Search Results (112)

Large-scale forest ecological restoration is commonly expected to improve habitat quality and promote population growth of forest-dependent herbivores. Yet, whether vegetation recovery facilitates or constrains herbivore growth and habitat use at local scales within nature reserves remains unclear, as vegetation recovery and canopy closure might alter forage availability and lead to ecological mismatch between vegetation features and population dynamic. Here, we used the endangered species South China sika deer as the study species, and its dominant distribution region—Qingliangfeng Biosphere Reserve—as the study area. Using decadal camera-trapping data (2015–2024) and extracted vegetation and other environmental variables, we quantified decadal trends in sika deer activity intensity and interannual variation in vegetation (leaf area index, LAI, and normalized difference vegetation index, NDVI). We incorporated topographic and anthropogenic disturbance variables and applied generalized linear mixed models and generalized linear models to analyze its habitat use. We found that: (1) Numbers of independent photographs and the relative abundance index of sika deer increased significantly and consistently from 2015 to 2024. (2) LAI exhibited substantial interannual variability without a stable trend. In contrast, segmented regression identified a clear temporal breakpoint in NDVI, with a significant increasing trend before 2021 followed by a pronounced decline thereafter. (3) In all years, distance to settlement had a significant and negative effect on activity intensity, whereas distance to road, elevation, and year had significant positive effects. LAI and NDVI showed negative and weak effects on sika deer activity intensity. In specific years, LAI had a significantly negative effect in early periods whereas NDVI became significantly negative in mid and late periods. Other environmental variables exhibited interannual heterogeneity. Our findings demonstrate that vegetation recovery within the reserve does not automatically improve habitats for forest-dependent herbivores and could lead to a potential ecological mismatch. Full article

Robotic manipulators were initially introduced to replace repetitive human labor and have since evolved to perform complex tasks in dynamic environments. In such systems, imitation learning and reinforcement learning models capable of real-time trajectory generation are widely applied. Among these approaches, imitation learning enables rapid training when high-quality datasets are available. However, it suffers from high costs associated with collecting expert demonstration data and significant performance variability depending on data quality. Recently, learning approaches utilizing large-scale datasets have been explored, but they often struggle to guarantee reliable performance in tasks requiring precise control and incur substantial computational costs for model construction, limiting their applicability as a general-purpose learning strategy. To address these limitations, this paper proposes an imitation learning framework that integrates sampling-based motion planning with a hybrid data curation strategy. The proposed method employs a sampling-based planner (e.g., RRT*) to generate diverse physically feasible trajectories, thereby reducing the cost of acquiring expert demonstration data. The generated trajectories are then curated through clustering-based grouping and rule-based filtering to select high-quality training samples from large-scale datasets. The proposed framework automatically generates physically feasible trajectories while selecting high-quality data from large trajectory pools, thereby improving training stability and reducing data-related costs. Experimental results demonstrate that the proposed method achieves an average success rate of 79.1% (95% CI: 74.3–83.2%) and produces trajectories with shorter trajectories, lower final distances, and reduced joint movements compared to conventional filtering methods. Full article

Purpose: The aim is to evaluate the effectiveness of artificial intelligence (AI)-based automatic segmentation as a predictive tool for clinical peer review in prostate cancer patients treated with hypofractionated radiotherapy. Methodology: A retrospective analysis was conducted on 62 patients treated across three Italian centers between 2020 and 2025. CT images were segmented using software based on 3D U-net models. Three workflows were compared: manual segmentation (C man), automatic segmentation (C AI), and AI-based segmentation adjusted by clinicians (C adj). Quantitative metrics used for comparison included the Dice Similarity Coefficient (DSC) and Hausdorff Distance (HDmax). Statistical analysis involved Welch’s t-test and Cohen’s d for effect size. Results: The results showed a significant improvement in agreement between C AI and C adj compared to C man. Median DSC for CTV increased from 0.80 (C man) to 0.92 (C adj), while HDmax decreased from 12.33 mm to 9.22 mm. Similar improvements were observed for the bladder and anorectum. All differences were statistically significant (p < 0.0001), with large effect sizes (Cohen’s d > 0.8). Discussion: AI use demonstrated a reduction in interobserver variability and segmentation time, enhancing workflow standardization. The C adj workflow, where the physician acts as a reviewer of AI-generated contours, proved effective and potentially integrable into clinical peer review. The predictive peer review refers to a preliminary support step in the clinical review process rather than a substitute for medical decision-making. Full article

