Sensors

34 pages, 3147 KB

Open AccessArticle

Multi-Objective Optimization for Data Center HVAC Systems Based on Edge–Cloud Collaborative Deep Reinforcement Learning

by Shichao Huang, Yibing Zhou and Yuan Liu

Sensors 2026, 26(16), 5031; https://doi.org/10.3390/s26165031 - 7 Aug 2026

The sustained growth of cloud computing and AI training workloads drives data center expansion. Optimizing their control is therefore critical for reducing operational costs. Edge real-time control is indispensable for guaranteeing thermal safety, data sovereignty, and offline availability. Yet deploying Deep Reinforcement Learning [...] Read more.

The sustained growth of cloud computing and AI training workloads drives data center expansion. Optimizing their control is therefore critical for reducing operational costs. Edge real-time control is indispensable for guaranteeing thermal safety, data sovereignty, and offline availability. Yet deploying Deep Reinforcement Learning (DRL) in production Heating, Ventilation, and Air Conditioning (HVAC) environments confronts cold-start risks, edge–cloud computational asymmetry, and multi-objective conflicts spanning energy efficiency, electricity cost, and thermal safety. To address these challenges, this paper proposes an edge-cloud collaborative physics-informed reinforcement learning framework for production data center HVAC control. The framework integrates a physics-informed cold-start solution using Adaptive Particle Swarm Optimization (APSO) to generate physically constrained initial policies on a gray-box digital twin without expert demonstration data, a three-time-scale edge–cloud architecture coordinating minute-level edge Soft Actor-Critic (SAC) real-time inference, weekly edge APSO online model identification, daily cloud Non-dominated Sorting Genetic Algorithm III (NSGA-III) thermal storage scheduling, and a constraint-aware safe projection layer that embeds thermal safety hard constraints directly into the neural network policy. The framework is validated through a seven-month production deployment spanning the complete summer-to-winter transition, comprising approximately

3.2

million sensor records and evaluated with rigorous statistical methods. Full article

(This article belongs to the Special Issue Edge Computing for Beyond 5G and Wireless Sensor Networks)

22 pages, 4292 KB

Open AccessReview

Sensor-Based Tracking and Localization of In-Line Inspection Tools in Oil and Gas Pipelines: A Review

by Jianfeng Zheng, Bingfeng Ju and Anyu Sun

Sensors 2026, 26(16), 5030; https://doi.org/10.3390/s26165030 - 7 Aug 2026

Abstract

Accurate tracking and localization of in-line inspection (ILI) tools are essential for mileage calibration, defect mapping, and blockage prevention in oil and gas pipelines. This review summarizes sensor-based approaches for external ILI-tool localization, emphasizing how sensing physics, deployment geometry, and signal interpretation determine [...] Read more.

Accurate tracking and localization of in-line inspection (ILI) tools are essential for mileage calibration, defect mapping, and blockage prevention in oil and gas pipelines. This review summarizes sensor-based approaches for external ILI-tool localization, emphasizing how sensing physics, deployment geometry, and signal interpretation determine practical performance. Extremely low-frequency (ELF) magnetic tracking is first examined through dipole modeling, sensor evolution, and weak-signal recovery under steel-pipe and soil shielding. Distributed fiber-optic sensing is then reviewed as a continuous-tracking alternative, with attention to fading mitigation, spatiotemporal denoising, and trajectory extraction from distributed acoustic sensing data. Acoustic arrays and hybrid schemes are discussed as complementary options for subsea or cable-free environments. Finally, the review assesses how data fusion and lightweight artificial intelligence (AI) can improve robustness while noting unresolved issues in field data availability, edge computing, and uncertainty quantification. The synthesis indicates that next-generation ILI tracking should combine heterogeneous sensing, physics-aware signal processing, and deployment-aware model design rather than rely on a single high-sensitivity sensor. Full article

(This article belongs to the Section Industrial Sensors)

34 pages, 5768 KB

Open AccessArticle

A TinyMLOps Pipeline for Coarse-Grained Plant Disease Classification in Precision Agriculture

by Hossein Aqasizade, Mattia Antonini, Massimo Vecchio and Fabio Antonelli

Sensors 2026, 26(16), 5029; https://doi.org/10.3390/s26165029 - 7 Aug 2026

Abstract

Identifying plant health conditions is an emerging precision-agriculture and food-security challenge, intensified by deploying deep-learning models on memory- and power-constrained edge devices. We present a TinyMLOps pipeline spanning model design, optimization, quantization, and deployment across diverse edge devices, evaluated under controlled laboratory conditions. [...] Read more.

Identifying plant health conditions is an emerging precision-agriculture and food-security challenge, intensified by deploying deep-learning models on memory- and power-constrained edge devices. We present a TinyMLOps pipeline spanning model design, optimization, quantization, and deployment across diverse edge devices, evaluated under controlled laboratory conditions. Using a dataset derived from the PlantVillage benchmark, 39 fine-grained classes are aggregated into three superclasses: healthy leaf, unhealthy leaf, and no leaves. The resulting system therefore performs plant health-status classification and background filtering rather than diagnosing specific diseases. We train a MobileNet-based convolutional neural network jointly optimized for classification accuracy and computational efficiency, adopting state-of-the-art hyperparameter optimization (HPO) tools. Five models are selected, four from the Pareto Front and one as the biggest evaluated model during HPO, converted to LiteRT and ONNX, and evaluated at float32 and post-training int8 precision on a Raspberry Pi Zero 2 W and an STM32H743ZI microcontroller. At float32, LiteRT is 1.87–2.65× faster than ONNX Runtime on the Raspberry Pi across all five models. Relative to their float32 LiteRT counterparts, the int8 LiteRT models are 2.83–3.67× smaller on disk and 21.3–31.2% faster on the same board, at a cost in F1-score of between 0.0010 and 0.0068. On the microcontroller, only the two smallest models deploy at both precisions; for these, the fully quantized int8 variants are

