly 45%, and secondary discharges increased throughout this range. These insights support the design of adaptive control strategies that adjust the rotation rate in response to detected phase changes, aiming to sustain machining stability. The findings support the development of dynamic control frameworks to improve EDM performance, particularly for mold fabrication using tungsten carbide. Full article

The accurate characterization of wear debris is crucial for assessing the health of rotating engine components and for conducting simulation experiments in debris detection. This study proposed an intelligent recognition method for ferrography wear debris images, leveraging several improved Mask Region-based Convolutional Neural Network (Mask R-CNN) algorithms to quantitatively calculate both the number of debris particles and their coverage areas. The improvement on the Mask R-CNN focuses on two key aspects: enhancing feature extraction through the feature pyramid network structure and integrating attention mechanisms. The most suitable attention mechanism for wear debris detection was determined through ablation experiments. The improved Mask R-CNN combined with the Convolutional Block Attention Module achieves the best Mean Pixel Accuracy of 87.63% at a processing speed of 7.6 frames per second, demonstrating its high accuracy and efficiency in wear particle segmentation. Furthermore, the quantitative and qualitative analysis of wear debris, including the number and area of debris particles and their classification, provides valuable insights into the severity of wear. These insights are essential for understanding the extent of wear damage and guiding maintenance decisions. Full article

Deadwood is a vital component of forest ecosystems, significantly contributing to biodiversity and carbon storage. Accurate mapping of deadwood is essential for ecological monitoring and sustainable forest management. This study introduces a method for downed deadwood mapping using a convolutional neural network (CNN) applied to very high-resolution UAV RGB imagery. The research was conducted in Hainich National Park, central Germany, aiming to enhance the precision of coarse woody debris (CWD) delineation in a dense and structurally diverse temperate deciduous forest. Key objectives included testing the deep learning (DL) model’s performance at area, length, and object levels and benchmarking its accuracy against a traditional object-based image analysis (OBIA) method. Deadwood volume was calculated from the mapping results. By implementing a U-Net architecture with a ResNet-34 backbone and utilizing data augmentation techniques, the model achieved very high classification performance (F1-scores between 73% and 96%). It provided precise delineation of individual CWD objects from the underlying ground, representing detailed stem forms. High precision values highlight the reliability of the mapping results, while lower recall values indicate that some CWD objects, especially smaller branches, were missed. The DL approach achieved higher accuracy values across all testing methods compared to the OBIA method. The study also addresses the challenges posed by spectral ambiguities in decomposed deadwood and recommends future research directions for enhancing model generalization across diverse forest types and acquisition conditions. Full article

Natural flood management strategies are increasingly recognized as sustainable alternatives to conventional engineered flood control measures. Among these, leaky dams, also known as woody debris dams or log dams, have emerged as effective nature-based solutions for mitigating flood risks while preserving essential ecosystem services. This review traces the historical evolution of leaky dams from ancient water management practices to contemporary applications, highlighting their development and adaptation over time. It presents a comparative examination of leaky dams and conventional flood control structures, outlining their respective strengths and limitations across ecological, hydrological, and economic dimensions. The review also introduces a conceptual classification of leaky dams into naturally occurring, engineered, hybrid, and movable systems, showing how each form aligns with varying catchment characteristics and management objectives. By synthesizing foundational knowledge and strategic insights, this paper establishes a theoretical and contextual framework for understanding leaky dams as distinct yet complementary tools in integrated flood management, laying the groundwork for further technical evaluations. The findings offer valuable insights for end users by highlighting the potential of leaky dams as integral components of sustainable flood management systems, elucidating their roles in mitigating flood risks, enhancing water retention, and supporting ecosystem resilience. Full article