Background: Cardiovascular magnetic resonance (CMR) is the clinical gold standard for assessing cardiac anatomy and function. However, the manual segmentation of cardiac structures and myocardial infarction (MI) is time-consuming, prone to inter-observer variability, and often depends on contrast-enhanced imaging. Although deep learning (DL) has enabled substantial automation, challenges remain in generalizability, particularly for MI detection from non-contrast cine CMR. Objective: This study proposes a comprehensive DL-based framework for automatic segmentation of cardiac structures and myocardial infarction using contrast-free cine CMR. Methods: The framework integrates multiple convolutional neural network (CNN) architectures for cardiac structure segmentation with an attention-based deep learning model for MI localization. Post-processing refinement using stacked autoencoders and active contour modeling is applied to improve anatomical consistency. Segmentation performance is evaluated using overlap-based and boundary-based metrics, including the Dice Similarity Coefficient (DSC), Mean Contour Distance (MCD), and Hausdorff Distance (HD). Results: The best-performing model achieved Dice scores of 0.93 ± 0.05 for the left ventricular (LV) cavity, 0.89 ± 0.04 for the LV myocardium, and 0.91 ± 0.06 for the right ventricular (RV) cavity, with consistently low boundary errors across all structures. Myocardial infarction segmentation achieved a Dice score of 0.80 ± 0.02 with high recall, demonstrating reliable infarct localization without the use of contrast agents. Conclusions: By enabling accurate cardiac structure and myocardial infarction segmentation from contrast-free cine CMR, the proposed framework supports broader clinical applicability, particularly for patients with contraindications to gadolinium-based contrast agents and in emergency or resource-limited settings. This approach facilitates scalable, contrast-independent cardiac assessment. Full article

The acoustic technique emerges as a highly promising, cutting-edge solution that can be effectively employed for extinguishing flames in locations where the access to classical fire-protection measures is limited, the available extinguishing agent is severely restricted, or the burning materials are difficult to suppress using currently known methods. The results of the experimental attempts confirmed that low-frequency acoustic waves containing higher even harmonics from the tenth to the sixteenth order (inclusive) can successfully extinguish flames, demonstrating both the feasibility and the novelty of the acoustic technique for fire protection. Moreover, statistical analysis was applied to identify operational boundary values and assess their variability, supporting the optimal selection of system parameters required for rapid and effective flame extinguishing. By integrating an acoustic extinguisher with optional intelligent sensors, including artificial vision, it becomes possible to rapidly detect flames at much greater distances than with conventional smoke and temperature sensors, as well as to automatically extinguish them. In this context, an integrated solution combining acoustic waves with an artificial intelligence module (smart sensor) may be employed for comprehensive fire management, encompassing both fire detection and flame extinguishing. Full article