4.3 \times

faster and 3.85× smaller in MCU flash footprint than the float32 counterparts. The mid-range model fits the 2 MB flash and 1 MB RAM budget only when quantized, while the two largest models exceed it in every configuration tested. Full article

(This article belongs to the Special Issue Machine Learning and Sensors Technology in Agriculture: 2nd Edition)

32 pages, 16745 KB

Open AccessArticle

Study on the Intelligent Recognition Algorithm for Open-Pit Mine Slope Fissures: Crack-YOLO with Texture and Semantic Enhancement

by Hongze Zhao, Hong Wei, Wei Liu, Haiyu Jia and Changbin He

Sensors 2026, 26(16), 5028; https://doi.org/10.3390/s26165028 - 7 Aug 2026

Abstract

Rock fissure parameters, such as length, width, and density, are essential for analyzing the progressive instability of open-pit mine slopes. Under the combined effects of engineering disturbance, geological conditions, and environmental factors, slope fissures continuously propagate and evolve. However, large variations in fissure [...] Read more.

Rock fissure parameters, such as length, width, and density, are essential for analyzing the progressive instability of open-pit mine slopes. Under the combined effects of engineering disturbance, geological conditions, and environmental factors, slope fissures continuously propagate and evolve. However, large variations in fissure scale, complex rock-surface textures, blurred boundaries, and weak micro-fissure features increase the difficulty of intelligent fissure segmentation, identification, and parameter extraction. Consequently, many mining enterprises still rely on manual interpretation, which is time-consuming and susceptible to subjective errors. To address these challenges, this study develops Crack-YOLO, a task-oriented fissure detection and instance-segmentation model based on YOLOv8-Seg. A total of 500 original UAV images were collected from multiple open-pit mines and processed to construct a dataset containing 3600 fissure image patches, including 3240 images for training and 360 images for testing. In Crack-YOLO, selected C2f modules are replaced with contextual semantic enhancement modules (CoT Blocks), and a texture information enhancement module (SM Block) is incorporated to strengthen contextual semantic representation and fine-grained texture-feature extraction. The model achieved segmentation precision, recall, mAP50, and mAP50:95 values of 0.896, 0.787, 0.854, and 0.392, respectively. For object detection, the corresponding values were 0.968, 0.862, 0.959, and 0.773, respectively. The segmentation results were further processed using K3M skeleton extraction and physical-scale calibration to quantitatively extract geometric parameters, including fissure length, equivalent average width, and azimuth. Validation using an image containing seven representative fissures yielded mean absolute errors of 0.016 m, 0.010 m, and 0.90° for fissure length, equivalent average width, and azimuth, respectively, indicating the feasibility of the proposed parameter-quantification workflow. In an application test conducted in a typical open-pit mine scene, the proposed workflow identified 196 fissures within approximately 22 s and quantitatively analyzed their geometric parameters and distribution characteristics. The results indicate that the proposed method has potential for fissure identification and geometric-parameter quantification in open-pit mine slopes and may provide quantitative data support for slope-fissure monitoring and stability analysis. Full article

(This article belongs to the Special Issue Defect Detection Based on Vision Sensors)

23 pages, 4052 KB

Open AccessSystematic Review

Development Boards and Microcontroller Platforms in Sports and Physical Activity: A Systematic Review

by Boryi A. Becerra-Patiño, Miguel Andrés Peñuela-Arguello, Cristián David Zapata-Piratova, Aura D. Montenegro-Bonilla, Rodrigo Yáñez-Sepúlveda, José Francisco López-Gil and José Pino-Ortega

Sensors 2026, 26(16), 5027; https://doi.org/10.3390/s26165027 - 7 Aug 2026

Abstract

Background: The development of data processing platforms makes it possible to monitor human behavior outside the laboratory, which facilitates decision-making regarding physical activity, health, and training. Objective: Analyze the available scientific evidence on data logging in the field of sports and physical [...] Read more.

Background: The development of data processing platforms makes it possible to monitor human behavior outside the laboratory, which facilitates decision-making regarding physical activity, health, and training. Objective: Analyze the available scientific evidence on data logging in the field of sports and physical activity using development boards and microcontroller platforms. Materials and Methods: A systematic review was conducted following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines. The selection and inclusion of studies in this review were based on the inclusion and exclusion criteria derived from the participants, interventions, outcomes (PIO) strategy. Methodological quality was assessed using the Methodological Index for Non-Randomized Studies (MINORS). Results: After all the information screening was completed, 23 documents met the eligibility criteria. The studies addressed a wide range of sports practices, including swimming, athletics, skiing, cycling, badminton, tennis, strength training, weightlifting, rehabilitation, outdoor route monitoring, and female contact-sport scenarios such as rugby, resulting in the implementation of microcontrollers in heterogeneous settings. Conclusions: For sports and health professionals, microcontroller-based systems can represent an effective way to implement objective and individualized monitoring, especially in high-performance contexts. For research, it is recommended that future studies focus on homogeneous comparisons with reference standards, larger and more diverse samples (including women), and longitudinal evaluations in real-world training and competition scenarios, so that these solutions evolve from functional prototypes into robust and transferable tools. Full article

(This article belongs to the Special Issue Sensing Functional Imaging Biomarkers and Artificial Intelligence)

► Show Figures

Figure 1

46 pages, 18762 KB

Open AccessArticle

A Physics-Informed Benchmarking Framework for Machine Learning and Tree-Based Ensembles in IIoT-Enabled Predictive Maintenance

by Yi-Kai Su and Chun-Jan Tseng

Sensors 2026, 26(16), 5026; https://doi.org/10.3390/s26165026 - 7 Aug 2026

Abstract

Reliable Predictive Maintenance (PdM) in Industrial Internet of Things (IIoT) environments is challenged by severe class imbalance, heterogeneous sensor variables, inconsistent experimental protocols, and deployment constraints. This study proposes a Physics-Informed Benchmarking Framework that integrates engineering-guided feature construction, Mutual Information (MI)-based feature relevance [...] Read more.

Reliable Predictive Maintenance (PdM) in Industrial Internet of Things (IIoT) environments is challenged by severe class imbalance, heterogeneous sensor variables, inconsistent experimental protocols, and deployment constraints. This study proposes a Physics-Informed Benchmarking Framework that integrates engineering-guided feature construction, Mutual Information (MI)-based feature relevance analysis, standardized model development, and deployment-oriented evaluation within a unified and reproducible workflow. Using the AI4I 2020 Predictive Maintenance Dataset, Logistic Regression, Isolation Forest, Random Forest, and Extreme Gradient Boosting (XGBoost) were evaluated using identical feature representations, train–test partitions, preprocessing procedures, and imbalance-handling strategies. The engineered feature space incorporates thermal, mechanical, interaction, and degradation-related information derived from the original sensor measurements. The results show that tree-based ensembles provide the strongest overall performance under severe class imbalance. Random Forest achieved an accuracy of 0.986, an F1-score of 0.722, and a ROC-AUC of 0.983, providing the best balance between failure detection and false-alarm control. XGBoost achieved an accuracy of 0.978, a recall of 0.853, and the lowest inference latency of 0.35 ms, indicating its suitability for latency-sensitive IIoT deployment. These findings demonstrate that combining engineering-guided feature representation with a standardized evaluation protocol enables fair comparison of representative learning paradigms while preserving engineering interpretability and deployment relevance. Full article

(This article belongs to the Special Issue Knowledge-Informed Machine Learning for Sensor-Driven Decision Making in Manufacturing)

► Show Figures

Figure 1

37 pages, 1221 KB

Open AccessArticle

A Reproducible Benchmark-Validity Audit and Calibration Study for Cross-Home Fault Diagnosis in Smart-Home Sensor Systems

by Norkobil Saydirasulovich Saydirasulov, Abror Shavkatovich Buriboev, Shuxrat Isroilov, Ryumduck Oh, Shavkat Buribayev, Abbos Abduvaytov, Jamshid Umirov, Jasur Ismailovich Badalov, Aziza Axmedova, Cheolwon Lee and Heung Seok Jeon

Sensors 2026, 26(16), 5025; https://doi.org/10.3390/s26165025 - 7 Aug 2026

Abstract

Diagnosing faults across different smart homes is hard: sensor names, layouts, and daily routines differ from home to home, so a model trained in one home rarely works in another. We study an ontology-guided framework for cross-home fault diagnosis, but our main contribution [...] Read more.

Diagnosing faults across different smart homes is hard: sensor names, layouts, and daily routines differ from home to home, so a model trained in one home rarely works in another. We study an ontology-guided framework for cross-home fault diagnosis, but our main contribution is a benchmark-validity audit—a systematic check of whether the datasets used to evaluate such systems actually measure fault detection. Using the public Center for Advanced Studies in Adaptive Systems (CASAS) smart-home datasets (homes hh101–hh110) and real household power data (HomeC, UMass Smart*), we show that much of the high cross-home accuracy reported on these benchmarks is an artifact of features that re-encode the labelling rules rather than evidence of transfer: when those features are removed, the macro-averaged F1 score (macro-F1) collapses toward the level obtained with randomly permuted labels. We therefore treat these datasets as semantic-transfer and benchmark-validity studies, not fault-detection results. The framework’s distinguishing component is a counterfactual calibration layer that returns a probability for its recommended intervention; on a controlled structural causal model with known interventions, it achieves a Brier skill score of

0.369

for intervention-success probabilities. Separately, on the simulation-derived LBNL Fan Coil Unit benchmark, a conventional gradient-boosted multiclass fault classifier achieves accuracy comparable to a random forest but about six times lower expected calibration error (

0.026

vs.

0.159

) under a scenario-matched split. This calibration advantage does not generalize to held-out simulation scenarios, where the calibration error rises to

0.372

; we report this negative result as a limitation. We are explicit about scope: the ontology reasoner and the real-stream causal graph are only partially implemented, and the counterfactual recommendations are validated only under controlled or simulated conditions, not in deployed homes. The results are intended for researchers who build or benchmark sensor-based fault-diagnosis models, for dataset curators, and for practitioners who need calibrated rather than merely accurate outputs. All code, the proxy-label rules, and the leakage audit are released. Full article

(This article belongs to the Special Issue Intelligent Sensors for Condition Monitoring, Diagnosis, and Prognostics, 2nd Edition)

14 pages, 795 KB

Open AccessArticle

Data-Efficient Unsupervised Recalibration of Calorimeter Sensor Arrays Using Wasserstein Adversarial Learning

by Saraa Ali, Vladimir Bocharnikov, Fedor Ratnikov, Mikhail Hushchyn, Artem Ryzhikov and Denis Derkach

Sensors 2026, 26(16), 5024; https://doi.org/10.3390/s26165024 - 7 Aug 2026

Abstract

Large distributed sensor arrays require repeated recalibration as radiation damage, material aging, gain variation, and readout drift alter channel responses. We studied a high-granularity calorimeter as a large sensor array and addressed unsupervised recalibration from two unpaired datasets: a nominal reference response and [...] Read more.