Accurate classification of precipitation phase (liquid, mixed, or solid) is essential in high mountain environments, where rapid changes in elevation can lead to abrupt phase transitions over short distances, significantly affecting hydro-meteorological, ecological, and socio-economic activities. However, most existing classification schemes have not been evaluated over long periods using real observational data, but mainly through simulations. This study addresses this gap by introducing a new methodology based on X-band polarimetric radar and by validating it against real precipitation events over an extended time period. The machine learning model is trained and tested using a four-year dataset including X-band radar, Micro Rain Radar, disdrometer, and temperature profile data from the Grenoble region (French Alps). To improve the classification accuracy, three temperature profile sources were tested: lapse rates obtained from automatic weather stations, interpolation of the temperature profile from the freezing level detected by the Micro Rain Radar, and temperature profiles from the operational AROME model forecast. Three different phase classification schemes were tested: two existing schemes based on fuzzy-logic, and the new method based on random forest. Results show that the random forest method, trained with radar polarimetric variables, AROME temperature profiles, and target labels derived from Micro Rain Radar observations, achieves the highest accuracy. Despite the overall good classification results, limitations persist in identifying mixed-phase precipitation due to its transitional nature and vertical variability. Feature importance analysis indicates that temperature is the most influential variable in the classification scheme, followed by reflectivity factor measured in the horizontal plane ($Z_e)$ and differential reflectivity ($Z_{dr})$. This methodology demonstrates the potential of combining machine learning techniques with multi-instrument observations to improve hydrometeor classification in complex terrain. The approach offers valuable insights for operational forecasting, water resource management, and climate impact assessments in mountainous regions. Full article

The article presents the analytic method for real-time detection of the stationary state of a vehicle based on information retrieved from 6 DoF IMU sensor. Reliable detection of stillness is essential for the application of resetting the inertial sensor’s output bias, called Zero Velocity Update method. It is obvious that the signal from the strapped on inertial sensor differs while the vehicle is stationary or moving. Effort was then made to find a computational method that would automatically discriminate between both states with possibly small impact on the vehicle embedded controller. An algorithmic step-by-step method for building, optimizing, and implementing a diagnostic system that detects the vehicle’s stationary state was developed. The proposed method adopts the “Mahalanobis Distance” quantity widely used in industrial quality assurance systems. The method transforms (fuses) information from multiple diagnostic variables (including linear accelerations and angular velocities) into one scalar variable, expressing the degree of deviation in the robot’s current state from the stationary state. Then, the method was implemented and tested in the dead reckoning navigation system of an autonomous wheeled mobile robot. The method correctly classified nearly 93% of all stationary states of the robot and obtained only less than 0.3% wrong states. Full article

Accurate phase association and arrival time picking are pivotal for reliable microseismic event location and source characterization. However, the complexity of downhole microseismic wavefields, arising from heterogeneous subsurface structures, variable propagation paths, and ambient noise, poses significant challenges to conventional automatic picking methods, even when the signal-to-noise ratio (SNR) is moderate to high. Specifically, P-wave coda energy can obscure S-wave onsets analysis, and shear wave splitting can generate ambiguous arrivals. In this study, we propose a novel multi-channel arrival picking framework based on Constrained Dynamic Time Warping (CDTW) for phase identification and simultaneous P- and S-wave arrival estimation. The DTW algorithm aligns microseismic signals that may be out of sync due to differences in timing or wave velocity by warping the time axis to minimize cumulative distance. Time delay constraints are imposed to ensure physically plausible alignments and improve computational efficiency. Furthermore, we introduce a Multivariate CDTW approach to jointly process the three-component (3C) data, leveraging inter-component and inter-receiver arrival consistency across the entire downhole array. The method is validated against the Short-Term Average/Long-Term Average (STA/LTA) and waveform cross-correlation techniques using field data from a shale gas hydraulic fracturing. Results demonstrate that the proposed algorithm significantly enhances arrival time accuracy and inter-receiver consistency, particularly in scenarios involving P-wave coda interference and shear wave splitting. Full article