Large distributed sensor arrays require repeated recalibration as radiation damage, material aging, gain variation, and readout drift alter channel responses. We studied a high-granularity calorimeter as a large sensor array and addressed unsupervised recalibration from two unpaired datasets: a nominal reference response and an aged response with attenuated cell-wise signals. Aging was modeled by a deterministic sensor-wise base field with reading-level stochastic variation; the base coefficients were used only for post-training evaluation. We evaluated a Wasserstein adversarial calibration field against an evaluation-only global-mean coefficient predictor and two non-adversarial estimators, a per-cell mean-energy ratio and independent per-cell Wasserstein matching. The adversarial objective supplied an adaptive event-level discrepancy over the full sensor array. In a hierarchical data-efficiency study with four reference/aged sampling-seed combinations and three configured-seed adversarial fits per seed combination, the adversarial method achieved an RMSE from

0.0238 \pm 0.0014

to

0.0173 \pm 0.0012

and a positive

R^{2}

from

0.828 \pm 0.022

to

0.910 \pm 0.013

across the tested event counts. Its RMSE was also below the approximately

0.0579

global-mean reference at every event count, demonstrating the recovery of cell-wise coefficient variation beyond the global mean. The mean-energy-ratio and Wasserstein-only estimators remained below the

R^{2} = 0

reference in this sparse benchmark. Full article

(This article belongs to the Special Issue Intelligent Sensor Calibration: Techniques, Devices and Methodologies)

23 pages, 1628 KB

Open AccessArticle

UAV-Based Thermal Inversion for Canopy Temperature Retrieval and Precision Irrigation

by Haoming Li, Wei Li, Chenchen Liu, Leilei Ji, Zhenbo Liu and Ramesh K. Agarwal

Sensors 2026, 26(16), 5023; https://doi.org/10.3390/s26165023 - 7 Aug 2026

Abstract

Accurate assessment of crop water status is critical for precision irrigation and sustainable water management in agriculture. This study develops a UAV-based thermal infrared inversion framework for high-resolution canopy temperature retrieval and irrigation decision support in tea plantations. The proposed approach integrates multi-frame [...] Read more.

Accurate assessment of crop water status is critical for precision irrigation and sustainable water management in agriculture. This study develops a UAV-based thermal infrared inversion framework for high-resolution canopy temperature retrieval and irrigation decision support in tea plantations. The proposed approach integrates multi-frame image mosaicking, threshold-based canopy extraction, and a gray–temperature calibration model to generate spatially continuous canopy temperature maps. Crop water stress was quantified using the Crop Water Stress Index (CWSI), and its reliability was further evaluated by analyzing its relationship with stomatal conductance. The framework further estimates soil moisture status and irrigation requirements based on a threshold-based irrigation strategy. The results show that the linear gray-temperature calibration model achieved a maximum absolute error of less than 0.3 °C and that the calculated CWSI and estimated irrigation requirement were strongly correlated with measured stomatal conductance, with R² up to 0.91. The proposed method provides a practical technical workflow from UAV thermal imagery acquisition to canopy temperature retrieval and quantitative irrigation decision-making, demonstrating its potential for precision irrigation management in tea plantations. Full article

(This article belongs to the Special Issue AI UAV-Based Systems for Agricultural Monitoring)

31 pages, 6877 KB

Open AccessArticle

Design, Fabrication, and Testing of a 3D-Printed Model Rocket with Integrated Telemetry Systems

by Philippos G. Moschidis, Petros S. Bithas and Florian Meyer

Sensors 2026, 26(16), 5022; https://doi.org/10.3390/s26165022 - 7 Aug 2026

Abstract

This study presents the design, fabrication, and experimental validation of the Hermes reusable model rocket platform integrating additive manufacturing, onboard sensing, and telemetry capabilities for low-cost aerospace experimentation. The rocket was manufactured using modular Polyethylene Terephthalate Glycol (PETG) components produced through fused filament [...] Read more.

This study presents the design, fabrication, and experimental validation of the Hermes reusable model rocket platform integrating additive manufacturing, onboard sensing, and telemetry capabilities for low-cost aerospace experimentation. The rocket was manufactured using modular Polyethylene Terephthalate Glycol (PETG) components produced through fused filament fabrication to achieve a lightweight and structurally robust configuration suitable for repeated flight operations. A custom flight computer based on a Raspberry Pi Zero 2W was developed to acquire in-flight data from an inertial measurement unit, barometric pressure sensor, and Global Positioning System module, while an onboard camera enabled post-flight trajectory assessment. Aerodynamic performance and stability were evaluated using OpenRocket simulations, and propulsion was provided by a cluster of Klima D9-5 solid rocket motors. Four experimental flights were conducted to evaluate the integrated system architecture, assess telemetry and sensor performance, and compare experimental flight data with simulation predictions. The recorded measurements successfully captured the primary flight phases, including launch, ascent, apogee, descent, and recovery. The experimental results showed qualitative agreement with the simulated flight profiles; however, deviations in apogee altitude, acceleration, and flight duration were observed due to aerodynamic drag, environmental disturbances, motor-performance variability, and implementation-related limitations. The flight campaigns additionally identified practical challenges associated with wireless telemetry reliability, GPS signal acquisition, electronic protection, and parachute deployment, leading to iterative system improvements. From a sensing perspective, the flight campaigns demonstrate the operation and limitations of a low-cost embedded acquisition architecture under dynamic conditions, including the effects of sampling rate, sensor calibration, synchronization, wireless-link interruption, and local data preservation on the quality of the recorded flight measurements. The presented platform demonstrates the feasibility of combining low-cost additive manufacturing techniques with commercially available embedded electronics for reusable aerospace testing and educational applications. The proposed system further provides a flexible experimental framework for flight-data acquisition, simulation validation, and iterative development in academic and amateur rocketry research. Full article

(This article belongs to the Section Remote Sensors)

► Show Figures

Figure 1

32 pages, 6439 KB

Open AccessFeature PaperArticle

Automatic Segmentation of Ischaemic Stroke Lesions Using Transformers and Convolutional Neural Networks Applied to Multimodal Neuroimaging

by Pablo Martínez Cegarra, Juan Francisco Zapata Pérez and Juan Martínez-Alajarín

Sensors 2026, 26(16), 5021; https://doi.org/10.3390/s26165021 - 7 Aug 2026

Abstract

Ischaemic stroke constitutes a leading cause of global disability. Rapid extraction of the infarct core from multimodal computed tomography perfusion (CTP) imaging guides reperfusion therapy and clinical decision-making. Deep learning algorithms automate this delineation, yet hospital translation is hindered by high-dimensional data, inter-scanner [...] Read more.

Ischaemic stroke constitutes a leading cause of global disability. Rapid extraction of the infarct core from multimodal computed tomography perfusion (CTP) imaging guides reperfusion therapy and clinical decision-making. Deep learning algorithms automate this delineation, yet hospital translation is hindered by high-dimensional data, inter-scanner variability, and the low contrast of early ischaemia. Architectural comparisons in the literature frequently carry methodological biases originating from disparate preprocessing protocols and data partitions. This study reduces these variables by evaluating three segmentation strategies under a shared preprocessing pipeline and an identical data partition using the ISLES 2024 dataset. Three models were trained on the same 133-patient partition using a shared preprocessing pipeline based on morphological skull-stripping and modality-specific clinical intensity ranges. The data, preprocessing, and partitions are held constant across models, while framework-dependent factors (optimiser, patch size, physical field of view, spatial resampling, augmentation policy, and model capacity) remain coupled to each architecture and are therefore treated as part of the compared strategy rather than as fully isolated variables. The first of these is a single-stage 5-channel nnU-Net, followed by a two-stage cascaded nnU-Net (2 and 7 channels) and a lightweight Transformer (SegFormer3D). Evaluation on a fixed 15-patient held-out test set isolated the architectural performance. The cascade model achieved the highest Dice Similarity Coefficient (0.224). The single-stage nnU-Net provided the most precise volumetric estimation, recording an Absolute Volume Difference (AVD) of 23.70 mL and a lesion-wise F1-score of 7.60%. On the other hand, SegFormer3D returned the lowest overall metrics (DSC 0.163, AVD 27.28 mL, F1 2.30%). In the small held-out cohort, paired statistical testing did not reveal significant differences between models, so the reported orderings describe the present dataset and experimental configuration rather than a general architectural law. Within these limits, the local inductive bias of the convolutional models retained an empirical advantage over the single Transformer evaluated when processing this moderately sized neuroimaging dataset, and complex cascade topologies offered only marginal gains compared with a well-calibrated single-stage network. Although the predictive segmentation of infarcted tissue at acute stages still demands computational improvements, these results suggest that preprocessing quality is at least as decisive for clinical impact as increasing the complexity of neural architectures. Full article

(This article belongs to the Special Issue Advanced Medical Sensing and Intelligent Processing for Precision Medicine and Clinical Decision Support)

24 pages, 3995 KB

Open AccessArticle

A Specimen-Separated Machine Learning Benchmark Toward Real-Time Tissue-Type Identification in Guided Surgery Using Ex Vivo Bovine Laser-Induced Breakdown Spectroscopy

by René Fernando Sosa-Santos, José Luis Arce-Diego and Félix Fanjul-Vélez

Sensors 2026, 26(16), 5020; https://doi.org/10.3390/s26165020 - 7 Aug 2026

Abstract

Real-time tissue identification during laser-guided surgery is a critical unmet need for collateral damage avoidance and margin delineation. Laser-Induced Breakdown Spectroscopy (LIBS) is compatible with pulsed laser surgical systems and offers rapid, label-free elemental analysis. This study presents a machine learning pipeline classifying [...] Read more.