A promising approach to advancing sustainable urban mobility is the increased use of light electric vehicles, such as e-cycles and their cargo-carrying variants: e-cargo cycles. These micromobility vehicles fall between e-cycles and conventional vehicles in terms of transport capacity, range, and cost. A key advantage of e-cargo cycles over their non-electrified counterparts is the electric powertrain, which enables them to carry heavier payloads, travel longer distances, and reduce driver fatigue. Since the primary use of e-cargo cycles is urban parchment deliveries, trip efficiency plays a critical role in their effectiveness within urban logistics. This efficiency is influenced by factors such as travel distance, traffic density, and the weight and volume of the delivery payload. While higher delivery capacity generally enhances efficiency, studies have shown that as the drop size increases, the efficiency of e-cargo cycle delivery trips tends to decline. A practical way to address this limitation is the use of cargo attachments, such as trailers. These micromobility solutions are already widely implemented globally and significantly enhance transport capacity. This paper reports the process of designing and testing the control algorithm of an electrical system for an experimental attachment demonstrator that can be used to convert most cycle vehicles into cargo variants. The system integrates two 250 W BLDC hub motors, two 576 Wh lithium-ion batteries, dual load-cell sensing in the coupling element, and an STM32-based controller to provide independent propulsion and synchronization with the leading cycle. The force-based control strategy enables automatic adaptation to varying payloads typically encountered in urban logistics, which is supported by the variable storage volume capable of transporting payloads of up to 200 kg. Full article

Background: Precise delineation of the rectum is crucial in treatment planning for cancers in the pelvic region, such as prostate and cervical cancers. Manual segmentation is also still time-consuming and suffers from inter-observer variability. Since there are meaningful differences in rectal anatomy between males and females, incorporating sex-specific anatomical patterns can be used to enhance the performance of segmentations. Furthermore, recent deep learning advancements have provided promising solutions for automatically classifying patient sex from CT scans and leveraging this information for enhancing the accuracy of rectal segmentation. However, their clinical utility requires comprehensive validation against real-world standards. Methods: In this study, a two-stage deep learning pipeline was developed using CT scans from 186 patients with either prostate or cervical cancer. First, a CNN model automatically classified the patient’s biological sex from CT images in order to capture anatomical variations dependent on sex. Second, a sex-aware U-Net model performed automated rectal segmentation, allowing the network to adjust its feature representation based on the anatomical differences identified in stage one. The internal validation had an 80/20 train–test split, and 15% of the training portion was held out for validation to ensure balanced distribution regarding sex and diagnosis. Model performance was evaluated using spatial similarity metrics, including the Dice Similarity Coefficient (DSC), Hausdorff Distance, and Average Surface Distance. Additionally, a radiation oncologist conducted a retrospective clinical evaluation using a 3-point Likert scale. Statistical significance was examined using Wilcoxon signed-rank tests, Welch’s t-tests, and Mann–Whitney U test. Results: The sex-classification model attained an accuracy of 94.6% (AUC = 0.98, 95% CI: 0.96–0.99). Incorporation of predicted sex into the segmentation pipeline improved anatomical consistency of U-Net outputs. Mean DSC values were 0.91 (95% CI: 0.89–0.92) for prostate cases and 0.89 (95% CI: 0.87–0.91) for cervical cases, with no significant difference between groups (p = 0.12). Surface distance metrics calculated on resampled isotropic voxels showed mean HD values of 3.4 ± 0.8 mm and ASD of 1.2 ± 0.3 mm, consistent with clinically acceptable accuracy. On clinical evaluation, 89.2% of contours were rated as excellent, while 9.1% required only minor adjustments. Automated segmentation reduced the average contouring time from 12.7 ± 2.3 min manually to 4.3 ± 0.9 min. Conclusions: The proposed sex-aware deep learning framework offers accurate, robust segmentation of the rectum in pelvic CT imaging by explicitly modeling sex-specific differences in anatomical characteristics. This physiologically informed approach enhances segmentation performance and supports reliable integration of AI-based delineation into radiotherapy workflows to improve both contouring efficiency and clinical consistency. Full article