Real-time tissue identification during laser-guided surgery is a critical unmet need for collateral damage avoidance and margin delineation. Laser-Induced Breakdown Spectroscopy (LIBS) is compatible with pulsed laser surgical systems and offers rapid, label-free elemental analysis. This study presents a machine learning pipeline classifying five ex vivo bovine tissue classes, plus one synthetic null-signal control class, from LIBS spectra, designed to control specimen-level data leakage and class imbalance bias. Key contributions are (i) a ‘peak max over baseline’ aggregation strategy suppressing shot noise while preserving emission peaks; (ii) a repeated, group-based cross-validation protocol (GroupShuffleSplit, N = 10) enforcing specimen-level separation; and (iii) a comparison of 30 configurations (10 classifiers × 3 pipelines). Extra Trees with normalization reached the highest weighted F1-score (0.934 ± 0.118); excluding the synthetic control, five-class scores fall to 0.875–0.915 and the ranking changes, so these are the reference figures for biological tissue discrimination. Support Vector Machines were less accurate but more consistent (0.917 ± 0.069). Acquisition takes approximately 3 s per point; inference is sub-millisecond. With five source animals, the best configuration chosen on the same outer splits, and inner tuning that was not group-aware, these estimates are an exploratory step toward real-time guided surgery. Full article

(This article belongs to the Section Biomedical Sensors)

► Show Figures

Figure 1

28 pages, 4152 KB

Open AccessArticle

Design Considerations, Field Validation and Perspectives of a Low-Power Wearable Sensor Collar for Continuous Monitoring of Ruminants

by Maria P. Nikolopoulou, Aikaterini-Artemis Agiomavriti, Dimitrios Loukatos, Dimitrios Grivas, Nikolaos Xotzempekoglou, Athanasios I. Gelasakis, Konstantinos G. Arvanitis, Konstantinos Demestichas and Thomas Bartzanas

Sensors 2026, 26(16), 5019; https://doi.org/10.3390/s26165019 - 7 Aug 2026

Abstract

Wearable sensors provide significant potential for continuous animal monitoring. However, the practical implementation of such technologies on livestock farms for animal monitoring raises significant challenges related to power efficiency, robustness, and reliability. In this research paper, we propose the design, implementation, and field [...] Read more.

Wearable sensors provide significant potential for continuous animal monitoring. However, the practical implementation of such technologies on livestock farms for animal monitoring raises significant challenges related to power efficiency, robustness, and reliability. In this research paper, we propose the design, implementation, and field testing of a low-energy, low-cost, multi-sensor wearable collar specifically designed for continuous monitoring of ruminants using LoRa. The proposed collar is based on a modular hardware platform that incorporates inertial sensing, temperature sensing, and wireless communication, with special attention to sensor choice, location on the body, casing, and mounting mechanism. Reduced weight, environmental protection, and long-term wearability without affecting animal behavior are the primary focus in the hardware design. Power optimization management strategies, including sleep mode and duty cycling functionality, are implemented to maximize autonomy and battery lifetime and are evaluated under realistic operating scenarios. Field deployment was conducted in a ruminant farm, where the wearable devices operated flawlessly for a long time period. Characteristic sensor data are collected, including accelerometer readings induced by animal movement, variability in received signal strength (RSSI), and animal temperature. The system demonstrates stable operation, satisfactory data completeness and consistency on multi-day basis. The knowledge acquired brings into focus the practical challenges and design issues associated with the use of wearable sensors in livestock and offers insights into designing efficient sensor collars for precision livestock farming. Full article

(This article belongs to the Special Issue Advanced Sensing and Electronic Measurement Systems for Animal Sciences and Wildlife Monitoring)

► Show Figures

Figure 1

13 pages, 6605 KB

Open AccessArticle

Assessing Neuromuscular Adaptation and Kinetic Symmetry in Elite Para-Kayakers: A Wearable Body Sensor Network Case Series

by Giovanni Saggio, Leonardo Pierucci and Luca Pietrosanti

Sensors 2026, 26(16), 5018; https://doi.org/10.3390/s26165018 - 7 Aug 2026

Abstract

Movement analysis is essential for high-level athletes to perform at their best. Such analysis necessarily involves studying the characteristics of one’s own body, whose adaptive processes are crucial for understanding how each athlete responds to sudden changes. In this preliminary exploratory work, the [...] Read more.

Movement analysis is essential for high-level athletes to perform at their best. Such analysis necessarily involves studying the characteristics of one’s own body, whose adaptive processes are crucial for understanding how each athlete responds to sudden changes. In this preliminary exploratory work, the paddling techniques of a limited, heterogeneous cohort of eight athletes were evaluated: four who were non-impaired and four who showed different classes of impairment (two KL3, one KL2, and one KL1). Each non-impaired athlete underwent three test series—one with a footrest, one without a footrest, and another with a modified footrest. KL3 athletes performed two tests, one with the prosthetic leg and one without, while KL2 and KL1 athletes completed a standard sprint. Measurements were obtained using a set of inertial sensors. The results showed a tendency toward greater symmetry in shoulder trajectories when the footrest was removed, and athletes improved their performance when sprinting under a modified configuration. These findings provide insight into how the body compensates for the absence of leg support, including changes in muscle recruitment and a tendency toward the better control of overall movement. Given the inherent constraints of the small sample size (n = 8), these results demonstrate the potential of IMU networks for personalized biomechanical assessment, providing descriptive, case-level insights rather than statistically generalizable trends. Full article

(This article belongs to the Section Intelligent Sensors)

► Show Figures

Figure 1

24 pages, 5179 KB

Open AccessArticle

Software-Only Registration and Cross-Spectral Classification of Unsynchronized RGB–LWIR Video: A Multisensor Benchmark for Conveyor-Based Waste Sorting

by Burak Akdemir and Seniha Esen Yuksel

Sensors 2026, 26(16), 5017; https://doi.org/10.3390/s26165017 - 7 Aug 2026

Abstract

Reliable multisensor perception is a key requirement for practical waste sorting, yet many low-cost sensor configurations cannot rely on hardware synchronization or carefully controlled acquisition. We present a pilot-scale multisensor waste-sorting testbed that combines an unsynchronized RGB camera with a long-wave infrared (LWIR) [...] Read more.

Reliable multisensor perception is a key requirement for practical waste sorting, yet many low-cost sensor configurations cannot rely on hardware synchronization or carefully controlled acquisition. We present a pilot-scale multisensor waste-sorting testbed that combines an unsynchronized RGB camera with a long-wave infrared (LWIR) camera for object classification on a continuously moving conveyor, and introduce ThermalRGBTrash, a new paired RGB–LWIR video dataset for this task. To enable fusion under asynchronous acquisition, we develop a fully software-based registration pipeline that combines SuperPoint–SuperGlue matching with an adaptive sliding-window strategy designed to recover from long-wave infrared sensor artifacts, including non-uniformity correction events. Across 281,439 matched frame pairs from 19 paired videos, the registration pipeline achieves a mean spatial alignment error of 2.27 pixels and matches 99.98% of attempted frame pairs. We then detect and segment objects with Mask R-CNN, track them across the conveyor, and classify each tracklet using frozen DINOv2 self-supervised Vision Transformer (ViT-L/14) features with a lightweight multilayer perceptron head. RGB and LWIR representations are combined through late fusion. On 550 tracklets under video-disjoint 10-fold cross-validation, the fused pipeline reaches a macro F1 score of 0.924, outperforming RGB alone (0.886) and LWIR alone (0.856). On a mixed-class test set of 351 tracklets reserved exclusively for final evaluation, fusion reaches a macro F1 score of 0.947. The fusion advantage persists across multiple backbone and pretraining choices, while ablation studies support the chosen temporal sampling and pooling design. These results show that accurate RGB–LWIR object classification is achievable without synchronization hardware, and establish ThermalRGBTrash as a benchmark for future work on practical multisensor perception in conveyor-based waste sorting. Full article

(This article belongs to the Special Issue Multisensor Image and Video Processing: Methods and Applications)

► Show Figures

Figure 1

20 pages, 8101 KB

Open AccessArticle

High-Resolution Forward-Looking Imaging Method for FMCW Radar Based on Sparse Sampling

by Qin Zhao, Xiaopeng Yan, Tao Zhang, Qingyu Hou, Qiang Liu, Jiawei Wang and Xinwei Wang

Sensors 2026, 26(16), 5016; https://doi.org/10.3390/s26165016 - 7 Aug 2026

Abstract

Platform-induced synthetic aperture is an effective approach to enhancing azimuth resolution in forward-looking radar imaging. However, for small platforms such as automobiles and unmanned aerial vehicles, the large volume of echo data required under continuous sampling, combined with the presence of Doppler ambiguity, [...] Read more.

Platform-induced synthetic aperture is an effective approach to enhancing azimuth resolution in forward-looking radar imaging. However, for small platforms such as automobiles and unmanned aerial vehicles, the large volume of echo data required under continuous sampling, combined with the presence of Doppler ambiguity, poses substantial challenges for high-resolution imaging. To address these issues, this paper proposes a forward-looking FMCW radar imaging method based on sparse sampling intervals. A uniform linear array is first employed to acquire measurements at different platform positions, and an initial range-angle image is obtained for each channel. Adaptive beamforming is then applied to impose nulls on false-alarm regions, including grating lobes and left-right ambiguity. Finally, coherent accumulation across channels yields a high-resolution range-angle image. Simulation and experimental results demonstrate that the proposed method achieves high-resolution forward-looking imaging while significantly reducing the volume of echo data. Full article

(This article belongs to the Section Sensing and Imaging)

► Show Figures

Figure 1

18 pages, 3326 KB

Open AccessArticle

Cooperative Monostatic and Bistatic Measurements for Low-Altitude UAV ISAC: System Implementation and Channel Characterization

by Nan Ming, Hanwen Xu, Kai Mao, Hanpeng Li, Mingqi Guo, Xiaomin Chen and Qiuming Zhu

Sensors 2026, 26(16), 5015; https://doi.org/10.3390/s26165015 - 7 Aug 2026

Abstract

Low-altitude unmanned aerial vehicle (UAV)-integrated sensing and communication (ISAC) channels are governed by rapidly evolving multipath. These dynamics arise from UAV motion, air–ground geometry, and scene-dependent scatterers, yet field evidence comparing monostatic and bistatic sensing links remains limited. We develop a cooperative monostatic–bistatic [...] Read more.