The efficient pre-incubation selection of duck eggs is essential to ensuring stable hatchability, but most existing optoelectronic and machine vision systems have been calibrated for chicken eggs and cannot be directly used for duck eggs because of their larger size, stronger reflectivity and wider morphological variability. This study proposes an optoelectronic method specifically adapted to Adigel duck eggs that combines load cell weighing, infrared distance sensing and dual-projection image processing in a single stationary setup. A total of 300 eggs were measured manually and automatically, and the results were statistically compared. The developed algorithm uses adaptive Gaussian thresholding (51 × 51, C = 2) and a median 5 × 5 filter to stabilize contour extraction on glossy and spotted shells, followed by ellipsoid-based volume estimation with a breed-specific correction factor (K_v = 0.637). The automatic system showed high agreement with manual measurements (r > 0.95 for mass and linear dimensions) and a mean relative error below 2%. Density, the shape index (If) and the shape coefficient (K₁) were computed to classify eggs into “suitable”, “borderline” and “unsuitable” categories. A preliminary incubation trial (n = 60) of eggs classified as “suitable” resulted in 92% hatchability, thus confirming the predictive value of the proposed criteria. Unlike chicken-oriented systems, the presented solution provides breed-specific calibration and can be implemented in small and medium hatcheries for the reproducible, non-destructive sorting of hatching duck eggs. Full article

This study presents the development and validation of an in situ monitoring method for the laser direct energy deposition (DED) process, utilizing an integrated optical camera (720 HD, 60 fps) to analyze melt pool imagery. The approach is grounded in an experimental framework employing Taguchi orthogonal arrays, which ensures a stable dataset by controlling process variability and enabling reliable extraction of relevant features. The monitoring system focuses on analyzing brightness distribution regions within the melt pool image, identified as specific clusters that reflect external process conditions. The method emphasizes precise segmentation of the melt pool area, combined with automatic detection and classification of cluster features associated with key process parameters—such as focus distance, the number of deposited layers, powder feed rate, and scanning speed. The main contribution of this work is demonstrating the effectiveness of using an optical camera for DED monitoring, based on an algorithm that processes a set of melt pool identification features through computer vision and machine learning techniques, including Random Forest and HistGradient Boosting, achieving classification accuracies exceeding 95%. By continuously tracking the evolution of these features within a closed-loop control system, the process can be maintained in a stable, defect-free state, effectively preventing the formation of common process defects. Full article

Aiming at the problems of low production efficiency and high manual dependence in the heat-sealing process of the pillow packaging machine, the existing optimization methods of process parameters were improved, and an intelligent decision-making model of the longitudinal sealing heat-sealing process based on the radial basis function neural network and orthogonal least square method was proposed to realize the efficient and accurate optimization of heat-sealing process parameters. By analyzing the fracture yield strength of the composite, the target heat-sealing strength range was determined. The heat-sealing temperature, heat-sealing speed, and heat-sealing plate distance were selected as key process variables, and the actual production data were used to train the model to accurately construct the nonlinear mapping relationship between heat-sealing process parameters and heat-sealing strength. On this basis, the genetic algorithm optimization framework with the model predictive output as the fitness function is designed to realize the rapid search of the optimal combination of process parameters. The optimization results were introduced into the pillow packaging machine for a verification test. The measured heat-sealing strength was stable within the target range. The maximum error of the optimization group was less than 10%, and the average error was less than 5%, which was significantly better than the effect of manual experience. The experimental results show that the proposed method can effectively improve the efficiency and consistency of process optimization under the premise of ensuring the quality of heat-sealing, meet the requirements of automatic production for high precision and low consumption of the heat-sealing process optimization, and realize the comprehensive improvement of efficiency, accuracy, and intelligent level in the longitudinal heat-sealing process of the pillow packaging machine. Full article