Low-altitude unmanned aerial vehicle (UAV)-integrated sensing and communication (ISAC) channels are governed by rapidly evolving multipath. These dynamics arise from UAV motion, air–ground geometry, and scene-dependent scatterers, yet field evidence comparing monostatic and bistatic sensing links remains limited. We develop a cooperative monostatic–bistatic measurement system for low-altitude UAV ISAC channel sounding. The system integrates a UAV-borne sensing node, a ground node, synchronized acquisition, and an offline processing chain. It enables UAV-borne monostatic sensing and air–ground bistatic sensing to be measured within the same low-altitude urban scenario. A field measurement campaign is conducted along a representative route containing buildings, trees, roadside facilities, and open ground. From the measured in-phase/quadrature (IQ) data, the system extracts channel impulse responses (CIRs) and power delay profiles (PDPs) as primary measurement products. Delay–Doppler processing, multipath-component extraction, and trajectory analysis are applied to compare link-dependent propagation behavior under matched environmental conditions. The measurement results demonstrate that the proposed platform can jointly capture monostatic and bistatic ISAC channel responses and support controlled and synchronized low-altitude UAV channel measurement experiments. This system provides an experimental basis for UAV ISAC channel modeling, measurement-platform assessment, and subsequent sensing algorithm evaluation. Full article

(This article belongs to the Special Issue Advanced UAV Communication and Sensor Technologies for Electromagnetic Environment Awareness and Channel Optimization)

► Show Figures

Figure 1

12 pages, 1515 KB

Open AccessCommunication

A Ground-Airborne Frequency-Domain Electromagnetic Rapid Imaging Method Based on Scalar Magnetic Field for Eliminating the Influence of Flight Attitude

by Shuxu Liu, Zhongming Li, Yanfu Tang, Hongyu Li, Yongqiang Yang and Junlin Li

Sensors 2026, 26(16), 5014; https://doi.org/10.3390/s26165014 - 7 Aug 2026

Abstract

The ground-airborne frequency-domain electromagnetic (GAFEM) method has the potential to detect underground anomalies at large depth ranges in areas with complex terrain. However, its specific configuration, which involves ground-based transmitting and airborne signal acquisition, inevitably introduces attitude noise into the measured magnetic field [...] Read more.

The ground-airborne frequency-domain electromagnetic (GAFEM) method has the potential to detect underground anomalies at large depth ranges in areas with complex terrain. However, its specific configuration, which involves ground-based transmitting and airborne signal acquisition, inevitably introduces attitude noise into the measured magnetic field vector data. The presence of this attitude noise alters the characteristics of the measured data, consequently compromising the accuracy of imaging methods that rely on vector data. To eliminate the influence of attitude on GAFEM, this study proposes a rapid imaging technique for GAFEM based on the scalar magnetic field. This method utilizes the total magnetic field magnitude at each measurement point along the survey line as the input for imaging parameter calculation, thereby circumventing the effects of attitude noise and enabling high-resolution detection of underground anomalies. This study begins by analyzing the mechanism through which flight attitude affects GAFEM. Using a previously published GAFEM imaging method based on vector magnetic fields, it is demonstrated that flight attitude severely degrades the accuracy of such methods. Furthermore, a new imaging approach is proposed that employs the scalar magnetic field, which is immune to variations in flight attitude. A detailed description of the method’s principles, physical basis, and computational procedures is provided. The feasibility of the method is validated using a synthetic GAFEM model. The results indicate that the proposed method can achieve high-resolution detection of underground anomalies without being affected by attitude noise. Finally, the performance of the method is further tested using a simulation model constructed from actual geological data. The results confirm that the proposed method possesses the capability to identify underground anomalies in a complex model. Full article

(This article belongs to the Special Issue Next-Generation Geophysical Sensing)

► Show Figures

Figure 1

38 pages, 13599 KB

Open AccessArticle

A Unified Evaluation Protocol and Late-Fusion System for Monocular Per-Object Distance Estimation in Indoor Scenes

by Adnan Ali and Yu Jun

Sensors 2026, 26(16), 5013; https://doi.org/10.3390/s26165013 - 7 Aug 2026

Abstract

Monocular per-object distance estimation aims to predict a metric distance for each detected object from a single RGB image. Although monocular dense depth estimation and monocular 3D object detection are well studied, indoor object-level distance estimation remains weakly standardized. In dense depth-based pipelines, [...] Read more.

Monocular per-object distance estimation aims to predict a metric distance for each detected object from a single RGB image. Although monocular dense depth estimation and monocular 3D object detection are well studied, indoor object-level distance estimation remains weakly standardized. In dense depth-based pipelines, object distance is commonly obtained by combining object detection with depth prediction and aggregating depth values within a region of interest. However, existing approaches differ in region selection, detection source, and aggregation strategy, limiting comparability across methods. This paper proposes a unified, deployment-aligned evaluation protocol for dense depth-based pipelines, where both predicted and reference distances are computed within the same predicted bounding box. This formulation standardizes region-of-interest selection, removes dependence on ground-truth boxes at inference time, and enables consistent evaluation across late-fusion methods. The framework integrates a curriculum-trained Depth Anything V2 ViT-S model for metric depth estimation with a YOLO11n detector for object localization. Under the same-box evaluation protocol on SUN RGB-D, the Depth Anything V2 ViT-S backbone achieves object-wise distance estimation accuracy of MAE = 0.1286 m, RMSE = 0.1817 m, AbsRel = 0.0662, and

δ_{1}

= 0.9785 using mean aggregation over valid box depths. Scaling the backbone from ViT-S to ViT-L further improves performance to MAE = 0.1059 m, RMSE = 0.1484 m, AbsRel = 0.0553, and

δ_{1}

= 0.9898, corresponding to approximately 17.7% lower MAE and 18.3% lower RMSE relative to ViT-S, and provides a standardized reference point for indoor object-level distance evaluation under the same-box protocol. Full article

(This article belongs to the Section Sensing and Imaging)

► Show Figures

Figure 1

Journal Description

Sensors

Latest Articles

Journal Menu

Journal Browser

Highly Accessed Articles

Latest Books

E-Mail Alert

News

Topics

Conferences

Special Issues

Topical Collections

Further Information

Guidelines

MDPI Initiatives

Follow MDPI