The segmentation and classification of color are crucial stages in image processing, computer vision, and pattern recognition, as they significantly impact the results. The diverse, hand-labeled datasets in the literature are applied for monochromatic or color segmentation in specific domains. On the other hand, synthetic datasets are generated using statistics, artificial intelligence algorithms, or generative artificial intelligence (AI). This last one includes Large Language Models (LLMs), Generative Adversarial Neural Networks (GANs), and Variational Autoencoders (VAEs), among others. In this work, we propose VitralColor-12, a synthetic dataset for color classification and segmentation, comprising twelve colors: black, blue, brown, cyan, gray, green, orange, pink, purple, red, white, and yellow. VitralColor-12 addresses the limitations of color segmentation and classification datasets by leveraging the capabilities of LLMs, including adaptability, variability, copyright-free content, and lower-cost data—properties that are desirable in image datasets. VitralColor-12 includes pixel-level classification and segmentation maps. This makes the dataset broadly applicable and highly variable for a range of computer vision applications. VitralColor-12 utilizes GPT-5 and DALL·E 3 for generating stained-glass images. These images simplify the annotation process, since stained-glass images have isolated colors with distinct boundaries within the steel structure, which provide easy regions to label with a single color per region. Once we obtain the images, we use at least one hand-labeled centroid per color to automatically cluster all pixels based on Euclidean distance and morphological operations, including erosion and dilation. This process enables us to automatically label a classification dataset and generate segmentation maps. Our dataset comprises 910 images, organized into 70 generated images and 12 pixel segmentation maps—one for each color—which include 9,509,524 labeled pixels, 1,794,758 of which are unique. These annotated pixels are represented by RGB, HSL, CIELAB, and YCbCr values, enabling a detailed color analysis. Moreover, VitralColor-12 offers features that address gaps in public resources such as violin diagrams with the frequency of colors across images, histograms of channels per color, 3D color maps, descriptive statistics, and standardized metrics, such as ΔE76, ΔE94, and CIELAB Chromacity, which prove the distribution, applicability, and realistic perceptual structures, including warm, neutral, and cold colors, as well as the high contrast between black and white colors, offering meaningful perceptual clusters, reinforcing its utility for color segmentation and classification. Full article

Fish stocks and their management are paramount for sustainable fisheries under the ongoing changes in atmosphere–sea interactions. The Aegean Sea, one of the composite seas influenced by different water masses, is characterized by a diverse ecosystem. Small pelagic fish are abundant and tend to form schools that vary in size. One of the most efficient and rapid techniques for sampling fish schools over a large area is the use of acoustic methods. Therefore, an acoustic survey was conducted in the coastal areas along the entire Turkish Aegean waters between June and August 2024, using a scientific quantitative echosounder equipped with a split-beam transducer operating at 206 kHz. During the survey, environmental parameters, including water physics, optics, and bathymetry, were measured at 321 stations. Additionally, satellite data were used to obtain water primary production levels for each sampling month across the entire study area. Using a custom computer algorithm written during the present study in MATLAB (2021a), fish schools were automatically detected to measure various morphological and acoustic features. Through a series of statistical analyses, three optimal clusters, validated with the total silhouette sum of distances (1317.38), were identified, each characterized by specific morphological, acoustic, and environmental variables associated with different areas of the study. School morphology and acoustic properties also varied with bottom depth. Cluster 1 was mostly found in open and relatively deep waters. Cluster 2 appeared in areas impacted by anthropogenic sources. Principal Component Analysis (PCA) revealed that the first component (PCA1) was correlated with school height from the bottom (HFB) and overall school height (SH), followed by minimum depth (MnD), maximum depth (MxD), and volume backscattering strength at the school edge (SvE). The second component (PCA2) was associated with school width (SW) and area (A). Cluster 1 was characterized by schools with large SW and A, and relatively high HFB and SH. Cluster 2 showed low HFB and SH, while Cluster 3 had high MnD and MxD and low SvE. Based on the descriptors for these clusters, each cluster could be attributed to fish species at different life stages inferred based on target strength (TS), namely sardine, horse mackerel, and chub mackerel, distributed along the entire Turkish Aegean coast. Full article